電子書 Media

Media
Infrastructures
and the Politics
of Digital Time
Essays on
Hardwired
Temporalities
EDITED BY
Ams te rdam AXEL VOLMAR

AND KYLE STINE
Uni ve r sit y
Press
Media Infrastructures and
the Politics of Digital Time
The book series RECURSIONS: THEORIES OF MEDIA, MATERIALITY, AND
CULTURAL TECHNIQUES provides a platform for cuttingedge research in the
field of media culture studies with a particular focus on the cultural impact of
media technology and the materialities of communication. The series aims to
be an internationally significant and exciting opening into emerging ideas in
media theory ranging from media materialism and hardware-oriented studies
to ecology, the post-human, the study of cultural techniques, and recent
contributions to media archaeology. The series revolves around key themes:
– The material underpinning of media theory
– New advances in media archaeology and media philosophy
– Studies in cultural techniques
These themes resonate with some of the most interesting debates in international
media studies, where non-representational thought, the technicity of knowledge
formations and new materialities expressed through biological and technological
developments are changing the vocabularies of cultural theory. The series is also
interested in the mediatic conditions of such theoretical ideas and developing
them as media theory.
Editorial Board
– Jussi Parikka (University of Southampton)
– Anna Tuschling (Ruhr-Universität Bochum)
– Geoffrey Winthrop-Young (University of British Columbia)
Media Infrastructures and
the Politics of Digital Time
Essays on Hardwired Temporalities
Edited by
Axel Volmar and Kyle Stine
Amsterdam University Press

Gefördert durch die Deutsche Forschungsgemeinschaft (DFG) – Projektnummer 262513311
– SFB 1187. Funded by the German Research Foundation (DFG) – Project-ID 262513311 – SFB
1187.
Funded through a Social Sciences & Humanities Research Council of Canada (SSHRC)
Connection Grant, file number 611-2015-0336.
Cover illustration: Pete Linforth (Pixabay), Pavlofox (Pixabay).

Cover idea: Julia Eckel
Cover design: Suzan Beijer

Lay-out: Crius Group, Hulshout
isbn 978 94 6372 742 6

e-isbn 978 90 4855 075 3
doi 10.5117/9789463727426
nur 757 | 811
Creative Commons License CC BY NC ND

(http://creativecommons.org/licenses/by-nc-nd/3.0)
All authors / Amsterdam University Press B.V., Amsterdam 2021
Some rights reserved. Without limiting the rights under copyright reserved above, any part of
this book may be reproduced, stored in or introduced into a retrieval system, or transmitted,
in any form or by any means (electronic, mechanical, photocopying, recording or otherwise).
Every effort has been made to obtain permission to use all copyrighted illustrations
reproduced in this book. Nonetheless, whosoever believes to have rights to this material is
advised to contact the publisher.
Table of Contents
Acknowledgments 7
Infrastructures of Time: An Introduction to Hardwired Temporalities 9

Kyle Stine and Axel Volmar
Part I Media Philosophies of Time Patterning
1. The Suspension of Irreversibility: The Fundamental (and

Futile) Task of Media 41
John Durham Peters
2. Time and Technology: The Temporalities of Care 55

Gabriele Schabacher
3. Problems of Temporality in the Digital Epoch 77

Yuk Hui
4. Suspending the “Time Domain”: Technological Tempor(e)alities

of Media Infrastructures 89
Wolfgang Ernst
Part II Microtimes
5. Infrastructuring Leap Seconds: The Regime of Temporal

Plurality in Digitally Networked Media 107
Isabell Otto
6. Life at the Femtosecond 125

Geoffrey C. Bowker
7. Artificial Intelligence and the Temporality of Machine Images 143

Andrew R. Johnston
8. Intervals of Intervention: Micro-Decisions and the Temporal

Autonomy of Self-Driving Cars 157
Florian Sprenger
Part III Lifetimes
9. Grounded Speed and the Soft Temporality of Network

Infrastructure 177
Nicole Starosielski
10. Unruly Bodies of Code in Time 191

Marisa Leavitt Cohn
11. Screwed: Anxiety and the Digital Ends of Anticipation 205

James J. Hodge
12. Beep: Listening to the Digital Watch 221

Sumanth Gopinath
Part IV Futures
13. Captured Time: Eye Tracking and the Attention Economy 243
Alexander Monea
14. Ahead of Time: The Infrastructure of Amazon’s Anticipatory

Shipping Method 263
Eva-Maria Nyckel
15. Artificial Neural Networks, Postdigital Infrastructuresand the

Politics of Temporality 279
Andreas Sudmann
16. Technics of Time: Values in Future Internet Development 295

Britt S. Paris
Index 309
Acknowledgments
This collection stems from two conversations started at McGill University in

Montréal, Canada, in 2015 and the University of Siegen, in Siegen, Germany,
in 2018. We express our gratitude to the Social Sciences and Humanities
Research Council of Canada (SSHRC), which generously funded the project
through a Connection Grant, as well as the Andrew W. Mellon Foundation,
the Dean of Arts Development Fund of the Faculty of Arts at McGill Univer-
sity, and McGill’s Post-Graduate Students’ Society (PGSS). We extend special
thanks to Jonathan Sterne for his support. Additionally, we thank Errol
Salamon, Maryse Ouellet, Anastasia Howe Bukowski, and Megan Mericle
for logistics and organizing; Caitlin Loney for designing our website, posters,
and flyers; and Kathleen Holden for administering the awards funds in the
Faculty of Arts. We also owe thanks to the Collaborative Research Center
“Media of Cooperation” at the University of Siegen, funded by the German
Research Foundation (Deutsche Forschungsgemeinschaft, DFG), and to
Jenny Berkholz and Anja Höse. Finally, we would like to extent special
thanks to the German Research Foundation for making this volume freely
available via Open Access.
Infrastructures of Time:
An Introduction to Hardwired
Temporalities
Kyle Stine and Axel Volmar
Abstract
The introductory essay to the volume proposes a framework for under-
standing the transformative and disruptive effects of digital time. It argues
for a multiscalar approach to the layers of temporality active in current
media infrastructures, which coordinate different magnitudes of time
from the microtemporal to the longue durée. Situating the phenomenon of
digital time within a trajectory of increasing materialization of temporal
relations, it provides a historical account of the becoming concrete in
technology of what were once relations between people and objects.
Keywords: media theory, digital time, infrastructures, materiality,

temporality
All machines, whether mechanical, electronic, or symbolic, are in a crucial

sense time machines. They pattern the movement of mechanisms, the flow
of electrons, or the operations of symbols to meet temporal demands such
as synchronism, succession, repetition, and pace. Media technologies thus
constitute not only material infrastructures, as has been a watchword in
recent media theory, but also temporal infrastructures, architectures, and
systems—materialities designed in and as time. An aspect of this patterning
of time that has received heightened scholarly attention is the ubiquitous
experience of technological and cultural acceleration. Temporal speed-up
has in fact emerged as a defining characteristic in accounts of modernity,
as Peter Conrad expresses in saying, “Modernity is about the acceleration
of time.”1 Recent works, from critical theory to the sociology of time, have
Volmar, A. and K. Stine (eds.), Media Infrastructures and the Politics of Digital Time: Essays on
Hardwired Temporalities. Amsterdam: Amsterdam University Press, 2021
doi 10.5117/9789463727426_intro
10 K yle Stine and A xel Volmar
emphasized this in terms of a cultural doctrine of accelerationism.2 However,

focusing on the temporal aspects of media culture reveals not only an
acceleration of life and communication systems but also complex temporal
relations within technologies, between technologies, and between human
time and technological time.
Take a smartphone, for example. A host of services are at one’s fingertips:
rides available for pickup at an exact address, rooms ready to be reserved and
digital keys accessed without interaction with any person, goods connected
to whole systems of order delivery through automated warehouses and
same-day transportation, and entertainment media set to stream on the go.
Such ease of access can give the impression of instantaneity and immediacy,
of time compressed to the zero degree. There are no lapses in programming,
such as when television stations of old shut down for the night, and no
closed signs to be flipped on shop doors.3 Yet obvious from experience is
that these conveniences are also subject to service interruptions, scheduled
maintenance, system lag, and downtime, not to mention rush hours and peak
pricing. Behind and beneath our real-time interactions with on-demand
media and services is a temporal geography as uneven as our social and
political geographies, in which slowness and waiting are produced and
distributed alongside every advance in convenience and speed.
Recent research in media and cultural studies has attended closely
to the widening gap in lived experiences of time across different social
geographies. Emily Keightley in this regard urges media scholars “to move
beyond a one-dimensional characterization in which speed and immediacy
monopolize accounts of how time is encountered and lived” and instead
address the “social temporalities of mediated experience.”4 Sarah Sharma
shows in her ethnography of business travelers that the experience of living
“fast,” as promoted throughout self-help literature on time management,
represents a luxury that draws on an entire service industry of workers
made to calibrate their bodies to the demands of global capitalism.5 In
the words of Markus Krajewski, modern service workers, or “servers,” to
express the comparison between people and technologies that are ordered
to stand by, have always been consigned to “idle time,” or waiting for a
bell, command, or instruction that positions them in relation to the time
they serve.6 In the essays collected in this volume, we follow these studies
by understanding time in its unequal distribution across our social and
political world, while emphasizing its particular relation to the problem of
technological acceleration, the stark departure of technological time from
lived human time, and in this way couple the themes of social inequality
and accelerationism in the analysis of time.
Infr astruc tures of Time: An Introduc tion to Hardwired Tempor alities 11
The scale of this inquiry cannot be addressed solely at the level of the
individual medium or technology, in that the defining characteristic of
digital time is that it coordinates multiple layers of technological time
within a comprehensive system. The analytical figure of digital time is for
this reason the infrastructure. On the one hand, infrastructures embody
temporal relations between technologies, as maintained through stan-
dards and protocols; on the other hand, they coordinate relations between
technologies and human beings, who serve as interlocutors, care givers,
proxies, and delegates. So while digital technologies construct an uneven
geography that accelerates life for some and impedes life for others, they
also profoundly transform and reorder temporal regimes and practices in
other ways that require critical attention.
A central argument of the volume is that the concerns of digital time
and contemporary media infrastructures exceed any one field of inquiry,
requiring cross-disciplinary perspective. Our objective is to account for
devices and processes whose spheres of action range from the microtemporal
to the geological, addressing for the time domain what studies have noted for
the spatial domain of international logistics, whose vectors range from the
flows of microscopic circuits to the paths of orbiting satellites.7 We pursue
this goal through a broadly inclusive range of media and infrastructure
studies, with perspectives from science and technology studies, cultural
studies, and the philosophy of time, while acknowledging the need for
future collaborations between the humanities, sciences, and engineering.
The volume in this way serves as a meeting ground between disciplines
and begins a cross-disciplinary conversation that will become increasingly
relevant and necessary as infrastructural systems extend yet further into
the social and political systems of everyday life.
The contributors study two aspects of the infrastructuring of time: the
infrastructures of temporality, namely the means of ordering time through
technologies and practices ranging from calendars to computers; and the
temporalities of infrastructure, or the specific, often incompatible temporal
orders of different technical milieu in media, science, business, and govern-
ment. As our title suggests, the central metaphor of our inquiry, drawing on
the predominant mode in which temporalities are inscribed and effectuated
today, is the hardwiring and rewiring of temporal orders, calling attention
to how stabilized temporalities, erected in infrastructures, exert pressures
of conformity and standardization on the temporalities of lived experience
and among different temporal infrastructures. Hardwired temporalities refers
to all the ways that time patterns become fixed in materialities: these can
occur through unplanned flows of habituation whose constancy eventually
sculpts a definable space, just as flowing water carves out a river bed that
endures even while remaining open to redirection; but they can also be
formed by more obdurate temporal governance, the imposition of temporal
patterns by top-down command. Temporal regimes refers to the guiding
principles of ordering time in a given locality and epoch. The overriding
temporal regime today is the imperative speedup of global capitalism, but
smaller scale temporal orders exist within this larger regime, such as the
orders of religious time and leisure time. Because networks so define social
space, we speak of hardwiring and rewiring temporal relations, rather than
using the more amorphous phrasing of temporal flows. Reorganization of
temporal patterns in infrastructures is not liquid—it cannot take just any
shape—but is nevertheless flexible and open to reconfiguration. Moreover,
these patterns coordinate different spheres of action. The temporal orders
of our digital culture involve infrastructural formations across multiple
temporal scales, from the microtemporal domains of manipulating, pro-
cessing, and transmitting information, through the temporal orders on the
meso scale of everyday life and lived temporalities, to the macrotemporal
scales of cosmological and geological deep time.
The collection speaks to and consolidates insights among three important
directions in media studies today, making contributions in this way also to
three broad pursuits in recent humanities and social science research. The
current social and political unrest in the neoliberal economies, following an
intensified concentration of wealth enabled by disruptive new technologies,
has prompted a turn toward material culture and a deeper consideration of
the technical specificities of the networks, devices, and programs used in our
daily lives. This turn toward the nonhuman and emphasis on new materialisms
for rethinking the relationship between human societies and technological
networks has significantly broadened the scope of inquiry and deepened the
scales of time considered.8 However, it also risks a certain spatializing bent.
Jane Bennett points to one aspect of this bias when she notes the tendency in
object-oriented ontology to disregard the relations between objects, in response
to which she suggests a way of thinking objects and their relations together.9
We argue that time is precisely the dimension that is lost in an approach to
objects that brackets out relations, and it is necessary to develop approaches
to materiality that analyze time, which is fundamentally relational.
Our inquiry comes at the question of digital temporality from two
directions. The first of these we have discussed in addressing the social
experience of time and its multiplicity, what we can characterize, for the
sake of comparison, as a culturally focused approach. However, we follow the
insights of science and technology studies to understand that no technical
system or device is value neutral. Any divergence between the temporal

patterns of human life and those of networked technologies is likely to
exacerbate inequalities already perpetuated by systemic discrimination. The
last two years have seen important interventions in the areas of technological
and algorithmic bias, whose insights point to ways scholars might further
interrogate the uneven distribution of time. The attention economy’s effort to
maximize engagement is an explicit program to monopolize people’s time.10
When biased algorithms filter results in discriminatory ways, as Safiya
Umoja Noble has shown, they not only misrepresent people and concepts;
they also misdirect people and consume their time.11 Charlton McIlwain and
Ruha Benjamin have pointed out how technological systems that present as
neutral means of problem-solving are cut through with racial biases. Early
computer systems, as McIlwain demonstrates, were explicitly intended to
aid police countermeasures against the Civil Rights Movement and were
further embedded in the carceral apparatus of the War on Drugs.12 Benjamin
has extended this insight in the deepest way to show that, even beyond the
point-of-the-sword biases of facial-recognition software and search tools
that predict ethnicity according to people’s names, technological systems
entrench racial hierarchies throughout their design in myriad ways that
are inescapable in their effects.13 The continual march of innovation, which
hardwires and rewires power relations, deserves further attention in analyses
of time, a framework to which the essays in this volume seek to contribute.
Complementary to the perspective on lived time is an approach that draws
insights from German media theory,14 media archaeology,15 and studies of
microtemporalities,16 with their attention to the design and inner workings
of the technologies themselves, to extend an analysis of time beyond repre-
sentational media, such as literature and film,17 to the nonrepresentational
media and programs that enable them, often invisibly. Following Wolfgang
Ernst, we understand that media studies must become “time-critical.”18
Being sensitive to the time-criticality of media technologies means being
attentive to temporal actions that are in a certain way “critical factors” for
the successful execution of a process.19 This includes real-time applications,
whose operations exist below the threshold of human perception, and more
generally the synchronization and coordination of different co-operative
speeds and time windows. The volume is thus informed by and speaks to the
current German discourse on understanding media less as means of repre-
sentation and transmission than as fundamental “conditions of cooperation,”
a conversation closely connected to the “practice turn” in media theory.20
Lastly, the volume benefits from and contributes to the growing interest
in infrastructures.21 We take our cue from Susan Leigh Star and Geoffrey
Bowker’s notion of “infrastructuring” as an active, ongoing process and

from Lauren Berlant’s recent broadening and refinement of the concept
of infrastructure to mean “the movement or patterning of social form.”22
Berlant’s choice of the word “patterning,” in its active, gerundial sense,
as opposed to the more static connotations of “pattern” and “form,” lays
emphasis on infrastructures as temporal processes of becoming. Social
patterns, however fixed they may seem, are only ever the circuit for move-
ments and temporal flows. Keller Easterling has shown how infrastructure
space, even when not “mediated” in the common sense by sensors and media
technologies, is an information technology where the mere mobilization of
form, in grids and containers, is “an operating system for shaping the city.”23
We might argue, in a similar way, that infrastructure time, the rhythm
and patterning of temporal order, is the very basis of information, which
never stands still but must be processed and transmitted. Indeed, recent
developments in network technologies and smart sensors, we argue, have
created a need to reexamine infrastructures particularly in terms of their
patterning of time. “To be modern,” as Paul Edwards puts it, “is to live within
and by means of infrastructures.”24 To be digital—or to be in an algorithmic,
networked culture—is to live within and by means of infrastructures that
are themselves monitored, maintained, and controlled by deeper data
infrastructures. The infrastructures of modernity, such as roads, bridges,
communication lines, and financial systems, have long been equipped
with cybernetic feedback infrastructures that monitor their operations,
make corrections, and, when needed, marshal workers to repair them.25
The temporalities of their operation and aging are now bound up in the
computational time of digital networks and, as such, submitted to the
control and surveillance of these networks. Yet at the same time, these
newer infrastructures rely on the older infrastructures of water, energy,
and human transit, as a nervous system relies on a circulatory system.
Infrastructures in this way are coordinating and synchronizing features of
multiscalar action, which, as we have mentioned, embody relations between
technologies and people, and are thus appropriate figures for thinking about
materiality, microtemporalities, and the social geography of time together.
Technologies as Consolidated Temporalities
A guiding thread in this regard involves the processes by which technical

systems consolidate temporality, in the most literal sense of their gathering
together disparate temporal processes and making them solid in physical
infrastructures. Scholars of the social construction of science and technology

teach us that even nonrepresentational technologies, which cultural analysis
long overlooked in favor of media and artistic works, embody social values
and relations of power. In Bruno Latour’s apt phrasing, technologies are
“full of people”: they concretize human expertise and function to advance
human goals.26 Just the same, technologies are full of time. Karl Marx
described one aspect of this when he argued that the value of the commodity
consists in “congealed labor-time.”27 Yet even beyond the actions necessary
for their immediate manufacture, technologies embed multiple histories and
spheres of temporal action. Michel Serres calls technologies “polychronic”
to express the multiple pleats of time that fold together “the obsolete, the
contemporary, and the futuristic”:
Consider a late-model car. It is a disparate aggregate of scientific and tech-

nical solutions dating from different periods. One can date it component
by component: this part was invented at the turn of the century, another,
ten years ago, and Carnot’s cycle is almost two hundred years old. Not
to mention that the wheel dates back to neolithic times. The ensemble
is only contemporary by assemblage, by its design, its finish, sometimes
only by the slickness of the advertising surrounding it.28
Comprising parts of different vintage, technologies also operate on multiple

scales of temporality. To use Serres’s example, a modern car going at 60 miles
per hour is likely to have wheels rotating at around 800 revolutions per min-
ute, an engine firing at 2,000 revolutions per minute, and a microprocessor
calculating at 2 GHz, or 120 billion cycles per minute. Meanwhile, the driver
must maintain a safe reaction time, at best 0.7 seconds, and endure the length
of travel, sometimes numbering several hours, interrupted by cyclical human
events such as stopping off for food or at a rest area. Human-scale actions
and technological actions, as this example suggests, are divergent along
many paths, guided by their own temporal logics, but they are importantly
coordinated and synchronized by means of cultural techniques of time
patterning.29 The frame for thinking this coordination must come from
temporal infrastructures, which by design integrate and mediate these
various human and nonhuman temporalities, drawing together the micro,
meso, and macro domains, and thereby enable systematization and diverse
forms of cooperation.30
Not only do temporal practices and processes tend to consolidate into
temporal infrastructures; they also reciprocally interact as hardwired
infrastructures with new kinds of pliable and adaptable systems, both
above and below: above, in the flexible software systems designed to

operate using new circuit generations and the various social practices
centered on new communication technologies; and below, in the flexible
economies and manufacturing processes at the base of fabricating these
technologies. This dialectic between rigidness and flexibility, structure and
versatility, predictability and unpredictability, as Serres notes, lies at the
very heart of temporality. It is for this reason that the Latin tempus gives us
not only the structured “temporality” of the clock but also the intemperate
fluctuations of “temperature,” the wild swings of “temperament,” and the
fierce unpredictability of “tempests”:
The French language in its wisdom uses the same word for weather and
time, le temps. At a profound level they are the same thing. Meteorolog-
ical weather, predictable and unpredictable, will no doubt someday be
explainable by complicated notions of fluctuations, strange attractors.
Someday we will perhaps understand that historical time is even more
complicated.31
John Durham Peters elaborates these etymological connections in broad

perspective:
In Latin, tempus means weather and time, giving English such words as
temporal and tempest, and French le temps and Spanish el tiempo, both of
which mean both time and weather; the Spanish al tiempo means both
“in season” (of fruits) or “at room temperature” (of drinks). Terms such as
temperature, tempering, tempo, and temperament show shared semantic
fields across heat, harmony, rhythm, and mood.32
Time, in this sense, is moody and multiplicitous, varied and in flux. The
goal of infrastructuring time, founded on time technologies and cultural
techniques of time management, is to tame these moody fluctuations and to
submit them to ordering—to hardwire them into lasting temporal regimes
or cultures.
By the same token, the term “hardwired” relates equally to embedded
systems and more flexible practices involved in rewiring temporal orders.
To use an example, computer components in the 1960s, whose electrical
layout would soon be characterized as hardwired, were at the time compared
favorably “soft” against the fixed-program analyzers of yesteryear. The first
recorded instance of the then-hyphenated “hard-wired” documents this
usage, when a contributor to the journal Nuclear Instruments and Methods
in 1965, comparing the two generations of computers, wrote: “Another trend

is the use of small computers instead of hard-wired analyzers.”33 In this way,
the conceptual ground of hardwiring is, in a deconstructive turn, precisely
the opposite of being fixed and immutable. Technical consolidation is instead
the product of a new kind of adaptive industry, a flexible economy founded
on the production of a new component, the integrated circuit, capable of
being hardwired and rewired across product generations. Moore’s Law,
formulated in the same year of 1965 to describe the regular doubling of
circuit complexity from one fixed pattern to another, has since turned into
a self-fulfilling prophecy, a stable temporality that serves to predict future
technological progress. In a system of planned obsolescence, hardwired
components come to enable the periodic “rewiring” of the systems they run.34
The hardwiring of space, as seen in the doubling of component density on
microchips, thus structures time in new ways as well, setting industry on a
regular course of introducing new product generations and creating a new
density of temporal intervals, or actions that can occur in a given period
of time, with increased processing speeds. It would not be a stretch to say
that these computational components and their infrastructuralization into
larger networks invented a new sphere of time in the same way that James
Carey saw the telegraph as instituting a new regime of time a century before.
In its ability to send messages faster than physical commodities, according
to Carey, the telegraph rendered obsolete the system of arbitrage—the
practice of buying low in one market and selling high in another—and
redirected financial speculation into commodity futures. “In a certain
sense,” Carey writes, “the telegraph invented the future as a new zone of
uncertainty and a new region of practical action.”35 That is, the telegraph
initiated a new domain of time by hard wires. Judy Wajcman stresses the
continued material pressures of such wires with the example of a recently
laid fiber-optic cable between Chicago and New York: “While previous cables
between the two cities had been laid along railway lines, the new cable takes
the shortest route possible, even drilling through the Allegheny Mountains.
It shaves 1.3 milliseconds off the transmission time of the earlier cables.”36
Likewise, contemporary developments in big data, artificial intelligence,
and machine learning can be regarded as a result of the “wires” that have
gone into the microprocessors and networked environments that enable
algorithms to operate over global information networks within ever smaller
temporal intervals.37
Historically, temporal practices have tended to consolidate into tech-
nical objects and from objects into structures and infrastructures. The
more organized the materialities, the more structured the temporalities.
Our project thus seeks to reassess material infrastructures as stabilized

structured temporalities, forms of patterned time sustained over given
periods of history in formations of technologies, practices, and conventions
that affect people’s actions and experiences and are themselves subject to
constant “rewiring.” The two terms that organize our thinking on the topic,
“hardwired” and “temporalities,” in this way name a dynamic interaction
between infrastructured temporalities and their continual interactions
with more pliable, flexible, mortal, human systems.
A Chronology
Temporality, as seen throughout history but especially in an age of global

digital networks, is palimpsestic. Concretized within temporal infrastruc-
tures and embedded in our experiences of temporality are multiple historical
regimes successively layered and combined. A modern smartphone, for
instance, which is subject to the timing of a processor clock and various
network synchronization protocols below the level of human awareness,
also remediates earlier temporal interfaces through apps such as calendars,
clocks, and stopwatches. To take but one example, Chinese, Hebrew, and
Islamic calendars are standard options on smartphones alongside the
western Gregorian calendar, while apps are available for Persian and Tibetan
calendars, among others, and even more remote historical calendars such
as the Maya calendar. Indeed, the applications of timekeeping in new media
are practically limitless. While not dismissing this multitude of applications
and their complex interactions, it is possible to outline four overarching
temporal regimes, or historical hegemonic temporal logics, that inhere
in modern technologies and continue to structure temporal techniques
and experiences, namely, calendar time, clock time, capitalist time, and
technological microtime.
We understand these temporal logics to have emerged from multiple
locally installed and trans-locally networked temporal orders rather than
from any single authoritative center. After all, even a regime as centrally
administered as medieval Christian time was not commanded by a single
source of timekeeping but was instead made possible by a multitude of
individual clocks and clock towers in every Christian settlement, involving
procedures for displaying the time, as for instance by the chiming of bells,
and people tasked with caring for the clocks and regularly setting the time.
Likewise, the concept of capitalist time, which introduced the virtue of
optimization, emerged from a proliferation of town bells and factory time
clocks, not from a single timekeeper, and in this way resulted in ever new
“rewirings” of procedures of production, organization, and cooperation.
An obvious way of understanding these four temporal orders, which
historically succeed one another, is in their progressive ref inement of
counting time. Each method of measuring time, which is also characteristic
of a historical epoch, sharpens the unit of measurement, while the epochs
themselves shorten. Calendar time, incipient with the invention of writing,
patterns longer intervals such as days and years, while Christian clock
time, beginning in the twelfth century, enables the standardization of
human practices within the span of a day through the divisions of hours.
Beginning in modernity, capitalist time, especially gaining momentum
in the nineteenth century, with its inherent logics of efficiency and ac-
celeration, increasingly focuses on the shorter measures of minutes and
seconds, building upon which technological microtime more finely divides
temporal measure below human sense thresholds. Temporal units and the
span of innovation in this way historically tighten. However, the command
of time, always directed toward the future, has progressively expanded,
as evidenced by contemporary megaprojects, including the decade-long
construction of CERN’s Large Hadron Collider and the nearly two-decade
construction of the Three Gorges Dam in China, which beyond taking a
long time to build, consumed billions of dollars of labor time, relying on
various specialized workforces.38 In a similar fashion, the historical record
available in each temporal regime has also expanded, evidenced nowhere
more conspicuously than in the introduction of deep geological time in
the eighteenth century, but also apparent in the heightened resolution of
historical data. In this sense, it is not enough to consider only the units of
temporal control and the length of future time under the command of the
present; we must also acknowledge the new scales of temporal complexity
within shorter intervals. Just as a computer performs more actions in a
second than is possible using conscious calculation, a megaproject facilitates
and coordinates more actions in the span of a week or year than was possible
in previous projects in previous eras. Put simply, technological time today
is denser and more vivid than past times; it contains more action moments
and has a much higher resolution. Within these denser frames of planning
and action must be coordinated the many temporal measures of the actors
involved, whether human or nonhuman, which operate across these four
temporalities.
In what follows, we attempt to delaminate these various layers of temporal
governance to better understand their historical sources and how they
interact, combine, over-pattern, and stabilize in durable infrastructures.
Calendar Time
Since the early beginnings of so-called “civilization,” the life of the vast
majority of people has been and still is governed not only by the natural
temporal rhythms of seasons and cycles of day and night but also by temporal
regimes, i.e., orders of patterned time sustained by technologies and practices
of timekeeping and temporal organization. John Durham Peters, for instance,
reminds us of the fundamental significance of the calendar as a cultural
technique of social order and governance that, through the science of astron-
omy and the politics of calendar making, provided a means of predicting and
determining recurring events, from yearly floods, as in Mesopotamia and
Egypt, to holy days.39 From their earliest uses to the present day, calendars
have served to track the succession of days by dividing the year into arbitrary
intervals of months and weeks. Their temporal divisions allow for repetition
and ritual and hence the coordination of social, economic, and religious
life into structured temporal schemes, both past and future. Alongside the
political and military control of space, or territory, as Harold Innis argued in
Empire and Communications, the cultural control of time based on common
cosmological, religious, or philosophical narratives and materialized into
different time media has played an equally important role in securing the
endurance of cultural-political entities. 40
A direct line extends from our present computational timekeeping
technologies back to the calendars of earlier empires. In an influential
essay on time and human language, Émile Benveniste explains that the
calendar owes its existence to a baseline computation. 41 Paul Ricoeur
explains Benveniste’s insight especially clearly: “the features common to
every calendar ‘proceed’ from the determination of the zero point of some
computation.”42 In this sense, the calendar can be viewed as an early form
of computing whose logistical functions issue from three basic conditions:
the establishment of an axial moment, e.g., in the common era of occidental
civilization marked by the birth of Christ; the determination of whether an
event occurred before or after the axis; and the measurement of intervals,
such as days, weeks, months, and years. The calendar, as Ricoeur puts it
elegantly, thus “cosmologizes lived time and humanizes cosmic time”43; it
is the first technique to organize these different spheres of temporality and
serve as a bridge between them.
Calendars, as tools of temporal social organization, are the first tech-
niques to introduce what Benveniste calls “chronic time,” a term he uses to
encapsulate both calendar and clock time for their ability to join together
interior subjective duration and exterior physical time within a coordinating
grid that locates personal experience within cosmic rhythms. 44 Out of

this logistical construction arises a seeming paradox in that chronic time,
which Benveniste says is the only time we generally encounter in our day-
to-day lives, does not move; it is instead arrested and, to repeat Hamlet’s
lament, “out of joint” with our inner experience of time, which constantly
slips away: “It might thus seem natural that the structure of chronic time
should be characterized by permanence and fixity. Yet, at the same time,
it must be realized that these characteristics result from the fact that the
temporal organization of chronic time is actually intemporal. This is not a
paradox.”45 Chronic time is a rigid atemporality, and only for this reason can
it situate passing events in relation to one another. To use the language of
our title, chronic time is the hardwired a priori of our more flexible everyday
experiences of and interactions with irreversible time. An insight that we
can draw from Benveniste is that just as the calendar presents a grid of
temporal reference for calculating events, today’s computer systems and
infrastructures extend this intemporal grid to new levels of complexity
and acceleration and thereby enable not only new modes of calculative
governance but also new variabilities in lived temporality.
Clock Time
In his foundational work in the history of technology, Lewis Mumford (1934)

demonstrates how the unified time of monastic life in the Middle Ages
precipitated the development of the mechanical clock and influenced the
subsequent temporal coordination of people and technologies that enabled
both the Scientific and Industrial Revolutions, a development that leads
him to argue: “The clock, not the steam-engine, is the key-machine of the
modern industrial age.”46 Alongside the calendar, which forges cultural
unity on the basis of common holidays and other social events throughout
the year, the clock enables within this initial computation of calendar
time more fine-grained coordinations of human activities and forms of
cooperation throughout the day and week, particularly the organization
and control of human labor.
Historically, the prime points of time reference were not calculated
abstractions but physical, often cyclical, work-based particularities. The
rising and setting of the sun marked the passage of days, the wilting of
flowers and spoiling of foods influenced cycles of work and gathering,
the passage of winters measured age, and the recurrence of ten moons
promised that an expecting mother would soon give birth. Noting among
other regularities of human life, such as “the beating of the pulse” and “the
breathing of the lungs,” Mumford cites practices of agrarian subsistence: “The

shepherd measures from the time the ewes lambed; the farmer measures
back to the day of sowing or forward to the harvest.”47 James Henry Breasted,
writing in 1935, noted how the lives of his contemporaries gained temporal
meaning by reference to seasonal fluctuations: “Among certain Swedish
peasants even at the present day a birthday may fall at the ‘rye harvest’ or
at the ‘potato harvest.’”48 From similar examples of celestial and earthly
timekeeping, Peters (2015) has argued that the movements of the skies and
earth themselves constitute “elemental media.”49
It would be a mistake, however, to suggest that the onset of regimented
clock time did away with these corporeal and more sensible temporal
measures. In his commanding work on the history of timekeeping, Eviatar
Zerubavel cites a striking example of time formulation without recourse to
calendar or clock from the opening of Kurt Vonnegut’s Cat’s Cradle: “When
I was a younger man—two wives ago, 250000 cigarettes ago, 3000 quarts of
booze ago.”50 Drawing on more functional examples, he reminds us that even
today it continues to be more appropriate “to designate the life expectancy
of tires and running shoes in terms of mileage—or that of children’s beds
in terms of the child’s weight—than in terms of years of use.”51 What has
changed though is that these time references take on new meaning in an era
when, as Mumford argues, “The modern industrial regime could do without
coal and iron and steam easier than it could do without the clock.”52 For,
now, all of these more variable temporal measures are caught within the
mesh of modern clock-based time.
The wresting of time away from personalized reference points has been
crucial in establishing intersubjective social realities. Time-counting devices,
from water clocks to later mechanical clocks, in situating individual actions
within a common social frame, have allowed for the organization of work, the
establishment of cultural identity through repetition, and the incorporation
of these identities and processes into larger cultural configurations such as
corporations and nations. The monastery, according to Mumford, was the
first instrument for calculating this form of social time:
Within the walls of the monastery was sanctuary: under the rule of the
order surprise and doubt and caprice and irregularity were put at bay.
Opposed to the erratic fluctuations and pulsations of the worldly life
was the iron discipline of the rule. Benedict added a seventh period to
the devotions of the day, and in the seventh century, by a bull of Pope
Sabinianus, it was decreed that the bells of the monastery be rung seven
times in the twenty-four hours. These punctuation marks in the day were
known as the canonical hours, and some means of keeping count of them
and ensuring their regular repetition became necessary.53
In a similar way, modern technical systems and infrastructures should be

thought of as instruments for calculating and managing time. As profound
as was the clock’s impact on social organization was its effect on mechan-
ical processes. For Mumford, the clock set “the regular collective beat and
rhythm” of a technical system by wedding together these regular mechanical
actions with the synchronized movements of people. More than a mere
counting device, it was also “a new kind of power-machine, in which the
source of power and the transmission were of such a nature as to ensure
the even flow of energy throughout the works and to make possible regular
production and a standardized product.”54 Timekeeping was, then, from
the beginning, a means of not only coordinating human actions but also
regulating and operating machinery. The clock’s qualities of standardization,
automatic action, precise gearing, accuracy, and reversibility allowed it
to divide time and conquer it. Even more significant, by taking on these
characteristics of space, time could be added, saved, and controlled, laying
the conditions for capitalist time.
Capitalist Time
Until recently, the temporal orders of calendar and clock have been subject
to a politics of time, struggles within societies based on conflicting interests
between the state, economic and religious actors, and individuals over
matters such as the recognition of holy days and the designation of work
times. The convergence of meanings on May 1 in cultures of the northern
hemisphere helps to illustrate these conflicting politics of time. First
celebrated in response to the astronomical event of spring, the day was
a seasonal festival of the return of the warm season. After being adopted
by the international workers movement to commemorate the Chicago
Haymarket massacre in the late-nineteenth century, the day became a
further palimpsest when, during the First Red Scare in the early 1920s, it
became a reactionary, unofficial holiday dubbed “Americanization Day”
that the US Congress would later inscribe into law as Loyalty Day during the
Second Red Scare in the 1950s. Concerns about the temporal politics of paid
labor continue unabated today in negotiations over how many hours per
day and per week employees should work, how much vacation time should
be allowed, how many sick days employers and health insurers should pay
for, how overtime should be compensated, and how the post-work life of
retirement should be managed. These struggles are largely crystallized in

agreements, contracts, laws, and other forms that set more or less specific
conventions for the patterning of human everyday life.55
In this regard, E. P. Thompson argued that the decisive change paving
the way for modern capitalism was the shift from task-oriented labor to
time-oriented labor.56 Labor focused on tasks such as fishing and harvesting
crops is embedded in the rhythms of the natural world, such as the rising and
falling of tides and the passage of the seasons, and thus ritually connected
to universal time. Such labor, rather than being set to the employer time
clock, is characterized by “alternate bouts of intense labor and of idleness.”57
The continuation of this labor pattern in the creative economy today leads
one to wonder, as Thompson himself wondered in the 1960s, whether “it is
not a ‘natural’ human work-rhythm.”58 But possessing one’s natural time
is largely at odds with capitalist economics. By uprooting time from one’s
personal experience, it becomes abstract and controllable, and in this way,
as Thompson explains, time becomes money: “Those who are employed
experience a distinction between their employer’s time and their ‘own’ time.
And the employer must use the time of his labour, and see it is not wasted:
not the task but the value of time when reduced to money is dominant.
Time is now currency: it is not passed but spent.”59 When time-oriented
labor becomes counted time, it makes work time accountable and evaluable,
with the historical side-effect that it renders forms of labor that are not
compensated monetarily, such as household and care work, traditionally
(and even today) largely performed by women, invisible.60 This alliance
between money and time, as Peters notes, rests on their being paradigm
cases of “logistical media,” or those media that “establish the zero points
of orientation.”61
In establishing a grid of temporal structure capable of containing and
coordinating diverse practices in time, the calendar and clock have func-
tioned to lift time out of the necessity of particular reference and produced
what French historian Paul Ricoeur calls “anonymous time.”62 Anonymous
time, for Ricoeur, is a mediating temporality between phenomenological
experience and worldly time; it functions to situate and compare subjective
and objective temporalities, giving temporal place both to inner experience
and external events. Reading Alfred Schütz’s influential phenomenological
account of intersubjectivity, Ricoeur argues that the division of social time
into anonymous categories of “contemporaries, predecessors, and succes-
sors” initiates a temporal logic that forms a bridge between “lived time and
universal time.”63 The succession of generations, socially experienced as the
“replacement of the dead by the living,”64 constitutes a third time between
inner subjective reality and the physical time of the world, out of which
time becomes a matter of roles. People in their radical singularity can never
replace one another in their phenomenological experience of time, a point
that Martin Heidegger emphasizes in calling phenomenological time one’s
“ownmost possibility,”65 but they can step into vacated stations. This ability
to assume specific temporal roles, first established in the psychological
relationship of contemporaries, predecessors, and successors, is accelerated
in the temporal coordination of industrial labor through the clock’s division
of work processes into replaceable tasks. Network temporalities today greatly
expand these logistical functions and their anonymizing proclivities, as can
be seen in ride-sharing services such as Lyft and Uber that, even as they
track individual riders and drivers, treat them as anonymous data points
to be algorithmically paired.
Indeed, anonymous time is crucial to timesharing companies, such as
Lyft, Uber, and Airbnb, which automate both monetary transactions and
clock and calendar time. Rides and rooms are not exchangeable with one
another totally but are instead exchangeable by categories, anonymously. The
driver is not treated as a singular, irreplaceable being but instead as a class
of car, a set of reviews, an anonymous anchor for a constellation of ratings.
Similarly, an Airbnb rental location is generalized, departicularized, made
anonymous, and submitted instead to ratings, reviews, and other data points.
Anonymous time is in this way constructed out of the unique possibility of
precise addressability. Although such anonymity seems merely coincident
with temporal organization, it proceeds from time management in a very
radical way. Its freedom to accept multiple diverse phenomenological actors
is founded on an exacting system of computed temporality.
Technological Microtime
Over the last two centuries, time media have increasingly come to operate
on microtemporal levels. In this process, temporal infrastructures have
come to more finely divide calendar and clock time, operationalize them,
and establish the structuring grids necessary for a new density of action
moments. For while calendrical moments, such as years, months, and days,
are efficacious in calculating events such as the rise and fall of empires, the
beginning and end of wars, or the course of a lifetime, they are inadequate
for calculating the clock time of hours, minutes, and seconds. More minute
and finely tuned temporalities require still more refined technologies of
temporal measure, such as Jimena Canales has explored in the nineteenth
century’s invention of the “tenth of a second.”66 Instruments such as
chronoscopes, myographs, and photographic cameras, which operated

beneath the temporal thresholds of human perception and reaction, carried
remarkable epistemological significance.67
In their attempts to study the processes of sense perception, nine-
teenth-century experimental physiologists turned equipment such as
telegraphs into instruments for measuring the microtemporal dynamics
of muscle activity and the transmission of nerve impulses within living
organisms. Hermann von Helmholtz’ measurements of the velocity of nerve
impulses and Matthäus Hipp’s reaction time experiments, for example,
gave rise to a new understanding of reality as being radically constituted
by the conditions of temporal perception.68 Such physiologists realized that
the temporal experience of living things is determined by their respective
temporal thresholds of perception and reaction, and thus the quality of
temporal perception came to be understood as a function of quantitative
values. Accordingly, the temporal category of the “present” was to be un-
derstood as determined by its appropriate sphere of action, the decisions
that calculate its order of magnitude in the interplay between reaction
times, transmission times, and processing times. Microtemporal actions
and technological speed-up would thus be seen as yielding less quantitative
than qualitative and thus aesthetic effects.
In 1860, the Baltic German entomologist Karl Ernst von Baer captured this
relation eloquently in a series of thought experiments by demonstrating the
relativity of temporal perception depending on the number of “moments” a
perceiving subject is able to distinguish in a given interval. For instance, he
stated that a human being taking in 1,000 instead of the usual ~10 moments
per second would perceive a waterfall as a quasi-stable object, comparable
to how we perceive the growth of plants, while events such as gunshots
would appear as traceable movements (similar to the perception of cinematic
slow-motion). On the other extreme, a subject with just one moment per
day or even per month would perceive the sun not as a slowly moving object
but, due to the extreme time-lapse, as a glowing ring, just like people would
normally perceive a piece of coal fixed to a string swirling around in a
circular fashion. Emerging from the same temporal regime of microtime,
cinematography produced out of the temporal succession of photographs the
illusion of visual movement.69 Notably, von Baer framed his ideas more than
a decade before the famous chronophotographic experiments by Eadweard
Muybridge and Étienne-Jules Marey, which precipitated motion pictures and
demonstrated the synthesizing effects of microtemporal technologies. Since
these experiments in the nineteenth century, the sciences of astronomy,
psychology, and microphysics, as well as the various media industries of
film, radio, and television, have constructed their own microtemporal

infrastructures to enable and govern the actions within these domains
beyond human sense.
The emergence of microtemporal technologies has correspondingly given
rise to another form of temporal politics, to what we refer to as the politics
of microtime. In order for technologies capable of recording, transmitting,
and reproducing sounds and images to become media, they need to be
articulated to one another and organized into larger social and economic
systems.70 Consequently, the political contestations of microtime are waged
over temporal machine standards and media formats, such as motion picture
frame rates, audio playback speeds, television line numbers, screen refresh
rates, and compression standards for both audio and video, all of which
pattern time on scales below the temporal resolutions of human perception
and cognition and yet perform the necessary work of rendering human
and nonhuman actions compatible.71 Another example might be the 60
Hz standard of the North American power grid, which allows different
technologies and devices to make use of the same resource and participate in
the same industry. In this way, temporal standards serve as what Susan Leigh
Star and James Griesemer call “boundary objects,” or artifacts that enable
and govern modes of cooperation between heterogeneous technologies
and social worlds.72
The necessity of coordinating human and nonhuman time can be
seen most fundamentally in the different ways traditional clock time and
microtime are counted. Time systems that have a direct human interface,
such as the calendar and clock, tend to use reference points that align
with human experience and the necessities of human reckoning, while
those time systems below the thresholds of human perception operate
on the metric system. The Russian and French Revolutions attempted to
institute nonreligious calendar systems that largely failed because of their
lack of intuitive connection with the celestial rhythms of people’s lives. The
explanation is relatively simple. While the metric system is excellent for
mathematical calculation, it is much less amenable to effortless counting
and subdividing by human minds unequipped with paper and pencil. Sex-
agesimal (based on 60) and duodecimal (based on 12) systems on the other
hand use superior composite numbers, with hours in a day being split into
two duodecimal halves divisible by 2, 3, 4, and 6, and sexagesimal minutes
and seconds adding divisors of 5 and 10, making them easily calculable
at a glance. It is on the basis of time systems having no necessary human
interface, however, that we owe much of contemporary technology. The
industrialized acceleration of digital switching over the last five decades
has not only transformed computers from machines of calculation and data
processing into media, in the traditional sense of audiovisual technologies,
but also rendered them increasingly as decision makers and thus autonomous
nonhuman actors in the temporal flow of the real world. On the operational
level, most general-purpose digital computers are based on technologies
of short-term memory as opposed to long-term storage, a temporal logic
Wendy Chun has termed the “enduring ephemeral,”73 which describes
random-access operations of reading, writing, and deleting information
that must be performed with considerable speed. The processual necessity
to refresh, which forms the temporal basis of computing from computer
memory and hard drives to bitmapped graphics and databases, makes
possible an endless process of reading and rewriting.74
The time-criticality of computers, their ability to act in real-time or
in the temporal flow of the world, is further predicated on technological
reconciliations between external time windows and internal processing
time. The principles and technologies of “timeliness,” reliant as they are
on the exponential growth of processing power captured in Moore’s Law,
are crucial to understanding digital temporalities, whether at play in the
hardware of computers or in the interactions of networks. Indeed, techno-
logical speedup forms the very basis of “smart machines” and other forms
of artificial intelligence applications. The range of qualitative tasks and
actions that an algorithm can perform, such as listening, speaking, playing
chess, detecting faces, or driving a car, largely depends on the number of
calculations that can be made in a given critical time window. Time is still
money, but in light of today’s digital capitalism, characterized by big data
analysis, algorithmic trading, and the mining of cryptocurrencies, it is
especially the investments in microtime and the massive exploitation of
data processing infrastructures that foster the contemporary imaginaries
of value extraction. Given that digital devices increasingly engage as non-
human actors and decision-makers in the real world, we need to consider
the temporalities on which their “smartness” stands.
The focus of this volume is on this unique extension of technological
microtime enabled by the universal medium of computation, which we
refer to as digitally networked time. Digital time is marked by its universality
and thus its ability to be extended into new domains of communication
and action. It is the baseline possibility of temporal coordination between
networked technologies that possess their own unique temporal orders.
Like the internet itself, this temporal network is distributed and layered;
it makes few restrictions on the types of time that can exist and enables
programs and apps to institute independent interfaces of time.
Layout of the Volume
The book is divided into four thematic sections, beginning with the holistic
concerns of media philosophy and passing into topical considerations of
temporal regimes on the micro, meso, and macro scale. The papers of the
first section, “Media Philosophies of Time Patterning,” investigate the specific
ability of media to suspend the course of time and how they pattern time to
make interventions, as nonhuman actors, in the present, past, and future,
on the level of both technical phenomena and human decision-making.
John Durham Peters takes up Kittler’s view of technical media as means of
“suspending irreversibility” and argues that this capacity for reversibility “is
the necessary condition of repeatability, transmission, and data storage.”
In an essay that builds on arguments from his book The Marvelous Clouds,
Peters ponders the phenomenological ironies of time’s irreversibility, which
exists, like music, in a constant dynamism of disappearing. Thus he reopens
the question of media ontology as crucially a question of time. Gabriele
Schabacher takes a media-theoretical approach to the question of care in
analyzing the energies and labor practices necessary to maintain technological
infrastructures, introducing a typology of four infrastructural care practices:
repair, maintenance, abandonment, and repurposing. Yuk Hui, drawing on
the phenomenology of Edmund Husserl and Martin Heidegger, as well as
their more recent take-up by Bernard Stiegler, seeks to understand the unique
new futurity coming into visibility with the rise of predictive technologies.
He argues, provocatively, that a new category of temporal experience is
currently taking hold, what he calls tertiary protention, or a future that issues
not from one’s own subjective projection but instead from a socially and
technologically constructed projection imposed from without in the form of
artificial intelligence. Wolfgang Ernst concludes the section by placing recent
developments within a deeper set of time-critical operations involved in media
infrastructures, focusing on “the basic layer of bit processing on the Internet.”
Delving into the operative dimensions of media infrastructures, Ernst directs
attention to the microtemporal processes that are their sine qua non, using
the example of the “Ping” signal as a time-critical signal of internet logistics.
Following from Kittler’s determination of technical media as technologies
that operate below thresholds of human sensory perception and cognition,
the papers of the second section, on “Microtime,” focus on media technolo-
gies that move beyond even the physiological and cognitive requirements
for displaying textual and audiovisual information to alter and manipulate
data in these inaccessible intervals. Isabell Otto shows in her analysis of
current debates about abolishing the leap second that digitally networked
media reveal “the fundamental relativity of each regularity of time.” Taking

as a starting point the video diary app Leap Second as a concrete example
of the plurality of time experience, she argues for an understanding of
the multiplicity of time measurements. The leap second makes a further
appearance in Geoffrey Bowker’s chapter “Life at the Femtosecond.” Going
back to Charles Babbage, Bowker roots the computer industry’s drive for
technological acceleration in the simple fact, as stated by Babbage, that
although machinery cannot be built into unlimited space, it can run through
unlimited time. Thus having not world enough, but time, computers traffic
in speed. Addressing operations taking place at the femtosecond, or the
very limits of technological microtime, Bowker asserts that, although
they fall well below human perception, they are nevertheless “real in their
consequences.” Florian Sprenger focuses in on a particular area where the
density of machinic action that can be performed in the blink of an eye
has created extraordinary new levels of complexity, tracing the logic of
microtemporal interventions in Tesla’s advanced driver-assistance systems.
These automated driving systems, by necessity, make decisions about future
events that escape human sense. As Sprenger puts it, “the autonomous car
brakes before the incident.” It calculates time and speed to predict possible
futures, such as a collision, opening up important questions about the politics
of machine decision in these inaccessible intervals. Andrew R. Johnston,
in his contribution on Google’s DeepMind project, notes that these new
levels of complexity come at a price. Researchers in machine learning are
increasingly confronted with the problem of rendering the computational
technologies they work with accessible. In particular, the efforts by Google’s
researchers toward accessibility have landed on the need for visualization,
turning to video games from the Atari 2600 system, such as Qbert and
Space Invaders, as ways of providing visual feedback for machine-learning
test runs. Moreover, the focus on test runs in the development of machine
learning systems reveals contemporary shifts in software engineering
where programmers enter into new relations of care and coaching with
increasingly autonomous algorithms.
Where the second section focuses on material changes and technical
objects, the third, “Lifetimes,” turns a lens toward the lived experiences of
human beings as they interact with and work to maintain network infra-
structures. Nicole Starosielski returns to the cable systems and network
infrastructures that formed the basis of her book The Undersea Network
to address the embodied experiences of the infrastructure operators who
maintain these systems. Two crucial new concepts emerge out of this
analysis: the idea that speed is always grounded speed in the sense that it
relies on the temporal rhythms and safe passages of human bodies, and the
problem of temporal irruptions that occur when the ground disrupts system
speed. Through these, Starosielski directs our attention to the everyday
experiences and politics that underlie digital networks, the often-omitted
“soft temporalities” of hardwired infrastructures. In a related investigation,
Marisa Leavitt Cohn argues that the overall emphasis in recent years on
the materiality of software has neglected “the temporal dimension of this
materiality—how software ages, decays, obsolesces.” Drawing on her ethno-
graphic work with engineers and software developers on a long-term space
project, Cohn examines how aging software becomes unmistakable in its
materiality and how it is feminized and pathologized for being material.
Software shows up as an “unruly body of code” in its passage through
different iterations, prompting a reckoning with its material history. In
this sense, the felt materiality of code is a product of time, revealing the
ideological forces that treated it as immaterial in the first place. In this way,
Starosielski and Cohn also connect back to Schabacher’s consideration of
human laborers as caregivers for nonhuman actors. James Hodge considers
how the temporal dynamics specific to network platforms open themselves
up to entertainment, both anxiety inducing and fun. Three online artifacts
come under Hodge’s watchful eye and incisive analysis: Brian Eaton’s artwork
the Memento Mori Clock, the “This Is Fine” meme, and a YouTube video
entitled I Put Wii Music over a Final Destination Death Scene. Through these
artifacts, Hodge considers how media creators and viewers reclaim the
demanding, machinic temporalities of contemporary infrastructures, which
through digital preemption both short-circuit anticipation and proliferate
experiences of anxiety, to make them humanly meaningful again. Con-
cluding the section, Sumanth Gopinath traces the emergence of the digital
wristwatch in the 1970s to show how designers used sound—in the form of
“beeps”—to connect these devices to the human sense realm, an industry
practice that has continued into the era of cell phones and smartphones.
The final section, “Futures,” concerns the ways technologies reach into
the future and order new regimes of time, commanding near-term actions
and provoking dystopic and utopic visions of their power. Alexander Monea
performs a media genealogy of the historical entanglement of vision and
attention in the discourse on eye tracking. Connecting this to contemporary
developments in eye-tracking technology, he argues that we may soon
experience very undesirable new arrangements of the attention economy
from digital platforms to smartphone apps where ads are able to stop playing
when you look away from the screen. Eva-Maria Nyckel studies Amazon’s
anticipatory shipping model through an analysis of the company’s patent
filings. Nyckel argues that efforts to reduce shipping latencies and better
forecast consumer demand have pushed logistical services to use predictive
technologies that not only speculate about future events but also serve to
build the infrastructure for their arrival. Amazon’s anticipatory shipping
model confirms that the industry’s adherence to the dictum “time is money”
is pushing it into developments where “time is media” and where effective
media are the future of the medium of money. Andreas Sudmann exam-
ines artificial neural networks (ANNs) as predictive systems to illustrate
the fundamental importance of analyzing this technology in terms of its
temporal dimensions. One aim of his essay is to show how an investigation
of the temporal infrastructures of modern ANNs also contributes to a more
substantial discussion of their political challenges, such as can be seen in
the labor of crowdworkers, hired via platforms like Amazon’s Mechanical
Turk, for labeling and producing the massive amounts of learning data for
ANNs. Concluding the volume with a reading between Kittler’s argument
that technical media reorder time independently of human input and
Stiegler’s stance that technically mediated time is central to the experi-
ence of human time, Britt Paris looks at how the NSF-funded Named Data
Networking (NDN) protocol—a possible successor to the current TCP/IP
network protocol—reconciles “social concepts of time with computational
and architectural constraints in network design.” Paris draws on firsthand
interviews with NDN researchers in her examination of how user-facing
temporal experiences take second place to the imperative speed-up of
information transmission.
Notes
1. Peter Conrad, Modern Times, Modern Places (London: Thames & Hudson,
1998), 9.
2. Hartmut Rosa and William E. Scheuerman, eds. High-Speed Society: Social
Acceleration, Power, and Modernity (Philadelphia: Penn State Universi-
ty Press, 2009); and Hartmut Rosa, Social Acceleration: A New Theory of
Modernity (New York: Columbia University Press, 2013). It is also worth
acknowledging Paul Virilio’s longstanding work on technological accelera-
tion. See, e.g., Virilio, Speed and Politics: An Essay on Dromology, trans. Mark
Polizzotti (New York: Semiotext(e), 1986). For works on accelerationism see
Robin Mackay and Armen Avanessian, eds., #Accelerate: The Accelerationist
Reader (Falmouth, UK: Urbanomic Media, 2014); and Nick Srnicek and Alex
Williams, Inventing the Future: Postcapitalism and a World Without Work
(London: Verso, 2015).
3. See, e.g., Jonathan Crary, 24/7: Late Capitalism and the Ends of Sleep (New
York: Verso, 2013).
4. Emily Keightley, ed. Time, Media and Modernity (London: Palgrave Macmil-
lan, 2012), 4, 201.
5. Sarah Sharma, In the Meantime: Temporality and Cultural Politics (Durham,
NC: Duke University Press, 2014), 20.
6. Markus Krajewski, The Server: A Media History from the Present to the Ba-
roque (New Haven, CT: Yale University Press, 2018), 339.
7. See Deborah Cowen, The Deadly Life of Logistics: Mapping Violence in Global
Trade (Minneapolis: University of Minnesota Press, 2014). Also, for an
excellent exposition of the scalar dimensions of media infrastructures, see
the introduction to Lisa Parks and Nicole Starosielski, eds., Signal Traffic.
Critical Studies of Media Infrastructures (Urbana: University of Illinois Press,
2015).
8. See, e.g., Jane Bennett, Vibrant Matter: A Political Ecology of Things
(Durham, NC: Duke University Press, 2010); Ian Bogost, Alien Phenomenol-
ogy, Or, What It’s Like to Be a Thing (Minneapolis: University of Minnesota
Press, 2012); Richard Grusin, ed. The Nonhuman Turn (Minneapolis: Univer-
sity of Minnesota Press, 2015); Bruno Latour, Pandora’s Hope: Essays on the
Reality of Science Studies (Cambridge, MA: Harvard University Press, 1999)
and Reassembling the Social: An Introduction to Actor-Network-Theory (Ox-
ford: Oxford University Press, 2005); and Jussi Parikka, A Geology of Media
(Minneapolis: The University of Minnesota Press, 2015).
9. Bennett, “Systems and Things: On Vital Materialism and Object-Oriented
Ontology,” in The Nonhuman Turn, 223–239.
10. This is evidenced, for instance, by how contemporary online platforms pro-
duce distinct forms of device-specific “realtimeness.” See Esther Weltevrede,
Anne Helmond, and Carolin Gerlitz, “The Politics of Real-Time: A Device
Perspective on Social Media Platforms and Search Engines,” Theory, Culture
& Society 31, no. 6 (2014): 125–150.
11. Noble reveals how search engine results reproduce racist and sexist atti-
tudes from the cultures in which they operate and reinforce those attitudes.
See Safiya Umoja Noble, Algorithms of Oppression: How Search Engines
Reinforce Racism (New York: NYU Press, 2018).
12. McIlwain breaks down his story into two parts, detailing the untold stories
of Black entrepreneurs and innovators, who developed “black software” for
the personal computing revolution and early years of the internet, and the
adverse side of this history where new computing technologies were put to
the task of fortifying racial hierarchies. See Charlton D. McIlwain, Black Soft-
ware: The Internet and Racial Justice, from the AfroNet to Black Lives Matter
(Oxford: Oxford University Press, 2019).
13. Benjamin covers a wide range of racial biases in technical systems, from
campus architectures to online mapping tools, and outlines a set of coding
practices to counteract the power of this “New Jim Code.” Ruha Benjamin,
Race After Technology: Abolitionist Tools for the New Jim Code (Cambridge,
UK: Polity, 2019).
14. See, e.g., Friedrich Kittler, “Real Time Analysis, Time Axis Manipulation,”
trans. and with an introduction by Geoffrey Winthrop-Young, Cultural Poli-
tics 13, no. 1 (2017): 1–18.
15. See, e.g., Eva Horn, “Editor’s Introduction: There Is No Media,” Grey Room 29
(2008): 7–13; Erkki Huhtamo and Jussi Parikka, eds., Media Archaeology: Ap-
proaches, Applications, and Implications (Berkeley: University of California
Press, 2011); Parikka, What Is Media Archaeology? (London: Polity, 2012).
16. Wendy Hui Kyong Chun, “The Enduring Ephemeral, or the Future Is a
Memory,” Critical Inquiry 35, no. 1 (2008): 148–171; Wolfgang Ernst, Digital
Memory and the Archive, ed. Jussi Parikka (Minneapolis: University of
Minnesota Press, 2013); Mark B. N. Hansen, Feed-Forward: On the Future
of Twenty-First-Century Media (Chicago: University of Chicago Press,
2015); Adrian Mackenzie, “The Technicity of Time: From 1.00 oscillations/
sec to 9,192,631,770 Hz,” Time & Society 10, no. 2–3 (2001): 235–257; Florian
Sprenger, The Politics of Micro-Decisions: Edward Snowden, Net Neutrality,
and the Architectures of the Internet, trans. Valentine A. Pakis (Lüneburg:
Meson Press, 2015); Axel Volmar, ed., Zeitkritische Medien (Berlin: Kadmos,
2009).
17. For literature, Paul Ricoeur, Time and Narrative, 3 vols., trans. Kathleen
Blamey and David Pellauer (Chicago: University of Chicago Press, 1984,
1985, 1988). For film, Gilles Deleuze, Cinema 2, trans. Hugh Tomlinson and
Robert Galeta (Minneapolis: University of Minnesota Press, 1994); and Mary
Ann Doane, The Emergence of Cinematic Time: Modernity, Contingency, the
Archive (Cambridge, MA: Harvard University Press, 2002).
18. Ernst, Digital Memory and the Archive.
19. See Volmar, Zeitkritische Medien, 25.
20. See Erhard Schüttpelz, “Infrastructural Media and Public Media,” Media in
Action 1, no. 1 (2017): 13–61; Ulrike Bergermann, Monika Dommann, Erhard
Schüttpelz, Jeremy Stolow, and Nadine Taha, eds., Connect and Divide: The
Practice Turn in Media Studies (Berlin and Zürich: Diaphanes, 2020).
21. Media scholars have shown heightened interest in infrastructure studies
in recent years in part because the growing ubiquity of digitally networked
platforms, the global distribution of information, and the media ecologies
of “smart cities” have shattered the idea of media being single devices or
products that can be understood without recognizing the larger systems or
environments that enable them.⁠ A survey of important texts includes Parks
and Starosielski, Signal Traffic; John Durham Peters, The Marvelous Clouds.
Toward a Philosophy of Elemental Media (Chicago: University of Chicago
Press, 2015); and Nicole Starosielski, The Undersea Network (Durham, NC:
Duke University Press, 2015).
22. Susan Leigh Star and Geoffrey C. Bowker, “How to Infrastructure,” in
Handbook of New Media: Social Shaping and Consequences of ICTs, ed.
Leah A. Lievrouw and Sonia Livingstone (London: SAGE Publications,
2002), 151–162; Lauren Berlant, “The Commons: Infrastructures for Trou-

bling Times,” Environment and Planning D: Society and Space 34, no. 3
(2016): 393.
23. Keller Easterling, Extrastatecraft: The Power of Infrastructure Space (New
York: Verso, 2014), 13.
24. Paul N. Edwards, “Infrastructure and Modernity: Force, Time, and Social
Organization in the History of Sociotechnical Systems,” Modernity and Tech-
nology, ed. Thomas J. Misa, Philip Brey, and Andrew Feenberg (Cambridge,
MA: MIT Press, 2003), 186.
25. For the recent interest in questions of maintenance and repair, see Ste-
ven J. Jackson, “Speed, Time, Infrastructure. Temporalities of Breakdown,
Maintenance, and Repair,” in The Sociology of Speed, ed. Judy Wajcman and
Nigel Dodd (Oxford: Oxford University Press, 2017), 169–205; and Andrew L.
Russell/Lee Vinsel, “After Innovation, Turn to Maintenance,” Technology and
Culture 59, no. 1 (2018): 1–25.
26. Bruno Latour, “The Berlin Key or How to Do Words with Things,” in Matter,
Materiality and Modern Culture, ed. P. M. Graves-Brown (London: Rout-
ledge, 2000), 10.
27. Karl Marx, Capital, vol. 1, trans. Ben Fowkes (New York: Vintage, 1977 [1867]),
130. Marx writes: “As exchange-values, all commodities are only greater or
smaller amounts of congealed labor-time.”
28. Michel Serres with Bruno Latour, Conversations on Science, Culture, and
Time, trans. Roxanne Lapidus (Ann Arbor: University of Michigan Press,
1995), 45, 60.
29. For more on cultural techniques of synchronization, see Christian Kassung
and Thomas Macho, eds., Kulturtechniken der Synchronisation, Kulturtech-
nik (München: Wilhelm Fink, 2013).
30. Our understanding of technological temporalities as being segmented
into micro, meso, and macro scales matches closely the tripartite division
of time advanced by Fernand Braudel. However, the differences between
our model and Braudel’s should also be seen to signal important shifts in
human and technological temporalities in a world dominated by networked
digital media. See Fernand Braudel, “History and the Social Sciences: The
Longue Durée,” in On History, trans. Sarah Matthews (Chicago: University of
Chicago Press, 1980 [1958]).
31. Serres, 58.
32. Peters, Marvelous Clouds, 244.
33. S. Hultberg, “Some Observations on Systems for Automatic Acquisition
and Reduction of Nuclear Data and a Preliminary Report on the Computer
System Trask,” Nuclear Instruments and Methods 34 (1965): 127.
34. Gordon E. Moore, “Cramming More Components onto Integrated Circuits,”
Electronics 19 (April 1965), 114–117.
35. James Carey, “Technology and Ideology: The Case of the Telegraph,” in
Communication as Culture: Essays on Media and Society, rev. ed. (New York:
Routledge, 2009), 168.
36. Judy Wajcman, Pressed for Time: The Acceleration of Life in Digital Capitalism
(Chicago: University of Chicago Press, 2015), 3.
37. Obviously, integrated circuits use “wires” only in a metaphorical sense.
38. Other imaginative efforts include Daniel Hillis’ Clock of the Long Now and
the performance of John Cage’s As Slow as Possible over the course of 639
years at a church in Halberstadt, Germany.
39. Peters, Marvelous Clouds. See also Thomas Macho, “Zeit und Zahl. Kalender
und Zeitrechnung als Kulturtechniken,” in Bild, Schrift, Zahl, ed. Sybille
Krämer and Horst Bredekamp, Kulturtechnik (München: Fink, 2003),
179–92.
40. Harold Innis, Empire and Communications (Oxford: Oxford University Press,
1950). See also Innis, The Bias of Communication (Toronto: University of To-
ronto Press, 1951). As Innis showed, “time-binding,” or durable, media have
played a considerable role in the growth and maintenance of large political
and economic bodies such as empires.⁠
41. Émile Benveniste, “Language and Human Experience,” Diogenes 13, no. 51
(1965): 1–12. Benveniste elaborates three conditions of calendars, namely
initiating, directing, and measuring, the first of which involves the calcula-
tion of an axial moment: “All calendars share common characteristics which
identify the basic conditions which they must fulfill. They take off from an
axial moment which provides the zero point of the computation” (5).
42. Paul Ricoeur, Time and Narrative, vol. 3. (Chicago: University of Chicago
Press, 1988), 107.
43. Ricoeur, 109.
44. Benveniste, 4. A common translation of Benveniste’s temps chronique,
as for instance in the translation of Ricoeur’s Time and Narrative, Vol. 3,
renders the term “chronicle time,” and in many ways this translation better
captures in English its sense of time as a succession of events. Benveniste
himself, however, suggested its translation as “chronic time.” Etymologically,
the term “chronic” entered late Middle English as a cognate of the French
chronique and meant simply “of time, concerning time,” a meaning reaching
back to the Greek khronikos, from kronos. Nevertheless, its later connota-
tions of disease and addiction might helpfully express the adverse under-
side of the temporal structure of events.
45. Benveniste, 6.
46. Lewis Mumford, Technics and Civilization (New York: Harcourt, Brace &
Company, 1934), 14.
47. Mumford, 15.
48. James Henry Breasted, 1935. “The Beginnings of Time-Measurement and the
Origins of Our Calendar.” The Scientific Monthly 41 (1935): 289–304.
49. Peters, Marvelous Clouds, 1.
50. Eviatar Zerubavel, “The Standardization of Time: A Sociohistorical Perspec-
tive,” American Journal of Sociology 88, no. 1 (1982): 11. See also his books
Hidden Rhythms: Schedules and Calendars in Social Life (Berkeley: Universi-
ty of California Press, 1985), The Seven Day Circle: The History and Meaning
of the Week (Chicago: University of Chicago Press, 1989), and Time Maps:
Collective Memory and the Social Shape of the Past (Chicago: University of
Chicago Press, 2012).
51. Zerubavel, “Standardization of Time,” 3.
52. Mumford, 17–18.
53. Mumford, 13.
54. Mumford, 15.
55. See Rosa, Social Acceleration.
56. E. P. Thompson, “Time, Work-Discipline, and Industrial Capitalism,” Past
and Present 38 (1967): 56–97.
57. Thompson, 73.
58. Thompson, 73.
59. Thompson, 61.
60. See Susan Leigh Star and Anselm Strauss, “Layers of Silence, Arenas of
Voice: The Ecology of Visible and Invisible Work,” Computer Supported
Cooperative Work (CSCW) 8, no. 1–2 (1999): 9–30.
61. John Durham Peters, “Calendar, Clock, Tower,” in Deus in Machina: Religion
and Technology in Historical Perspective, ed. Jeremy Stolow (New York: Ford-
ham University Press, 2013), 42.
62. Ricoeur, 112.
63. Ricoeur, 105.
64. Ricoeur, 109.
65. Martin Heidegger, Being and Time, rev. ed., trans. Joan Stambaugh (Albany:
SUNY Press), 252.
66. Jimena Canales, A Tenth of a Second: A History (Chicago: Chicago: University
of Chicago Press, 2009), 3.
67. See also Soraya de Chadarevian, “Graphical Method and Discipline: Self-Re-
cording Instruments in Nineteenth-Century Physiology,” Studies in History
and Philosophy of Science 24, no. 2 (1993): 267–291.
68. See Laura Otis, “The Metaphoric Circuit: Organic and Technological Com-
munication in the Nineteenth Century,” Journal of the History of Ideas 63,
no. 1 (2002): 105–128; Henning Schmidgen, “Of Frogs and Men: The Origins
of Psychophysiological Time Experiments, 1850–1865,” Endeavour 26, no. 4
(2002): 142–148.
69. Karl Ernst von Baer: Reden gehalten in wissenschaftlichen Versammlungen
und kleinere Aufsätze vermischten Inhalts, Erster Theil. Reden (St. Petersburg:
Schmitzdorff, 1864), 237−284. Reprint in Volmar, ed., Zeitkritische Medien,
45–59.
70. See Jonathan Sterne, The Audible Past: Cultural Origins of Sound Reproduc-
tion (Durham, NC: Duke University Press, 2003).
71. See Marek Jancovic, Axel Volmar, and Alexandra Schneider, eds., Format
Matters. Standards, Practices, and Politics in Media Cultures (Lüneburg:
meson press, 2020).
72. Susan Leigh Star and James R. Griesemer, “Institutional Ecology, Trans-
lations, and Boundary Objects: Amateurs and Professionals in Berkeley’s
Museum of Vertebrate Zoology, 1907–39,” Social Studies of Science 19, no. 3

(1989): 387–420. See also Elaine K. Yakura, “Charting Time: Timelines as
Temporal Boundary Objects,” Academy of Management Journal 45, no. 5
(2002): 956–970.
73. Chun, “Enduring Ephemeral,” 148.
74. See also Ernst, Digital Memory and the Archive.
About the Authors
Kyle Stine teaches Film and Media Studies at Johns Hopkins University.
His writings on cinema and technology have appeared in Critical Inquiry,
Discourse, Grey Room, and the Journal of Cinema and Media Studies.
Axel Volmar is a Postdoctoral Fellow at the Collaborative Research

Center “Media of Cooperation” at the University of Siegen. He is author of
Klang-Experimente: Die auditive Kultur der Naturwissenschaften 1761–1961
and co-editor of Format Matters: Standards, Practices, and Politics in Media
Cultures and Interrogating Datafication: Towards a Praxeology of Data.
Part I
Media Philosophies of Time Patterning
1. The Suspension of Irreversibility:
The Fundamental (and Futile)
Task of Media
John Durham Peters
Abstract
This essay explores the media-theoretical and time-critical implications
of the facts that we all will die and that we move irreversibly forward
through time. It ranges broadly across several thinkers (such as von Baer,
Eliot, Kittler, and Borges) and situations (thermodynamic decay, being
stuck in traffic, and the rarity of randomness).
Keywords: media theory, time, thermodynamics, Kittler, von Baer,

irreversibility
We all know what time is, said St. Augustine, until we start to ask what it
is, in which case we really have no idea. Such a mixture of self-evidence in
the long shot and bafflement in the close-up is not usual in philosophical
inquiry. Most things that we know or think we know crumble under closer
inspection. Socrates enjoyed terrorizing Athenian notables by demonstrating
this fact. We all know that all people are mortal—that point is easy to get.
But that I am going to die—that is much harder. Everyone must die; I am not
everyone; therefore I am not going to die—this is the sort of warped syllogism
that we all live by! The more basic the theme, the more the philosophical
and existential stakes diverge. I want to explore the media-theoretical and
time-critical implications of the facts that we all will die and that we move
irreversibly forward through time.
Let’s start with the hypothesis of Karl Ernst von Baer, the nineteenth-cen-
tury Estonian nobleman and founder of the field of embryology, that there
is a correlation between an organism’s length of life and its sense of time’s
doi 10.5117/9789463727426_Ch01
42 John Durham Peters
passage and duration. In a lecture on, among other things, entomology from
1860—and thus too soon to take on board the Darwinian revolution—von
Baer notes the ineradicable human habit of positing ourselves as the measure
of time and space and of using our bodily proportions to serve as cosmic
measurements, with the result that we can never get big enough to see
everything.1 A second, he conjectures, was originally a heartbeat, and he
thinks that our experience is potentially dividable into six to ten events per
second, though he renounces the effort to specify the universal minimal
interval for sensory experience.2 Here he picks up a theme beloved of
phenomenologists and psychologists: how long is the now?
From this heuristic of sensory allotment—six to ten impressions per
heartbeat—von Baer then goes on to consider rabbits and cows. Here the
hypothesis gets a little goofy. Since cows have a pulse that is twice as slow,
and rabbits twice as fast, their experience must run slower and faster,
respectively, than ours. From this he weaves to the general point that any
organism’s experience of nature is a function of its physiology and lifespan.
Thus creatures with different life spans would have a very different physics
and astronomy than ours: a person with a thousand-fold acceleration of
experience (and thus consequent shortening of life) would live through one
phase of the moon; nights would stretch out into seasons enduring almost
a year; trees (if in summer) would never lose their leaves and water never
grow firm into an icy solid. A person with an even shorter lifespan of 42
minutes would never know that grass and flowers were not everlasting,
and the “entire organic world would appear to be lifeless” (262). Von Baer
is interested in varying the axis of temporal perception such that our eyes,
and especially our ears, could stretch in their powers of discernment. (His
time-lapse and slow-motion fantasy is obviously resonant for the pre-history
of cinema.) If our life were short enough—that is, if our perception of fine
slices of time were acute enough—we might even be able to hear light,
though of course we would no longer be able to hear what we now hear. (He
seems to assume that sensory acuity covaries with temporal perception; he
doesn’t imagine our music software that allows pitch-shifting while holding
the time axis invariant.) A universe of vibrations currently inaccessible to
us would appear if the temporal structure of our organisms changed: we
might see sound, hear heat, or even listen to the music of the spheres as the
planets oscillate (263).
Von Baer not only speeds up the elementary unit of perception but also
slows it down. What sort of universe would a person live in who had a pulse
a thousand times slower than ours—i.e. someone destined to live not 80
but 80,000 years? Von Baer notes that an 80,000 thousand-year person
The Suspension of Irreversibilit y: The Fundamental (and Futile) Task of Media 43
would barely perceive the shuttling between day and night and might never
recognize that it gets completely dark, since an entire year would take place
within “31 1/2” heart beats (you have to love the precision). The rumbling bass
tones that move the universe along would finally become audible. In his
Principles of Psychology (1890) William James riffs on the time-lapse effects
of von Baer’s speculations about the slow-pulse person: “mushrooms and
the swifter-growing plants will shoot into creation so rapidly as to appear
instantaneous creations; annual shrubs will rise and fall from the earth like
restlessly boiling water springs; the motions of animals will be as invisible
as are to us the movements of bullets and cannon-balls; the sun will scour
through the sky like a meteor …”3
The punch-line of von Baer’s speculations is that such varieties of temporal
experience are not merely hypothetical but probably already exist in the
animal kingdom of protozoa and insects, pigeons and bees. Just as there are
huge ranges in spatial perception—a microscope can transform a puddle
into a universal ocean—so there are radically varying scales of temporal
experience. The best standard, says von Baer, is the biggest—everything
is always too small. In some ways our knowledge is bound by our temporal
mode.
George Eliot (Mary Anne Evans) clearly knew of von Baer as both an
acquaintance of one of his most dogged disciples (Herbert Spencer) and as
one of the best informed people about German thought in nineteenth-cen-
tury England. In her masterwork Middlemarch (1871–1872) she describes the
insulation of our self-knowledge as a kind of existential comfort. “If we had
a keen vision and feeling of all ordinary human life, it would be like hearing
the grass grow and the squirrel’s heartbeat, and we should die of that roar
which lies on the other side of silence. As it is, the quickest of us walk about
well-wadded with stupidity.”4 In contrast to von Baer’s discontent with our
inevitably puny point of view, she found something to commend: sensory
dullness saves us from a fatal cosmic roar. Charles Sanders Peirce took
the same lesson, that a relatively short lifespan provides us with narrative
materials for meaning-making, such as beginnings and endings, and prevents
us from the wreckage that infinity would inevitably bring. “If man were
immortal he could be perfectly sure of seeing the day when everything in
which he trusted should betray his trust, and, in short, of coming eventually
to hopeless misery. He would break down, at last, as every good fortune, as
every dynasty, as every civilization does. In place of this we have death.”5
If we weren’t mortal, the universe might not make sense.
Such embedment in the finite definitely makes it hard to get a fix on
the historical strangeness of the world that we live in today in comparison
44 John Durham Pe ters
with the past. We have a necessarily myopic historical sense. As James

said of a human life accelerated into a duration lasting less than an hour,
“if born in winter, we should believe in the summer as we now believe in
the heats of the Carboniferous era.” A day blurs into a season blurs into
an epoch depending on your quotient of mortality. This scale-dependent
perception describes our historical near-sightedness quite well. We live
within a rare geological moment within a rare geological moment, the
upper holocene boundary. A pleasant spell of nice weather, known as the
holocene, has enabled civilization as we know it, by providing cooperative
natural conditions for settlement, agriculture, and large populations over
the past ten thousand or so years. Since around 1800, human population,
capacity, wealth and carbon have exploded. From 1900 to 2000, cropland
doubled, the human population quadrupled, the pig population went up
ninefold, energy use went up sixteen times, and industrial output went up
forty times.6 To von Baer’s 80,000-year person, the industrial era would
spring into being almost like a spontaneous creation, and to people born
in the middle of it like us, it might seem second nature.
We live in THE moment, the human apex of planetary domination. The
fact that we are alive is a tribute to the favorable natural and anthropogenic
conditions for the flourishing of large human populations. If we were to
roll the dice, chances are we would be born when we were. Somewhere
between 5 and 10% of all human beings who have ever lived are alive right
now, depending on your estimate of the total number of humans who’ve
lived on earth (taking 107 billion as a standard estimate). In his novel 2001,
which parallels the film, Arthur C. Clarke suggested that there were thirty
dead people who had once lived for every person alive: “Behind every man
now alive stand thirty ghosts.” He was writing in 1968, when the human
population was under 4 billion. If he were writing today when it is pushing
8 billion, he’d have to change it to fifteen ghosts. In five decades, the odds
of being alive in this moment have doubled.
This fact has subtle warping effects on our perceptions. Our lives, our
bodies, our minds, our visions of the world have all been shaped by time
on a planet when there is a vast herd of domesticated human beings—who
weigh about one third of the total vertebrate matter on earth, the other two
thirds being accounted for almost entirely by animals that humans keep in
order to eat.7 Those birds chirping outside, the whales in the sea, the feral
dogs of Athens are all only a tiny fraction of vertebrate biomass compared
with the anthropocentric part. Never has the planet been arranged like
this. To us it seems as natural and eternal as daylight or dandelions would
be to a 42-minute human.
Consider a puzzler. Why is it that when you are sitting in traffic, you
feel like you are always in the slowest moving of all the lanes? You feel that
way in part because you probably are in the slowest moving lane. This
perception is not a result of egocentric woe-is-me bias; the odds are that it
is highly accurate. The slowest lane is the lane with the most cars in it. The
odds are that you will be in the lane with the most cars. What you think of
as your private perception or a spell of bad luck is in fact the infrastructural
fact of your connection to larger populations and conditions than you can
perceive. (In the famous words, being precedes consciousness.) Extraordinary
geological conditions profoundly shape our everyday perceptions but our
lives are too short to take those conditions as anything but normal. Our
moment is both unique in human history, since for only a couple centuries
has anything like this been possible, but also widely shared by billions. We
have lots of synchronic, but little diachronic solidarity.
Media theory has a mixed mission. Partly, it aims to make the unconscious
conscious, to bring out the thisness of a that and the thatness of a this (as
Kenneth Burke said). But media theory also aims to defy any strong split of
subject and object. Our consciousness converges with our conditions. The
frustrated driver in the slow lane’s view of the traffic is a function of where
they are in it in the same way that our view of history and of the universe is
a product of the history of the universe. The anthropic principle, to simplify,
suggests that the conditions necessary to produce an observer predestine the
kind of observable universe there could be.8 What we can know is deeply tied
up with the processes that have produced us as observers. (This argument
updates the German idealist claim found especially in Hegel and Schelling
that history enables the collusion of subject and object since they are its
common progeny.) The only universe in which we could exist, and period
of the universe in which we could exist, is one that is biased—to invoke the
ever-useful term from Harold Innis—in systematic ways. It would have to be
old, spread out, and cold, but not too much. It would have to have a certain
kind of chemistry, in our case, organic, which presupposes supernovas,
the cosmic furnaces that forged by nucleosynthesis every element more
complex than hydrogen and helium. The size and scale of the cosmos are
in some way complicit or at least correlated with our ability to know them.
Or, maybe better put, our minds are shaped by the same conditions that
we are trying to observe.
On a cosmic scale, we live in a relatively brief moment in the history
of the universe, and a brief corner of its diaspora, in which things stick
together, in which houses cohere and trees thrive, on an earth in which
matter is 1030 times more concentrated than the cosmic average. In the
eons, the cosmos will degrade into diffuse boring nothingness, unable to
stick together in any way. When the universe is 1018 seconds old, as it is now,
a certain kind of knowledge is possible. The entropic cost of observation
is not prohibitive. It is possible to discern causal chains. By the time the
universe is 10107 seconds old, it will be out of thermodynamic and cognitive
gas: assuming continued expansion (against a possible Big Crunch) each
particle will be as far from every other as the current universe is wide.9
Maybe von Baer should have applied his sliding time scale not to organisms
but to the cosmos. If you imagine yourself with a millennial, million- or
billion-year heartbeat, you can feel the future pulling the universe apart,
and even your body; some of us, indeed, can already feel that. If we lived in
a much later era of the universe, the nature of things would have changed
enough to require a different kind of knowledge, mode of observation, and
observer. The physical limits of the knowers are shared in some ways by
the physical limits of the things to be known.
These reflections grow out of thermodynamic debates about order since
the later nineteenth century. Does order belong to things or to mind? The
point of Maxwell’s demon, the famous thought experiment, was to turn en-
tropy uphill by sorting out gas molecules into fast and slow. Maxwell’s demon
somehow knew where all the molecules were without exerting any effect
on their behavior. Mind was invisible, aloof and yet somehow everywhere
without interfering—rather like a parody of the god of mathematicians or
of Descartes’ thinking substance without extension, a point without materi-
ality.10 The stakes were that this smart but physically non-intervening being
could push entropy uphill and thus save the universe from the inevitable
“heat death” of universal degradation. Clearly there was a cultural element
here: the morality of husbanding energy well, masculinity under threat
(as it chronically is), the waning coal supply, etc. for worried Victorians.11
Quantum physics put an end to the dream of a noninterventional omni-
science. Knowledge and the behavior of subatomic particles turned out to
be woven from the same fabric. How an electron behaves depends on how
we look at it. As John von Neumann said, “An observation is an irreversible
process.”12 The more a system is monitored, the more it is tampered with.
You can distort the behavior of a system by watching it—a fact that is not
only true in culture, as Stanley Milgram, Michel Foucault and many others
have shown, but in nature as well. The particles in the quantum panopticon
monitor their monitors. There is no representing without intervening, to
invoke Ian Hacking.13 Knowledge is physical. Norbert Wiener, who must
count as one of the great thinkers of hardwired temporality, noted: “In
nineteenth century physics, it seemed to cost nothing to get information.”14
We know now just how expensive information is, both thermodynamically

and in a more ordinary sense in our digitally-scraped lives.
Time, of course, is just as mysterious an entity as information. Is it a
feature of the subject—time flies while you’re having fun, or while your
heart beat is faster, as von Baer thought—or of the object—relentless clock
time? Is time’s arrow, its irreversible flow, just a function of entropy, of the
fact that everything in the universe tends to move from a more ordered to a
less ordered state? Is time the medium of entropy or its expression? Is time’s
irreversibility necessary or just highly probable? There is no physical law,
in fact, saying that you cannot scoop up the precise same moles of water
molecules that your cup just poured into the sea, but the odds are crushingly
minute that you will. You could reconstruct the house that burnt down in
the fire from its ashes: there is no dictatorial law saying it is impossible, only
absurdly slight in probability. Maybe tables really do occasionally rise in
the air due to fortuitous conjunctions of Brownian motion and maybe time
occasionally, in coincidental convergences, lurches backwards or hovers
briefly in slight hiccups.
I once wrote: “Ironically enough, just as physics was discovering irre-
versibility media engineered reversibility” but I no longer think it ironic.15
Time-reversibility in media shows the time-irreversibility of life. The great
analog media of time-axis manipulation, phonography and film, showed
that optical and acoustic data, once arranged into series, could be sped
up, slowed down, or reversed in playback, but they also showed with fresh
clarity that some natural processes cannot be reversed. It is remarkable how
often cinema and sound-recording pop up in discussions of irreversibility.
Wiener, for instance, in the famous opening chapter to Cybernetics on
Newtonian and Bergsonian time, uses a nice film metaphor. He argues
that astronomy is reversible: “The music of the spheres is a palindrome,
and the book of astronomy reads the same forward as backward” (31). That
is, a film of the planets, played backwards, would be possible according
to Newtonian mechanics. “On the other hand, if we were to take a mo-
tion-picture photograph of the turbulence of the clouds in a thunderhead
and reverse it, it would look altogether wrong. We should see downdrafts
where we should expect updrafts, turbulence growing coarser in texture,
lightning preceding instead of following the changes of cloud which usually
precede it …” (32).
A few years later, Karl Popper objected, saying that Newtonian physics
also allows irreversible processes, using the example of “a film taken of a
large surface of water initially at rest into which a stone is dropped.” The
reversed film will show the waves increasing instead of dissipating, flowing
in rather than out, and culminating in the sudden appearance of a placid

surface. As he drily notes, “This cannot be regarded as a possible classical
process.”16 Popper’s point seems to be that entropy is not necessary, only
highly probable, in keeping with his view of an indeterminate universe.
Under some highly special or bizarre conditions, a stone might be able to
spring out of a turbulent body of water, leaving it completely smooth: it
is possible, only extremely improbable. Film is the sine qua non of both
Wiener’s and Popper’s staging of reversibility. (Perhaps Newton’s use of
calculus to slice motion into asymptotically small intervals is already cinema
avant la lettre, or its mathematical antecedent at least.)
Fortunately for media scholars, we have a robust tradition of theorizing
the relationship between entropy and media, time and mortality that goes
back to Friedrich Kittler’s incandescent texts on time-axis manipulation
and lightning and thunder. The first, “Real Time Analysis, Time Axis Ma-
nipulation” (1990), starts with the observation that the alphabet was the
first technology of time-axis manipulation. Kittler notes that nature, unlike
writing, does not recognize the copresence of full and empty slots. There are,
that is, no yes–no, i.e., discrete, machines in nature—no placeholders, no
decimal points or zeroes, no spaces between words. You can rearrange letters
along their syntagmatic axis but you can’t do that to natural phenomena.
“You can certainly reverse the word LEBEN [life] and logically get the word
NEBEL [fog], but not life itself, to say nothing of fog itself.” (In English we
can think of golf and flog or live and evil.) His examples are not by chance
the core objects of the entropic imagination: life and weather. Life doesn’t
regenerate itself once it is dead; thunder does not go before lightning. These
are one-way events. In discussing Georges Méliès’s pioneering 1895 film
on the mechanical butcher that shows a sausage turning into a pig, Kittler
heralds: “Und die Auferstehung des Fleisches ward Anschauung.” Geoffrey
Winthrop-Young translates: “Behold the resurrection of the flesh!” This is
an admirable solution to translating a phrase that echoes both the Gospel
of John and the chorus mysticus from Goethe’s Faust: “And the resurrection
of the flesh became perception.” Unfortunately for Kittler’s argument, no
such thing occurs in the film, but as usual in Kittler, we shouldn’t let the
facts get in the way of an interesting idea: where only theology heretofore
has dared contemplate such reversibility, now media theory steps forward.
In a historically unprecedented way, analog media in the late nineteenth
century “made for the first time contingent time-series events recordable.”
As with writing and the alphabet before, recording media opened up a
realm of reversibility and play that is briefly exempt from the relentless
grinding of real time.
The Suspension of Irreversibilit y: The Fundamental (and Futile) Task of Media 4 9
In a lecture on “Lightning and Series—Event and Thunder” first published

in 2003 Kittler continues in his transcendental way. This is one of the most
sublime and baffling texts I know on any subject. In it he tells a tale of three
moments in the history of acoustic media: Zeus’s thunder in ancient Greece,
Father Mersenne’s experimental cannonball blasts in the seventeenth
century, and modern computers directly playing with the universe of
frequencies thanks to Fourier analysis. (Sometimes it sounds as if Kittler
has been reading von Baer.) The single stroke of the event/lightning cannot
be perceived by humans due to the slowness of nervous propagation in
our bodies; we have access only to the rumble of the vibrations, the series/
thunder that follows. The gods had lightning, but humans had thunder. But
modern computation and mathematics bring together extremes that were
separated in antiquity. We can’t fly like Icarus into the Olympian range
where gods read frequencies, but our computers can. Below the sensitivity
of our ears lies thunder’s frequencies; above the sensitivity of our eyes
lies lightning. If what Kittler calls our “filters (eyes and ears, etc.)” could
ascend or descend the ladder, they could play with light as we do with
sound, with lightning as we do with thunder. (He gushes about the recent
sonification of an earthquake in Kobe, Japan.) Any still image of sound such
as a spectrogram necessarily implies a temporal dimension, typically along
the x-axis. In contrast, many images appear frozen in time. An image does
not necessarily require a time-axis (though images can imply one, such as
a blurred photograph). To the gods, however, every image would require a
time-axis. If our eyes operated at god-like—computer-like—speeds, we’d
know that every still visual image is also a slice of time. Pictures only pretend
to freeze time for beings with slow sense organs. The old division of spatial
image and temporal sound is just a species-specific result of the speed of
our sense organs. Other species would see the time in every image or freeze
sound into shape. Such a species, for Kittler, is the computer.
Kittler goes theological, or at least proposes what Foucault once called “an
analytic of finitude”: “as nature has consigned us to a finitely broad spectrum
in her immeasurable range of frequencies so the old gods consigned us to
finitely long lives in the time domain.” Here human sensory limitation and
mortality again become the marks of our condition, defined both thermo-
dynamically and media-theoretically. Note the two sentences that begin the
lecture: “Nothing is denied more to us temporal beings than to know time.” The
second one—the lecture was given to an audience of graduate students—is
practical: “When have you read your sources enough to be confident of them?”
Starting from a primal ban on human knowledge—his set-up reflects the
Edenic scenario of knowledge the transgression of whose limits would make
humans dangerously or excitingly like God—Kittler shifts to wondering

when we will ever have the confidence to call our studies done. For him, our
finitude in perception and knowledge go together with the irreversibility of
nature and reversibility of media. As he concludes, with relevance to both
any research project as well as to Heisenberg’s uncertainty principle, “Every
possible choice between being and time comes at the cost of knowledge.”
Sybille Krämer astutely sees in time-axis manipulation the very heart
of Kittler’s thinking. She offers my thesis in a nutshell: “the most basic
experience of human existence—and this is relevant because the human
is, after all, a physical being—is the irreversibility of the flow of time. Tech-
nology is precisely the attempt to ward off (or charm) this irreversibility. In
media technology, time itself becomes one of several variables that can be
manipulated.”17 In life, time goes in one direction only; in media, time can
be stretched, compressed, reversed, yo-yoed, in short, edited. With media
we can hopscotch through time, but in our bodies, we get one second older
every second. Our techniques allow parallel processing, but we live according
to the strict law of serial processing.
Hartmut Winkler brilliantly builds on Kittler and Krämer as well as
Bernard Vief to make the point that temporal reversibility in media owes
to spatialization. As he puts it in his book on processing, “media in general
are machines that transcode space into time and time into space.”18 Time
axis manipulation operates by the transposition of sequentially occurring
events into spatially ordered symbols, which then can be rearranged without
cost. LEBEN becomes NEBEL without danger or expense. What Winkler calls
“the geometry of time” occurs this way in alphabetic writing, in magnetic
tape, in a vinyl LP, and internet streaming—in short, in any kind of data
storage that enables a temporal playback of serially accessible data. Space
here serves time. In the first of these, speech, which is a linear stream in
time, is projected onto the space of the writing surface with the divisible
corpuscles of letters. This is the fundamentally unnatural act of putting
placeholders into a linear system. (Nothing is discrete in nature.) This
argument is implicit in Kittler, specifically in his praise—which goes further
in his grandiose final books on music and mathematics—of the ancient
Greeks for having “placed the time of tones into space.”19 He was thinking of
how on a monochord, an ancient instrument, or Odysseus’s bow, you could
point to places—like frets on a guitar—that sounded the octave, fifth, and
fourth. Here space marks sound. The only way to jump out of real time is
to take advantage of space’s ability to transcend time.
Winkler builds on the Brothers Grimm tale about the hedgehog and the
hare who challenge each other to a race. The hare, though faster, always
loses. The hedgehog has placed a second hedgehog, his wife who looks just
like him, at the end of the furrow in which the race is to occur. It takes
no time for the doubled hedgehog to get from one end to the other. The
hedgehogs taunt the hare as the latter runs like lightning toward the finish
line, “I am already here.”20 The hedgehog is the principle of simultaneity,
or of multiple copies pre-distributed in space, of parallel processing which
costs no time, while the hare is the principle of transmission, or of serial
processing, which always costs time. (Marisa Leavitt Cohn shows this
acutely well in chapter 10.) The work of past time is recorded and available
for instant access without the requirement of scrolling in real time. A nice
point clinching Winkler’s argument is buried in a footnote. A crowd is
gathered. Two medics and a doctor are working on a man lying on the ground.
Someone, when asked what happened, reports: he was dead.21 Making death
past tense is the ultimate in time axis manipulation, and Kittler was right
to think about resurrection here. To say that someone was dead is not just
a grammatical puzzle, but the essence of TAM. No question, save perhaps
the incarnation, is more important to Christian theology—or to Kittler’s
media theory.
It’s important to note that we can’t actually manipulate real time. Tempo-
ral and spatial fixity are not symmetrical. Playback is still new every time.
Time is always mercilessly and blessedly disappearing. This is the point
about time-axis manipulation being ultimately futile. You can’t record sound;
you can record instructions that allow you generate it anew every time in a
form that is functionally but never completely identical. Sound as pressure
or vibration will always be slightly different in each new instantiation.
Kittler quotes Hegel’s “merciless sentence” that sound exists by disappearing
or disappears by existing. Recording media don’t hold sound—they hold
scripts that inscribe patterns for sounding-devices to perform, but once
called forth, sound will dissipate once more. Sound is dissipation—no way
around it. Light is too, but we are too dull to know better.
I recently proposed that cetaceans—dolphins, porpoises, and whales--
model intelligence consigned to nonmaterial expression by their aqueous
environment and lack of manipulative limbs. They live in matter, but
cannot mold it. What matter is to dolphins, time is to us. Dolphins lack
three-dimensional modeling; we lack four-dimensional modeling. Unlike
such smart marine mammals, whose durable engineering is limited to
moving mud around on the ocean floor or pulling things around, humans
have developed a vast array of durable media, material moldables into which
we can imprint our schemes. Among the greatest of all human technical
achievements is the ability to record the data of happenings in spatial form
and then spin them back later into real-time. The playback of course takes
time and occurs in time but we cannot, in the end, capture time. No material
ultimately catches and molds real time. We can only impose symbols serially
onto spatially manipulable media. Any time control erodes just as quickly as
the dams the dolphins build. We cannot get anything to stick for good in real
time; at best we use symbolic machines to enable reversibility under special
conditions. We have to convert time into space to manage it. We can only
“write” events on some lasting spatial substrate that play them back with
some kind of fidelity. The best we can do is substitute space for time, and
firm for fluid matter. We lack any medium of four-dimensional plasticity.
What if we could mold real time? It might sound like bliss to be able to
recall our sweetest moments in all their fullness, and have memories as
strong as the experience. It also might be a complete hell, if pain would recur
again and again in memory as intensely as it took place in experience. (The
name for such repetition is trauma, a psychological fact intimately connected
with playback technology.22) The lack of four-dimensional manipulability
is at once a cursed and handsome condition. That we hurtle forth serially
in time, irreversibly, means that we are compelled to choose. Irreversibility
is actually tied up with conditions of meaningfulness. (This echoes Peirce’s
point that death is our defense against complete annihilation.)
Everything we do is completely improbable. It is not that our deeds lack
meaning but that it is impossible for us to act in anything but a meaningful
way. Every act is a choice against infinitesimal odds. Take the act of writing.
At the most minimal, take 26 letters plus a space: 27 options are found for
every parking spot on the line of writing (a typical keyboard allows more
like 90 characters, if you include the shift key, and then there are all the
special characters.) Type 56 characters (=27 to the 56th power or about 10 to
the 80th) and you have already met the number of protons in the universe;
type 70 characters (about 10 to the 100th, that is, a googol) and you are well
beyond the highest estimate of how many elementary particles there are in
the universe (10 to the 97th). Type 2000 characters, around a double-spaced
page, and you have “unimaginably” succeeded the size of the universe. This
of course is Borges’s Library of Babel.23 The possibilities multiply no less
staggeringly with speech. With 44 phonemes in English, for instance, as soon
as you articulate 61 of them you surpass a googol of possibilities. (There is
no grammar for action, but the same vastness prevails.) There are so many
options, but each choice narrows the next. If you write the letter Q, odds are
that the next letter will be a U. We can write Iraqi or QWERTY and still make
sense, but the possibilities for a meaningful choice shrink to a needle’s eye
the further you go. Thermodynamics says that there are many more ways
that things can be chaotic than organized. Muddle has near infinite versions,
tidiness only a few. We can’t help but be meaningful. The burden we bear
is not, as the pop existentialists thought, the universe’s lack of meaning,
but the inability to stop making sense. Our symbolic machines—spoken,
written, optical and acoustic, digital—lift us out of time’s flow at the price
of being obedient to their rules.24 We walk through time unconsciously
crushed by the barometric pressure of uncountable possibilities. And we
can only say, do, see, or hear anything in the moment between past and
future. O tempora, O media!
Notes
1. “Welche Auffassung der lebenden Natur ist die richtige? und wie ist diese
Auffassung auf die Entomologie anzuwenden?” (1860), Reden gehalten in
wissenschaftlichen Versammlungen und kleinere Aufsätze vermischten Inhalts
(St. Petersburg: Verlag der Kaiserlichen Hofbuchhandlung H. Schmitzdorff,
1864), 3 vols. Facsimile reprint 2 vols. (New York: Arno, 1978), 1:239–284. The
media-theoretical implications of this lecture are nicely treated by Stefan
Rieger, “Der dritte Ort des Wissens: Das Gedankenexperiment und die
kybernetische Grundlagen des Erhabenen,” Zeitkritische Medien, ed. Axel
Volmar (Berlin: Kadmos, 2009), 61–80, esp. 66–72.
2. He was in tune with the later nineteenth-century interest in the tenth of a
second as an interval of perception: see Jimena Canales, A Tenth of a Second
(Chicago: University of Chicago Press, 2009).
3. The Principles of Psychology (1890; New York: Dover, 1950), 1:639.
4. Middlemarch (1871–2; New York: Penguin, 1994), 194.
5. Charles Sanders Peirce, “The Doctrine of Chances,” in Chance, Love, and
Logic, ed. Morris R. Cohen (Lincoln: University of Nebraska Press, 1998), 72.
6. John R. McNeill, Something New Under the Sun: An Environmental History of
the Twentieth-Century World (New York: Norton, 2000), 360–361 and passim.
7. Vaclav Smil, Harvesting the Biosphere (Cambridge, MA: MIT Press, 2013).
8. John D. Barrow, The Constants of Nature (New York: Vintage, 2002), chap-
ter 8.
9. William Poundstone, The Recursive Universe: Cosmic Complexity and the
Limits of Scientific Knowledge (Mineola, NY, 2013 [1985]), 163.
10. See James Gleick, The Information: A History, A Theory, A Flood (New York:
Pantheon, 2011), chap. 9.
11. Crosbie Smith and M. Norton Wise, Energy and Empire: A Biographical
Study of Lord Kelvin (Cambridge: Cambridge University Press, 1989).
12. In Poundstone, Recursive Universe, 67.
13. Ian Hacking, Representing and Intervening (Cambridge: Cambridge Univer-
sity Press, 1983).
14. Norbert Wiener, The Human Use of Human Beings: Cybernetics and Society,
2nd ed. (Boston: Houghton Mifflin, 1954), 29.
15. John Durham Peters: The Marvelous Clouds (Chicago: University of Chicago
Press, 2015), 310.
16. K. R. Popper, “The Arrow of Time” (Letter), Nature 177 (17 March 1956): 538.
17. Sybille Krämer, “The Cultural Techniques of Time Axis Manipulation: On
Friedrich Kittler’s Conception of Media,” trans. Geoffrey Winthrop-Young,
Theory, Culture, and Society 23 (2006): 93–109, 96, trans. slightly modified.
18. Hartmut Winkler, Prozessieren: Die dritte, vernachlässigte Medienfunktion
(Munich: Fink, 2015), 12.
19. Friedrich A. Kittler: “Blitz und Serie – Ereignis und Donner,” in Zeitkritische
Medien, ed. Axel Volmar (Berlin: Kadmos, 2009), 157; see also Musik und
Mathematik. Vol. 1: Hellas, Part 1: Aphrodite (München: Fink, 2006).
20. “Der Hase und der Igel,” Die Märchen der Brüder Grimm: Vollständige Aus-
gabe (Augsburg: Goldmann, 1981), 559–61.
21. Winkler, Prozessieren, 237n356. This brilliant chapter is available in English
translation at homepages.uni-paderborn.de/winkler/hase-e.pdf.
22. Amit Pinchevski, Transferred Wounds: Media and the Mediation of Trauma
(New York: Oxford University Press, 2018).
23. William Goldbloom Bloch, The Unimaginable Mathematics of Borges’
Library of Babel (New York: Oxford University Press, 2008). See also my “Re-
semblance Made Absolutely Exact: Borges and Royce on Maps and Media,”
Variaciones Borges 25 (2008): 1–23, https://www.borges.pitt.edu/bsol/docu-
ments/Peters.pdf.
24. The guarantee of encryption in non-quantum computing—i.e. the inability
to step sideways out of time for hedgehog like parallel calculations—is the
same as the proof of insincerity in interpersonal ties: finitude, that is, the
scarcity of time.
About the Author
John Durham Peters, Maria Rosa Menocal Professor, Department of English,

Yale University, New Haven CT, author most recently, with the late Kenneth
Cmiel, of Promiscuous Knowledge: Information, Image, and Other Truth
Games in History (Chicago, 2020).
2. Time and Technology:
The Temporalities of Care
Gabriele Schabacher
Abstract
By focusing the temporalities of care, the chapter analyzes a special relation
between time and technology that underlies the making and persisting of
media and infrastructures. I propose to differentiate between four types
of care practices with corresponding different temporal patterns that are
highly relevant for the functioning of technological systems in the past
and present. First, the retrospective response to unforeseen interruptions
(repair); second, the prospective routine procedure to prevent all forms of
disorder (maintenance); third, a neglect of care that leads to devaluating
infrastructure (abandonment) as well as—fourth—forms of revaluation
in changing contexts (repurposing). Taking the new Berlin airport BER as
an example, it will be shown that infrastructures exhibit different layers
of temporality formed by these cyclic and repetitive processes of care
and their transforming effects. Thus, even the performance of the most
“hardwired,” late modern technology systems is crisscrossed by temporal
regimes that stem from older, non-modern temporalities of care.
Keywords: repair, maintenance, care, infrastructure, cultural techniques,

abandonment
In this article, I analyze the relations between time and technology that
underlie the making and persisting of media and infrastructures by focusing
on the temporalities of care. I take up the notion of the hardwiredness of in-
frastructures in the sense of their consolidation and material embeddedness,
raising the question of the practices responsible for achieving and maintain-
ing technical structures. This implies understanding hardwiredness not as a
fixed state of being of materially “wired” artifacts or digital infrastructures,
doi 10.5117/9789463727426_Ch02
56 Gabriele Schabacher
but rather as a network effect with relative duration brought about by

specific types of labor. Seen in this light, the effect of being “hardwired”
appears not only in processes of industrialization and standardization since
the nineteenth century but also in older, premodern cultural techniques
responsible for the conservation of things. I understand caring activities
in the sense of a certain type of constant, repetitive, and comparably slow
work that tends to be invisible and naturalized in the mundane routines of
everyday life. The advantage of such a perspective for media studies is that
it enables insight into the various temporalities of infrastructures, which, as
I argue, form a vertical layering of different ages within one and the same
infrastructure, leaving several strata to be uncovered.
Recent research in science and technology studies, urban studies, ar-
chitectural studies, and the history of technology has shifted attention to
processes of upkeep and repair as highly relevant to the functioning and
stabilization of organizations and infrastructures.1 However, questions
of repair and upkeep are less addressed with respect to the functioning
of digital infrastructures.2 For this reason, I take up the insights of repair
studies, especially as developed in the fields of architecture and urban
studies, in order to show their relevance for a discussion of hardwired
infrastructures under digital conditions. Repair studies analyzes the care
of physical infrastructures, such as buildings and transport systems, and
acknowledges the materiality of those things, people, and codes out of which
also “the digital” is made, whether cables, workflows, or programs. Such a
perspective has already been productive in studying the work places and
labor practices of early computing, for example in analyzing the so-called
software crisis of the 1960s as “essentially a maintenance problem.”3
The concept of care has recently been addressed with regard to ethical
implications and the type of work it generates, together with the question
of who cares for whom and the affective relationships that go along with
this practice. According to Maria Puig de la Bellacasa, the three dimensions
of care—“labor/work, affect/affections, ethics/politics”—are distributed in
all relational situations. 4 Care work can refer to people (including the care
for oneself),5 things and artifacts, but also to machines and infrastructures.
Thus, the gender bias6 care work implies, the “invisible work”7 it goes along
with and the “tacit”8 knowledge of carers are relevant aspects for an analysis
of care work, whether this relates to humans or the realm of material objects.9
Care is in general closely linked to concern10 and draws on affective relations
between human and nonhuman actors. Latour’s study of the transport
system Aramis discloses the affective relationship with technology (the love
of technology) as the most important aspect for accomplishing or giving up
Time and Technology: The Tempor alities of Care 57
on a project.11 Only on the basis of an attachment to a certain thing can an

affective relation of care be built.12
Studies of care stress two kinds of temporalities involved in caring
practices. On the one hand, they refer to the fragility of the material world
and processes of decay; on the other hand, they point to the work side of
the problem, that is, the array of continual “reordering micro-processes.”13
Steven Jackson claims that the temporalities of maintenance and repair blend
“the unruly timelines of things.” Drawing an analogy between the care of
the human body and the repair and maintenance of objects, he describes a
temporal relation in which the care of things involves “a staying with in time
and place,” that means, an adjusting of one’s own time to “other temporal
flows and processes, including the temporalities of breakdown and decay.”
Repair-as-care, according to Jackson, implies opening and tying oneself to
the “rhythms, flows, and timeliness of another.”14
Considering these ideas, I propose to differentiate between four types of
practices that concern the care for technology and, correspondingly, four
different types of temporal patterns. First, the activities of restoring—es-
pecially practices of repair—that respond to unforeseen events, ranging
from catastrophic situations to ordinary breakdowns and malfunctioning,
and rebuild a previous state of affairs. With respect to their temporal logic,
these activities can be understood as a retrospective form of care. Second, the
practices of maintenance, upkeep, and regular service of technical systems
that are intended to prevent all forms of disorder by routine and planned
procedures of control. Insofar as these practices are directed toward the
future of technology, I will understand them as a prospective form of care.
Third, the devaluation of infrastructures through decay and deterioration.
These processes exhibit an inversion or negative form, a lack of care char-
acterized by neglect and indifference. Fourth, and sometimes as a result
of decay, processes of revaluation and repurposing that I understand as a
redirection of care by reusing an existing infrastructural setting.
Although the distinctions between the four different types of caring
activities are not clear-cut, I take them to be useful for analyzing temporal
regimes of hardwiredness. This offers a perspective on care practices that
concentrates on their status as cultural techniques and demonstrates that
maintenance carries forward a non-modern notion of care stemming from
the cultural techniques in the sense of cultura. In doing so, the article wants
to accentuate two general aspects. First, in light of these various temporal
patterns of care, hardwired infrastructures appear less as linear, enduring,
stable objects and systems than as cyclical and repetitive processes of
formation and transformation. Even the most hardwired, late-modern,
high-technology systems are crisscrossed by different temporal regimes,

especially on the level of micro practices that stem from older, non-modern,
and, to some degree, more organically and culturally embedded temporalities
of care. This becomes obvious in particular with regard to what I will call
“maintaining-as-waiting” (warten). Second, different layers of temporality
and care practices exist within any single infrastructural setting. These
various temporalities and “ages” of infrastructures produce effects of inter-
ference, accumulation, and repetition because regimes of work, affective
relationships, juridical decisions, lifetimes of components, to name only a
few, are not distributed in a coordinated way with regard to actors and goals.
Layered Temporalities: The BER project
The different layers of temporality inherent in a given material infrastructure

can be illustrated by the well-known example of the Berlin Brandenburg
Airport. The BER had been under construction since 2006, after its initial
planning in 1995. The opening had been postponed several times since 2011
(see Table 2.1). The last target for the official opening date was October 2020,
but this again had been called into question because of an internal (leaked)
TÜV Rheinland report of 2019 revealing serious safety def iciencies in
important technical installations and suggesting an opening not before
2021.15 In the public, the BER is known as a case of severe mismanagement,
corruption, and poor construction planning and execution. The already long
list of construction flaws became longer every day (the TÜV report lists 11,519
technical flaws altogether), and the costs for not bringing the airport into
operation amounted to about 36 million euro per month. The Flughafen
Berlin Kosten website displayed the rising costs per second in real time.16
By 2020 the whole enterprise had cost the taxpayer up to 5.3 billion euro
already and would have increased to an estimated 7 billion euro by 2025.
Yet from an infrastructural perspective, the BER project demonstrates not
only mismanagement but also the heterogeneity of the socio-technical “ac-
tor-world” in question in a paradigmatic way,17 bringing together materials,
political parties, contracting companies, reports, and investigations,18 as well
as considerations of environmental impacts, increases in passenger traffic,
design flaws, and costs. The BER project shows the failure to come to mutual
and simplifying translations to align all these heterogeneous elements in
one direction, namely the completion of the construction project. Instead,
each entity operates as a full mediator of the whole process, following its
own directions.19
Table 2.1. Postponement of BER opening dates
Announcement Opening
September 5, 2006 (original) October 30, 2011

June 25, 2010 June 3, 2012
May 7, 2012 March 17, 2013
October 27, 2012 October 27, 2013
January 1, 2013 on/after 2014
January 8, 2014 on/after 2015
February 24, 2014 on/after 2016
May 14, 2014 on/after 2017
December 1, 2014 2nd half of 2017
January 21, 2017 July 10, 1905
December 15, 2017 October 31, 2020
From Wikipedia, Berlin Brandenburg Airport, https://en.wikipedia.org/wiki/Berlin_Branden-

burg_Airport, accessed November 11, 2020.
With regard to the temporalities of infrastructures, the BER is interesting in

at least three respects. First, it demonstrates the difficulty of consolidating
and stabilizing a project at all, that is, coming to a (temporary) state of
“hardwiring.” The BER project failed even to arrive at this point of upkeeping
an achieved stability, as the heterogeneity of citizens, organizational actors,
laws and prescriptions, environmental organizations, and material and
technical components could not be aligned long enough to complete the
project in the first place. Second, the time it took to build the infrastruc-
ture itself encompassed processes of decay. The passenger information
monitors already had to be replaced because they had reached the end of
their lifetime.20 In the airport’s underground train station, several “ghost
trains” had to be run each day to provide fresh air and prevent mold from
accumulating.21 Regimes of planned obsolescence, as seen in the case of
the monitors, interfered with the longevity of the construction process;
cycles of maintenance had to be executed to simulate usage, although the
structures were not officially open to the public. Third, the time that passed
itself altered perspectives on the future of the project. New calculations of
projected passenger traffic, for instance, now demand a restructuring of the
whole airport area. The so-called “master plan 2040”22 calls for a rewiring,
so to speak, of the whole project before it was even completed, merging the
current areas of the Berlin Schönefeld Airport with the unfinished BER and
adding a “midfield” and other buildings to create an even bigger airport. The
map combines existing buildings, buildings planned until 2030, and desired
future buildings, thereby showing the site as a “contested gathering of many
conflicting demands” where processes of design constantly interfere with

what already has been built.23
On a more general level, the example shows different layers of temporality
within a single infrastructure. According to Steward Brand, in How Buildings
Learn, the very idea of architectural permanence is misleading. Instead,
the adaptability of any architecture has to be seen as a continual flow of
transformations. “All buildings grow” and age, says Brand, but “different
parts of buildings change at different rates.”24 Brand differentiates between
the layers of “site,” “structure,” “skin,” “services,” “space plan,” and “stuff.”25
Whereas the geographical setting, or site, of a building is long-lasting, the
layer of services, such as electrical wiring, plumbing, sprinkler system,
and ventilation, must be renewed every seven to fifteen years. This level of
services caused severe problems in completing the BER, as monitors, cables,
and even safety helmets became obsolete during the process. The “shearing
layers of change” have the effect of the building “tearing itself apart.”26 With
regard to the BER, we can see shearing layers of different time scales not
only with respect to the physical dimensions of the infrastructure but also
their socio-technical aspects, such as bureaucratic procedures, prescriptions,
investigations, and planning.
The temporal regimes involved in the BER example concern, on the one
hand, infrastructures of temporality, that is, the governing of time through
plans, management routines, and maintenance cycles. On the other hand,
the example illustrates the temporalities of infrastructure, that is, the
different lifespans of the involved entities, the transforming effects of
flaws, mistakes, and corruption on the project as a whole and the (organic)
processes of decay and repurposing. A notable example of repurposing is
that the uncompleted airport became a tourist attraction. One could book
tours via bus or bike and, as of March 2017, 1.55 million visitors had been
shown around the airport.27 A perspective on the relations of technology
and time can therefore provide for a fruitful perspective on infrastructural
networks, adding to spatial, topological, and environmental approaches
insights on the ephemerality of their existence.
Repair
As we have seen with respect to the BER project, different layers of tem-
porality exist within the same infrastructural setting. In the following, I
describe in more depth the four practices of care outlined above, starting
with the practice of repair.
How one assesses the importance of repair in the present it is first of

all a question of perspective. Under conditions of scarcity, repair has been
ubiquitous for a long time in history,28 and it still is in countries of the
Global South, whereas in Western industrialized countries “the continuities
between the hidden and ongoing cultures of repair that characterize urban
life outside of catastrophic states … tend to be dramatically underplayed.”29
As a consequence, infrastructures tend to be regarded as “hardwired,” that
is orderly, fixed, and stable. However, repair activity is not less relevant in
the present, but only less visible. The international repair movement (repair
cafés, the website iFixit, for example),30 while claiming a fundamental
“right to repair” and taking up ideas from discourses on sustainability,
commonism, and the do-it-yourself-scene,31 focuses on household items
(toasters, printers) or consumer products (iphone, ipad) and their planned
obsolescence rather than on machines and infrastructures. More funda-
mentally, Steven Jackson calls for a “broken world thinking,” a perspective
on the principal fragility and vulnerability of the material world that takes
its “always-almost-falling-apart” as the starting point for a rethinking of
repair activities.32
With respect to the temporalities of infrastructure, two aspects have to
be emphasized. First, the constitutive ex-post-character (Nachträglichkeit)
of repair (re-parare, in the sense of a reconstruction33), which is responsible
for the epistemic quality of repair, as knowledge of breakdowns or accidents
only becomes evident in the aftermath of an irritation.34 For this reason,
I call this practice retrospective care. Second, and nevertheless, the grade
of belatedness and visibility differs. Roughly speaking, Western societies
believe and imagine themselves to be living in more stable environments
(despite all the makeshift solutions in daily life), whereas people in the
Global South engage openly in repair as a ubiquitous, mundane activity.
Regarding the BER example, the obsolescence of products forced repair
work. Besides the 750 monitors mentioned, 16,849 fire detectors had to be
exchanged at a cost of 1.6 million euro because 90 percent of them had
reached the end of their permitted service life. This turned out to be quite
lucrative for the respective firms, as they could build the airport several
times, so to speak.35
Abandonment
Abandonment refers to the processes of devaluation that technologies and

infrastructures undergo as a result of a lack of care.36 When infrastructures
are not cared for, they decay. Or more precisely, other processes take over.
While the BER was not used as an airport, grass could grow in places where
there was otherwise intense traffic. The heterogeneous actors were not
aligned in one direction, but follow their own paths.
The concept of decay raises questions concerning the status of materiality
and its “hardwiredness.” Regarding the material ecology of subway signs, for
example, the workers did not experience the signs’ “materiality” but were
immersed in a “malleable material flux” of different material properties,
including the walls, metal brackets, and signboards, as well as screws, plugs,
and glue cement.37 From an architectural and urbanist perspective, such a
flux can be understood as “architecture’s ‘life’ and ‘death.’” Instead of relying
on architecture’s attested “material durability,” this means to concentrate
on the relation of architecture to “decay, deterioration, and destruction.”38
In reference to Michael Thompson’s analysis of the complex processes of
valuation and devaluation of things,39 Stephen Cairns and Jane M. Jacobs
highlight “matter” and “mattering,” that is, the dimension of materiality and
the processes of valuation, as two concepts to explain the “relative durability”
of built structures, which in their opinion is linked to a specific temporality:
Architecture’s relative durability does not exempt it from the principle of

mutable value, but it does ensure that architecture generally “circulates”—
via processes of reinvestment, restoration, and revaluation—more slowly
through its ebb and flow. As a consequence, buildings are regularly out of
time—unused, unloved, unappreciated, devalued—but still very much
in place. 40
Infrastructural decay should not be seen simply as a natural by-product of

the time passing. In her study on the aging infrastructure of a NASA space
project, Marisa Cohn shows that it requires active work to bring about the
end-of-life of such a huge system. Infrastructural decay, here, “is composed of
multiple lifetimes of different parts of the system—hardware, software, code,
organizational processes, programming languages, institutions, careers—all
of which are entangled and are aging or obsolescing at different rates.”41
With respect to the logic of abandonment, the specific obduracy of built
structures seems interesting, as it serves to make visible the lack of care.
An obsolete building can be understood as an “obduracy-in-obsolescence,”
being “in place but out of time.”42 It cannot be made to disappear from sight
in simple ways: “Unlike other waste objects, which can be managed or
rendered invisible by being pushed into a garbage bin, stored in the attic,
compacted in a landfill, or biodegraded, buildings often, resolutely and
publicly, stay in view and in place regardless of their economic and public
evaluations.”43 This can also be seen at the BER. The buildings are physically
there, they cannot be easily removed, although the expected increase in the
number of passengers would demand another, bigger airport. So we only
see the demolition (Rückbau) of selected parts. 44
The distinction between different time scales of infrastructure provides
a further perspective on the obduracy and stability of infrastructures. In his
“multiscalar approach,” Paul N. Edwards refers to the different temporalities
of human life, history, and geology. 45 Infrastructures exhibit stability and
durability only on the level of human and historical times, whereas on
geological or “long-term historical” scales infrastructures and nature present
themselves as less discernible from each other, up to the point of their iden-
tification where “[N]ature is … in some sense the ultimate infrastructure.”46
Nevertheless, Edwards’ view on infrastructure’s fragile, ephemeral qualities
is informed by an understanding of technical malfunctioning (he uses
the concepts of “irregularity,” “breakdown,” etc.) rather than an interest in
processes of decay and deterioration or material relics and ruins, which are
already beyond a logic of functioning and purpose. However, certain organic
processes represent symptoms of abandonment, such as the grass growing
between the concrete joints at the BER, and can be seen as the interference
of other time regimes within the logic of infrastructural becoming.
Processes of decay shift our attention from disturbance and disaster, as
more or less discontinuous and abrupt events, to slower and often unnoticed
temporal processes of change. Nevertheless, as any amateur gardener knows,
nature reconquers man-made structures steadily. Processes of abandonment
can therefore be seen as an inverted or negative form of care, characterized
by indifference and a lack of concern.
Repurposing
In his autobiography, Roland Barthes recalls the allegory of the Argo as an

object of continual, transformative processes of care:
A frequent image: that of the ship Argo (luminous and white), each piece
of which the Argonauts gradually replaced, so that they ended with an
entirely new ship, without having to alter either its name or its form. The
ship Argo is highly useful: it affords the allegory of an eminently structural
object, created not by genius, inspiration, determination, evolution, but
by two modest actions (which cannot be caught up in any mystique of
creation): substitution (one part replaces another, as in a paradigm) and

nomination (the name is in no way linked to the stability of the parts). 47
What Barthes invokes in the ship that the Argonauts permanently rework
is the perspective on a processuality of technology and architecture that is
barely noticeable, in that it draws on not radical upheaval or individual ge-
nius, but steady transformation. Although Barthes argues from a structuralist
perspective that underlines the analogous and language-based quality of
substitution within the same paradigm, it is nevertheless possible to think
of a ship’s material quality and the necessity to improvise with respect to
repair and maintenance work due to scarce resources on the high seas.
This transformative aspect has also been addressed as technology’s
“fluidity” and “adaptability.”48 With regard to mobile phone repair in Dhaka
and Kampala, Steven Jackson argues, “the phone that emerges at the end
is demonstrably not the same device.” The work done on the mobile phone
changes it: “The phone has become in effect a different object: new but not
radically new, separated from and connected to its past by the forms of
breakdowns, maintenance, and repair through which it has passed.”49 The
same is true for all the processes connected to the reworking of software
and digital infrastructures, such as updating, and jumpering.50 This work of
transformation generally implies the need to improvise, that is, workarounds
and makeshift solutions that include moments of bricolage, artisan tinkering,
and creativity.51
Regarding the BER, one could say that the project did not stay the “same”
over the time of its construction. Even more, it had not been “unique”
from the beginning. For Latour, “to design is always to redesign,” there is
“something medial in design” so that it never creates ex nihilo.52 Rather,
artifacts are conceived as “complex assemblies of contradictory issues,”
disputed matters of concern, that we are still unable to design in all their
complexity.53 Against this background, one could even say that the BER from
the beginning is a project of re-designing, as it starts from “something that
exists first as a given, as an issue, as a problem.”54 And this “given” is Berlin’s
“airport situation,” which the new BER is supposed to change for the better.
A further aspect regarding the transformative nature of care concerns
the re-direction of its goals, the “creative” processes of adaptation and
repurposing of artifacts and technologies toward other contexts.55 As the
emphasis on first design neglects “the extraordinary life stories” of objects
and technologies in other cultures, there is the need to account for ‘creole’
technologies” as fundamental parts of change: “[M]ost change is taking
place by the transfer of techniques from place to place.”56
Interestingly enough, such transformations not only solve problems but

also generate new ones, which can be illustrated by a historical example.
With reference to the difficulties of the shipment of plants at the end of the
eighteenth century, Maren Mayer-Schwieger has shown that the legendary
mutiny on the HMS Bounty was more or less a direct result of transformations
on the ship. In an effort to prevent the breadfruit plants being shipped
from dying, as had been the case in previous attempts at transporting
live plants, radical measures were taken. Primarily, the captain’s cabin
became the storage space for hundreds of flowerpots. This arrangement
severely disturbed order aboard the Bounty, in particular, the hierarchical
priority of the captain, and the resulting mutiny again led to a failure of
plant transportation.57
Concerning the specific type of care in question, three aspects can be
emphasized. First, processes of re-structuring necessarily belong to the usage
of infrastructures and artifacts. Second, they can generate radical effects
and conflicts because of the involved processes of de- and revaluation of
certain elements of the network, such as the captain’s cabin. And third, to
focus on processes of re-structuring and re-purposing calls into question the
assumption that there can be something like a “new” artifact. Something is
new only in relation to what is already there. As for the temporal dimensions
of care for technology, re-structuring processes can therefore be understood
in terms of the redirection of care to new goals by re-using an existing and
potentially devaluated infrastructural setting.
Maintenance
If we now take a closer look at the set of caring activities involved in main-
tenance,58 upkeep, and regular service of technical systems, we can see
that caring for technology takes a prospective form, here, as it is directed
to stabilizing the uncertain future of an artifact, technology, or system.
Caring in this respect demands a focus on not only the functional relations
to technology but also the “moral relations.”59 It seems as if the question
of care can only be linked to technology if the latter is situated in an area
“beyond” the realm of mere technical functioning.
Forms of maintenance often go along with an affective relationship to the
maintained object, in particular on the level of micro practices. The subjects
involved regularly develop a certain “sense” for the thing cared for, which
provides a sort of diagnosing tool. The service operator responsible for Paris’
water management system (SAGEP), for example, claims that the sluices
of Paris’ water supply had “talked” to him via the old servo-controls.60 In a
similar way, some of the engineers who worked for decades within the aging
Saturn mission spacecraft infrastructure were “accustomed to computing
work that would be difficult to find still practiced elsewhere.”61 What these
examples illustrate is the fact that although control may be digitalized, the
stabilization of the systems as a whole is dependent on procedures that refer
to its material substrate as well as to the acquired skills of the caretakers
(which again implies an essentially temporal dimension).
To elaborate on the specific temporality of maintenance and upkeep, the
etymology of the word waiting (German warten) is of special interest. The
verb warten in German has two meanings: first, to maintain, to upkeep; and
second, to wait, to hold out. As I argue, it is this dimension of waiting that
underlies and determines the temporal logic of maintenance processes.62
In the Deutsches Wörterbuch, the most comprehensive dictionary of the
German language, begun by the Grimm Brothers in 1854, we find an entry
on warten of no less than 42 columns. The article describes the usage in the
sense of waiting (“to await what is coming”) only in the final part, while the
preceding pages are centered on the relation between directed attention and
the practices of guarding, watching, and caring. Interestingly, people are
also said to be maintained (“Personen warten”) in the sense of being cared
for or administered.63 This includes caring for the sick, children, animals,
gardens, and—more common to today’s usage—objects. To speak of care
with respect to things, the Grimm Brothers note, means to care for or just
to deal with them constantly.64 This last mentioned aspect establishes a link
between temporal and affective dimensions of maintenance, as this type
of practice creates and is, therefore, embedded in a sort of habit.
The process of maintaining-as-waiting in the sense of the German warten
denotes a specific form of concern characterized by attentiveness toward the
object of care, be it a thing, a person, an animal, or, we can add, a machine.65
This attention takes the form of a close and constant contact with objects
cared for. We thus deal with an activity that derives its temporality from
the objects it follows in an effort at preservation. Referring to Heidegger’s
notion of dwelling as “sparing and preserving,”66 one could say that Being-
in-the-World in the sense of dwelling can be understood as form of care for
the environment as such.
In general, maintaining-as-waiting (warten) always proceeds in the
present, but it is, at the same time, directed prospectively toward the future
of the object cared for. This can be seen as a fundamental difference to
the practice of repair that responds to a disturbance or, at least, irritation,
and to forms of neglect of care that leave the former object of care to itself,
whereas forms of repurposing may again involve a future for the object
cared for. The practices of maintaining-as-waiting (warten) cultivate a
type of concern tuned to preventing harm by regularly caring for objects
and people, such as through the oiling and cleaning of machines and the
proper nutrition and hygienic measures of people. The lifetimes of artifacts
and people, in effect, coincide with the processes of their maintenance.
Cost-intensive infrastructures, such as weapon systems, aircrafts, and
busses, thus tend to live “eternally,” if they are cared for in a proper way.
This can be seen by the continued usage of taxis, busses, and even bicycles
sorted out in Western societies in the Global South,67 or in the necessity
to actively end their lifetimes.68 However, this longevity as an effect of
maintenance practices reminds us of the basic fragility of things. Things
do not “exist” in an uncomplicated way but only because of the work and
care invested in them.
Although maintenance work can be described with respect to temporal-
ities of different infrastructural entities, it is of course also relevant to point
to the infrastructures of temporality that are in turn involved in processes
of maintenance. Two aspects can be distinguished.
First, and in contrast to the possible longevity of artifacts and machines,
consumer products have implanted limits that are known as planned
obsolescence, and are conditioned to become cyclically out fashioned in
increasingly shorter intervals. This phenomenon belongs to the logic of
capitalist value creation, for which it would be fatal if objects were cared for
in a relevant sense. This is one reason why today’s call for “repairability”69
and “maintainability,”70 that is, the production of maintainable products
(e.g., the Fairphone with interchangeable parts),71 did not manage to gain a
deeper impact until now. Accordingly, the so-called life cycle management
might be taken to amount to nothing more than the attempt to exploit each
phase of life of the object (planned obsolescence inclusively). Although it
may seem at first glance as the revival of older practices of re-using, known
from an economy of scarcity, the idea of life cycle management is informed
by cost-effectiveness and not by a logic to spare things.72 Nevertheless, also
in the economic field, there are tendencies to see maintenance not as a
“necessary evil,” but rather as a potential to minimize material and energy
consumption,73 controlled by “life cycle ‘big data’ analytics.”74
Second, processes of technical maintenance are often informed by
infrastructures of temporality. As we saw with the Berlin Airport exam-
ple, these procedures follow certain cycles (for example, the necessary
ventilation runs of the airport train). They are structured according to
the logic of intervals of time and therefore dependent on a certain type
of clocking that organizes their execution, repetition, and control. Seen

in this light, maintenance procedures might also belong to the realm of
planning, logistics, and management, and they might thus be adapted to a
cybernetic logic of control.
Conclusion
In analyzing the hardwiredness of infrastructures, I have distinguished

several layers of temporalities that go along with four different practices
of care. I have discussed repair as an activity that aims to restore order
retrospectively. While this represents the normal state of things in the Global
South or under preindustrial conditions, Western industrialized societies
often exhibit a belief in the stability of infrastructures. The phenomenon
of decay, in turn, has been addressed as a form of abandonment that results
from devaluation and lack of care, whereas processes of transformation and
repurposing have been analyzed as complementary forms of revaluation
and therefore redirections of care. For the activity of maintenance, I have
assumed a dialectics of non-modern concepts (the logic of concern as a type
of attentiveness to things in the form of waiting) and modern concepts (that
integrate maintenance into logistical cycles of management).
Considering practices of maintaining-as-waiting (warten) of devices,
machines, and large technical systems in particular, we get closer to basic
processes within the realm of the organic, where cultura in the sense of care
and concern for the soil in agriculture is important for the flourishing of
plants, animals, and humans. Such a perspective highlights the logic and
importance of cultural techniques to our modern understanding of culture,
society, and technology.75 Maintaining-as-waiting (warten) would then refer
to a specific “Eigenzeitlichkeit,” or temporality, of things, demanding that
one holds out and allows things to take their time.76
This also implies a non-modern understanding of maintainability, for
which the general adaptivity of things is central: “Age plus adaptivity is
what makes a building come to be loved.”77 The possibility of modification
is responsible for the adjustment of buildings and infrastructures to the
changing conditions of time.78 Consequently, in terms of care, “a concern
for maintainability translates material permanency into sociotechnical
sustainability, and recognizes that stability, like reality, is ‘an active
verb.’”79 We have to develop an understanding of “appropriate technology,”
where “fluid object[s]” remain open to change and dismantling during
their development.80 Taken as matters of concern, things are on the move
in the very basic sense of “moving project[s].”81 As I have argued, main-

tenance in the sense of waiting could then be seen as a set of extremely
undervalued practices of care and concern that are important for the
continued existence of culture, society, and technology, including the
continued existence of digital cultures. From this perspective, hardwired
infrastructures appear less as stable and durable systems than as ongoing
processes of transformation. Even high-technology infrastructures,
thus, are consisting of and crisscrossed by different temporal regimes
that include non-modern, more organically and culturally embedded
temporalities of care.
Notes
1. See for example Christopher R. Henke, “The Mechanics of Workplace Order:

Toward a Sociology of Repair,” Berkeley Journal of Sociology 44 (2000): 55–81;
Stephen Graham and Nigel Thrift, “Out of Order: Understanding Repair
and Maintenance,” Theory, Culture & Society 24, no. 3 (2007): 1–25; Steven J.
Jackson, “Rethinking Repair,” in Media Technologies. Essays on Communica-
tion, Materiality, and Society, ed. Tarleton Gillespie, Pablo J. Boczkowski, and
Kirsten A. Foot (Cambridge, MA/London: MIT Press, 2014), 221–239; Stefan
Krebs, Gabriele Schabacher, and Heike Weber, eds. Kulturen des Reparierens.
Dinge – Wissen – Praktiken (Bielefeld: transcript, 2018); Andrew L. Russell/
Lee Vinsel, “After Innovation, Turn to Maintenance,” Technology and Culture
59, no. 1 (2018): 1–25.
2. For notable exceptions see Neil Pollock et al., “Post Local Forms of Repair.
The (Extended) Situation of Virtualised Technical Support,” Information
and Organization 19, no. 4 (2009): 253–276; Jason Farman, “Repair and
Software: Updates, Obsolescence, and Mobile Culture’s,” continent. 6, no. 1
(2017): 20–24. With respect to new media’s logic of constant actualization,
see Wendy Chun, Updating to Remain the Same. Habitual New Media (Cam-
bridge/London: MIT Press, 2016).
3. Nathan Ensmenger, The Computer Boys Take Over. Computers, Program-
mers, and the Politics of Technical Expertise (Cambridge, MA/London: MIT
Press, 2010), 225. See also Thomas Haigh, “The Chromium‐Plated Tabulator.
Institutionalizing an Electronic Revolution, 1954–1958,” IEEE Annals of the
History of Computing 23, no. 4 (2001): 75‐104.
4. María Puig de la Bellacasa, Matters of Care: Speculative Ethics in More Than
Human Worlds (Minneapolis/London: University of Minnesota Press, 2017), 5.
5. Foucault, Michel, The Care of the Self. The History of Sexuality 3 (Harmond-
sworth and others: Penguin Books, 1990); see also Annemarie Mol, Ingunn
Moser, and Jeannette Pols, eds. Care in Practice. On Tinkering in Clinics,
Homes and Farms (Bielefeld: transcript, 2010).
6. Emily K. Abel and Margret K. Nelson, eds. Circles of Care: Work and Identity
in Women’s Lives (New York: State University of New York Press, 1990).
7. Susan Leigh Star and Anselm Strauss, “Layers of Silence, Arenas of Voice:
The Ecology of Visible and Invisible Work,” Computer Supported Cooperative
Work (CSCW) 8, no. 1–2 (1999): 9–30.
8. Michael Polanyi, The Tacit Dimension (New York: Doubleday Company,
1966).
9. Jérôme Denis and David Pontille, “Material Ordering and the Care of
Things,” Science, Technology & Human Values 40, no. 3 (2015): 338–367; Nicky
Gregson, Alan Metcalfe, and Louise Crewe, “Practices of Object Mainte-
nance and Repair: How Consumers Attend to Consumer Objects within the
Home,” Journal of Consumer Culture 9, no. 2 (2009): 248–272.
10. María Puig de la Bellacasa, Matters of Care, 42–48; see also Bruno Latour,
“From Realpolitik to Dingpolitik—An Introduction to Making Things
Public,” in Making Things Public. Atmospheres of Democracy. Catalogue of
the show at ZKM, ed. Bruno Latour and Peter Weibel (Cambridge, MA: MIT
Press, 2005), 4–31.
11. Bruno Latour, Aramis, or the Love of Technology. Transl. Catherine Porter
(Cambridge, MA, London: Harvard University Press, 1996).
12. Antoine Hennion, “Attachments, You Say? … How a Concept Collectively
Emerges in One Research Group,” Journal of Cultural Economy 10, No. 1
(2017): 112–121.
13. Denis and Pontille, “Material Ordering and the Care of Things,” 355 and 353.
14. Steven J. Jackson, “Speed, Time, Infrastructure. Temporalities of Breakdown,
Maintenance, and Repair,” in The Sociology of Speed, ed. Judy Wajcman and
Nigel Dodd (Oxford: Oxford University Press, 2017), 169–205, here: 179 and
183. On the interference of temporal regimes of embodied human expe-
rience and the materialities of technology, see also the contributions of
Nicole Starosielski and Marisa Leavitt Cohn in this volume.
15. Thorsten Metzner, “Hauptstadtflughafen. TÜV hält BER-Terminplan
für stark gefährdet,” Der Tagesspiegel, April 11, 2019. Available at https://
www.tagesspiegel.de/berlin/hauptstadtflughafen-tuev-haelt-ber-termin-
plan-fuer-stark-gefaehrdet/24210992.html (accessed June 15, 2019). The TÜV
(Technischer Überwachungsverein) Rheinland is a technical inspection as-
sociation that provides inspection and product certification services accord-
ing to international standards. The BER indeed opened on October 31, 2020.
16. The calculation is based on an extrapolation from the last reliable data and
the assumption of a linear progression of costs. See the website “Flughafen
Berlin (BER) Kosten,” https://www.flughafen-berlin-kosten.de/ (accessed
June 15, 2019).
17. See Michel Callon’s analysis of the unsuccessful project to develop an elec-
tric vehicle in the 1970s, Michel Callon, “The Sociology of an Actor-Network.
The Case of the Electric Vehicle,” in Mapping the Dynamics of Science and
Technology: Sociology of Science in the Real World, ed. Michael Callon, John
Law, and Arie Rip (Basingstoke: Macmillan, 1986), 19–34.
18. A parliamentary committee of inquiry investigated the mismanagement

between 2012 and 2016 and published a report of 1.269 pages in June 2016,
Bericht des 1. Untersuchungsausschusses des Abgeordnetenhauses von
Berlin – 17. Wahlperiode – zur Aufklärung der Ursachen, Konsequenzen und
Verantwortung für die Kosten- und Terminüberschreitungen des im Bau
befindlichen Flughafens Berlin Brandenburg Willy Brandt (BER), Druck-
sache 17/3000, June 14, 2016. Available at https://www.parlament-berlin.
de/C1257B55002AD428/vwContentbyKey/206D70E8863C06F9C-
12580C60051E19A/$FILE/d17-3000.pdf (accessed May 27, 2018).
19. See also Bruno Latour, “Trains of Thought: Piaget, Formalism, and the Fifth
Dimension,” Common Knowledge 6, no. 3 (1996): 170–191, here: 175; Gabriele
Schabacher, “Mobilizing Transport. Media, Actor-Worlds, and Infrastruc-
tures,” Transfers. International Journal of Mobility Studies 3, no. 1 (Spring
2013): 75–95.
20. “Berlin: 750 Monitore am BER schon veraltet,” Spiegel Online, http://www.
spiegel.de/wirtschaft/unternehmen/berlin-750-monitore-am-ber-schon-
veraltet-a-1198359.html (accessed November 8, 2019).
21. “S-Bahn-Lüftungsfahrt zum Airport BER,” Berliner Verkehrsblätter 60, no. 4
(2013): 67.
22. See “Master Plan BER 2040,” https://www.berlin-airport.de/de/presse/
publikationen/geschaeftspartner/2018/2018-masterplan-ber-2040.pdf
(accessed June 15, 2019).
23. Bruno Latour and Albena Yaneva, “‘Give me a gun and I will make all build-
ings move’: An ANT’s View of Architecture,” in Explorations in Architecture:
Teaching, Design, Research, ed. Reto Reiser (Basel: Birkhäuser, 2008), 80–89,
here: 86.
24. Stewart Brand, How Buildings Learn: What Happens After They’re Built (Lon-
don: Penguin, 1994), 10, 12.
25. Brand, How Buildings Learn, 13.
27. Sonja Gurris, “Führung auf dem Flughafen BER. Der Airport, über den Berlin
lacht,” NTV, December 2, 2017. Available at https://www.n-tv.de/reise/Der-
Airport-ueber-den-Berlin-lacht-article20156588.html (accessed June 15, 2019).
Interestingly enough, this experience was not offered on the international site.
28. Reinhold Reith and Georg Stöger, “Einleitung. Reparieren – oder die Leb-
ensdauer der Gebrauchsgüter,” Technikgeschichte 79, no. 3 (2012): 173–184.
29. Graham and Thrift, “Out of Order,” 9. As Jörg Potthast has shown with
regard to the work of air traffic controllers, even the seemingly “normal
case” exhibits high degrees of disorder, Jörg Potthast, “Ethnography of a
Paper Strip: The Production of Air Safety,” Science, Technology & Innovation
Studies 4, no. 1 (2008): 47–68, here: 51f.
30. Sigrid Kannengießer, “Repair Cafés—Reflecting on Materiality and Con-
sumption in Environmental Communication,” in Environmental Commu-
nication Pedagogy and Practice, ed. Tema Milstein, Mairi Pileggi, and Eric
Morgan (London: Routledge, 2017), 183–194.
31. For a historical perspective, see Daniela K. Rosner and Fred Turner, “The-
aters of Alternative Industry: Hobbyist Repair Collectives and the Legacy of
the 1960s American Counterculture,” in Design Thinking Research: Building
Innovators, ed. Hasso Plattner, Christoph Meinel, and Larry Leifer (Berlin/
Heidelberg/New York: Springer, 2015), 59–69.
32. Jackson, “Rethinking Repair,” 221 and 222.
33. Entry “repair, v.2,” in Oxford English Dictionary, Oxford University Press 2019.
Available at https://www.oed.com (accessed November 8, 2019).
34. For the media history of accidents and their reconstruction, see Christian
Kassung, ed. Die Unordnung der Dinge. Eine Wissens- und Mediengeschichte
des Unfalls (Bielefeld: transcript, 2009).
35. For the Bosch company that was contracted to build the fire detection
system and safety technology for about 12 Million Euro the contract has
multiplied up to 66.4 Million Euro in 2018, because of supplements due
to the postponed opening and pricing on hourly basis (Thorsten Metzner,
“Die Mängelliste wird kürzer. Wo es jetzt noch hakt auf der BER-Baustelle,”
Der Tagesspiegel, June 5, 2019. Available at https://www.tagesspiegel.de/
berlin/die-maengelliste-wird-kuerzer-wo-es-jetzt-noch-hakt-auf-der-ber-
baustelle/24420576.html (accessed June 19, 2019)).
36. See also Gabriele Schabacher, “Abandoned Infrastructures. Technical
Networks beyond Nature and Culture,” Zeitschrift für Medien- und Kulturfor-
schung 9, no. 1 (2018): 127–145.
37. Denis and Pontille, “Material Ordering and the Care of Things,” 352.
38. Stephen Cairns and Jane M. Jacobs, Buildings Must Die. A Perverse View of
Architecture (Cambridge, MA/London: MIT Press, 2014), 2 and 1.
39. Michael Thompson, Rubbish Theory: The Creation and Destruction of Value
(Oxford and others: Oxford University Press, 1977).
40. Cairns and Jacobs, Buildings Must Die, 49, 58.
41. Marisa Leavitt Cohn, “Convivial Decay: Entangled Lifetimes in a Geriatric
Infrastructure,” CSCW ’16. Proceedings of the 19th ACM Conference on Com-
puter-Supported Cooperative Work & Social Computing 2016: 1511–1523, here:
1511 and 1513.
42. Cairns and Jacobs, Buildings Must Die, 111 and 103. On obduracy in urban
contexts, see Anique Hommels, Unbuilding Cities. Obduracy in Urban
Sociotechnical Change (Cambridge, MA/London: MIT Press, 2008). On ruins
and ruination, see Gastón R. Gordillo, Rubble. The Afterlife of Destruction
(Durham/London: Duke University Press 2014).
43. Cairns and Jacobs, Buildings Must Die, 58.
44. Metzner, “Hauptstadtflughafen. TÜV hält BER-Terminplan für stark gefährdet.”
45. Paul N. Edwards, “Infrastructure and Modernity,” in Modernity and Technolo-
gy, ed. Thomas J. Misa, Philip Brey, and Andrew Feenberg (Cambridge, MA:
MIT Press, 2003), 185–225, here: 185 and 194.
46. Edwards, “Infrastructure and Modernity,” 195 and 196.
47. Roland Barthes by Roland Barthes. Transl. Richard Howard (Berkeley/Los
Angeles: University of California Press, 1977), 46.
48. Marianne de Laet and Annemarie Mol, “The Zimbabwe Bush Pump:
Mechanics of a Fluid Technology,” Social Studies of Science 30, no. 2
(April 2002): 225–263, here: 225 and 226.
49. Jackson, “Speed, Time, Infrastructure,” 179.
50. Farman, “Repair and Software.”
51. On workarounds see Gabriele Schabacher, “Im Zwischenraum der Lösun-
gen. Reparaturarbeit und Workarounds,” in Workarounds. Praktiken des
Umwegs, ed. Holger Brohm, Sebastian Gießmann, Gabriele Schabacher, and
Sandra Schramke (Berlin: Philo Fine Arts, 2017), xiii–xxviii; on bricolage see
Claude Lévi-Strauss, The Savage Mind (London: Weidenfield and Nicolson,
1966), 16f.; on artisan tinkering Simon Schaffer, “Easily Cracked. Scientific
Instruments in States of Disrepair,” Isis 102, no. 4 (2011): 706–717, here: 708;
on improvisation Henke, “The Mechanics of Workplace Order,” 66–69.
52. Bruno Latour, “A Cautious Prometheus? A Few Steps Toward a Philosophy
of Design (With Special Attention to Peter Sloterdijk),” in: Proceedings of
the 2008 Annual International Conference of the Design History Society –
Falmouth, 3–6 September 2009, ed. Fiona Hackne, Jonathn Glynne and Viv
Minto (Boca Raton: Universal Publishers, 2008), 2–10, here: 4 and 5.
53. Latour, “A Cautious Prometheus?,” 4, 11 and 12.
54. Latour, “A Cautious Prometheus?,” 5.
55. See Ben Jervis and Alison Kyle, eds. Make-do and Mend. Archaeologies of
Compromise, Repair and Reuse (Oxford: Archaeopress, 2012); Paula Jarzab-
kowski and Trevor Pinch, “Sociomateriality is ‘the New Black’: Accomplish-
ing Repurposing, Reinscripting and Repairing in Context,” M@n@gement 16,
no. 5 (2013): 579–592; David Keller and Maria Dillschnitter, eds. Zweckentfre-
mdung. ‘Unsachgemäßer’ Gebrauch als kulturelle Praxis (Paderborn: Fink,
2016).
56. Edgerton, Shock of the Old, 38, 43 and 209.
57. Maren Mayer-Schwieger, “Umwege auf See. Zur Pflanzenverschiffung Ende
des 18. Jahrhunderts,” in Workarounds. Praktiken des Umwegs, ed. Holger
Brohm, Sebastian Gießmann, Gabriele Schabacher, and Sandra Schramke
(Berlin: Philo Fine Arts, 2017), 146–156.
58. On the history and historiography of maintenance, see Russell and Vinsel,
“After Innovation, Turn to Maintenance.”
59. Jackson, “Rethinking Repair,” 231.
60. Bruno Latour and Emilie Hermant, Paris ville invisible (La Décourverte:
Les Empêcheurs de penser en rond, 1998), 26. On the holistic knowledge
involved in care practices, see also Douglas Harper, Working Knowledge. Skill
and Community in a Small Shop (Chicago: The University of Chicago Press,
1987); Stefan Krebs, “‘Dial Gauge versus Senses 1–0’. German Car Mechanics
and the Introduction of New Diagnostic Equipment, 1950–1980,” Technology
and Culture 55, no. 2 (2014): 354–389.
61. Cohn, “Convivial Decay,” 1513.
62. In English one can still find this meaning of caring and guarding in the
word waiter (entry “waiter, n.,” in Oxford English Dictionary (OED). Oxford
University Press 2019. Available at https://www.oed.com (accessed Novem-

ber 8, 2019)) as well as in older – today obsolete – meanings of the verb wait
as “to (keep) watch” (entry “wait, v.1,” OED 2019). The verb maintain, in turn,
refers not only to the upkeep of things, but also in general to the actions of
supporting, assisting, defending, continuing, prosecuting, preserving, and
providing for the sustenance, for example, of an animal (entry “maintain,
v.,” OED 2019). See also Gabriele Schabacher, “Waiting. Cultural Techniques,
Media and Infrastructures,” in Cultural Techniques. Spaces, Texts, Collectives,
ed. Jörg Dünne, Kathrin Fehringer, Kristina Kuhn, and Wolfgang Struck
(London/New York: De Gruyter, 2020), 71–84.
63. Entry “warten,” in Jakob und Wilhelm Grimm, Deutsches Wörterbuch, vol.
XIII (Leipzig: Hirzel, 1922), column 2137.
64. Grimm, Deutsches Wörterbuch, column 2141.
65. In Being and Time Heidegger describes concern as the fundamental way of
Being-in-the-world, Martin Heidegger, Being and Time [1962]. Transl. John
Macquarrie & Edward Robinson (Oxford/Cambridge: Blackwell, 2001), 56.
66. Martin Heidegger, “Building Dwelling Thinking,” in Martin Heidegger,
Poetry, Language, Thought. Transl. Albert Hofstadter (New York and others:
Harper & Row, 1971), 145–161, here: 149.
67. Hans P. Hahn, “Das ‘zweite Leben’ von Mobiltelefonen und Fahrrädern.
Temporalität und Nutzungsweisen technischer Objekte in Westafrika,” in
Kulturen des Reparierens. Dinge – Wissen – Praktiken, ed. Stefan Krebs, Ga-
briele Schabacher, and Heike Weber (Bielefeld: transcript, 2018), 105–119.
68. Cohn, “Convivial Decay.”
69. Daniela K. Rosner and Morgan Ames, “Designing for repair?: Infrastructures
and Materialities of Breakdown,” CSCW ’14. Proceedings of the 17th ACM
Conference on Computer Supported Cooperative Work & Social Computing,
Baltimore, Maryland, USA — February 15–19, 2014: 319–331, here: 319.
71. See the respective website “Fairphone | The phone that cares for people and
planet,” https://www.fairphone.com/en/ (accessed June 20, 2019).
72. John Stark, “Product Lifecycle Management,” in Product Lifecycle Manage-
ment (Volume 1). 21st Century Paradigm for Product Realisation, third ed., ed.
John Stark (London and others: Springer, 2015), 1–29.
73. Shozo Takataa et al., “Maintenance: Changing Role in Life Cycle Manage-
ment,” CIRP Annals 53, no. 2 (2004): 643–655, here: 643.
74. Rajkumar Roy et al., “Continuous Maintenance and the Future–Founda-
tions and Technological Challenges,” CIRP Annals 65, no. 2 (2016): 667–688,
here: 667.
75. Bernhard Siegert, Cultural Techniques. Grids, Filters, Doors, and Other Artic-
ulations of the Real. Transl. Geoffrey Winthrop-Young (New York: Fordham
University Press, 2015).
76. See Michael Gamper and Helmut Hühn, eds. Zeit der Darstellung. Ästhe-
tische Eigenzeiten in Kunst, Literatur und Wissenschaft (Hannover: Wehr
hahn, 2014).

78. See also Christopher Alexander, The Timeless Way of Building (New York:
Oxford University Press, 1979).
80. de Laet and Mol, “The Zimbabwe Bush Pump,” 235 and 226.
81. Latour and Yaneva, “Give me a gun,” 80.
About the Author
Gabriele Schabacher is Professor of Media and Culture Studies at the

Johannes Gutenberg University of Mainz, Germany. Her research interests
include the media history of traff ic, mobility, and infrastructures, the
cultural techniques of repair, and the media history of seriality. Recent
co-edited volumes focus on the cultures of repair (2018) and on the practices
of workarounds (2017).
3. Problems of Temporality in the
Digital Epoch
Yuk Hui
Abstract
This essay approaches the question of time by considering how digital
media have reconstructed the relations between past, present, and
future. It proposes that the past and present are becoming more and
more determined by future events that have not yet happened but are
paradoxically already there. Building on Bernard Stiegler’s analysis of
primary, secondary, and tertiary retention, it argues that the missing
term in current understandings of time is tertiary protention, or the
phenomenon of preemption.
Keywords: philosophy of technology, preemption, interobjectivity, time,

futurity, China
I have chosen a monstrous title, “Problems of Temporality in the Digital

Epoch,” in that it is difficult if not impossible to treat the subject of tem-
porality in general because every attempt to clarify temporality inevitably
becomes lost in confusion, as Saint Augustine once said in his Confessions:
“What then is time? If no one asks me, I know what it is. If I wish to explain
it to him who asks, I do not know.” Every attempt to seize hold of what time
is leads to ekstasis, or being outside of itself, which recursively extends to
the indefinite. I do not intend to construct an ontology of time here, since it
will be doomed to failure; but concerning the monstrosity of the concept of
temporality, I also cannot let complacency triumph by taking for granted the
present transformation of temporality. Therefore, I would like to talk about
problems here. To focus on these problems, I will discuss only individual
time and historical time, limiting myself to the question of the future. If I
posit the future as a sphere of problems (with the understanding that for
doi 10.5117/9789463727426_Ch03
78 Yuk Hui
certain people the future is always a problem on account of its uncertainty),

it is because I would like to suggest a need to strike free from such futures,
which is nothing posthuman but too human.
Orders of Temporal Magnitude
In general, we can approach the question of time in terms of a threefold

temporality, namely past, present and future, according to different orders
of magnitude. In the first instance, we understand time from an objective
geometrical point of view, that is to say, the spatialization into geometrical
form divided by second, minute, and so on so forth. For the ancient Greeks,
time is thus characterized as “between” (metaxu); for the Stoics, it is “interval”
(diastama); and for Augustine, it is sentimus intervalla temporum. In the
second case, we approach time in terms of living experience, in which
its thickness cannot be reduced to geometrical points but rather must
be grasped as indivisible totality, for example, what Henri Bergson calls
durée. Lastly, we approach time from the perspective of history, as a past
that we have never lived but that belongs to us and remains the condition
of our experience. Rather than being merely different points of view, how-
ever, these notions of time correspond to different orders of magnitude.1
Between these different orders of magnitude, as I argue, we can locate
forms of technological mediation that complexify what Heidegger calls the
temporalization of past, present and future. In turn, this means that the
mediation between these orders of magnitude is subject to an analysis of
the evolution of technology.
This brings us to the modes of mediation in the digital epoch. Digital
technology has brought a tremendous challenge to the previous under-
standing of time by reconstructing the relations between these orders
of magnitude. For example, we know that with digital technology it is
possible to divide a second of sound into milliseconds of microsound,
as demonstrated by the granular synthesis of Curtis Roads and Iannis
Xenakis.2 Such works open a new relation to indivisible human experience,
in that a listener is not be able to perceive the millisecond as an individual
sound but instead hears only the whole that is already synthesized, as
when Leibniz says that when we hear the roar of the ocean, we do not
perceive the petites perceptions of the wave but only a synthesis en gros.
Nevertheless, in and between these two orders of magnitude, that is to
say, in and between the micro and the phenomenological, interobjective
relations can be established through digital technology that are available
Problems of Tempor alit y in the Digital Epoch 79
to visual display. By interobjective relations I mean those relations that

exist independently of associations made by the human mind and that are
rather discovered in their constant process of materialization according to
new epistemologies and technologies. In On the Existence of Digital Objects, I
tried to demonstrate that the history of technology could be considered as a
series of discoveries and masteries of interobjective relations (or discursive
relations).3 In this sense, we can see that temporality is always material;
it is hardwired in the technologies that constitute it, and this process
of hardwiring will continue indefinitely. Interobjective relations in their
prevailing form today—data—are those which reconstitute the dynamic
of temporalization.
This brief summary is meant to show that technological evolution
alters the question of temporality precisely because it constitutes new
interobjective relations that penetrate into different orders of magnitude
that have hitherto been unrealized. It is not that technology reconfigures
physical time; rather, it brings about new dynamics in what we might call,
following Heidegger, the temporalization of past, present, and future.
My emphasis, as mentioned, will be on the future. I would like to put my
arguments in a simple proposition: our past and present are becoming
more and more determined by future events that have not yet happened
but are paradoxically already there. The future is a temporality that ex-
ceeds what is “hardwired”; however, “being hardwired” remains its most
important condition (this is the reason I distinguish discursive relations
from existential relations).
The ultimate future for every individual is of course death; the ultimate
future of humankind is the solar explosion. But it is also limits that give
weight and meaning to temporality instead of pure becoming. Human
being anticipates its own death, which is a future already given and is called
destiny. Such is Martin Heidegger’s famous notion of being-towards-death,
in which the philosopher ties finitude to the authenticity [Eigentlichkeit]
of human existence, which he calls Dasein. Heidegger’s privileging of
authenticity consists of two gestures: first, he emphasizes the privation of
the They [das Man], a fictional future that separates Dasein from its own
time; second, he understands time as self-constituting, by aff irming a
future that is transcendent, namely death, and refusing a future that blinds
Dasein from truth. 4 However, at stake is the exact nature of this limit. The
digital epoch, as I want to argue, is characterized precisely by this fictive
futurity, not only because the gap between the world and science fiction is
diminishing but also because the future is becoming a metaphysical force
and is in the process of liberating itself in all domains.
80 Yuk Hui
The Problem of Tertiary Protention
Today, technological development, with the combination of big data and

predictive algorithms, has rendered what I call tertiary protention more
visible and made the relation between the future and the They more
explicit than ever. In Husserl’s phenomenology of time-consciousness, he
distinguishes primary and secondary retentions and protentions.5 To assist
this with an example, let us consider Johann Strauss’s An der schönen blauen
Donau. When I listen to the piece for the first time, I retain every now of
the melody because every now is always no longer; this is called primary
retention. At the same time, I also anticipate the coming melody, without
which I would be unable to comprehend the phrasing and would hear not
music but only sound; this anticipation of the coming “now,” the “not-yet,”
is primary protention. If tomorrow, I remember the Blauen Donau, it is no
longer a temporarily retended “now” but rather a recollection, namely, a
memory or secondary retention; and since I already have memory of the
music, I am able to anticipate the end of every phrase and the end of the
piece as well, as secondary protention. Building on Husserl’s concepts of
primary and secondary retention, Bernard Stiegler has suggested what he
calls tertiary retention, or artificial memories.6 For example, my secondary
retention of the Strauss piece is not reliable in that it vanishes over time,
but a CD may help me to recover that memory. The gramophone (analogue)
and the CD or now MP3 (digital) are tertiary retentions that in some way
invoke our primary and secondary retentions and protentions like Proust’s
madeleine. I propose to add to this schema a tertiary protention, a form
of anticipation that is no longer my own subjective projection but rather
a projection that is imposed on me and proceeds ahead of me. Protention
cannot be reduced to any form of retention or any combination of retentions
because protention is ontologically different from retention. Otherwise, there
would be no différance, in the sense of Derrida, because différance is the
“retention and protention of differences.”7 I argue that tertiary protention is
the missing term necessary to bring forward a full hermeneutic circle and
analysis of the contemporary politics around attention, desire, and capital.
Everyone using social media knows very well that we are dealing with
recommendations all the time, as if the future was always already present
in the form of multiple choices. A typical example used for decades by
computer scientists working in the field of ubiquitous computing is the
following: you just finished work and are feeling tired while driving home,
and when you arrive, you find a cup of warm, freshly brewed coffee waiting
for you, with exactly the type of coffee bean you have in mind. This example
is only the tip of the iceberg of the increasing power and phenomenon of
tertiary protention, described by many as preemption. Through preemptive
algorithms, we are witnessing a mode of marketing based no longer only
on psychoanalysis, as what Edward Bernays, the father of public relations
and nephew of Sigmund Freud, used, but rather on detailed statistical study
of behaviors in order to capture the future through the past and impose it
on the present. Here, we observe the repetition of the future, functioning
as a new synthesis apart from those Gilles Deleuze analyzed in Difference
and Repetition.8 Deleuze discusses three syntheses of time: the Humean
repetition of habits, the Freudian return of trauma, and the Nietzschean
eternal recurrence of the same in which the future is already at the present.
In this fourth synthesis of time that we call tertiary protention, the future
is also at the present, but it does not carry the same sense as Deleuze’s third
synthesis; in fact, it is a return to a mode of repetition that is complex but
mechanical and homogenous. In our digital milieu consisting of smart
objects, smart homes, smart cities, and social networks, the time of each
individual is fragmented and reorganized so that tertiary protention is
ahead of us.
What could have been called an insight decades ago is today no more
than common sense in view of the becoming of technical systems. (I take
the term “system” here in the sense of Jacques Ellul, who in the 1970s already
clearly described the totalization and autonomy of a technical system in
view of the computer’s capacity for data processing.9) In On the Existence
of Digital Objects, I suggest to liberate time from the increasing threat
of the determination of tertiary protention by inventing new temporal
organizations through creation of alternative technological architectures
and algorithms.10 This involves fundamentally a reorganization of interob-
jective relations. For example, from 2011 to 2012, I worked with the computer
scientist Harry Halpin to develop a prototype of a social network based on
groups instead of individuals, with each group being based on a common
project and each project referring also to a projection, or an investment of
time.11 Additionally, in 2013, I worked with a group of computer scientists
from Princeton University and Deutsche Telekom Laboratory to propose a
new recommender system based on groups. If the new form of interobjective
relations has already acted upon us in our daily lives and constitutes what
Antoinette Rouvroy calls “data behaviourism,” then it extends much further
than individual time, constituting instead a collective imagination of the
future.12 It is because of this futurality powered by the algorithmic processing
of data that I would like to proceed to the second problem concerning our
collective past and its becoming, which we call world history.
82 Yuk Hui
The Problem of the Digital Future
It is not only “personal time” that is hardwired but also historical time.
Through the hardwired geometrical time that we call synchronization,
we arrive at a historical time that we call globalization. It is in the axis
of time in which the historical temporalities of all cultures converge that
we can articulate a world history. The danger of synchronization is that
with digital technology, especially the data economy, we are all moving
towards one future, the pinnacle of humanism. This future is associated
with technological singularity, artificial intelligence, machine learning, and
the realization of the homo deus. Yuval Noah Harari, the author of Homo
Deus, embraces a human-algorithm reduction, proposing in the name of
“life science” the following:
1. Organisms are algorithms, and humans are not individuals—they are
‘dividuals,’ i.e. humans are an assemblage of many different algorithms
lacking a single inner voice or a single self.
2. The algorithms constituting a human are not free. They are shaped by
genes and environmental pressures, and take decisions either deter-
ministically or randomly—but not freely.
3. It follows that an external algorithm could theoretically know me
much better than I can ever know myself…. Once developed, such an
algorithm could replace the voter, the customer and the beholder. Then
the algorithm will know best, the algorithm will always be right, and
beauty will be in the calculations of the algorithm.13
The transhumanist tone, claiming insight from “life science,” has already point-
ed to a future of humanity that can be reduced to an artificial intelligence
governed by a super intelligence, which knows anything and everything. This
is a future that will arrive at us. Questioning the “we” is delicate in that we
may easily fall into one of two sides. One is ethnocentrism and ethnofuturism,
which often fall prey to proto-fascism. The other is a universalism that affirms
the realization of humanity through technical advancement. But it is because
of the delicacy of this question that I think we cannot avoid it in our reflection
on the digital future or futures, lest we lose sight of the issue of locality. To
tackle this question, we will need to reassess the history of technology and
challenge the universalist conception of it. I use the term “universalist” to
refer to the unreflective belief that there is only one type of technology, no
matter if it is Chinese, Indian, Amazonian, or European, with differences
among these contexts being ascribed to levels of advancement. Nevertheless,
it is by the advancement of technology that the future merits its name.
Therefore, I will go a step further to suggest that we must liberate

ourselves from this deterministic view of history in the digital epoch,
called forth by transhumanist futurisms and prometheanisms, such as
accelerationism. The problem is not that the transhumanist’s proposal
of immortality or a-mortality problematizes the “being-towards-death”
of Heidegger; rather, the realization of the homo deus, or technological
singularity, as end of history expresses the desire and the delirium of the
modern. This competition for technological singularity will characterize
the geopolitics of the coming decade. Technological competition, which
historically took shape largely as military competition, has now become
a domestic competition in terms of automated technologies, artif icial
intelligence, deep learning, nanotechnologies, and genetic technologies
that define the sovereignty of the nation state. China released a whitepaper
in August 2017 that states its wish to become the leading country in AI by
2030, with AI being introduced into primary school education and onward
in the meantime. In September 2017, Russian President Vladimir Putin gave
a speech to the Russian children in which he proposed that “whoever leads
in AI will lead the world.” Earlier, during his 2016 US presidential campaign,
Donald Trump proposed to bring back jobs that China had “stolen,” which
could only be done, as some economists suggested, by implementing auto-
mation on a large scale, so that the competition would be between Chinese
robots and American robots. The geopolitics of the coming decades could
be understood as a competition towards singularity under names such as
sinofuturism and Asian futurism.
It is within the desire for technological singularity, which conceals the
ultimate dream of all forms of prometheanism, that we observe histories
converging further on the same axis of time, moving towards a single
direction and the realization of the homo deus. Coincidently, this conver-
gence resonates with the progress of history as theodicy in the sense of
Leibniz, as Hegel famously claimed. Digital studies must not relinquish
this question too quickly and analyze only phenomena produced by new
industrial technologies. Instead, it is necessary to open new directions for
the development of digital technology in order to imagine a future with
technical diversities.
As with the first problem, in which the future is already determined
through calculation, which commands τἄ πάντα like the lightening of
Heraclitus, time in the second problem is constrained by a world history
presented as a single time axis, in which the present is always already
determined by a future imposed as necessity. World history, in this un-
derstanding, follows a defined order: premodern, modern, postmodern,
84 Yuk Hui
apocalypse. I would like to argue that modernization is a synchronization

based on a universalist concept of technology that confuses universalism
with universalization, with the latter being inseparable from colonization
qua technological globalization. We are confronted with a homogenous
technological future, which nurtures fanaticism as well as the fanaticism
of speculation. This unquestioned universalism of technology must be
exposed to its limit so that we can imagine a plurality of technological
futures and therefore reopen the question of technology and hence imagine
a new geopolitics and a new globalization to come.
Antinomy of the Universality of Technology
Conceiving the possibility of techno-diversity, I suggest, is possible only

by breaking away from the synchronization described above to envisage
different technological futures. This, in turn, is possible only if we recognize
the plurality of cosmotechnics, a term that I coined to distance us from
the impression that technology is universal. Here is a primary definition
of cosmotechnics: the unification of the cosmic and moral orders through
technical activities. This conceptualization is based on an antinomy of
the universality of technology addressed in my recent book, The Question
Concerning Technology in China: An Essay in Cosmotechnics, which could
be stated:
Thesis: technology is an anthropological universal, understood as the

exteriorization of memory and liberation of bodily organs, as some
anthropologists and philosophers of technology have formulated.
Antithesis: technology is not anthropologically universal, being enabled
and constrained by particular cosmologies, which go beyond function-
alities and utilities.14
The peculiarity of the Kantian antinomy is that each thesis holds on its
own but opposes the other; such an antinomy must be resolved by a form
of thinking beyond universality and particularity. Synchronization relies
on the thesis and undermines the antithesis. To answer the question
concerning futurity, we must clarify the antithesis before a resolution can
emerge. This is why I propose that each culture should develop its own
history of cosmotechnics by systematically rediscovering and formulating
its epistemologies and tracing the history of its epistemes in response to
our current historical moment. The aim of conceptualizing cosmotechnics
is to reopen the question of technics that was unfortunately closed down

in past centuries.
Following the analysis in Heidegger’s 1949 lecture later published as “Die
Frage nach der Technik,” we find two concepts and essences of technics.15 The
first is the Greek technē, which means poiesis, or bringing forth (Hervorbrin-
gen), and the second is modern technology, the essence of which is no longer
the Greek technē but rather enframing (Gestell), meaning that everything
is considered calculable and exploitable as resources (Bestand). Gestell
stands as the symbol of the modern, in which cosmologies are replaced
with astrophysics. However, this discourse of technology is fundamentally
European. While it is difficult, if not impossible, to position other kinds
of technics—for example, the Chinese, Indian, or Amazonian—without
reducing them to Greek technē, it is self-evident that they are not “modern”
technologies.
Therefore, we need a new way of understanding technologies outside
of Europe because they are neither technē nor modern technology. It is
intriguing to hear non-European philosophers, when explain the origin
of technologies in their cultures, refer immediately to the mythology of
Prometheus, as if the Greek technē were the origin of all technics. This
reference is symptomatic in the sense that it means a disorientation, in
the double sense of the word. Historians of technology often compare the
advancement of a particular technology in different cultures, for example,
papermaking in China and Europe, or the dynamic of technical systems
in China and Europe, as the French historian Bertrand Gilles has done.
However, it is also possible that these comparisons easily circumvent the
antinomy that we have raised above, namely the cosmological and meta-
physical specificity of cosmotechnics. The question of technological futures
must be approached through historical and metaphysical investigations of
cosmotechnics—a task I have suggested carrying out by reconstructing a
Chinese technological thought.
I attempted to do so by tracing the relation between dao (道) and qi (
器) in Confucianism, Neo-Confucianism and New Confucianism through a
characterization of the dynamics between these two metaphysical categories
as different epistemes. Dao literally means path, and qi, utensils (not to
be confused with the qi or ch’i [氣] usually translated as energy). Dao is a
moral cosmological thinking that situates humans as cosmological beings
and leads them toward the good. The unification of qi and dao is neither
a formula nor a static being; rather, it is dynamic and corresponds to the
episteme of the epoch. I won’t be able to elaborate on this development.
And if I have attempted to diversify the concept of technics, it is because
86 Yuk Hui
I want to put forward the following questions: Is it possible to take this

historical moment to rethink digital technology, not only to understand
its transformative power in new social phenomena but also to open up
the question of technology and technodiversity? Is it possible to nourish
digital technologies with new epistemologies and epistemes in light of the
concept cosmotechnics?
Let us return to the two problems of temporality that we have aimed to
sketch out in this paper. We see how digital technology interacts with other
orders of temporal magnitude and have attempted to problematize these
observations. I emphasize problems because these are what allow us to
invent. The future already contained in preemptive algorithms is a battlefield
for the studies of ethics and technology; to engage with the question posed
by this future, I think it is important to unfold the epistemological and
ontological presuppositions of these algorithms in order to put forward
alternatives. We have also challenged the universalist concept of technology
and proposed that in order to break from such technological determinism,
it is necessary to reopen the question of technodiversity, which is also the
condition of noodiversity. Therefore, we must be critical with notions of
the future of humanity presented as the realization of homo deus or the
progress of modernity. By rejecting the linear path attached to the image
of technological progress, we also reject the politics of acceleration as the
only option available for resolving social and political problems. We seek to
allow multiple futures to emerge and remain irreducible to the global axis
of time perceived as world history since what is called modernity.
Notes
1. See Yuk Hui, On the Existence of Digital Objects (Minneapolis: University of

Minnesota Press, 2015), 29–32.
2. See Curtis Roads, Microsound (Cambridge, MA: MIT Press, 2004).
3. For more on interobjective relations, see my chapter “The Time of Technical
Systems,” in On the Existence of Digital Objects, 151–186.
4. See Martin Heidegger, Being and Time, trans. John Macquarrie and Edward
Robinson (New York: Harper & Row, 1962).
5. Edmund Husserl, The Phenomenology of Internal Time Consciousness, trans.
J. S. Churchill (Bloomington: Indiana University Press, 2019).
6. See Stiegler, Technics and Time, 2: Disorientation, trans. Stephen Barker
(Stanford, CA: Stanford University Press, 2009).
7. Jacques Derrida, “Différance,” in Margins of Philosophy, trans. Alan Bass
(Chicago: University of Chicago Press, 1982), 15.
8. Gilles Deleuze, Difference and Repetition, trans. Paul Patton (New York:
Columbia University Press, 1994).
9. See Jacques Ellul, The Technological System, trans. Joachim Neugroschel
(Eugene, OR: Wipf & Stock, 2018).
10. See “Logic and Time,” in On the Existence of Technical Objects, 221–252.
11. For more, see Yuk Hui and Harry Halpin, “Collective Individuation: The
Future of the Social Web,” in Unlike Us Reader: Social Media Monopolies and
Their Alternatives, ed. Geert Lovink and Miriam Rasch (Amsterdam: Insti-
tute of Network Cultures, 2013), 103–116.
12. Antoinette Rouvroy, “The End(s) of Critique: Data Behaviourism versus
Due Process,” in Privacy, Due Process and the Computational Turn (London:
Routledge, 2013), 157–182.
13. Yuval Noah Harari, Homo Deus: A Brief History of Tomorrow (London: Vin-
tage, 2016).
14. Yuk Hui, The Question Concerning Technology in China: An Essay in Cosmo-
technics (Cambridge, MA, and London: MIT Press, 2019).
15. Martin Heidegger, “The Question Concerning Technology,” in The Question
Concerning Technology and Other Essays, trans. William Lovitt (New York:
Garland, 1977), 3–35.
About the Author
Yuk Hui is the author of On the Existence of Technical Objects (Minnesota,

2016), The Question Concerning Technology in China: An Essay on Cosmotech-
nics (MIT, 2017), and Recursivity and Contingency (Rowman & Littlefield,
2019).
4. Suspending the “Time Domain”:
Technological Tempor(e)alities of
Media Infrastructures
Wolfgang Ernst
Abstract
Any media event is a time function of signals. In favor of a diagrammatic
definition of technological media, media archaeological investigation is
not only concerned with their structural “hardwired” level but with their
operative unfolding-in-time as well. Such an understanding of techno-tem-
poralities does not focus on phenomenal effects of media on humans but
primarily refers to the microregimes within technological devices. In that
sense, “hardwired temporality” refers to the infrastructuring of time by
technologies and to temporal structures which are revealed from within
techno-logical knowledge itself. From that arises an epistemology of
technical processuality beyond the conventional notion of “time.”
Keywords: Media archaeology, techno-temporality, clocking, synchro-

nization, UNIX-time, time of non-reality
The Nonhuman Temporality of Technological Knowledge
In favor of a diagrammatic definition of technological media, media archae-

ology is concerned not only with their “hardwired” structure but also with
their operative unfolding-in-time. This vector places the field close to signal
analysis. Any media event is a time function of signals.1 In addition, the
radical media archaeology of the present media-infrastructural condition
requires analysis of its mathematical, algorithmic operations.
Such an understanding of techno-temporalities does not focus on phenom-
enal effects of media on humans but primarily refers to the microregimes
doi 10.5117/9789463727426_Ch04
90 Wolfgang Ernst
within technological devices. Complementary to discourse analysis, it listens

to the implicit epistemic articulations and enunciations of infra-technical
operations. In that sense, “hardwired temporality” refers to the infrastruc-
turing of time by technologies and to temporal structures revealed from
within techno-logical knowledge itself. From that arises an epistemology
of technical processuality beyond the conventional notion of “time.”
Clocking as the Anonymous Time Base of Modernity
The time base of present information society (for which actor–network

theory admits nonhuman members) turns out to be a critical focus of
analysis. High-frequency trading as data exchange, in the electronic stock
market as well as in the communication sphere of the internet, surprisingly
recalls late medieval monasteries. A true media archaeology of technically
temporalized infrastructures starts here and not in the short memory of the
recent present. Benedictine monks who needed periodically exact clocking
for prayers according to their monastic rules invented the escapement-driven
wheeled clock. Regular oscillation, subdividing movement into equal quanta,
is a precondition not only for industrial production but also “social media”
communication. Within the von Neuman architecture of current computers,
the heartbeat of the time base enables exact synchronization of cycling
units in data processing.
The escapement-driven clock, providing occidental culture with its equi-
distant beats, oscillations, and synchronisms, was invented anonymously.2
Such temporality cannot be reduced to the timeline of cultural history but
rather finds “itself.” There is a rather nonhuman temporality at work in the
infrastructuring of techno-logical knowledge.
According to Martin Heidegger, long before computing in the strict sense,
information society already started with the modern Weltbild: with the
measuring and numerical (scientific) approach to physical nature.3 An
archaeology of the contemporary therefore starts with what in German
is appropriately called “Neuzeit,” or modern time, and in a double sense,
because the new epoch starts with the mechanical clock itself. Marshall
McLuhan, in media-epistemological terms, recedes even further below the
clock technology, to the “technologizing” of the spoken word by the ancient
Greek vocal alphabet itself. 4 It was alphabetic writing that cognitively
induced the analysis and synthesis of oral speech flow into discrete, digital
units. Only when the letter (reduced to the minimum as a binary information
unit) becomes radically meaningless in itself does it enable its full range of
Suspending the “ Time Domain” 91
storage, transfer, and symbol manipulation (processing) of meaning. Once

more, it is epistemologically remarkable that there is no way to inscribe
this cultural technique on the historic timeline because there is neither a
precise date nor a known inventor.5 Technological culture takes place in
anonymous temporality.
Enduring Oscillations versus Transient Phenomena
Both the pulled string of an instrument and articulated speech, by their

very fading out, reveal the endlichkeit (the temporal limit) of any physical
signal event. The moment a string is pulled or a sound is articulated, like a
breaking wave, the chord already anticipates and senses its very end, almost
instantaneously but strictly temporally. A string of an instrument, once
struck, never simply articulates the pure sine tone of its proper frequency
but carries with it a fading memory of its violently abrupt, sudden, transient
start. In the real—that is, physical—world, the picked string fades out
due to mechanical loss of energy (Heidegger’s “being-toward-death” in
its mechanical sense), like a swinging pendulum, a reverberation in the
sonosphere.
Any pleasure of a musical performance by real humans in real time
and space (be it oral poetry as sung by Homer or a musical concert) is
always already accompanied by a melancholic anticipation (that is, irony)
of its ending. On the contrary, audiovisual machine recordings, such as
gramophone discs and videotapes, can be replayed with no internal sense
of ending. To modify Walter Benjamin’s analysis on “The Work of Art in the
Age of Its Technological Reproducibility,” with technical re-petition, the
temporal aura (which is based on the allegorical awareness of ending) is lost.6
But the reentry of temporality happens within the very signal processing
of technological media.
Even electromagnetic waves, such as those calculated by James Clerk
Maxwell, propagate very fast but not instantaneously. In 1879 Hermann von
Helmholtz initiated a prize by the Berlin Academy of Sciences to answer
whether, following Maxwell’s mathematical equations, light is in fact part of
the electromagnetic wave spectrum and thereby subject to temporality, the
literal “speed of light” limit. Radio waves, as turned out by Heinrich Hertz’
experiments on the very media-archaeological level (that is, before becoming
part of a mass-medium called “radio”), have a sense of ending on their very
electrophysical-event level; the secondary level is the modulation of such
principally timeless carrier waves by the proper radio program, which by
92 Wolfgang Ernst
its very program structure introduces a symbolical order of beginning and

ending by arbitrary cultural and media-economical decisions.
The out-fading oscillations in early spark-driven telegraphy become
transformed into stabilized, equi-dynamic, in principle, “eternal,” oscillations
only by the electronic vacuum tube in feedback circuits. Electronics turned
electric signals from ephemeral into regenerative, Bergsonian duration.
Only this allowed for the radio transmission of carrier-wave-modulated
speech and music, resulting in the pulse modulated digital transmission.
The Non-Sense of “Time” for Technological Analysis
Technological media know processual realities but not “time”; therefore, the
neographism “tempor(e)alities” is preferred in the subsequent argument,
which unfolds such media tempor(e)alities on three levels: first of all, within
technologies, when the micro-timing of signal transduction is crucial for the
event to succeed at all; next, in the human sense of temporal experience,
when media events induce phenomenal affects and irritations; finally,
“deep” media temporality (logic and machine), which turns out to be rather
autonomous from the cultural imaginary of imaginary time a.k.a. “history.”
In that context, the very notion of hardwired temporality is a relief. Its
ambition is not deep philosophical questioning of the nature of time; tem-
porality is rather a term that names a couple of signal functions, such as the
Delta-t, and functions of signal processualities, such as transfer, delay, and
even storage. For the close analysis of contemporary media technologies, the
arché is not in “time,” but a cooriginality where “time” is suspended. Media
archaeology is “radical” in the sense that it looks for roots not in “time” but
in the technological event. There can be no infrastructuring of temporal
orders, only infrastructures triggering notions of “time” as secondary effect.
Here, the notion of time as a priori condition of perception (as defined by
Immanuel Kant) does not count. If time counts here, it is in the precise sense
of clocking which generates what Heidegger denounced as “vulgar” technical
time. In that sense, technical infrastructures are not “time” based, but
“time-basing.” What has been decisive for digital computing in the individual
machine counts more interconnected computers. Communication is not
only the material infrastructures of cables, antennas, and data-processing
devices but also events temporally infra-structured. Synchronization of
internet traffic creates an artefactual tempor(e)ality, just as the introduction
of a synchronizing master clock was the condition of a universal “time” in
the nineteenth century. The master clock, in networked computing and
communication, has imploded into a myriad of local temporal agencies. This

temporal pattern is multiple and invites the replacement of transcendent
signifiers with a plurality of more precise termini technici, practicing a
language of analysis that bypasses outdated notions like “time” in the
mediasphere.
Walter Benjamin defined the nineteenth century as the antiquity of
the “now.” Indeed, the “present” can already be seen in 1839 when Karl
August Steinheil designed a timekeeping system for the synchronization
of electromagnetic external clocks by a mechanical central clock whose
pendulum triggers alternating positive and negative poled “time impuls-
es”—an interlacing of timekeeping and telegraphy. Today, the radio time
signal transmitter DCF77 (77,6 kHz) at Mainflingen synchronizes radio
clocks wirelessly.
In a radically media-archaeological understanding, techno-logical
infrastructures extend from material (technical) to mathematical (log-
ical) grids, to the algorithms embedded in ubiquitous microchips. The
imaginary unifying time axis is decomposed into a symbolic concatenation
of programmed processes, patterned and interrupted by moments of the
tempoReal. Aristotle was radically media-archaeological in book IV of his
Physics, defining “time” as resulting from measuring movement sequentially
by numbers. Even etymologically, only from incisive cuts does the notion
of a temporal order arise.
Synchronization and Functional Timing
Rigid analysis of technology does not know metaphysical “time” but rather
an ensemble of temporal operators. There is no given preexisting “time”
but rather an enforced timing, as expressed in the very constructive term
of synchronization.
Edmund Husserl once phenomenologically described the mechanism of
the inner sense of subjective time.7 Such a temporal horizon unfolding as
extended present between re- and protention is radically “grounded” and
de-metaphysicized when it comes to technological ensembles. Lewis Fry
Richardson’s once designed a (human) computing “forecast-factory” for
real-time calculation of weather, connected to local weather stations in
telegraphic “instantaneity.” A central official maintains “a uniform speed
of progress” for the individual partial calculations, “like the conductor
of an orchestra in which the instruments are slide-rules and calculating
machines.” Panoptically, “he turns a beam of rosy light upon any region
94 Wolfgang Ernst
that is running ahead of the rest, and a beam of clue light upon those who
are behindhand.”8
In the meantime, human synchronization of human “computers” has been
replaced by the cybernetic diagram of feedback circuits, replacing the central
time control agency by a flexible automation where human monitoring itself
becomes integrated into the circuitry. For a moment, technological utopia
has become reality: the “Opsroom” for monitoring and control in Stafford
Beer’s Cybersyn computing structure in Salvador Allende’s Chile 1971–73 for
national feedback and control of economic data from state-owned factories.
The data flow was based on a “Cybernet” teletype network, to be calculated
by a central mainframe IBM 360 computer, Cyberstride, and monitored by
a collective of humans for positive or negative feedback. Human wetware
is wired into the symbolic infrastructure for the option to interrupt and
correct: suspending the time series generated by data.9 The melodic re- and
protentional “inner time consciousness” of the human individual (Husserl)
is replaced by the radically “discrete” temporality of counting.
The Emancipation of Technical Timing from Natural “Time”
Media culture is not shaped by a transcendent timeline, but technologies

themselves “shape” time.10 In the 1930s, the crystal quartz clock emancipated
the time base of technology from the astronomical “natural” reference—an
epistemological breaking point. Since 1971, the 555 timer chip has been the
most popular integrated circuit for its ability to generate a steady oscillation
independent of supply voltage.
Against phenomenological suggestions of endurance, a truly archaeolog-
ical analysis of temporality is time-discrete. This imperative conditions the
rhythmic bias of digital culture. As once expressed by Alan Turing himself,
clocking is still the heartbeat of structuring data processing from within.11
Against its meaning at first sight, “real-time” data processing is not about
being close in time but its actual negation. It replaces the familiar “live” signal
transmission by a re- and protentional microtime window of the present;
telepresence (Heideggerian “Ent-fernung”) becomes predictive coding, a
mathematical in-between of humans and their physical environment.
Since 1970, Unix time has described time for operating systems and
file formats in terms of the number of seconds that have elapsed since
00:00:00 Coordinated Universal Time (UTC), Thursday, January 1, 1970. This
starting date is appropriately called “The Epoch,” and thus the epoch of the
computational present starts here. Unix subdivides the day exactly into
86,400 seconds. But this chronotechnical regime does not itself tolerate the
interpolation of leap seconds according to atomic clock–generated timekeep-
ing, where a second must be added to the UTC system to coordinate with the
variability of the Earth’s rotation. Occasionally, when computers have been
forced to use the leap-second friendly UTC, such ruptures of the tempoReal
have crashed websites and confused airline departures.12 Similarly, the
Millennium Bug was a reminder at the end of the 1990s that time in digital
computers is logical and mathematical, not intuitive (“Bergsonian”) time.
The tempor(e)ality of “online” timing actually escapes the historiograph-
ical timeline. Accurate timekeeping systems such as the cesium atomic
clock embedded in GPS signal traffic are rather independent. In computing
science, so-called “real-time,” or the physical clock measuring physical
time as hardware, differs from the logical clock as software; this causes the
necessity to synchronize, in intervals, the real-time clock in computers with
external time, by requesting time from time servers and then, by intelligent
algorithms, equaling the time delay in the Network Time Protocol, based
on the Internet Protocol and time synchronization software.
In October 1998, the Swiss watchmaker Swatch announced “Internet
Time,” which undoes the familiar differential time zones. Every day is
divided into 1,000 “beats,” creating a new meridian in Biel, home of Swatch
itself: the Biel Mean Time (BMT) as universal reference to Internet Time.13
This is “an indifferent time, no longer the vectorial time of chronology.”14
At that point, the familiar historic timeline graph fails.
The Fuzzy Present of Internet Communication
Any media archaeological analysis of technological communication in-

frastructures takes their increasingly time-critical nature into account.
If “time” is not understood in a unifying metaphysical sense but in its
Aristotelian definition, it turns out as a functional plurality of signifying
incisions and intervals. What was once electromagnetic “live” radio or TV
signal transmission in telecommunication has been replaced by “real time”
computation which dissimulates the belatedness of complex calculations
of discrete pulse trains so as to trap human perception of the “present.” In
digital online communication, there is no transmission in time but a timing
of data packets (datagrams) that are numerically time-stamped to avoid
internet traffic congestion.
In 1972, Bob Metcalfe developed a program called PING for testing the
interoperability of networked devices on what would come to be called
96 Wolfgang Ernst
Ethernet. The program would open a connection through the physical and
logical network topology to test whether the addressee actually reacts. From
that technical implementation of a time-critical test, Vint Cerf developed the
Transmission Control Protocol for the ARPANET in 1975, which preceded the
actual Internet. So-called time-to-live and ping-to-death are articulations
of internet temporality. The past is not “imperfect” anymore but becomes
“historical perfect,” residually enduring within the present. The “residual” is
still active, “not at all as an element of the past, but as an effective element
of the present.”15
Metaphors like “streaming media” are misleading in their suggestion
of a temporal flow. Even with respect to the signal carrier (the “flow” of
electricity), information depends on the digitally coded electrons. Where
humans believe themselves to be communicating messages via digital
channels, a nonhuman temporality is at work between computers. Any
political criticism of the microphysics of power has to focus on time-critical
events on the most physical level of the OSI network model.
Network culture is less about modernist clock time than about latencies.
The delayed present stems from the “hyperbolic temporalities of digitality.”16
Speculative media theory asks experimental questions: What if data packets
were humans, how (if at all) do they experience time? In Web 2.0 packet
switching, before any kind of “social memory” is triggered, intermediary
storage is a decisive and integral part of the technical transmission itself.
The age-old contradiction between archive and transmission collapses
in the delayed present. Before there can be any moments of short-time
virtual communities (“crowds,” or even societies), data networks consist of
distributed sparks of ultra-short retentions and protentions.
Time-to-Live and Ping-to-Death
In the times of internet protocols McLuhan’s thesis that the pace of electronic
media changes the patterns of temporal perception deserves a closer reading.
Time-critical processes take place in the most media-archaeological sense,
that is, on the basic layer of bit transfer in the internet, the physical layer.
This layer represents the interface between symbolic transfer and material
channel of communication and thus embodies very concretely the interlacing
of logi(sti)cs and matter which is already implied in the term “technology.”
It is on this layer that the voltage level of what is meant to represent a logic
“zero” and a logic “one” is being defined. The function of this bit-transfer
layer is to transform signals from the physical channel into information to
be passed on to level two of the OSI system.17 This identification of signals

happens within the time-critical field, such as signal frequency and signal
duration, synchronous or asynchronous clocking, and the decision on serial
or parallel data transfer.
In communication networks, topological systems are being appropri-
ately expressed by hypertextual links, whereas time-critical processes
rarely become apparent. The answer to this is to find a new term that does
not nominate a new medium but declares the temporal mode of a mode
its decisive media-theoretical criterium. “The real-time web is a set of
technologies and practices which enable users to receive information as
soon as it is published … rather than requiring that they or their software
check a source periodically for updates.”18 The communicative practice of
instant messaging belongs to this temporal field; in McLuhan’s sense, the
message of the medium here is immediacy serving to create the illusion
of a pseudo-copresence. This recent form of web economy is being defined
by communication within the time-critical realm; cyberspace as what Ted
Nelsen calls “docuverse” is being replaced by the extreme speeding up of
information processing in cybertime.19 The internet thus turns out not to
be just a topological extension of a generalized archive, but equally as a
chrono-technical expression of time.
The time-critical message of the internet requires a close look at the
time-critical operations on its physical and logistical levels, such as seen
in the “Ping” signal. Each data packet into which a document has been
sliced is being observed individually; its transfer happens independent of
its preceding or successive packets. This procedure is radically time-critical
because it takes place within the so-called “time to live” field that defines
the maximal temporal duration in seconds that an IP packet is allowed
to exist in the internet. A counter is progressively reduced during this
routing; if the TTL-counter reaches zero before the packet has reached
its destination, it is annihilated.20 Media time is not endless. In TCP/IP
as fundamental network program, techniques of synchronization meet
a deadly economy of time. “Time to live” means that each data packet is
assigned a given lifespan; “time to die” thus becomes a crucial signature
of the information age.
Whereas Husserl’s phenomenology expresses neuro-cognitive perception
of the present in “time diagrams” between retention and protention, media
archaeology tries to precisely identify the rather different operations of
micro-technical signal transduction and algorithmic data processing.
Humans and machines are different in their operative signal processing,
resulting in different tempor(e)alities.
98 Wolfgang Ernst
(A)Temporalizing Radio
The present can be analyzed only when it starts to recede into the past.
The concept of a prehistory of the present technological condition refers
not only to a “before” in its temporally sequential, historical sense but also
to its technological preconditions. This prestructuring “before” endures
or re(oc)curs in the present in nonlinear modes. Media archaeological
analysis, besides its apparent meaning in a search for “origins,” refers to a
structural argument: the arché which is the insisting, essential features of
a technological system. Heuristically, this means analytic reduction to the
essential functions, the elementary bits, a rarefication against discursive
redundancies.
All of a sudden, the “recent” is not past but a concealed retreat, the
hidden, still copresent ground behind the apparent visible. The techno-
logical conditions take place in intervals (epoché) where an established
infrastructure remains valid across all apparent political, historical, and
cultural changes, just as analog AM radio has endured for more than 80
years almost technologically unchanged. Such intervals, as Delta-t, endure
anachronistically (even achronically) when compared with the historical
timeline. Public radio, in Germany, dates back to October 1923. As an inde-
pendent media format, based in autonomous technological implementation,
it apparently dies these years in its familiar AM / FM analog technology.
A historic “timeline” representation of the heroic radio age is misleading.
There is a reentry of “radio” into mobile communication, not as a program
format and “broadcasting,” but in its purest form as technical medium:
wireless (German “Funk”) electro-magnetic waves, this time digitally
modulated, as in mobile telephony and WLAN internet access. In present
mobile communication, there is more radio than ever, even if dissimulated
as a condition of possibility.
The infrastructure of wired and wireless networks increasingly becomes
interlaced. “Rather than wireless cities of wireless networks, it might be more
accurate to speak of the rewiring of cities through the highly reconfigurable
paths of chipsets.”21 In wireless communication, the infrastructure has
become mobile itself. McLuhan described the wireless transmission of
analog signals as a sphere as “acoustic space” because its inherent message
(its implicit “sonicity”) has been the wave as temporal form. “Electric speed
is approximately the speed of light, and this constitutes an information
environment that has basically an acoustic structure.”22 In digital com-
munication, this almost Heideggerian “Being” of electromagnetic spacehas
been inverted from melodic tuning to pulsed rhythm.23 Eleni Ikoniadou’s
The Rhythmic Event conceptualizes digital events not as binary sequences

of zeros and ones but instead as relational pulsation.24 This correlates with
the notion of “algorhythmics”—the rhythmic processes of computational
algorithms.25 In that sense, hardwired temporalities can never be reduced
to (infra-)structures, but in reverse, they temporalize such structuring
conditions (l’archive, in the sense given by Foucault) themselves, “posting”
infrastructures.26
While data-processing microchips are still hardwired infrastructures in
themselves, the wireless signal transfer has become a dynamic infrastructure
consisting of extremely volatile temporal objects: multiple radio-frequency
waves transduced by devices using digital signal processing (DSP) embedded
within mobile phones and other wireless communication devices. To exhaust
the available channels for signal transmission in parallel (known from
George Antheil and Hedy Lamarr’s invention of “frequency hopping”), “the
designers of contemporary wireless DSP chipsets usually supply a palette
of different hardwired algorithms alongside generic processors.”27 Just as
in integrated circuits, the integration of hardwired and wireless networks
allows for algorithms to operate. Radical mathematization transforms
conventional communication engineering into intelligence in its double
meaning.28 While most of the physical layer of information networks “is
quasi-hardwired into semiconductor chips,”29 in order to facilitate algo-
rithmic intelligence (or “cognitive radio”) to unfold its micro-time-critical
efficiency, its dynamic intra-structure is not based on “time” anymore in
the ontological sense but on an asynchronous grid of operative temporal
actions, such as compressing movements into ultra-short slots. Such a mobile
network replaces the immobile cables of the internet or telephone lines
with a dynamic, in fact temporal grid for time-critical signal processing.
Wireless communication is not simply a form of bridging spatial distances by
electromagnetic waves between antenna and receiver anymore, but becomes
primarily a function of intelligent shaping of signal events in between.
The traditional linear channel becomes itself dynamical and tempo-
ralized, composed of algorithms that “generate waveforms that support
conjunctive pathways”—that is, a mobile infrastructure, a new kind of
ether.30 Although this almost amorphous infrastructure is still grounded in
conventional communication engineering and technologies, in its essence it
becomes more radically mathematized than even conceived by Shannon.31
While the traditional radio wave has been a time signal, in digital pro-
cessing it becomes decomposed in frequency values in order to be treated
computationally. The time axis itself is techno-mathematically suspended
when a waveform that varies over time is transubstantiated into a set of
100 Wolfgang Ernst
component frequencies by Fourier transform; at the same time it allows

for switching back from the frequency domain into the time domain. Such
communication infrastructures oscillate between temporal and atemporal
moments. Fourier transform is challenged by a structural incertitude either
toward temporality or spatiality; in so-called wavelets, temporal linearity
itself is suspended. For more efficient transmission streams, digital signals
themselves become superimposed and enclosed in a signal envelope that
looks like white noise, which is finally filtered back into a data stream.
Temporality is just one (itself “ephemeral”) function within such chains
of operations. Efficient coding (convolution) has been developed to match
signal errors and erroneous signal intrusions: “The stochastic character of the
Viterbi algorithm … alters the terrain on which machine time moves.”32 What
once was conceived as moments “in time” becomes a function of irruptions
of the tempoReal (an escalation of Norbert Wiener’s notorious term “time
of non-reality” for the switching momentum between binary states33). Far
beyond the world of communication engineering, knowledge culture has
to acknowledge this nonsymbolic temporal quality, learning from media.
Contemporary “Media Ecology” Identified from within Its

Temporal Infrastructures
In media-archaeological terms, computational technologies, in the very

essence of both components (techné as well as lógos), consists of both “hard-
wired” temporality on the very infrastructural level of microchip circuitry
and “soft-wired” temporalities resulting from what drives such machines:
source code which concerns time-critical operations in computational
languages such as assembly language. Micro-temporal physicality (the “real”)
is as much the object of media-archaeological analysis as “the symbolic”
ordering of signal series, the algorithmic logic of digital culture—bypassing
the imaginary alias “history.”
The often debated “technological determinism” in media-archaeological
analysis refers to infrastructures by necessity: a reconnaissance of both hard-
wired (materially embedded) and softwired (algorithmic) structures which
govern media temporalities from within. Media archaeology is a consciously
anachronistic identification of contemporary predefining states and layers of
electronic media culture. The technological infrastructure of AM radio, for
instance, endured for almost the whole twentieth century, notwithstanding
the political and cultural catastrophes occurring within that century. The
endurance of insisting hardwired and softcoded technologistics creates a
temporal interval of its own. As long as such an epoché is still in operation, its
media are excepted from the transience of the historical event. Such a theory
of media time not only concerns emphatic “deep temporality” on a grand
scale but also inversely re(oc)curs within the microscale of technological
timing in the concrete circuitry of electronics itself.
Current “speculative design” theory registers a shift from progressive
modernity to an epoché of contemporaneity where time is not an empty
duration unaffected by the processes that happen within its technologies.
On the extensive and microlevel of technological infrastructures, there
is no homogeneous “time” but multiple and asynchronous tempor(e)
alities, reminiscent of Ernst Bloch’s notion of “non-contemporaneous
contemporaneities” (“die Gleichzeitigkeit des Ungleichzeitigen”). The
infrastructures of technological contemporaneity are not a coming
together of data in time but a coming together of functional timings.
Technological machine times challenge historicist notions of accumulative
continuity. The concept of “media ecology” in that sense refers to temporal
environments, to being in a clocked world, Heidegger’s “Zeit des Weltbilds”
taken literally.
Notes
1. “Zeitfunktionen der Signale”: Karl Küpfmüller, Die Systemtheorie der elek-

trischen Nachrichtenübertragung (Stuttgart: Hirzel, 1974), 393.
2. See Gerhard Dohrn-van Rossum, History of the Hour (Chicago: University of
Chicago Press, 1996).
3. Martin Heidegger, “The Age of the World Picture,” in The Question Con-
cerning Technology, trans. William Lovitt (New York: Harper & Row, 1977),
115–154.
4. Walter Ong, The Technologizing of the Word (London: Methuen & Co., 1982).
5. See Barry Powell, Homer and the Origin of the Greek Alphabet (Cambridge:
Cambridge UP, 1991).
6. Walter Benjamin, The Work of Art in the Age of Its Technological Reproduc-
ibility and Other Writings on Media, ed. Michael William Jennings, Brigid
Doherty, and Thomas Y. Levin (Cambridge, MA: Harvard University Press,
2008).
7. Edmund Husserl, On the Phenomenology of the Consciousness of Internal
Time (1983–1917), ed. and trans. J. B. Brough (Dordrecht: Kluwer Academic
Publishers, 1991).
8. Lewis Fry Richardson, “A Forecast-Factory,” in Collected Papers of Lewis Fry
Richardson, vol. 1, Meteorology and Numerical Analysis, ed. P. G. Drazin
(Cambridge and New York: Cambridge University Press, 1993), 219.
102 Wolfgang Ernst
9. See Manuela Garretón and Diego Gómez Venegas, “Towards an Archaeology

of Media and Visual Languages,” in The Counterculture Room: Pavillion of
Chile at the London Design Biennale 2016, ed. FabLab Santiago (Barcelona:
Ediciones Poligrafa, 2016), 108–127.
10. George Kubler, The Shape of Time: Remarks on the History of Things (New
Haven, CT: Yale University Press, 1962).
11. Alan Turing, “Lecture to the Mathematical Society on 20 February 1947,” in
A. M. Turing’s ACE Report of 1946 and Other Papers, ed. B. E. Carpenter and R.
W. Doran (Cambridge, MA: MIT Press, 1986), 111.
12. However, Linus Torvald himself argued in favor of such time-critical synco-
pic interpolations. See Robert McMillan, “Linux’s Creator Wants Us All to
Chill Out about the Leap Second,” Wired, January 9, 2015, http://www.wired.
com/2015/01/torvalds_leapsecond/.
13. Geert Lovink, Dark Fiber: Tracking Critical Internet Culture (Cambridge, MA:
MIT Press, 2002), 152. For more, see the chapter “Net.Times, Not Swatch
Time: 21st-Century Global Time Wars,” 142–159.
14. Lovink, Dark Fiber, 143.
15. Raymond Williams, Marxism and Literature (New York and Oxford: Oxford
University Press, 1977), 122. See Shannon Mattern, “Deep Time of Media
Infrastructure,” in Signal Traffic: Critical Studies of Media Infrastructures
(Urbana, Chicago, Springfield: University of Illinois Press, 2015), 71–93.
16. As stated in a lecture by Jussi Parikka, “Of Queues and Traffic: Network Mi-
crotemporealities,” on occasion of the symposium Digital/Social Media and
Memory, University of Glasgow, April 17, 2013.
17. Christoph Neubert, “Elektronische Adressenordnung,” in Die Adresse des
Mediums, ed. Stefan Andriopoulos, Gabriele Schabacher, and Eckhard
Schumacher (Köln: DuMont, 2001), 34–63, 41.
18. Wikipedia, s.v. “Real-Time Web,” accessed January 20, 2010, http://en.wikipe-
dia.org/wiki/Real-time_web.
19. “Früher ging es um die Schaffung von Räumen … heute geht es um die Zeit
selbst, um Chronos, um die Kunst der longue durée.” Geert Lovink, “Was
uns wirklich krank macht,” Frankfurter Allgemeine Zeitung, June 21, 2010,
27.
20. Othmar Kyas, Internet: Zugang, Utilities, Nutzung (Bergheim: Datacom,
1994), 65.
21. Adrian Mackenzie, Wirelessness: Radical Empiricism in Network Cultures
(Cambridge, MA: MIT Press, 2010), 65.
22. Letter to Barbara Ward, February 9, 1973, in Letters of Marshall McLuhan,
ed. Matie Molinaro, Corinne McLuhan, and William Toye (Oxford: Oxford
University Press, 1987), 466.
23. See Rainer Bayreuther, “Phänomenologische Grundlegung,” in Heideg-
ger-Handbuch: Leben, Werk, Wirkung, ed. Dieter Thomä, Katrin Meyer, and
Hans Bernhard Schmid (Stuttgart and Weimar: Metzler, 2013), 509–512.
24. Eleni Ikoniadou, The Rhythmic Event: Art, Media, and the Sonic (Cambridge,
MA: MIT Press, 2014).
25. Shintaro Miyazaki, “Algorithmics: Understanding Micro-Temporality in

Computational Cultures,” Computational Culture 2 (2012).
26. Michel Foucault, The Archaeology of Knowledge, trans. A. M. Sheridan Smith
(New York: Pantheon, 1972).
27. Mackenzie, Wirelessness, 72.
28. This becomes apparent from topics like “Algorithms and Modeling for
Tracking and Locating Mobile Users” in the call for papers to the 10th
International Conference on Wireless and Mobile Network (WiMo 2018),
May 26–27, 2018, Vienna.
31. Claude E. Shannon and Warren Weaver, The Mathematical Theory of Com-
munication (Urbana: University of Illinois Press, 1949).
33. See Claus Pias, “Time of Non-Reality: Miszellen zum Thema Zeit und
Auflösung,” in Zeitkritische Medien, ed. Axel Volmar (Berlin: Kadmos, 2009),
267–279.
About the Author
Wolfgang Ernst is Full Professor for Media Theories at the Institute for
Musicology and Media Science at Humboldt University in Berlin. Having
been academically trained as a historian and classicist (Latin Philology and
Classical Archaeology) with an ongoing interest in cultural temporalities,
Ernst grew into the emergent technology-oriented “German school” of media
science. His academic focus has been on archival theory and museology,
before attending to media materialities. His current research covers “radical”
media archaeology as method, the epistemology of technológos, the theory of
technical storage, the technologies of cultural transmission, micro-temporal
media aesthetics and their chronopoetic potentials, and sound analytics
(“sonicity”) from a media-epistemological point of view. Ernst’s latest booklet
in English is entitled Technológos in Being. Radical Media Archaeology and
the Computational Machine (2021).
Part II
Microtimes
5. Infrastructuring Leap Seconds: The
Regime of Temporal Plurality in
Digitally Networked Media
Isabell Otto
Abstract
The chapter pursues the hypothesis that the plurality of time in an age of
digital interconnectivity imposes itself as a time regime to human and
nonhuman entities. By looking at user practices, conventions of time
measurement, and temporal operations of digital technologies it is argued
that an infrastructural/infrastructuring process consists of the continuous
weaving of a relational assemblage between different temporalities, which
does not harmonize them, but makes them relevant to each other in their
heterogeneity. Thus, the time regime of digitally networked media does
not consist of the power constellation of an absolute, “true,” measurable
time, but of a fundamental plurality, which becomes visible on the basis
of invisible processes and by that challenges all practices of temporal
ordering.
Keywords: time regime, time measurement, temporal ordering, interfac-

ing, practices, leap second
“Enjoy the moments of your life.” With this slogan, the video app Leap
Second promises to keep a special kind of diary: App users are invited to
create one-second videos, select the best second for each day, and compile
the individual seconds of the day into video diaries. This way they are
able to review the days of a month or a whole year in seconds, and finally
share these quotidian, yet outstanding moments of their lives via social
media. On Instagram the app is advertised with small example videos:
In seconds, outdoor and indoor shots alternate, slower and more eventful
doi 10.5117/9789463727426_Ch05
108 Isabell Otto
images, landscapes and portraits; shots of people toasting to each other at

dinner are replaced by video snaps from an airplane; streets and houses
abruptly change to shots of a workplace at home. The videos present a
colorful and varied mixture of lived time.1
I would like to follow up this example with observations on three aspects
that will interest me in the following chapter. First, the one-second videos
refer to the relativity of human perception of time. The “second-days” seem
to take different lengths. The moving image to which they are compiled
appears homogeneous and clocked, but at the same time discontinuous,
depending on the very different processes and situations in the individual
videos, such as movement, actors, sound, color, light, and weather condi-
tions. The homogeneous timing results from the technical settings of the
app: The moments of the day are fitted into the almost imperceptibly fast
succession of 100 centiseconds. In relation to the discontinuous moving
images, the second appears as a reliable, inexorable, even absolute measure
of a technically clocked time that forms and orders human perception of
time.
But the app’s name, Leap Second, probably quite unintentionally, indi-
cates that such a dichotomous juxtaposition does not meet in any way the
socio-technical and infrastructured temporalities under the conditions of
digitally networked media. This is my second observation: the leap second is
a phenomenon that not only emphasizes the meaning of a second, but also
reveals the relativity of all systems of chronology and time measurement. As
I will explain in more detail below, the pluralization of time measurement
systems, which inevitably goes hand in hand with this relativity, becomes
particularly precarious under the digital condition of temporality.
Thirdly, as Leap Second shows, all this has to do with the temporality of
infrastructures and the infrastructures of temporality. The visual interface
of the app bundles disparate and diverse processes of an ordering reference
to temporality. The videos refer to natural time cycles, in which people are
involved, when they picture the change from day to night or from season
to season. The seconds (or centiseconds), on the other hand, refer to the
socio-technical timing of clock time. Below these visible processes, however,
the fabrication of different temporalities takes place on an infrastructural
level, on the basis of micro-temporal and time-critical software processes.2
Beyond the displayed encounters of temporalities, an “inter-facing” between
human users and nonhuman software and hardware components is taking
place.3
An interfacing in this sense is less a spatial and temporal intersection,
less a fixed thing in between than a process that fabricates the togetherness
Infr astruc turing Leap Seconds 109
of disparate entities, or precisely: the being together “in time” (and “in
permanence”) of entities, each with its own proper time. This infrastruc-
tural and infrastructuring process consists of the continuous weaving
of a relational assemblage between disparate entities and their different
temporalities, which does not harmonize them, but makes them relevant
to each other in their heterogeneity and plurality, in the first place. 4 Apps
such as Leap Second thus indicate a problem which they equally offer to
solve.
With these preliminary considerations in mind, I would like to pursue
the following questions: How are human, natural, and technical temporali-
ties confronted with each other under the condition of digital technologies?
And how are they becoming identifiable and problematic for each other
in a processual temporality of infrastructuring? I pursue the hypothesis
that the plurality of time in an age of digital interconnectivity imposes
itself as a time regime to human and nonhuman entities. It is not only
human and social temporalities that are plural. We can f ind a similar
plurality in all orders of measured time. In a first step I would like to look
at the plurality of digitally conditioned temporality from the perspective
of user practices.
Demands of Digitally Infrastructured Temporalities and

Resistant Tactics
In his critical reckoning with the early utopias of the net culture in Zero
Comments, Geert Lovink sketches a differentiated picture of so-called
“Internet time.” On the one hand, he states, there are the practices of
internet users who ideally behave “indifferent to time,” when they spend
time online or rather surrender themselves to the “luxury to get lost” and
losing time as “data dandies” strolling through the net, contradicting all
imperatives of effectiveness. On the other hand, Lovink observes the work
processes of the IT industry, whose cooperation extends to different time
zones. Global cooperation shows most clearly that there is “no simple
synthesis of the local with the global.”5 An “enhanced global time awareness”
is necessary, “an awareness of other times.”6 Lovink proposes to examine
the requirement or even the demand to be confronted with a plurality of
different times under the condition of digitally networked work as “time
regimes under which today’s Internet user are actually operating.”7 This
temporal plurality not only arises in the cooperation across time zones
and in the spatiotemporal delimitation of work, which the internet makes
110 Isabell Otto
possible and demands, but is also evident in the everyday practices of

digital networking:
The online session is perhaps the best time unit to express what time on
the Internet could look like. Think of more sessions happening simulta-
neously, such as chatting, talking on Skype, surfing MySpace, watching
videos, following blog links, reading and answering incoming e-mails,
and conducting a search. When you are online all the time (with a DSL
broadband Internet connection), it is the bundle of these never-ending
sessions that defines the Internet experience.8
Lovink believes, in 2008, that digitally networked time can be limited to

time units of being online. Digital time in this perspective is confined to
a temporal refugium of the internet in which different proceedings run
simultaneously but are separated from the “offline-world.” However, he
designs the scenario of “never-ending sessions,” which is more appropriate
for the current situation, because the technological condition of digital
interconnectivity allows users to be online all the time.
It is exactly this state that current countermeasures of various “digital
detox” movements want to deal with. “Participants gain insight into personal
lifestyle techniques and practices that keep them grounded and connected
even in the most stressed, overwhelming and technologically driven times,”
says the invitation of a digital detox organization, which has dedicated
itself to the goal of (re)establishing some sort of balance in the digital age
and offers device-free events, workshops, and retreats in nature, with the
slogan “Disconnect to Reconnect.” A recreational holiday at a California
camp promises, according to its rules, a withdrawal on several levels: “No
Digital Technology—No Networking—No Phones, Internet or Screens—No
Work-Talk—No Clocks—No Boss—No Stress—No Anxiety—No Fomo
(fear of missing out).”9
It is striking how much the digital detox movement refers to temporal as-
pects in its diagnosis of the current situation: We suffer from the compulsion
to have to respond immediately to messages, to constantly check the input
of new messages; we have to keep pace with the speed of networked com-
munication without finding time to draw breath. The plurality of constant,
simultaneous, and far-too-fast processes is, according to these diagnoses,
a characteristic of a new time regime of digitally networked media. There
is no doubt that digital detox does not offer the prospect of a renunciation
and definitive liberation from a digitally networked life. Disconnection is
carried out for the purpose of better reconnecting afterwards. The temporary
voluntary exclusion from a networked community follows the logic of

permanent connectivity and affirms it, as Urs Stäheli has shown.10 “Digital
detox” aims at the formation of subjects who can better adapt themselves
to the demands of a “hardwired time” of digital networks and meet them
in a self-regulatory way.
The same can be seen in the advice literature on self- and time-manage-
ment. Typical here is the recommended way of dealing with the synchronous
and asynchronous forms of communication that characterize a digitally
networked workplace and that time management literature wants to
optimize. This is where the plurality of diverse temporalities becomes
apparent—especially in the description of badly handling emails and
instructions for a correct way to do so. The time management literature
suggests strategies for getting a grip on the “flood of e-mails” that charac-
terizes every workplace. “Why am I not able to work because of all these
e-mails?” A first answer to this question comes from a time-management
guide using the Microsoft Off ice Outlook mail and calendar program,
which argues for protecting the worker’s proper time: “Don’t be distracted
all the time. Answer consciously and deliberately, instead of always reacting
immediately to every message.”11
The educational program of time-management literature is not about a
complete correction of a work situation characterized by too many temporal-
ly diverging and accelerated processes, disturbances, and interruptions, but
about a better adaptation of the working subject to new technological (work)
environments. The addressing of the subject is neoliberal and governmental:
the aim that the subject voluntarily concerns itself with a safeguarding of
proper relaxation and recovery times, which guarantee a better integration
into a technological-economic power constellation. The supposed resistance
strategies of digital detox and time-management are rather strategies of
regulation, which refer to a techno-environmental condition without being
able or wanting to change it. But how is the “ecology” of these practices to
be characterized,12 the environment of human and nonhuman, natural
and technological procedures that surround the practices and are regu-
lated to govern subjects or enable self-government? What processes and
constellations of power are inscribed in the regime of plurality? What are
the conditions of possibility for the regime of plurality?
To determine this more precisely, a software developer is assumed whose
practices of digitally networked collaboration across time zones could be
directed by self-governmental regulation, such as digital detox or time
management. In the course of a working day, when she not only writes the
code for a new application, she communicates with designers, customers,
112 Isabell Otto
fellow developers, and hardware companies scattered around the planet

in different time zones. She not only has her own physical temporality
and energy phases to consider, as Lovink cites an observer of working in
the IT outsourcing industry, the “diurnal cycle of the human animal,”13
but also has to develop an increased awareness of the temporalities of her
colleagues. The plurality of time is also conveyed by the parallel processes of
synchronous and asynchronous processes of analogue and digital working
communication through email, video chat, letter post, inhouse messaging,
memo, meetings or telephone calls, in which different analogue and dig-
itally networked devices (computers, smartphones, tablets, watches) are
included. Finally, temporal plurality is conveyed to her in the practices of
programming, which have to be oriented toward the different simultaneous
processes of the computer.
The multiplicity of time is not necessarily tied to digitally networked
media and could also be described from the sociological perspective of
Barbara Adams, who sees a variety of other times included in the shaping
of social time, “a multitude of times which interpenetrate and permeate our
daily lives,” including memories, anticipations, travel, and mobile working
hours, as well as weather conditions and temporalities of the involved
media.14 This multiplicity confronts the software developer just as it did an
accountant around 1900. But I want to argue, following Lovink’s observations,
that under the condition of digitally networked media the plurality of time
becomes a time regime on which regulating strategies of the adaptation of
subjects orientate themselves, because of a specific relationship of visibility
and invisibility that characterizes the temporal ordering of human subjects
confronted with the temporalities of digital infrastructures. This means that
digital time cannot be realized at all as a liberation from the dictate of time
measurement, as imagined in visions of “network time.”15 Instead, social
temporalities are structured and challenged by the (micro-)temporalities
of digital infrastructures. Additionally, there can just as little be a new
standard time established that tames this plurality of time. The plurality
of time cannot be suspended.
We are dealing with a specif ic form of (in)visibility of plural space-
time systems: The time regime of digitally networked media consists in a
visibility of different spatiotemporal orders that run simultaneously but
independently of each other and that are perceptible for human users in
processes of interfacing. This visibility, which is by no means limited to
visuality and thus to graphical user interfaces, is conditioned and made
possible by the constitutive invisibility of digital processes that take
place in discontinuous pulsing below the threshold of human perception
and thus evoke the impression of “real time” without identifying the
synchronization and coordination necessary for producing visibility and
continuity.16 In digital processes, the transmission time of a time signal
that connects two independently running space-time systems is (for
human observers) imperceptibly short. This is precisely the basis of the
universal time fictions from the early internet era. For example, Swatch
has for some time pursued the goal of establishing a globally uniform
time order measured in beats. But the time regime of the digital does not
consist of a standard time.
Digital interconnectivity brings independently running time orders into a
relationship of mutual visibility and disturbability. In this way, the relativity
and contingency of any time system become recognizable. The time regime
of digitally networked media does not consist of the power constellation of
an absolute, “true,” measurable time,17 but of a fundamental plurality, which
becomes visible on the basis of invisible processes and by that challenges
all practices of temporal ordering and synchronization. In this sense, the
time of digital media is not characterized by a multiplicity of time, which
enables new creative developments, but by a time regime that requires an
increased sensitivity for the relativity and plurality of time.
Our imagined software developer, based in Vienna, who works under the
condition of digital networked working environments, knows that for her
colleague in Australia, with whom she is having a Skype call, different space-
time conditions prevail than for herself. Daytime and season are completely
different. But the imperceptible processes enabled and conditioned by the
infrastructures of digital networks are what make this other time visible
and audible on her device in a process of interfacing, a space-time system
that appears simultaneous to her own spatio-temporality, but nevertheless
is perceptibly different. This demands a temporal plurality from her and
challenges her to adapt her practices—perhaps when her meeting is in the
morning of Sydney local time and she has to fight tiredness because for her
it is 11 p.m. Simultaneous temporal orders are no longer independent of each
other under the condition of digital interconnectivity, but become visible
and relevant to each other. The relativity and contingency of temporal
orders become apparent, the locally and diurnally different observations
and experiences of time. As will still be seen, this applies not only to human
perceptions of time and practices of temporal order, but also to the technical
processes of digital connectivity. However, it is fundamental that relativity
and contingency are inscribed in every measurement of time. I will further
explore this point using the example of the measurement of the second in
the next step of my argumentation.
114 Isabell Otto
The Measurement of the Second and the Fleeting Stability of

Time Orders
Social time orders in the form of clocks, work plans, and calendars form time
regimes that demand self-regulating adaptation of subjects to economic
structures.18 However, the clock is not a once and for all stable technical
timepiece that regulates the social realm, a determining, inanimate time
technique that threatens and destroys living, subjective times. Rather,
each time measurement is based on a “technicity” in the sense of Gilbert
Simondon as the “degree of the object’s concretization,”19 which stabilizes
a spatiotemporal coordination of socio-technical collectives, but also
constantly keeps them open for restructuring and, despite an increasing
precision of technical time measurement and standardization, can only
establish a temporarily stable structure.20 The fact that the continuous
restructuring of time regimes is taking place on the basis of a changing
technological condition becomes particularly clear in the current debate
on the leap second.
Since 1972, an additional second has been inserted at irregular intervals
into the Coordinated Universal Time (UTC) to compensate for fluctuations
and a gradual slowdown in the earth’s rotation. Due to its relation to the sun’s
position, UTC is still regarded as “natural” and “appropriate” for living beings
on earth. Measured on the basis of atomic clocks, UTC is slowed down by the
leap second in such a way that it never deviates by more than 0.9 seconds
from a time measurement oriented at the position of the sun, the rotation
of the earth, or the orbit of the earth. This deviation is determined by the
International Earth Rotation and Reference System Service by constant
observation or measurement. The service then decides whether the day is
one second longer at the end of June or December, i.e., a 61st second—23h
59m 60s—is inserted into the UTC at the end of the day. Without the ir-
regular insertion of leap seconds, according to a fear of the unpredictable
development of the difference between the Earth’s rotation and atomic
clocks, the deviation could be four hours in 2000 years. Even further in the
future, the clocks might indicate noon when it is in the middle of the night.21
It is important to recognize that the leap second problem has arisen only
from timekeeping practices that are part of a continuous restructuring of
clock time and its basal unit of measurement, the second. The leap second
is a metastable remnant that results in a “supersaturation” of the current
standard time system and requires its restructuring.22 The leap second
thus arises in the course of a technicity of measured time, which inscribes
into each time regime an openness to restructuring: With the beginning
of time measurement by atomic clocks, the “power/knowledge” regime of

clock time, the political and institutional competence shifts as well as the
expertise between astronomy and atom physics.23 A decisive aspect of this
shift is the refinement and stabilization of time measurement by atomic
clocks. From the mid-1950s onwards, the physical determination of time by
measuring the transitions between levels of the atom’s ground state was a
new way of separating the temporal order from planetary constellations.
Until 1956, the second was determined by the Earth’s rotation around its
own axis, i.e., as a fraction of the mean solar day, and was then replaced by
the ephemeris second, which is oriented at the Earth’s orbit around the sun
and which was considered to be more stable, then. Quite in contrast to the
irregularities of astronomical time measurement, the period duration of an
electromagnetic radiation absorption or emission in the transitions in the
ground states of an atom, proves to be—at least in principle—constant.
Physicists in the middle of the twentieth century, using the cesium atom,
determined the length of a second that remains valid until today, defined
by the Bureau International des Poids et Mesures (BIPM): “The effect of this
definition is that the second is equal to the duration of 9 192 631 770 periods
of the radiation corresponding to the transition between the two hyperfine
levels of the unperturbed ground state of the 133Cs atom.”24
The new precision of time measurement entails some consequential
problems because the accuracy of the measurement makes it obvious that
every time measurement is relative and situational. In the atomic age of
time measurement the irregularities of universal world time become vis-
ible: not only the relativity of time, but also the irregularities of each time
measurement that is oriented at the position of the sun, the rotation of the
earth, or the orbit of the earth. Basically, inaccuracy also applies to atomic
time measurement. However, the determination of the second, which is
nevertheless accurate to microseconds, can show that the mean solar day of
the astronomically calculated UT1 (the current equivalent of Greenwich Mean
Time) is 2 milliseconds longer than the day calculated from atomic seconds: it
comprises 86,400.002 instead of exactly 86,400 seconds. The atomic clock thus
becomes a (temporarily) more stable and precise clock only in comparison
with other clocks, which can represent the course of time less constantly.
This difference between the atomic and astronomical time scales, visible in
the precision of atomic time, gives rise to the UTC regime and, with it, the
need to introduce leap seconds that keep this difference within a tolerance
range in practices of continuous balancing. In favor of a uniform time order,
the leap second prevents two time scales from drifting apart, but at the
same time it inscribes a discontinuity into this uniform time order, which
116 Isabell Otto
in the case of digital networking is clearly recognizable as a problematic

uncertainty. As a sign of its technicity, the leap second keeps universal time
open and changeable. Its potential to interconnect the macro-cosmological
environment of the astronomical and the micro-cosmological environment
of atomic time measurement can thus only temporarily lead to a stable order
of time. A stability that, under the condition of digitally networked media, is
currently about to turn into instability. While the leap second has guaranteed
a coordination with, or adaptation of, socio-technical to planetary-organic
processes for more than 40 years, thus enabling a collective of human and
nonhuman entities to be together “in time,” it now fails in the confrontation
with digital media and processes that are not only time-dependent, but
time-critical.25 A restructuring of the time regime of a universal world time
is necessary, which in the interaction of political, economic, and scientific
interests will probably find a new, but just as temporary answer to this
problem. In the third step of this chapter I will argue, that the plurality of
time not only challenges social practices of time ordering but also digital
devices and systems.
Leaping Seconds and Digital Interconnectivity
The plurality of the now coexisting different orders of time measurement

is particularly visible in the problems of networked computer systems. One
of the oldest internet protocols was developed by David L. Mills, who calls
himself an “Internet timekeeper.”26 To this day, the so-called Network Time
Protocol is used and ensures clock synchronization on the internet. It is
based on the coordinated world time and inserts leap seconds. The Network
Time Protocol is part of the decentralized power structures of the internet
as described by Alexander Galloway.27 It can be described as a kind of time
management guide for digital devices because it allows computers to cope
with the requirement of time plurality.
The protocol provides for a tree-like structuring from servers to clients,
based on computer servers whose clocks are synchronized via precise
atomic clocks. These in turn can be used as a reference for the coordinated
world time. When a leap second is introduced, the time elapsing according
to the Network Time Protocol is frozen for one second. Immediately after
the introduction of the leap second, the system clock continues to run as if
nothing had happened. It “forgets” the introduction of the leap second as
well as the introduction of all previous leap seconds. Each new leap second
generates a new time scale and thus pluralizes internet time successively.
There are as many Network Time Protocol time scales as leap seconds
historically introduced since 1972; therefore, each time the system’s past is
accessed, the time scale must change and the corresponding leap seconds
must be subtracted again.28 The discontinuity of the coordinated world time
thus multiplies in the historical course of the internet time based on the
Network Time Protocol, which adapts to UTC again and again.29 The question
arises as to what happens to computer processes within the “paused time”
of the leap second, whether the processuality of computational time, which
is based on discontinuities and caesuras,30 but is nevertheless constantly
ongoing, can be stopped at all for one second. This is hard to imagine for
complex, digitally networked systems, according to the argumentation of
developers of the company Google:
Very large-scale distributed systems, like ours, demand that time be

well-synchronized and expect that time always moves forwards. Com-
puters traditionally accommodate leap seconds by setting their clock
backwards by one second at the very end of the day. But this “repeated”
second can be a problem. For example, what happens to write operations
that happen during that second? Does email that comes in during that
second get stored correctly? What about all the unforeseen problems that
may come up with the massive number of systems and servers that we
run? Our systems are engineered for data integrity, and some will refuse
to work if their time is sufficiently “wrong.” We saw some of our clustered
systems stop accepting work on a small scale during the leap second in
2005, and while it didn’t affect the site or any of our data, we wanted to
fix such issues once and for all.31
The solution for Google is to “smear” the leap second: an adjustment by

milliseconds over a day. But would it not be desirable to abolish the leap
second and move on to a continuous time measurement that completely
detaches itself from the sun as the central timer? This question has been
increasingly discussed since the beginning of the 2000s. As a trigger, an
increased time sensitivity due to the (feared) “Millennium Bug” of numerous
computer systems is very likely. The camp of supporters is growing steadily.
In view of the changing technological conditions, a collective of authors
that brings together the physical, astronomical, and geopolitical expertise
of different institutions concludes that we should not hesitate to establish
a binding time system that adapts to the modern technologies and needs of
accurate time measurement in space travel, satellite navigation, metrology,
telecommunications, and synchronization of networked computers.32
118 Isabell Otto
The fact that the International Telecommunication Union took up the

debate and examined the feasibility of a continuous time scale at the World
Radiocommunication Conference in 2015 is a clear signal for the shift in
the time regime resulting from the current technological condition. Judah
Levine, of the US National Institute of Standards and Technology (NIST),
clearly voted as early as 2013 to refrain from inserting leap seconds in
the future. Keeping the difference between coordinated world time and
astronomically measured time as small as possible is too high a price to
pay, given the massive difficulties that leap seconds entail for digitally
networked systems:
The problems of time-ordering, causality and the ambiguity of time

intervals in the vicinity of a leap second are not easily remedied because
they arise in a fundamental way from the interaction of the binary
representation used for time stamps and the occurrence of a positive
leap second. During a leap-second correction, the time servers operated
by NIST will receive approximately 150 000 time requests when the
time transmitted by the server is 23:59:59, and the increasing number
of financial transactions that depend on millisecond-level timing are
sure to be affected.33
However, no decision was made at the 2015 World Radiocommunication

Conference. The evaluation of further studies and the consideration of
a new time order were postponed until 2023. The abolition of the leap
second could be identified as a subjection to a “regime of technology,” as
a “harder hardwiring” of temporality, that now dominates all natural and
social processes; as an overhand gain of techno-economic processes that
sets the pace and the need for precision for a binding world time. A world
time without leap seconds—would that be an order of time that makes a
(planetary, organic) outside of technological processes irrelevant? A closer
look at the ecology of the leap second, its integration into a network of
atomic, planetary, organic, social, and technological relations, has shown,
however, that this description would be too short-sighted. In a final step I
would now like to specify my thesis of temporal plurality.
The Time Regime of Plurality
What the perception of different temporalities in (work-)processes of global

interconnectivity makes just as clear as the drifting apart of astronomical
and atomic orders of time, is the fundamental relativity of each regularity

of time. Influenced by the theory of relativity, which he received as a “new
theory” in 1920, Alfred North Whitehead already clearly summed up this
fact at the beginning of the twentieth century in his natural philosophic
work Concepts of Nature and warned against a confusion:
According to the new theory, there are an indef inite number of dis-
cordant time-series and an indefinite number of distinct spaces. Any
correlated pair, a time-system and a space-system, will do in which to fit
our description of the Universe. We find that under given conditions our
measurements are necessarily made in some one pair which together form
our natural measure-system. The difficulty as to discordant time-systems
is partly solved by distinguishing between what I call the creative advance
of nature, which is not properly serial at all, and any one time-series. We
habitually muddle together this creative advance, which we experience
and know as the perpetual transition of nature into novelty, with the
single-time series which we naturally employ for measurement.34
If one avoids this misunderstanding and distinguishes between process—as

the term Whitehead uses in his major work Process and Reality, instead of
“creative advance of nature”35—and (measured) time, one can thus doubt
that two observers mean the same thing when they determine space and
time from their own perspectives. Each measurement of time must therefore
produce a different order of time. If one assumes, with Whitehead, that
space and time (in the measurable sense) are only possibilities to express
certain truths about the relations between constantly becoming entities
within the basic process of all existing things, but that there are numerous
truths corresponding to the numerous space-time systems, time orders
such as clock time or coordinated world time must be particularly powerful
and momentous orders that are temporarily capable of forming regular
time regimes, time regimes that combine chronopolitical with geopolitical
interests and form and sustain cultural or social sequences and practices.366
However, the technicity of time measurement, which becomes recogniz-
able by the leap second, introduces time as a fundamental process of becom-
ing and passing into every order of time and prevents its complete fixation. It
thus focuses on the condition of power relations and normalization processes
of a unified temporal order. Each practice of time ordering is to be viewed in
the context of its “ecology of practices”377 and develops its own truth there,
which is always only one within a plurality of other “true” time orders taking
place in parallel, which this practice must blank out and from which it must
120 Isabell Otto
detach itself in order to justify itself as “true.” A critique of power in the sense
of a “cosmopolitan politics” has to return the practices of time measurement
to their situational interdependencies of human and nonhuman processes
and their mutual dependencies, relations and affiliations.388 Those who want
to secure access to time “in itself” through operations of measurement must
make productive what Whitehead characterizes as a misunderstanding: the
identification of time as a temporal regularity that is temporarily stabilized,
and time as a fundamental processuality that permeates every stabilization.
The implementation of a standard time with a universal claim can thus be
described as a gesture of power, as a power/knowledge regime in the sense
of Michel Foucault, which helps a temporal order to gain hegemony and
which—as in the case of clock time and its standardization in a universal
world time since the end of the nineteenth century—represents the condition
for a normalization and naturalization of this one possibility of temporal
order, detached from its situational contexts of measurement. This procedure
is, however, supported by the socio-technical production of a measuring and
abstracting-calculating access to time itself, which changes on the basis of
the changing technological condition, but which must suppress this change
in favor of a universalization of time. “Physicists feel weak and they protect
themselves with the weapons of power, equating their practice with claims
of rational universality.”39
Digital infrastructures provide a constellation in which the relativity of
temporal regularity becomes visible and the assertion of an identity between
time order and “natural” time “in itself” is no longer a necessary argument
for establishing a binding standard time. The coupling of power and truth
is replaced by a combination of power and neoliberal economic expediency
that knows about its contingency. The abolition of the leap second and the
introduction of a universal time running constantly over atomic seconds
would not be a final solution, not an order of time that would be adequate
for a digital temporality once and for all, but only a temporary stabilization,
another time regime that has emerged from an ecology of atomic physics,
astronomy, and IT practices and that differs from previous regimes (e.g.,
Greenwich Mean Time) by a clear reference to the relativity of time and to
the plurality of possible time measurements. However, the debate about
the leap second shows a circumstance that is of highest relevance for the
investigation of a temporality of infrastructures: The temporal processes
in systems of digital data transmission do not take place in a refugium
that completely excludes them from their cosmological environment—the
radiation emissions of the cesium atom or the gravitational f ields and
rotations of the earth. Rather, digital processes are part of this environment,
they are infiltrated into it, shape and alter it, and determine the orders
of time that can temporarily stabilize on its basis. 400 In a time of digital
interconnectivity, temporality is conditioned by a technologically shaped
environment, by a media ecology from which users, technical objects, and
data networks emerge with their respective orders of time, by a web of
relations that runs through and crosses every socio-technical order of time.
The discrete processuality of digital infrastructures as a condition of
possibility for the simulation of perceptible digital objects justifies speaking
of a specificity of a digital time that characterizes our present condition
through digitally networked media that permeate all areas of life. This
characteristic is insufficiently captured with a reference to the multiplicity of
digital time. For the description of an experiencability of manifold inherent
times—of the user, the device, the software, the network—all of which are
related to one another and perceived as a multi-temporal web, leaves open
why this should be new or special under the condition of the digital. The
differentiation of a time of digital interconnectivity within the fundamental
plurality of time, according to my thesis, lies in a specific visibility of or
disturbability by the plurality of time at the level of technical operations
und user practices. The digital process causes the perceptible appearance of
a symbolized time, i.e., the perceptibility of images, sounds, or text elements
on displays, in a way that is imperceptible to human beings, and thus under-
mines the difference in their spatial-temporal orders by the speed of digital
processing undermining the perceptible low-frequency range. However, this
happens without cancelling the difference of spatiotemporal orders. It is
rather reduced to an imperceptibly small “in-between.” Thus, time orders
are confronted with each other that would otherwise run independently
of and undisturbed by each other.
The time of digital networks thus does not multiply the temporalities
themselves but the constellations in which different times are confronted
with each other. A webcam image during a Skype call shows a different
spatiotemporal situation, but it shows (simulates) it here and now as simul-
taneous and relevant to the practices of the user viewing it. This becoming
visible for each other—or better: the becoming relevant of different time
orders—takes place on the level of not only human perceptibility but also the
infrastructure: the manifold hardware and software processes and system
times. The relevance of temporal plurality is particularly apparent in the
change of uniform time orders, such as the coordinated world time, or in
the coordination of human and technical proper times. The confrontation
with the plurality of different temporalities does not lead to a (harmonious)
temporal fabric. Rather, the plurality of time under the condition of digitally
122 Isabell Otto
networked media requires a constant sensitivity to the relativity of time and

results in an increased necessity for coordination or (re-)ordering. This can
be seen in the debate about the leap second as well as in the possibilities of
the records of lifetime as promised by the app Leap Second. The diversity of
time, visible in its manifold confluence, becomes a regime of plurality that
constantly challenges the practices of ordering time anew.
Notes
1. See Jon Andersen @leapsecond, Instagram, https://www.instagram.com/

leapsecond/?hl=de, October 22, 2019.
2. Wolfgang Ernst, Chronopoetics: The Temporal Being and Operativity of Tech-
nological Media, trans. Anthony Enns (London and New York: Rowman &
Littlefield, 2016).
3. Alexander Galloway, The Interface Effect (Cambridge and Malden, MA:
Polity Press, 2012).
4. For more on infrastructuring, see Susan Leigh Star and Geoffrey C. Bowker,
“How to Infrastructure,” in Handbook of New Media: Social Shaping and
Consequences of ICTs, ed. Leah A. Lievrouw and Sonia Livingstone (London:
SAGE Publications, 2002), 151–162.
5. Geert Lovink, Zero Comments. Blogging and Critical Internet Culture (New
York and London: Routledge, 2008), 120.
6. Lovink, 121, 126.
7. Lovink, 118.
8. Lovink, 122f.
9. Digital Detox Retreats, http://digitaldetox.org/retreat, October 22, 2019.
10. Urs. Stäheli, “Entnetzt euch! Praktiken und Ästhetiken der Anschlusslo-
sigkeit,” Mittelweg 36 22, no. 4 (2013): 3–28.
11. “Lassen Sie sich nicht dauernd ablenken. Antworten Sie bewusst und über-
legt, anstatt auf jede Nachricht immer sofort zu reagieren.” Lothar Seiwert,
Holger Wöltje, and Christian Obermayr. Zeitmanagement mit Microsoft
Office Outlook. Die Zeit im Griff mit der meistgenutzten Bürosoftware—Strate-
gien, Tipps und Techniken (Versionen 2003–2010), 8th ed. (Unterschleißheim:
Microsoft Press, 2010), 13.
12. See Isabelle Stengers, “Introductory Notes on an Ecology of Practices,” Cul-
tural Studies Review 11, no. 1 (2005): 183–196.
13. Lovink, 120.
14. Barbara Adams, Timewatch: The Social Analysis of Time (Cambridge: Polity
Press, 1995), 12.
15. Robert Hassan, “Network Time,” 24/7: Time and Temporality in the Network
Society, ed. Robert Hassan and Ronald E. Purser (Stanford, CA: Stanford
University Press, 2007), 37–61.
16. See Friedrich Kittler, “Real Time Analysis—Time Axis Manipulation,”

Zeit-Zeichen: Aufschübe und Interferenzen zwischen Endzeit und Echtzeit, ed.
Georg Christoph Tholen and Michael O. Scholl (Weinheim: Acta humanio-
ra, 1990), 363–377; and Esther Weltevrede, Anne Helmond, and Carolin
Gerlitz, “The Politics of Real-Time: A Device Perspective on Social Media
Platforms and Search Engines,” Theory, Culture & Society 31, no. 6 (2014):
125–150.
17. Clark Blaise, Time Lord: Sir Sanford Fleming and the Creation of Standard
Time (New York: Vintage, 2002).
18. Colin Symes, “Chronicles of Labour. A Discourse Analysis of Diaries.” Time
& Society 8, no. 2 (1999): 357–380.
19. Gilbert Simondon, On the Mode of Existence of Technical Objects (Minneapo-
lis, MN: Univocal, 2017 [1958]), 72.
20. Adrian Mackenzie, “The Technicity of Time: From 1.00 oscillations/sec to
9,192,631,770 Hz,” Time & Society 10, no. 2–3 (2001): 235–257.
21. Robin McKie, “Time Running Out for ‘Leap Second’ That Has Kept Us in
Step with Our Slowing Planet,” Guardian, January, 15, 2012, http://www.
theguardian.com/science/2012/jan/15/time-leap-second-atomic-clock.
22. See Gilbert Simondon, “The Position of the Problem of Ontogenesis,” Par-
rhesia 7 (2009 [1989]): 6.
23. See Michel Foucault, Power/Knowledge. Selected Interviews and Other Writ-
ings, ed. Colin Gordin (New York: Pantheon Books, 1980).
24. Bureau International des Poids et Mesures (BIPM), Measurment Units,
https://www.bipm.org/en/measurement-units, October 26, 2019.
25. See Ernst, Chronopoetics.
26. David L. Mills, “A Brief History of NTP Time: Confessions of an Internet
Timekeeper,” ACM Computer Communications Review 33, no. 2 (2003), http://
www.eecis.udel.edu/~mills/database/papers/history.pdf.
27. Alexander R. Galloway, Protocol: How Control Exists after Decentralization
(Cambridge, MA, and London: MIT Press, 2004).
28. See Mills, “Brief History of NTP Time.”
29. As can be seen in RFC 8633 from July 2019, the Network Time Protocol
is still a standard recommendation. It might be that the software clock
synchronization system HUYGENS, which allows nanosecond accuracy, will
change the fleeting stability of NTP in dealing the discontinuities of leap
seconds. Yiling Gent, et al., “Exploiting a Natural Network Effect for Scal-
able, Fine-grained Clock Synchronization,” Proceedings of the 15th USENIX
Symposium on Networked Systems Design and Implementation, April 9–11,
2018, Renton, WA, USA, https://www.usenix.org/system/files/conference/
nsdi18/nsdi18-geng.pdf. I thank Armin Beverungen for this hint.
30. Georg Christopher Tholen, Die Zäsur der Medien: Kulturphilosophische
Konturen (Frankfurt am Main: Suhrkamp, 2002), 141, 132.
31. Christopher Pascoe, “Time, Technology and Leaping Seconds,” Google
Official Blog, September 15, 2011, http://googleblog.blogspot.de/2011/09/
time-technology-and-leaping-seconds.html.
124 Isabell Otto
32. Robert A. Nelson, et al., “The Leap Second: Its History and Possible Future,”
Metrologia 38 (2001): 509–529.
33. Judah Levine, “Impact of Leap Seconds on Digital Time Services: Internet
Time Servers,” ITU News 7 (2013), https://itunews.itu.int/En/4276-Impact-
of-leap-seconds-on-digital-time-services-BRInternet-time-servers.note.
aspx.
34. Alfred North Whitehead, The Concept of Nature (New York: Cosimo, 2007
[1920]), 178.
35. Whitehead, Process and Reality: An Essay in Cosmology, Corrected Edition
(1927–28), ed. David Ray Griffin and Donald W. Sherburne (New York: Free
Press, 1978).
36. See Ian Klinke, “Chronopolitics: A Conceptual Matrix,” Progress in Human
Geography 37, no. 5 (2013): 673–690.
37. Stengers, “Introductory Notes on an Ecology of Practices.”
38. Stengers, 188f. See also Cosmopolitics 1. I. The Science Wars. II. The Invention
of Mechanics. III. Thermodynamics. (Minneapolis and London: University
of Minnesota Press, 2010); and Cosmopolitics 2. IV. Quantum Mechanics. V. In
the Name of the Arrow of Time. VII. The Curse of Tolerance (Minneapolis and
London: University of Minnesota Press, 2011).
39. Stengers, “Introductory Notes on an Ecology of Practices,” 196.
40. See Erich Hörl, General Ecology: The New Ecological Paradigm (London:
Bloomsbury, 2017).
About the Author
Isabell Otto is Professor of Media Studies at the Department of Literature,

Art, and Media Studies, University of Konstanz. Currently she researches on
temporalization, media theory and digitally networked mobile communities.
Her book Prozess und Zeitordnung. Temporalität unter der Bedingung digital
vernetzter Medien is published by Konstanz University Press (2020).
6. Life at the Femtosecond
Geoffrey C. Bowker
Abstract
This chapter argues that we are creating a fundamental new ontological
layer which has far reaching social and political consequences. In this
new “present,” we need to account for ever higher time frequencies—since
this is where decisions are being made which affect us all. In particular
it argues that whereas the prior forms of capitalism concentrated on
colonizing space, newer forms involve colonizing time.
Keywords: computer time, ontology, science studies, temporality,

acceleration
Par la vitesse, le présent se transforme en éternité.

—François Hartog 1
Time, loosely put, is the direction in which physics tells its best stories.
—Craig Callender2
And so, Ladies and Gentleman, as the Southwestern train of time is delayed by
the points failure of predictability and the pissed off Poole-based passenger of
perpetuity becomes trapped in the out-of-order train toilet of eternity, I notice
it’s the end of the show.
—Jack Dee3
It is a well-trodden historical path to take Charles Babbage, through his

design for an Analytical Engine, as the progenitor of the modern digital
computer. This hindsight can make it harder to see an origin of computing
in the production of machinery—the Engine was to be a vast machine,
deploying techniques learned in his tour of machinery and manufactures
doi 10.5117/9789463727426_Ch06
126 Geoffrey C. Bowker
in England in the 1820s (weaving and watch production) in the heat of the
industrial revolution and expressed not in binary code but in a symbolic
language describing the functioning of machines.4 He conceived it not only
in terms of the time compression we associate with computing, but rather
its opposite: “It is impossible to construct machinery occupying unlimited
space but it is possible to construct finite machinery, and to use it through
unlimited time. It is this substitution of the infinity of time for the infinity
of space which I have made use of, to limit the size of the engine and yet to
retain its unlimited power.”5 This infinite logical time was also—as this new
language and as his analysis of time taken for operations in watch production
in the 1820s—complemented by time compression: the ineluctable drive
to make things happen faster and faster: “whenever the Analytical Engine
should exist, all the developments of formula would be directed by this
condition—that the machine should be able to compute their numerical
value in the shortest possible time.”6
Babbage’s new technology, then, ushered in its own infinite time at the
very moment when geology in the 1830s invented a new infinity of time7—a
consonance of which Babbage was deeply aware: variations in time scale
from the fleeting life of the mayfly to the eons of geology were central to his
imaginary.8 One of the tricks of the trade for digital computers is to collapse
potentially infinite serial time through synchronization: running many
operations at the same time and coordinating the results.9 Commands to
go parallel, spawn processes, and sync results sit on top of the underlying
serial structure.10 An extreme formulation of the substitution of space
for time here is David Deutsch’s description of quantum computing as
a way of allowing tasks to be performed collaboratively across multiple
universes11—he reckoned that to factor a 250 digit number we would need
to deploy 10500 different universes (the operation would run parallel in each,
and the result would be given by interference) in almost no time—where
Donald Knuth had estimated that it would take over a million years using
a million computers. Time was for Deutsch a budgeting issue:
Evolution would never have got off the ground if the task of rendering
certain properties of the earliest, simplest habitats had not been tractable
(that is, computable in a reasonable time) using readily available molecules
as computers. … What computations, in other words, are practical under
a given time and under a given budget?12
The parallelism of computing today—whether based in the cloud or under

the hood of a PC in the form of multicore processors—trades between two
Life at the Femtosecond 127
temporalities, the speed of messaging (spatial) and the remorseless ticking

of the computer clock (temporal; not highly scalable above current limits).
The fastest computers now can do about 16 petaflops (floating point
operations) per second—that is 16 by 1015 operations. That is a whole lotta
flops, even for an industry that romances the decimal point.13 It means
that in fewer than one hundred seconds, it could do the equivalent of one
calculation per second since the putative Big Bang. Roughly. On the other
hand, in a deeply meaningless calculation, it has been asserted that the
human brain performs between 10 and 30 times as many.14 The reason
why this comparison is even a question is that the holy grail of artificial
intelligence needed to replicate the work of many complex professions, such
as scientist and psychiatrist—known as AI-complete programs—require
complete emulation of the human brain in the interests of efficiency.15 If we
succeed: “then it will become feasible for machines to carry out such jobs,
and to do so more cheaply and more effectively than humans.”16 Indeed: “if
and when human-level AI is achieved, superintelligence will soon follow. …
Even if human-level AI is achieved by the most conservative means—by slav-
ishly copying nature—the resulting liberation from the speed restrictions
inherent in biology is enough.” The romance does not stop there—Murray
Shanahan fantasizes: “A theoretically perfect computer with a mass of 1 kg
and occupying a volume of 1 liter would perform 5.4x1050 logical operations
per second on 1031 bits.” This is 39 orders of magnitude greater than today’s
computers. In the future, we might see a portion of space, as Shanahan
quotes Hans Moravec, “‘rapidly transformed into a cyberspace, [wherein
beings] establish, extend, and defend identities as patterns of information
flow … becoming finally a bubble of Mind expanding at near lightspeed.’”17
We do not generally think of these kinds of speeds—our lives seem to
flow at the rate of less than one thought per second and connectivity for
many is so fast that it’s basically just instantaneous, and so invisible. (The
days of watching an email message unfurl painstakingly over a modem
attached to a telephone line are long over.) And yet they affect us in our daily
lives. They can irrupt—in the form, say, of the flash crash of 2010, where
trading algorithms were making and changing bets at an astonishing rate,
leading to a trillion dollar dive on Wall Street that lasted all of thirty-six
minutes. More insidiously, very fast computing times are needed for the
forms of machine learning being used by casinos, states, and advertising
companies such as Facebook and Google to create the nudge technologies
that cosset you along the path of life mostly without your being aware.18
Thus, Armin Beverungen and Ann-Christine Lange cite Katherine Hayles
on the missing half second between perception (registering an event)
and consciousness (processing it): “‘This cost [the delay] … assumes new
importance when cognitive nonconscious technical devices can operate at
temporal regimes inaccessible to humans and exploit the missing half-second
to their advantage.’”19
Timothy Mitchell asserts that we are pillaging the past of its biological
productions; he estimates that a single liter of petrol needs about 25 metric
tons of ancient marine life, and that we burn 400 years of entire plant and
animal life production a year.20 In our brave new age, we are also burning
time at an ever faster rate—and this is just as ontologically, politically, and
socially significant. Ian Osborne and Daniel Clery observed in 2004 that
the “most important commodity” today is time.21 And the control of events
at the very small scale is becoming central—as Paul Hegarty argues in his
discussion with Gary Genosko of Google’s attempt to “own” the smear of the
leap second needed to cope with the irregularities of the earth (that perfect
clock for the Enlightenment scientists): “Google’s execution of time-critical
processes establishes its mastery over the measurement and manipulation
of humanly imperceptible micro-temporal events.”22 And it all adds up—by
one estimate, the world will have spent a billion years online in 2018.23 As
Wolfgang Ernst says, we need to take machine time seriously.24
The argument here will be twofold: first that we are creating new onto-
logical layers of the socionatural through the development of computing
technology and second that the associated temporalities are richly textured,
that is, not all about speed.
Speed, Acceleration and Delay
High-frequency trading is always a good spot to find computers and networks

working at their fastest. A network switch made by Metamako “allows a trade
order to be placed in the time it takes a photon to travel 90 feet”25—and
there is a market, these switches in 2016 were selling at 100 units a month.
We need shorter times. This can reach fantasy proportions: “When in 2012
scientists briefly thought they’d detected neutrinos that could travel faster
than light, for example, high-frequency traders pondered how they might
build a system that would execute trades that would, theoretically, occur
in the past.”26 In this current imaginary, it really is only a step from high
energy physics and cosmology (two modes of inquiry exploring the nature of
physical reality) and computing:27 a short circuit revealing of the ontological
significance of computer time—“If you understand the universe as a giant
computer constantly calculating its way through time, it’s always easier—less
resource-intensive—for things to flow forward (cause, then effect) than

backward (effect, then cause). This idea is called the “arrow of time.”28 As
Sybille Krämer says, time moves in this imaginary from being “a universal
form of our perception or existence” to “a universal form of technological
accessibility.”29
Even without tachyons to help, the “light cone”30 of events in computing
can seem to violate causality. In Flash Boys, Michael Lewis describes an
interesting move that high-frequency traders in dark pools made. Brad
Katsuyama from the Royal Bank of Canada was wondering why when he put
in a “buy” order, the shares he wanted to buy (which were listed) evaporated
before he could complete the trade—even though he saw the price and
the offer in “real time.” It gets baroque, but it turns out the reason was that
there were several different places to do trades clustered around NY and
New Jersey, and some folks had learned how to manipulate millisecond
differences in trading in such a way that offers and prices were manipulated
before his orders went through. So he wrote an algorithm that deliberately
slowed his messages to some of the trading centers so that all his messages
arrived at each of the centers at the same time. This produced a realignment
between the “ticker tape” prices and offers and his trades. So the ticker tape
(central to the temporality of the market since the nineteenth century)31
did not show anything real until the delays were brought in—a real “live”
market could only be produced in this way.32 The role of delay at very high
speeds in order to create the appearance of real-time is central to Kittler’s
canonical essay on time-axis manipulation—he observes that there is no
“real-time” analysis, delay is needed in order to assemble the parts (“only
that which is switchable is at all”) to give the impression of events passing
in real time.33 This new reality is only buildable because “time exists as
quantified and synchronized packets whose size approaches zero.”34 As
Espen Ytreberg remarks in the context of the production of live events on
television, the present is an outcome of the manipulation of display.35
These time-manipulations, at speeds far surpassing human perception,
are real in their consequences: fortunes are lost and won; real-time events
are created on television sets. And as David Mills asserts, “Reliable and
accurate computer time is necessary for any real-time distributed computer
application, which is what much of our public infrastructure has become.”36
What of the human in this? Well, for one thing, the micro-temporality of
the computer can only be supported through the macro-temporality of
constant care.37 Humans have been in the loop since the earliest days of
computing—Babbage envisaged needing them for logarithmic functions
(when the machine needed a log function it would ring a bell; if the attendant
brought the wrong card containing the function, it would just ring loud-
er).38 Beverungen and Lange discuss a high-frequency trader matching his
circadian rhythm to the cadences of his algorithms—the futures market
he traded in was open 23 hours a day; he only slept four hours a night and
even then got up every 45 minutes to tend his flock of “algos” (they were
pre-programmed to rewrite themselves and interact in complex ways, so
one could never be sure what decisions they were making).39
Where game theoretic economists gave us homo economicus (the person
making “rational” choices defined by self-interest), one dream now is of
machina economicus, where algorithms working at the speed of light could
make rational choices for us40; tellingly Parkes and Wellman anoint this “a
new species of machine.”41 A recent textbook on Blockchain (the technology
behind Bitcoin) promulgated the vision that we would need to delegate more
than our rational qualities to machines, but also our ethics and our policy;
the future would see “a public open distributed ledger with general purpose
rational agents (Machina Economicus) running on blockchain, making
decisions and interacting with other intelligent autonomous agents on behalf
of humans and regulated by code instead of law or paper contracts,” and the
new technology would “impact every industry including but not limited to
finance, government, and media.”42 By this vision, there would be a whole
level of life going on at speeds far exceeding any possible human perception:
One solution might be to introduce intelligent bots or AAs [autonomous

agents] or even contracts that are programmed with regulatory logic
embedded within them. They are most likely programmed by regulators
and law enforcement agencies and live on the blockchain as a means to
provide governance and control. 43
The coinage “live” is interesting here; for they will certainly have to be
autonomous and adaptive—the comforting myth of their being programmed
by humans as unreal as in the case of high-frequency trading: if this were
to come to pass, they would need to be designed by algorithms to evolve
by circumstance. Larry Lessig, who went from being a contract lawyer to
writing about code as law, would surely recognize this world. 44 In general,
it has been claimed, human knowledge and cognition are progressively
more marginal in markets. 45
Where is the real in all this? It is certainly the case that staggering
resources are being put into making computers (and the algorithms they
spawn) faster. Equally certain, new kinds of events are occurring in the world
spawned by new kinds of entities which act using much the same register
as we have ascribed to intentional agents in the past. If we take ontology

as being about what kinds of things the world is peopled with, there are
indeed new kinds of entities, operating at new times.
Just as evolutionary theory looks very different if taken at the temporality
of the virus or the human (even current attempts to find selection within one
human generation—such as lactose tolerance—are necessarily concerned
with multiple years46); so does history look very different if taken at the
temporality of these new entities. 47 The real question is whether it is just
the world inside the computer, or if something very different is going on at
this new ontological level—will we find, as Jacques Revel observed in the
context of microhistory over short time scales, that “the change in the scale
of observation revealed not just familiar objects in miniature but different
configurations of the social.”48
Linear Time
The formal structure of time in a computer is relatively simple: it all comes

down to carrying out one instruction per tick of the clock (I am not covering
here modifications which partially work around this; they do not make an
analytic difference)—the faster the clock ticks, the faster the computer
performs. The trouble is that clocks can only tick so fast:
For many years, we could count on processor clock speeds increasing at

a steady rate. Physical limitations present a fundamental roadblock to
ever-increasing clock speeds, however: because power density increases
superlinearly with clock speed, chips run the risk of melting once their
clock speeds become high enough. In order to perform more computations
per second, therefore, chips are being designed to contain not just one
but several processing “cores.”49
As we have seen, it really is serial despite the language of parallelism. In

a process which might be called descentralization, each set of commands
in parallel processing can be converted into a serial set (one darned thing
after another):
the memory behaves as if the instructions were executed sequentially

according to some global linear order that preserves the individual
orders in which each processor issues its own instructions. For dynamic
multithreaded computations, which are scheduled onto processors
automatically by the concurrency platform, the shared memory behaves

as if the multithreaded computation’s instructions were interleaved to
produce a linear order.50
It is hard work maintaining this remorseless linearity—both the Therac-25

radiation therapy deaths51 and the North American Blackout of 2003 were
caused by so-called race conditions. These are when time gets out of joint
and a new value is posted to a variable when the old value was still needed:
“A determinacy race occurs when two logically parallel instructions access
the same memory location and at least one of the instructions performs a
write.”52 The difficulty arises because most of the time the program will
run well (which makes it very hard to test for), but occasionally the unlikely
but logically possible race will occur and prove disastrous. Again, events at
the micro level have macro effects.
There is an analogy here with quantum mechanics. At the level of
our day-to-day perception (being, as we are, the measure of all things) a
Newtonian physics is good enough to go by; at the level of the very small
and very fast, it seems as if there is a new set of rules—a different kind
of physics. Our generic relationship with computers works pretty well if
we assume they work in the same temporality as ours, just scaled down.
When it gets interesting is when the nested scaling breaks down—as in a
determinacy race—and we need to recognize that time becomes a different
thing both sociotechnically (it becomes pure commodity) and socionaturally
(it becomes granular, discontinuous—taking on just the property that Kittler
hailed as central53). And the temporalities are consonant—in datestamp
format: “The second can be represented to about 500 attoseconds, or about
a tenth of the time light takes to pass through an atom.”54 We need that
shorter time.
Historical Time
Computers have come to define new historical epochs—somewhat helter

skelter but with a determined drive: “With the invention of bitcoin in 2008
the world was introduced to a new concept that is now likely to revolutionize
the whole of society. It’s something that has promised to impact every
industry including but not limited to finance, government, and media.”55
Not bad for an algorithm—the mining of bitcoins is now, alas, no longer
“democratic” but under the control of large organizations that can afford
the suites of graphics cards that can mine most efficiently.
Within the computer, all kinds of historical time are reproduced. For the
Ethereum blockchain algorithm, the oil companies have reappeared, with
two key variables being gasPrice and gasLimit.56 In his book on computer
time synchronization, David Mills revels in the investing of human historical
temporality into the computer. He tells of the need to separate “the ‘truec-
himers, whose clocks gloriously tell the truth from the falsetickers, whose
clocks lie viciously.”57 These terms are not his florid invention—they are
terms of the art in the field of computer time. All the pageantry of human
history can be found at the level of the femtosecond:
Even under peacetime conditions, the truechimers surviving the select

algorithm might have somewhat different time offsets due to asymmetric
delays and network jitter. Various kinds of cluster and combine algorithms
have been found useful to deliver the best time from the unruly bunch.
The one used in NTP sorts the offsets by a quality metric, then repeatedly
discards the outlier with the worst quality until further discards will
not reduce the residual error or until a minimum number of servers
remain. The final clock adjustment is computed as a weighted average
of the survivors.58
This war of clock against clock is needed to prevent stocks from being sold
before they are bought or the evening news coming on at midnight.59 At stake
is nothing less than the principle of cause and effect (it is notable how often
the troubling of cause and effect comes up in discussions of computing).
Thus, in real-time systems, in order to avoid “anomalous behaviors (e.g., by
actions that bypass the system’s normal operation and could violate the
cause-effect relation by “making holes” in the “light cone” of events, the
interactions between the system and the environment must be explicitly
take into account.”60 For Mills, the internet is a growing child, with its
development measured in its need for and colonization of ever shorter time
intervals—100 microseconds when it was “teething” to a few microseconds
in its current “adolescent” state.61 He looks forward to the glorious rule, pace
E. P. Thompson, of time, work-discipline, and postindustrial capitalism: “The
ultimate accuracy can be achieved only when the clock can be disciplined
with exquisitely intimate means. In practice, this requires the discipline
algorithm, normally implemented in the NTP software, to be implemented
in the operating system kernel.” Just as we have been disciplined over the
past few hundred years out of lives lived by diurnal and seasonal rhythms
into lives run on schedule, by calendar, so have our computers undergone
their own temporal discipline over the past thirty years.
We have seen the importance of voting models. It need not be about

industrial discipline under democratic values, though. Professor of Cognitive
Robotics Murray Shanahan reckons that in superintelligent (meaning beyond
human intelligence) AI might become a “willing intellectual slave who never
eats or sleeps and wants nothing more than to work” and that this would be
“many corporations’ idea of the perfect employee, especially if they don’t
require wages.”62 He does wonder, though, whether they could be treated
as slaves with “moral impunity.”63 In turn, self-reproducing superintelligent
machines could “colonize the galaxy.”64 One popular methodology is the
imperialist competitive algorithm, which models “social evolution.”65 It
involves assimilating colonies, the occasional revolution, and of course
intense warfare between empires. So former economic orders don’t go away,
they get buried in machines operating in temporalities humans cannot
directly access. There is an analogy here with the kind of historicity explored
by Kathleen Davis.66 She points to the cocreation of slavery and colonial
power (extended in space) and the myth of the middle ages and feudal
order (extended back in time). The computer spawns multiple historical
times which we thought were over (ontogeny recapitulating phylogeny).67
There is a trope in the history of science that Copernicus left us with
an earth occupying no privileged position in the universe, Darwin left
humans with no privileged position in time, and Freud left reason with no
privileged position in the mind. As early as 1985, Michael Arbib argued that
the computer left humans with no privileged position in society.68 Shanahan
leaves us with no privileged position in history, arguing that, possibly, real
history is about the history of matter gaining complexity.69 His position is
reminiscent of Sybille Krämer’s whether this disappearance of humans from
the story was “the intellectual property of a media-technological version
of eschatology.”70 From this decentered position, it makes a sort of sense
that we project onto matter—with a far greater temporal range than our
own—the course of (non)human history.
Present (Future/Past)
Much writing about computers inhabits the future tense—it’s as pervasive as

the passive voice of policy documents. Babbage concluded of his Analytical
Engine in 1852 that “nothing but teaching the engine to foresee and then to
act on that foresight could lead me to the object I desired.”71
Prediction has been core to the development of computing in the twen-
ty-first century. It often works on a form of neutral hypothesis—it is easiest
if one assumes that the future is like the present: we are constantly “updating
to remain the same,” in Wendy Chun’s lovely locution,72 because the power
of, say, recommender systems is to make us cleave to our present profile
in the future. The work of a “discriminant” in machine learning (such as
credit risk) is based on the assertion that “the future is like the present.”73
We are creating a new “temporeality.”74
The colonization of the future by the present is a theme explored by
Hartog historiographically: “The present became something immense,
invasive, and omnipresent, blocking out any other viewpoint, fabricating
on a daily basis the past and the future it needed.”75 He ties it to a shift
away from the future being seen as holding promise (as in the Victorian
age of certainty 76 or the French revolutionary fervor for a better society)
to the future as being one of threat and risk (environmental degradation,
biodiversity loss, and so forth). A common trope in computing holds similar
fears. Sometime in the next twenty to forty years, the world as we know it
will change radically and frequently not for the better—it depends on how
you understand the coming “singularity.” Is it a nightmare or an optimistic
vision to pose the following: “Between 2020 and 2050: Artificially Intelligent
DAOs [distributed autonomous organizations] will prevail on blockchains
that will make rational decisions on behalf of humans”?77 It is certainly a
world in which “humans have become increasingly irrelevant.”78
Jürgen Schmidhuber, who sees full artificial intelligence as a simple
function of whether or not we can compress time into small enough units in
the computer, takes a trope that was created by Ray Kurzweil and recently
developed by Geoffrey West.79 Counting down from the omega point, he
argues that we have witnessed (for West, we require) ever faster rates of
breakthrough in modes of life—starting from humans leaving Africa at
29 lifetimes, we advance, step by logarithmic step, down to the Greeks
inventing democracy and science at 25 lifetimes, the Age of Enlightenment at
22 lifetimes, PCs at half a lifetime, and the upcoming revolution projected to
occur in a quarter of a lifetime.80 (Schmidhuber is somewhat tongue-in-cheek
here—he does wonder at the end of the paper whether this compelling
scale is more a feature of the way we remember—giving more weight to
recency—than of the nature of the record.) In this vision we need deal only
with ever-shorter futures until the future disappears through machine
learning into the unchanging, untheorized singularity.
Similarly, there the past is colonized by the present to the point of
disappearance. We can see this through the language of the Internet of
Things. Pioneering cybernetician W. Ross Ashby asserted in the 1950s that
a well-configured cybernetic system did not need to hold its own memory.
The world, and the machines used to analyze it, could be constructed in
terms of closed Markov chains—meaning that all one ever needed to know
to understand the state of either world or machine was to know its present
state and the set of rules for its change. Remembering what a former state had
been was simply irrelevant. This has been a major temporal metaphor since
the Enlightenment—for Laplace, the universe could be uniquely described
by a set of first-order differential equations, meaning that knowing the state
of the world right now meant that you could predict its future state to any
degree of required accuracy.
As Ashby developed his work especially in Design for a Brain,81 he made
much of a theory of interchangeability, whereby contingent history just
did not matter—the past was irrelevant. We might “lose” Hampshire to a
devastating accident, but the Britain would auto-regulate and keep func-
tioning in the future pretty much as in the past—Britain was a system
with interchangeable parts. Similarly, the death of an individual human
just did not mean much in terms of the temporality of the system. Humans
are interchangeable with each other and with machines—“really” we are
just talking about a set of operations carried out in metronomic fashion.
When Neil Gross, a pragmatist sociologist, wrote a piece in 1999 entitled
“The Earth Will Don an Electronic Skin” presaging the Internet of Things,
he wrote: “Hundreds of thousands of PCs working in concert have already
tackled complex computing problems. In the future, some scientists ex-
pect spontaneous computer networks to emerge, forming a ‘huge digital
creature.’”82 His image was of a Leviathan, matching Hobbes’ (the famous
frontispiece to whose work pictures a colossus made of humans forming
the body of the king/state). The PCs are working in concert—they are
synchronized together. He goes on:
It will use the Internet as a scaffold to support and transmit its sensations.
This skin is already being stitched together. It consists of millions of
embedded electronic measuring devices: thermostats, pressure gauges,
pollution detectors, cameras, microphones, glucose sensors, EKGs, elec-
troencephalographs. These will probe and monitor cities and endangered
species, the atmosphere, our ships, highways and fleets of trucks, our
conversations, our bodies—even our dreams.
This image of the colossus well exceeds human temporal scales—it exists
in the eternal present of the responsive machine. It is a fundamentally new
fact about human existence that our human temporality is now that of the
sociotechnical world we have created.
There is a whole new ontological level which works at ever small time
scales and yet which can have large-scale effects. Humans drop out of the
equations in much the same way as objects drop out of the equations of
quantum mechanics. They are not quite the self-contained things we used
to imagine: for both, we need new sets of rules to study in fine the operation
of the world. We have both constructed physically and constituted socially
new temporalities and new understandings of objects that just do not work
at the rate of human perception.
Conclusion
Time is at the center of our political economy—displacing space, which

was central to colonial power: the demotion of the meter to the distance
per second traveled by light (1/299,792,458) is but a symptom of this new
form. Network time protocol is enforced in “space, on the seabed … and on
every continent, including Antarctica.”83
The control of space was the control of other countries in far-flung em-
pires. But it was already about the control of time. When the oil companies
came into Venezuela, they were able to interact beneath the state—in
the subsurface—manipulating temporalities that had never explicitly
been part of Venezuelan economic life.84 This knowledge enabled them
to help supercharge the metabolism of the human species by tapping into
low-entropy stored solar energy and converting it into commodities and
waste at a higher entropy.85 Computer time is central here: the Schlumberger
company was at the fore of the development of expert systems to deal with
microtemporalities (reading seismic waves), which revealed the macro
structure of the earth. As Paul Prudence says, “Algorithms can become a
Rosetta Stone for the lithic scryer, for if the writing of stones are hieroglyphics
for Caillois then the language of algorithms is a demotic script that can
bridge the gap between geological noise and intelligible information.”86
Learning to work temporalities outside of our immediate sensory expe-
rience has been central to the working of capitalism. Alfred Sohn-Rethel’s
assertion that Galilean absolute space and time was basically about the free
flow of the commodity form in an ideal, frictionless market gains traction
in this context. The faster the money-commodity-money cycle, the more
surplus value can be created—in other words, the more we can work at
higher temporalities, the better for the system.
The remorseless, rapid ticking of the computer clock enables control at
a wide range of temporal scales. The rapid ticking of the truechimers is
a source simultaneously of the speed and acceleration we experience in

the present and the extension of the present into the immediate past and
future, creating Hartog’s “monstrous present.”87 There is no historical time
as modelled by historiographers on the one side and computer time on the
other. At all temporal scales they interpenetrate. This happens at the scale
of the longue durée: the computer grew up with and is modeled on factory
production as industrialization took command of world economies (from the
pillaging of natural resources by the colonizing powers to the organization
of manufacturing hubs in the developed world). The event of the short durée
is increasingly mediated by machine learning. And at the very shortest
scales—the femtosecond—a variety of historical times play out.
We saw above Revel’s assertion that scaling down temporally produced
new configurations of the social and the political.88 In computer time, even
while all of these temporal scales interpenetrate, different configurations are
being developed. But they are made of the same stuff as standard historical
time. Inside computers, we find colonialism, slavery, resource control, and
the manufactured event beloved of media theory—as well as geological
and evolutionary time. The technological here is not different from other
spheres of existence; it is fully coeval with them.
This can only really make sense if we see that the sociotechnical and
the socionatural work together. There is no nature “out there” for the
computer—we understand nature through its workings at the same time
as we model the computer on that understanding.
Life at the femtosecond is consequential for our ways of being in, appre-
hending, and acting in the world. Real-time is ineluctably real time, and
for us real time is now real-time.
Notes
1. François Hartog, Régimes d’historicité: Présentisme et experiences (Paris:

Seuil, 2003), 120. “Through speed, the present transforms itself into eternity”
(my translation).
2. Craig Callender What Makes Time Special (Oxford and New York: Oxford
3. I’m Sorry I Haven’t a Clue, Season 70, Episode 2.
4. Charles Babbage, “On a Method of Expressing by Signs the Action of
Machinery,” Philosophical Transactions of the Royal Society of London 116,
no. 1–3 (1826), 250–265.
5. Charles Babbage, Passages from the Life of a Philosopher, ed. Martin Camp-
bell-Kelly (New Brunswick, NJ: Rutgers University Press, 1994 [1864]), 94.
6. Babbage, Passages, 90.

7. See Charles Lyell, The Principles of Geology, vol. 1 (Chicago: University of
Chicago Press, 1990 [1830]).
8. Charles Babbage, The Ninth Bridgewater Treatise: A Fragment, 2nd ed. (Lon-
don: John Murray, 1838).
9. This temporal collapse is perhaps pointed to by Kittler’s use of “a scream-
ing comes across the sky” (16)—the opening and closing lines of Gravity’s
Rainbow, encapsulating the extensive time of the novel into the moment of
the dropping of a rocket.
10. Thomas H. Cormen, Charles E. Leiserson, Ronald L. Rivest, and Clifford Stein,
Introduction to Algorithms, 3rd ed. (Cambridge, MA: MIT Press, 2009), 774.
11. David Deutsch, The Fabric of Reality (London: Penguin, 1997), 155.
12. Deutsch, Fabric of Reality, 200, 196–7.
13. I borrow this felicitous phrase from Robert F. Service, “Time’s Romance of
the Decimal Point,” Science 306, no. 5700 (November 19, 2004): 1310–1311.
14. AI Impacts, https://aiimpacts.org/brain-performance-in-flops/, accessed
November 25, 2019. Kurzweil gives 1016.
15. “Messy emulation” is just doing neuron-for-neuron mapping; other forms
would develop new kinds of intelligence. (It is surprising that such a limited
view of the workings of the brain—neuron firing—persists in computer
science, as well as the willful ignorance of the distributed, embodied nature
of cognition.)
16. Murray Shanahan, The Technological Singularity (Cambridge, MA: MIT
Press, 2015), 152–3.
17. Shanahan, Technological Singularity, 157.
18. Dow Schulman, xxx.
19. Armin Beverungen and Ann-Christine Lange, “Cognition in High-Frequency
Trading: The Costs of Consciousness and the Limits of Automation,” Theory,
Culture & Society 35, no. 6 (2018): 80.
20. Timothy Mitchell, Carbon Democracy: Political Power in the Age of Oil (Lon-
don: Verso, 2011), 15.
21. Ian Osborne and Daniel Clery, “Higher Standards,” Science 306, no. 5700
(November 19, 2004): 1307.
22. Gary Genosko and Paul Hegarty, “Where Has Become of Time? Temporal
Smearing and Media Theory,” SemiotiX, https://semioticon.com/semi-
otix/2018/03/where-has-become-of-time-temporal-smearing-and-media-
theory/.
23. Simon Kemp, “Digital in 2018: World’s Internet Users Pass the 4 Bil-
lion Mark,” We Are Social, January 20, 2018, https://wearesocial.com/
blog/2018/01/global-digital-report-2018. This is spread over 4 billion users,
and interesting more for the order of magnitude rather than as a realistic
assessment.
24. Quoted in Jussi Parikka, “Introduction,” in Wolfgang Ernst, Digital Memory
and the Archive, ed. Jussi Parikka (Minneapolis: University of Minnesota
Press, 2014), 15.
25. Michael Reilly, “High-Frequency Trading is Nearing the Ultimate Speed Lim-
it,” MIT Technology Review, August 9, 2016, https://www.technologyreview.
com/s/602135/high-frequency-trading-is-nearing-the-ultimate-speed-limit/.
26. Reilly, “High-Frequency Trading.”
27. See, for example: Rafi Letzter, “How Quantum Computers Could Kill
the Arrow of Time,” Live Science, July 27, 2018, https://www.livescience.
com/63182-quantum-computer-reverse-arrow-time.html.
28. Letzter, “Quantum Computers.”
29. Sybille Krämer, “The Cultural Techniques of Time Axis Manipulation: on
Friedrich Kittler’s Conception of Media,” Theory, Culture & Society, 23,
no. 7–8 (2006): 93–109
30. Fabio A. Schreiber, “Is Time a Real Time? An Overview of Time Ontology in
Informatics,” in Wolfgang A. Halang and Alexander D. Stoyeno (eds.), Real
Time Computing, NATO ASI Series, Series F: Computer and Information
Systems 127 (1994): 294.
31. Alex Preda, “Socio-technical Agency in Financial Markets: The Case of the
Stock Ticker,” Social Studies of Science 36, no. 5 (2006): 753–782. Speed has been
central to the market since at least the late eighteenth century—Chartier
(xxx) points to five issues of a paper being put out daily to describe market
changes. We often forget this time compression from pre-digital days—in
London in the early nineteenth century there would be four or five postal
deliveries a day—making the service more like email than the more recent in-
stantiations we tend to project onto the past (under the retronym “snail mail”).
32. Michael Lewis, Flashboys: Cracking the Money Code (London: Allen Lane,
Penguin. 2014).
33. Friedrich Kittler, “Real Time Analysis, Time Axis Manipulation,” Cultural
Politics 13, no. 1 (2017), 1–17.
34. Kittler, “Real Time Analysis,” 10.
35. Espen Ytreberg.
36. David L. Mills, Computer Network Time Synchronization: The Network Time
Protocol on Earth and in Space, 2nd ed. (Boca Raton, FL: CRC Press, 2016),
xxiii.
37. See Gabriele Schabacher in this volume.
39. Beverungen and Lange, “Cognition in High-Frequency Trading,” 88–89.
40. Martin Prause, “On the Trail of Machina Economicus,” Info Insights, 2017.
41. David C. Parkes and Michael P. Wellman, “Economic Reasoning and Artifi-
cial Intelligence,” Science 349, no. 6245 (July 17, 2015): 267–272. doi:10.1126/
science.aaa8403.
42. Imran Bashir, Mastering Blockchain: Distributed Ledgers Decentralization
and Smart Contracts Explained (Birmingham-Mumbai: Packt, 2018), 1.
43. Bashir, Mastering Blockchain, 475.
44. Lawrence Lessig, “Law regulating code regulating law,” Loyola University
Chicago Law Journal 35 (2003): 1–14.
45. Beverungen and Lange, “Cognition in High-Frequency Trading,” 80.
46. Elizabeth Pennisi, “Humans Are Still Evolving—And We Can Watch It

Happen,” Science, May 17, 2016, https://www.sciencemag.org/news/2016/05/
humans-are-still-evolving-and-we-can-watch-it-happen.
47. Thanks to Bodhisattva Chattopadhyay for this observation. See Bodhisattva
Chattopadhyay and Geoffrey C. Bowker, “Ant Network Theory,” NatureCul-
ture 5 (2019): 26–49.
48. Jacques Revel, introduction to Histories: French Constructions of the Past, ed.
Jacques Revel and Lynn Hunt, trans. Arthur Goldhammer (New York: New
Press, 1995), 46.
49. Cormen et al., Introduction to Algorithms, 10.
51. Nancy G. Leveson and Clark S. Turner, “An Investigation of the Therac-25
Accidents,” Computer 26, no. 7 (1993): 18–41.
53. Kittler, Discourse Networks 1800 / 1900, trans. Michael Metteer (Stanford, CA:
Stanford University Press, 1992).
54. Mills, Computer Network Time Synchronization, 18.
60. Fabio A. Schreiber, “Is Time a Real Time? An Overview of Time Ontology in
Informatics” in Wolfgang A. Halang and Alexander D. Stoyeno (eds.), Real
Time Computing, NATO ASI Series, Series F: Computer and Information
Systems 127 (1994): 283–307.
294.
61. Schreiber, “Is Time a Real Time?,” 9.
62. Murray Shanahan, The Technological Singularity (Cambridge, MA: MIT
Press, 2015), 93. See E. P. Thompson, “Time, Work-Discipline, and Industrial
Capitalism,” Past & Present 38 (1967): 56–97.
63. Shanahan, The Technological Singularity, 153.
64. Shanahan, The Technological Singularity, 157.
65. https://en.wikipedia.org/wiki/Imperialist_competitive_algorithm. Accessed
December 21, 2018.
66. Kathleen Davis, Periodization and Sovereignty: How Ideas of Feudalism and
Secularization Govern the Politics of Time (Philadelphia: University of Penn-
sylvania Press, 2008), 42–50.
67. Ernst Haeckel, who was so influential in establishing the tree metaphor for
biological descent also gave us the resounding tocsin “ontogeny recapitu-
lates phylogeny.” The two are closely linked when taken in terms of infor-
mation organization and access in computer science terms; and in terms of
historiographical analysis in the nineteenth century in humanist terms.
68. Michael A. Arbib, “On Being Human in the Computer Age,” in Impacts of
Artificial Intelligence, ed. Robert Trappl (Amsterdam: Elsevier, 1985).
69. Shanahan, The Technological Singularity, xxii–xxiii.

70. Krämer, “The Cultural Techniques of Time Axis Manipulation,” 107.
72. Wendy Chun, Updating to Remain the Same (Cambridge, MA: MIT Press,
2016), 79–85.
73. Ethem Alpaydin, Introduction to Machine Learning, 3rd ed. (Cambridge, MA:
MIT Press, 2014).
74. Jussi Parikka, “The Underpinning Time: From Digital Memory to Network
Microtemporality,” in Digital Memory Studies: Media Pasts in Transition, ed.
Andrew Hoskins (New York: Routledge, 2018), 156–172.
75. François Hartog, Regimes of Historicity: Presentism and Experiences of Time
(New York: Columbia University Press, 2015), 185.
76. Asa Briggs, A Social History of England (New York: Viking, 1984).
78. Kittler, “Real Time Analysis,” 13.
79. Ray Kurzweil, The Singularity Is Near (New York: Penguin, 2005). Geoffrey
West, Scale: The Universal Laws of Growth, Innovation, Sustainability, and
the Pace of life, in Organisms, Cities, Economies, and Companies (New York:
Penguin, 2017).
80. Jürgen Schmidhuber, “New Millennnium AI and the Convergence of Histo-
ry,” ArXiv, June, 29, 2006.
81. W. Ross Ashby, Design for a Brain (London: Chapman & Hall, 1954).
82. Neil Gross, “The Earth Will Don an Electronic Skin,” Bloomberg, August 30,
1999, https://www.bloomberg.com/news/articles/1999-08-29/14-the-earth-
will-don-an-electronic-skin.
84. Geoffrey C. Bowker, Science on the Run: Information Management and In-
dustrial Geophysics at Schlumberger, 1920–1940 (Cambridge, MA: MIT Press,
1994).
85. Alf Hornborg, “Machine Fetishism, Value, and the Image of Unlimited
Good: Towards a Thermodynamics of Imperialism” in Man, New Series,
Vol. 27, No. 1 (Mar., 1992), pp. 1–18.
86. Paul Prudence, “The Algorithmic Writing of Stones: A Cybernetics of Geolo-
gy,” SubStance 47, no. 2 (2018): 71–83: 82.
87. I borrow this phrase from Helge Jordheim, personal communication.
88. Revel, Histories, 206.
About the Author
Geoffrey C. Bowker is Donald Bren Chair at the School of Information and

Computer Sciences, University of California at Irvine. Together with Leigh
Star he wrote Sorting Things Out: Classification and its Consequences. His
most recent book is Memory Practices in the Sciences.
7. Artificial Intelligence and the
Temporality of Machine Images
Andrew R. Johnston
Abstract
DeepMind, a recent artificial intelligence technology created at Google,
references in its name the relationship in AI between models of cognition
used in this technology‘s development and its new deep learning algo-
rithms. This chapter shows how AI researchers have been attempting
to reproduce applied learning strategies in humans but have difficulty
accessing and visualizing the computational actions of their algorithms.
Google created an interface for engaging with computational temporal-
ities through the production of visual animations based on DeepMind
machine-learning test runs of Atari 2600 video games. These machine
play animations bear the traces of not only DeepMind‘s operations, but
also of contemporary shifts in how computational time is accessed and
understood.
Keywords: artificial intelligence, animation, digital media, video games,

technology, history
DeepMind, a recent artificial intelligence technology created at Google,

references in its name the relationship in AI between models of cognition
used in this technology’s development and the deep learning algorithms
that make up its layers of nonlinear processing units. With DeepMind
Google has been attempting to reproduce patterns of visually based re-
inforcement learning strategies observed in human subjects through the
use of big data and probability scenarios in neural networks. Like other AI
that were created and used in speech recognition or machine translation
applications, researchers at Google encountered two problems: 1) how
to model reactive sensory systems and 2) how to visualize and access
doi 10.5117/9789463727426_Ch07
144 Andrew R. Johnston
their mechanics when the systems operate below thresholds of human

sensory perception and cognition. The second question is one endemic
to contemporary computational processes and sets of big data generally,
whose weighty flows of indeterminate scales of information are born out
through representational tropes or descriptions that refer to their inacces-
sibility. Researchers managed both problems by designing an AI system
that could train to win emulated Atari 2600 video games such as Qbert,
Space Invaders, Breakout, and 46 others, each played 600 times in order to
visualize the mechanics of the AI. The gameplay videos, or machine runs,
of these training sessions offer representational data points for researchers
who map how reinforcement learning strategies are employed over the
course of various gameplay scenarios, revealing in visual space how the AI
adapts to different contexts and evaluation criteria through operations in
time that are otherwise inaccessible.
This out of reach character of digital technologies is commonplace.
Materialist analyses of digital media, like the ones included in this volume,
have brought attention to the disjuncture of temporalities present in contem-
porary technological systems and regimes of experience. Running at speeds
faster than that which sensation can detect and through machinations
outside the purview of consciousness, digital technology seems to break
traditions of technological synchronization that match human experience
to the temporal actions of machines. Descriptions of this historically
antecedent organization of time and technics can be seen in the work of
Mary Ann Doane, who, following E.P. Thompson, argues that a nineteenth
century reorganization of time was engendered by industrial revolution
machines and clocks that enabled it to be measured and divided. This
also externalized temporality as a “surface phenomenon” that individuals
attempted to capture through mediated representations.1 That chase was
precisely what Paul Virilio worried over, as he envisioned how the mutual
accelerations of technology and understandings of time could erode the
limits of human history, knowledge claims, and order.2
The link between human sensation and machinic temporality is what
scholars such as Mark Hansen and Wolfgang Ernst argue has been severed
by contemporary digital technologies. Ernst emphasizes how digital tech-
nologies materially instantiate a time outside of consciousness through the
calculation and flow of signals so that even digital “storage is nothing but a
limit value of transfer … storage is a transfer across a temporal distance.”3
This articulation of time and space is microscopic in scale and employed by
digital media for the abstractions of signal transmissions, which allow for
later formal actions that are made available to the senses by these objects.
Artificial Intelligence and the Tempor alit y of Machine Images 145
Hansen makes a similar point, arguing that twenty-first-century media,

or digital objects and network infrastructures, “operate predominantly,
if not almost entirely, outside the scope of human modes of awareness
(consciousness, attention, sense perception, etc.).”4 The resulting human
experiences with these media and their environments are therefore only
indirect relative to their machinations; since one cannot witness their signal
pulses, to experience them is to be at a remove from their technical work.
Hansen argues that because of this arrangement contemporary digital media
has a predictive relation to our bodies in a feed forward architecture “that
literally mediate the data of causal efficacy (as measured, calculated, and
analyzed by twenty-first-century media) for future consciousness to factor
into its activity-to-come.”5 Since digital technologies are built upon indirect
engagements with their inner workings, their signal transmissions are
structured through predictions of future action in a race with consciousness
and experience to produce illusions of interactivity.
All these epistemological formulations are rooted in contemporary digital
media. And for good reason, since digital technology used for computer
graphics or other applications in the 1960s and 1970s was notoriously slow
relative to analog computers and often relied on paper storage.6 But the above
authors also emphasize how the operations of contemporary digital media
are at such a remove that they appear opaque, a black boxing of technology
that Ernst, as well as Bruno Latour, warn against in understanding material
renderings of experience brought about by shifting technical regimes.7 The
seeming opacity of digital media is certainly a cultural phenomenon that
generates anxiety, often manifest in new genres or mediated by horror
narratives, arguments persuasively made by scholars such as James J. Hodge,
Shane Denson, and Adam Hart.8 That said, this perspective is not totalizing
and, as the strange periodizing of Hansen and Ernst reveal, is a shifting
process born of multiplicity and contradiction rather than singularity
and uniformity. As Ernst emphasizes, “a media-archaeological view of
the temporal modes of media almost inevitably leads to a critique of the
totalizing collective singularity of ‘time’ itself.”9
To perform this media archeological analysis is to open the black boxes
of contemporary digital technologies that can appear so opaque. Doing
so performs an epistemological reverse engineering that makes manifest
the impact of technologies in the production of knowledge through
their mechanisms of sensual and information presence. 10 And it also
clarifies the contours of that epistemology and mediations of temporal
experience. While the phenomenal actions of computational media may
index obscurity, they also point towards and create new arrangements
and networks of activity. Such changes may induce anxiety in cultural

narratives, a reaction found during other periods of shifting temporal-
ities born of technological change, exemplif ied in Alvin Toffler’s 1970
book Future Shock.11 But as past studies of these moments have shown,
such transitions are often fractured with multiple layers of technicity
interwoven with one another to both manage these changes and develop
use strategies.12 In this light the machine run videos of DeepMind’s Atari
play have particular significance, since they are not created as a formal
articulation that gestures towards an unknowable obscurity, but instead
attempts to quantify and abstract the temporal and computational actions
of the AI for later modification. In Google’s published work, the abstract
neural network of DeepMind is displayed with corresponding screenshots
from gameplay to provide illustrations of what they argue is machine
learning. These gameplay animations index machine action and are a
means through which Google can refine and develop computational deep
learning. Doing so, I argue, positions animation as an epistemic object and
tool to make nonhuman cognition and time sensually present. Functioning
as an investigatory device, it maintains the aesthetic wonder that animation
is often associated with, but simultaneously reveals different technical
operations buried in the AI system that produce these articulations.13
These unseen microtemporal actions are animations unto themselves,
since the system analyzes, abstracts, and then synthesizes images into
taxonomic frameworks or movements acted upon in a different layer within
its feed forward architecture. Though operating at different scales of time,
the qualities of technical abstraction and synthesis in other animations
are still present, though addressed for a system whose visual syntheses
have been modeled from human perception and that we cannot see. But
these actions can also become the means through which phenomenal
images are generated that then makes sensually present information for
human users, a secondary animation layered onto the others. I argue that
together this stack of animation produces the interface for working with
and manipulating networked infrastructures of AI, an interweaving of
what Ernst, quoting Gilbert Simondon, calls different levels of temporal
knowledge: “a micro-level of physical and techno-physical processes,
[and] a meso-level of psychic-cognitive processes.”14 Layering them onto
one another both enables speculative knowledge about the AI system’s
temporal modulations and about our relation to these new actions of time.
The remainder of this essay will work through and open this AI interface,
moving up and down its stacks to reveal animation’s role in the diverging
temporalities that mark these technologies and this moment.
AI emerged as a field in the 1950s out of cybernetics and one of its ear-
lier and most long-lasting research areas was visual pattern classification
and the development of machines that would have the ability to detect
predetermined objects within given fields. Though operating in fits and
starts over several decades, this area of AI has recently become a center
of inquiry, in part due to the accelerated growth of graphics processing
units used in feed forward neural networks created after what many call
an AI winter in research during the 1970s.15 Unlike other AI information
oriented platforms, DeepMind has action-oriented goals achieved through
rewards defined within specific environments. It is designed to create
software agents that “take actions in an environment so as to maximize
some notion of cumulative reward.”16 Taken as a whole, this is a feedback
system that explores and modifies the agent’s environment and like earlier
AI systems uses behaviorist psychology frameworks. But its modeling of
the ways individuals solve problems through feedback mechanisms is put
in conjunction with a hierarchical sorting and categorization of sensory
data. Previous AI systems were built around applications set in motion
by particular stimuli. IBM’s Watson, for instance, is a natural language
and information retrieval system, whose functions would be triggered by
spoken questions.17 Because of this, the system does not explore or test its
environment through sensory data, but instead operates through an if-then
Boolean logic so that if particular inputs exist, then a functioning action
within the platform will result. Using a behavior tree for executed actions,
the platform is dependent upon language, accent, and the information
database accessible to it.
The exploration of an environment through visual data was still the
most prized area of AI research and much of its success has emerged from
investigations into vision performed by Nobel prize winning neurobiologists
David Hubel and Torsten Wiesel. Beginning in 1958 Hubel and Wiesel traced
the mechanisms of action within the visual cortex and showed how the
brain processes visual information from the retina through a columnar
architecture filled with simple and complex cells, each of which is sensitive
to particular visual features for specific parts of the visual field.18 In their
experiments they determined how the visual cortex of a cat, and later other
animals, uses simple nerve cells to locate edges or borders of contrast along
with other formal features and that the cortex employs complex nerve cells
to combine input from simple cells to gauge motion and other actions. Finally
developing a topographical map of the visual cortex, Hubel and Wiesel
showed how different neurons from the retina operate as inputs to produce
binocular perception and the visual pathway of information in the brain.19
Importantly, Hubel and Wiesel’s experiments also illustrated the

feed forward mechanisms through which visual perception functions,
which rather than having a singular feedback loop for input and outputs,
nests multiple layers of inputs and outputs that don’t cycle or loop back
to each other but move in a linear fashion. Fundamentally, the nodes
in this system constantly evaluate or categorize the input they receive,
each making slight transformations before outputting to the next node
in a unidirectional manner through time. This design became the model
for large computational feed forward networks, which were successfully
employed to analyze and categorize images by computer scientists like
Yann LeCun in the late 1980s. He developed convolutional neural networks
that had many f ilter layers for the categorization and identif ication of
image features, enabling the system to work through the logic and pattern
differentiation problems that plagued AI image research in the past. This
requires using Hubel and Wiesel’s analysis of edge detection for images but
in a large enough network that the subsampling of images continues until
the system can accurately identify objects. Breaking the images up many
times over, the network races to abstract the image, compare those pieces
to others it has analyzed before, and then categorize and take action on
that decision. Citing the neurobiologists as a precedent, LeCun describes
how this convolutional neural network can be used for handwriting and
optical character recognition machine learning, each of which has been
employed by the US postal service and several banking institutions.20
Google’s DeepMind team references and builds from LeCun’s model for
“exploiting the local spatial correlations present in images” in order to
take into account viewpoints that the AI employs while exploring visual
space.21 Once again, for DeepMind, the AI does not just seek particular
data from sensory inputs, but attempts to organize all the information
presented to its inputs within a specif ied domain of action. It explores
but categorizes according to specific criteria. Like other contemporary
AI systems, DeepMind is a sophisticated bot technology and is an object
categorization machine using a feed forward network agent. All the visual
information presented to the AI is sorted as quickly as it can through its
matrix of subsampling nodes in order to then provide actionable informa-
tion within the machine’s environment.
For those designing AI, the speed of this sorting is the problem. Exper-
iments with machines that could perform these actions took place almost
immediately after Hubel and Wiesel’s experiments. In the late 1950s Frank
Rosenblatt created a pattern classification algorithm called a perceptron
that attempted to categorize visual images. Put simply, the perceptron would
evaluate image information presented to it and determine, through if-then

logical statements, whether or not the image had features of a preassigned
category. By utilizing multiple inputted weights, or parts of an image,
the algorithm could examine the image several times to compare these
weighted values to one another in the overall assessment of the image as
belonging or not belonging to the pre-assigned category. After a number
of tries, certain inputs are deemed more important than others in the
image for determining whether or not it belongs to a category, such as “cat
pictures.”22 At the time, the adaptive quality of the algorithm’s operations
produced massive speculation about the future of AI in the popular press,
especially after Rosenblatt created an analog computer for the perceptron
that employed 400 photocells for inputs whose resistance parameters were
controlled by motors.23 Thinking he had successfully determined whether
a tank was in a photograph or not, an application sought after by the U.S.
military, Rosenblatt postulated that his machine had “original ideas” and
that “as a concept, it would seem that the perceptron has established,
beyond doubt, the feasibility and principle of nonhuman systems which
may embody human cognitive functions.”24 This fantasy, however, was
short-lived since Rosenblatt was actually guessing at what his machine was
classifying, learning later that the perceptron was not identifying a tank
in photographs, but rather the time of day.25 For Rosenblatt, not having an
interface through which he could make phenomenal the microtemporal
actions of the machine resulted in not knowing what the machine could
or would do, a blindness to both its rhythms of time and its output of
information.
Animation solves this problem by both completing a visualized feedback
loop for researchers and projecting fantasies of autonomy onto its actions
in an environment. But the question of reinforcement learning and its
relation to mediations of animation loomed large for Google’s research
team, since DeepMind both responds to and changes the environmental
conditions through which it receives sensory information. In short, Google
wanted to display an interface of time through animation so it could show
consequences of computational temporalities and their effect on a phe-
nomenal world. In attempting to develop this behaviorist project, Google
isolated the visual input available for the system and its actionable output.
Using emulations of the Atari 2600 games, they coded DeepMind’s sensory
inputs to be an 84 x 84 pixel gameplay grid and game scores, the latter of
which would function as evaluation criteria. Structuring the inputs this
way not only isolates ideas of reinforcement learning and processing to
visual systems, but also limits specific visual operations, namely motion
detection. Google wanted to develop an AI algorithm in DeepMind that

could be used on large neural networks and therefore isolated various
parameters of the system at its genesis. Emphasizing a particular action in
reinforcement learning, DeepMind excels when encountering puzzles or
obstacles that require a measured degree of planning. As Google explains,
in Breakout “the agent learns the optimal strategy, which is to f irst dig
a tunnel around the side of the wall allowing the ball to be sent around
the back to destroy a large number of blocks” but also admits that “games
demanding more temporally extended planning strategies still constitute
a major challenge.”26 Throughout their description of DeepMind it becomes
clearer as to why Google selected Atari 2600 games. Not only is the game
environment limited spatially and temporally, but the Atari’s sprites are
also abstract objects with clear f igure/ground relations whose spatial
correlations require less processing than other platforms. Such reductions
focalize the action of the AI while also conflating the rules or logics of each
game with the mechanics of the system, an operation that Stephanie Boluk
and Patrick LeMieux identify as prevalent within video games.27 To be
clear, it is the mechanics of the specific game that DeepMind categorizes,
since the system cannot abstract the principles of gameplay from one game
to another, but instead must begin again and learn the design logics and
operations of each game from scratch. The researchers functionally bar the
AI from accessing the code and other algorithmic functions that produce
the Atari games, a more efficient means through which the neural network
could learn the inputs and outputs associated with this gameplay. Instead,
they develop an artificial barrier that attempts to mimic a naive player,
a sort of DeepMind noob, whose ultimate precociousness is revealed in a
dazzling technological display. It is a brilliant marketing move, but also
shows how the AI is learning the mechanics available to objects in the game
more than learning how to “play.” Google’s conflation, like the one Boluk and
LeMieux identifies, of video game logic with object mechanics, nonetheless
produces a projection of mastery onto the machine as it sorts through
various values, actions, and temporal relations over the 600 iterations
of play for each Atari game. Since the AI operates through conditions of
gameplay dependent upon display projections, the resulting actions run
as animations and visually mark the system’s performance through the
game’s challenges.
As such, the animations that make present the operations of the AI exist
within a well-known genre that performs a similar function: tool-assisted
speed runs. These are programs that exploit the serial interface of emulated
video games to reveal the most efficient combination of inputs into the
game to finish or “win” it. Through an almost linear editing of possible

commands within the game’s interface, a recorded animation of action
emerges and makes phenomenal the operations of the software. Creating
an AI interface like this makes familiar both unseen temporalities of the
system as well as the aesthetic and informational form of its output for
people familiar with the Atari games. The animations make relatable the
AI’s operations while acutely marking its differences from human users.
Though competitive speed runs performed by human players have been
popular within gaming communities for some time, the contrast of these
with tool-assisted speed runs, as LeMieux explains, more generally “reveals
the alienating effects of digital seriality and dramatizes the distinction
between human and machine scales of temporality.”28 With DeepMind,
the alienation builds over time since the spectacular failure of the AI is put
on display over the first hundred iterations of gameplay. By limiting the
AI’s inputs to the screen and score, Google blinds DeepMind to the serial
interface of the Atari games, but simultaneously creates visual access to
DeepMind’s processing operations for researchers. In this way, the machine
run animations function as a heuristic tool within the laboratory, standing
in not only for ideal human cognition in reinforcement learning scenarios,
but modulations of time as well. Google’s comparison of gameplay between
the AI and professional game testers emphasize this point, showing how
in games like Road Runner, Star Gunner, or Assault, DeepMind operates
at or above what it characterizes as “human-level” gameplay, boasting
that it “performs at a level that is broadly comparable with or superior to
a professional human games tester … in the majority of games.”29 Thus,
through DeepMind’s limited ocular mimicry to analyze its reinforcement
learning architecture, the resulting animations bridge forms of human and
nonhuman subjectivity by manifesting phenomenal, aesthetic action that
both subjects could produce.
Again, without losing its aesthetic contours, animation in this context
becomes an instrument of speculation, or what historians of science describe
as an epistemic object: an experimental form characterized by indetermi-
nacy within a historical field, but whose instantiations mark patterns of
emerging knowledge and categories. Karin Knorr Cetina calls them “objects
of knowledge [that] appear to have the capacity to unfold indefinitely … like
open drawers filled with folders extending indefinitely” since their technical
composition is in a constant state of change.30 Epistemic objects’ changes
speak to historic patterns that are dependent upon knowledge formations.
Like Lisa Gitelman’s examination of documents as having a know-show
function, that is, of having “the kind of knowing that is wrapped up with
showing, and showing wrapped up with knowing,” animation here works

in the same way: as a means of knowing the temporal modulations buried
in the AI’s operations, and showing the ways in which these machinic
temporalities interface with human cognition and sensation.31 No longer
printing results like computational outputs of the past, AI and contemporary
computer science research uses animation to index and explore interfaces
of time. By making sensually apparent the temporal vectors of processing
in DeepMind and pitting those against familiar logics of time in game
mechanics, Google manifests a laboratory of AI time, acknowledging that as
they develop their algorithms, they will turn to games with “more temporally
extended planning strategies.”
Positioning animation as tool of knowledge production stacked into
an interface with multiple directions of address aimed at nonhuman and
human actors punctuates its importance in contemporary digital culture.
As Hodge emphasizes relative to the historicity of digital media, “animation
captures the time-based volatility of digital media” and reveals forms of
historical experience constituted through digital technology’s obscured
articulations of time.32 Similarly, the stacks of animation in AI constellates
another point in animation’s history of the ways in which humans and
nonhumans are suffused or intertwined in its constitution, even when
seemingly separate from a distance as in the animations of DeepMind’s
machine runs. Google’s animations serve as a means of assessment for
DeepMind, a practice which, as they allude to above, has continued in the
development of this AI into new gameplay arenas and platforms, such as
Starcraft II.33 Inviting professional Starcraft II game players to compete
against DeepMind, the uncertainty and anxiety over the system’s refined
iterations is captured not only by the screen recordings of gameplay, but
also by reaction shots of DeepMind’s development team. Functioning as
an epistemological agent, animation for these researchers offers the best
means through which traces of the algorithmic functions of sorting and
categorizing can be found. It is a communicative medium that indexes an
action rhetorically compared to human intelligence and projected as being
autonomous, or seemingly alive. This sense of independent action that these
machine runs take, and that so much visually based AI has, is a legacy of
equivalences of sight with mastery and intelligence, a story often rehearsed
and recently challenged by many.34 As Bernard Geoghegan illustrates, by the
mid-twentieth century this visual emphasis had become fully integrated into
both cybernetics and the development of computer visualization systems.35
This led to the development of AI that could, in the popular press, make
claims to possessing forms of intelligent learning because of actions built
around the identification, categorization, and analysis of objects within

puzzle environments.
Visualizing these actions and making them seen through an interface was
equally as important. All the projections of intelligence and senses of life onto
DeepMind’s animations are not simply bound to the moving images, but also
to the technical artifacts that generate them, whose mechanisms seemingly
contain an energy that produces actions. By creating a phenomenal action
of DeepMind’s reinforcement learning processes, Google’s research team
both employ and index an articulation of animation defined through its
ability to make visible modulations of time within contemporary digital
media. Its stacks contain prisms of time, with multiple addresses, but these
technical articulations make intelligible digital networks while projecting
fantasies of agency onto them. The wonder of animation, its projection of
movement that theorists like Sergei Eisenstein and Alan Cholodenko have
characterized as lifelike, becomes a means of making visible algorithmic
forms of time to project notions of intelligent action that can be understood
across human and nonhuman subjects.
Notes
1. Mary Ann Doane, The Emergence of Cinematic Time: Modernity, Contingen-

cy, the Archive (Cambridge, MA: Harvard University Press, 2002), 9.
2. Paul Virilio, Politics of the Very Worst, trans. Michael Cavaliere, ed. Sylvère
Lotringer (Los Angeles: Semiotext(e), 1999). Grant Bollmer provides an
excellent analysis of Virilio’s ideas relative to nineteenth century technics
in Grant Bollmer, Materialist Media Theory: An Introduction (New York:
Bloomsbury, 2019).
3. Wolfgang Ernst, Digital Memory and the Archive, ed. Jussi Parikka (Minneap-
olis: University of Minnesota, 2013), 100.
4. Mark B. N. Hansen, Feed-Forward: On the Future of Twenty-First-Century
Media (Chicago: University of Chicago Press, 2015), 5.
5. Hansen, Feed-Forward, 58.
6. See Matthew G. Kirschenbaum, Mechanisms: New Media and the Forensic
Imagination (Cambridge, MA: MIT Press, 2008); Nick Montfort, “Continuous
Paper: The Early Materiality and Workings of Electronic Literature,” paper
presented at the Modern Language Association conference, Philadelphia,
PA, December 2004, http://nickm.com/writing/essays/continuous_pa-
per_mla.html; Tung-Hui Hu, A Prehistory of the Cloud (Cambridge, MA: MIT
Press, 2015); Andrew R. Johnston, “Models of Code and the Digital Architec-
ture of Time,” Discourse: Journal for Theoretical Studies in Media and Culture
37, no. 3 (Fall 2015): 221–246.
7. See Wolfgang Ernst, “Media Archaeology: Method and Machine versus

History and Narrative of Media,” in Media Archaeology: Approaches, Appli-
cations, and Implications, eds. Erkki Huhtamo and Jussi Parikka (Berkeley:
University of California Press, 2011), 239–255; Bruno Latour, Pandora’s Hope:
Essays on the Reality of Science Studies (Cambridge, MA: Harvard University
Press, 1999).
8. See James J. Hodge, Sensations of History: Animation and New Media Art
(Minneapolis: University of Minnesota Press, 2019); Shane Denson, Discor-
related Images (Durham: Duke University Press, 2020); Adam Hart, Mon-
strous Forms: Moving Image Horror across Media (Oxford: Oxford University
Press, 2019).
nological Media, trans. Anthony Enns (London: Rowman & Littlefield, 2016),
205.
10. See Ernst, “Media Archaeology: Method and Machine,” 239.
11. Alvin Toffler, Future Shock (New York: Random House, 1970).
12. See especially Pamela Lee, Chronophobia: On Time in the Art of the 1960s
(Cambridge, MA: MIT Press, 2004).
13. For more on wonder in animation, see Alan Cholodenko, “Introduction”
to The Illusion of Life: Essays on Animation, ed. Alan Cholodenko (Sydney:
Power Publications in association with the Australian Film Commission,
1991), 9–36; Tom Gunning, “Animating the Instant: The Secret Symmetry
between Animation and Photography,” in Animating Film Theory, ed. Karen
Beckman (Durham, NC: Duke University Press, 2014), 37–53; Colin William-
son, Hidden in Plain Sight: An Archaeology of Magic and the Cinema (New
Brunswick, NJ: Rutgers University Press, 2015).
14. Ernst, Chronopoetics, 4.
15. For more on the collapse of interest, funding, and research in AI from this
time, see Daniel Crevier, AI: The Tumultuous Search for Artificial Intelligence
(New York: Basic Books, 1993).
16. Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Andrei A. Rusu, Joel
Veness, Marc G. Bellemare, Alex Graves, Martin A. Riedmiller, Andreas Fi-
djeland, Georg Ostrovski, Stig Petersen, Charles Beattie, Amir Sadik, Ioannis
Antonoglou, Helen King, Dharshan Kumaran, Daan Wierstra, Shane Legg
and Demis Hassabis. “Human-Level Control through Deep Reinforcement
Learning,” Nature 518 (2015): Methods.
17. Terrence J. Sejnowski, The Deep Learning Revolution (Cambridge, MA: MIT
Press, 2018), 171–172.
18. See D.H. Hubel and T.N. Wiesel, “Shape and Arrangement of Columns in
Cat’s Striate Cortex,” The Journal of Physiology 165.3 (1963): 559–68; David H.
Hubel, Eye, Brain, and Vision (New York: W.H. Freeman, 1988).
19. Hubel, Eye, Brain, and Vision, 131.
20. Yann LeCun, Léon Bottou, Yoshua Bengio, Patrick Haffner, “Gradient-Based
Learning Applied to Document Recognition,” Proceeding of the IEEE 86.11
(November 1998): 2283–2284.
21. Mnih, et al., “Human-Level Control,” 529.

22. See Frank Rosenblatt, “The Perceptron: A Perceiving and Recognizing
Automaton,” Report 85-60-1, Cornell Aeronautical Laboratory (Buffalo, New
York: 1957); and Frank Rosenblatt, “The Perceptron: A Probabilistic Model
for Information Storage and Organization in the Brain,” Psychological Review
65, no. 6 (1958): 386–408.
23. “New Navy Device Learns by Doing: Psychologist Shows Embryo of Com-
puter Designed to Read and Grow Wiser,” New York Times, 8 July 1958, 25.
24. Frank Rosenblatt, “The Perceptron: A Probabilistic Model for Information
Storage and Organization in the Brain,” quoted in Rudolf Seising, “The
Emergence of Fuzzy Sets in the Decade of the Perceptron—Lotfi A. Zadeh’s
and Frank Rosenblatt’s Research Work on Pattern Classification,” Mathe-
matics 6, no. 7 (2018): 110.
25. Sejnowki, 47.
27. Stephanie Boluk and Patrick LeMieux, Metagaming: Playing, Competing,
Spectating, Cheating, Trading, Making, and Breaking Videogames (Minneap-
olis: University of Minnesota Press, 2017), 8.
28. Patrick LeMieux, “From NES-4021 to moSMB3.wmv: Speedrunning the Seri-
al Interface,” Eludamos: Journal for Computer Game Culture 8, no. 1 (2014): 12.
30. Karin Knorr Cetina, “Objectual Practice” in The Practice Turn in Contem-
porary Theory, ed. Theodore R. Schatzki, Karin Knorr Cetina, and Eike von
Savigny (London and New York: Routledge, 2001), 190.
31. Lisa Gitelman, Paper Knowledge: Toward a Media History of Documents
(Durham: Duke University Press, 2014), 1.
32. Hodge, Sensations of History, 16.
33. For details about DeepMind’s use in Starcraft II see Oriol Vinyals, Timo
Ewalds, Sergey Bartunov, Petko Georgiev, Alexander Sasha Vezhnevets,
Michelle Yeo, Alireza Makhzani, Heinrich Küttler, John Agapiou, Julian
Schrittwieser, John Quan, Stephen Gaffney, Stig Petersen, Karen Simon-
yan, Tom Schaul, Hado van Hasselt, David Silver, Timothy P. Lillicrap,
Kevin Calderone, Paul Keet, Anthony Brunasso, David Lawrence, Anders
Ekermo, Jacob Repp and Rodney Tsing, “StarCraft II: A New Challenge
for Reinforcement Learning,” ArXiv abs/1708.04782 (2017): n. pag.; Vi-
nicius Zambaldi, David Raposo, Adam Santoro, Victor Bapst, Yujia Li,
Igor Babuschkin, Karl Tuyls, David Reichert, Timothy Lillicrap, Edward
Lockhart, Murray Shanahan, Victoria Langston, Razvan Pascanu, Matthew
Botvinick, Oriol Vinyals, Peter Battaglia, “Deep Reinforcement Learning
with Relational Inductive Biases,” International Conference on Learning
Representations, 2019.
34. See, for example, Jonathan Sterne, The Audible Past: The Cultural Origins
of Sound Reproduction (Durham: Duke University Press, 2003) and David
Parisi, Archaeologies of Touch: Interfacing with Haptics from Electricity to
Computing (Minneapolis: University of Minnesota Press, 2018).
35. Bernard Dionysius Geoghegan, “An Ecology of Operations: Vigilance, Radar,

and the Birth of the Computer Screen,” Representations 147, no. 1 (Summer
2019): 59–95.
About the Author
Andrew R. Johnston is Associate Professor in the Department of English, the

Film Studies Program, and the Communication, Rhetoric, and Digital Media
Program at North Carolina State University. His book Pulses of Abstraction:
Episodes from a History of Animation (University of Minnesota Press, 2020)
examines abstract animation in cinema and computational media from
the 1950s-70s.
8. Intervals of Intervention: Micro-
Decisions and the Temporal Autonomy
of Self-Driving Cars
Florian Sprenger
Abstract
Self-driving vehicles do not simply translate algorithmic def initions
of their interaction with the environment into material actions. In the
implementation of microdecisions, temporality itself becomes an element
of the success of operations. Taking the fascination for a non-human and
distributed capability of decision-making as a starting point, the paper
explores how the temporality of microdecisions is integrated into technical
systems that interact with their surroundings. On the basis of a media
archaeology of these temporalities, it develops a heuristic of autonomous
technologies that explores the role of micro-decisions. With self-driving
cars, terms such as agency (based on algorithms), temporality (in different
intervals of intervention), decision (in reference to alternative scenarios),
and autonomy achieve new meanings worthy of a re-interpretation.
Keywords: self-driving cars, autonomy, microdecisions, microtemporality
In September 2016, car manufacturer Tesla rolled out update 8.0 for its
autopilot operating system including a new radar processing algorithm.
Manufactured by Bosch, the mid-range radar sensor (MRR) installed below
the car is also used by other car companies, but Tesla was the first to integrate
a new function: with algorithms developed by machine learning, the on-
board computer system, Nvidia’s DRIVE PX 2, enhanced its capacity to
analyze data about the movement of two cars in front.1 Beginning with this
update, the system utilizes the fact that the radar signal bounces between
the vehicle driving in front and the underground of the road in order to
doi 10.5117/9789463727426_Ch08
158 Florian Sprenger
detect the shape and movement of objects ahead of the car driving in front,
even if they are invisible to the driver and the car’s visual sensors.
This new technology reconfigures the autonomous car’s intervals of
intervention, that means the temporality between the registration of an
event and the according reaction. Two months after the update, two cars
crashed on a Dutch highway at high speed. Fortunately, no one was injured
in this accident. A video recorded by a dashcam in an uninvolved Tesla Model
X and published on YouTube by the driver—a Tesla enthusiast—shows how
the Tesla’s driving assistants react to the accident of the car driving in front
and the car that was two cars in front of the Tesla before it actually happens.2
The Tesla brakes autonomously before the driver even has the chance to
recognize that something is about to happen, much less to intervene. In
the video, a warning signal is emitted and the car starts to reduce speed,
but at that moment nothing unusual is visible on the highway. Only a few
seconds later, we recognize that the car, by means of the new processing
algorithm, predicted the collision before it happened by calculating the
speed and movement of a car invisible to the driver. If it hadn’t reacted
autonomously in the short interval available, it might have crashed too,
because traffic was fast and the accident came out of nothing. This interval
of intervention was only available to the car, but not to the driver. After the
fact, we understand that the time in which a reaction was possible remained
below the threshold of human attention and that the car anticipated a
crash that became visible to the driver only in its consequences. The car’s
and the driver’s intervals of intervention did not overlap. The vehicle had
to take into account the probability of a future in which it would crash. In
an extremely short interval—shorter than human reaction time—it had
to decide autonomously between this future and a reaction that might
prevent it.
This video shows computational agency in action and raises a certain
uncanniness. Specifically, there are two sources of uncanniness: the car’s
autonomous reaction to an approaching collision which remains invisible,
and the temporal interval between the (invisible) event demanding a reaction
and its initiation. The consequence of invisibility is the impossibility of a
human reaction, and the speed of the car’s reaction demonstrates an impasse
of human and non-human agency. Both amounts to an uncanniness that
forces us to reflect the temporality of human capabilities. Perhaps the
driver could have reacted if the collision had been visible. They might even
have anticipated the accident. But it is likely that the interval between
anticipation and motoric reaction would have been too long. The autonomous
car brakes before the incident, it reacts to a potential event, because its
Intervals of Intervention 159
so-called Advanced Driver Assistance Systems (ADAS) compare speed and

direction of all three cars and calculate the probability of a collision.
The accident is preempted because in the available interval, the car
decides autonomously between different scenarios depending upon their
calculated probability and initiates a motoric reaction. This whole process
is a microtemporal act of decision, and this paper sets out to explore the
implications of this temporality. Self-driving vehicles do not simply translate
algorithmic definitions of their interaction with the environment into
material actions. In the implementation of what I suggest we call micro-de-
cisions, temporality itself becomes an element of the success of operations.
The term micro-decision should be used as a heuristic, as a language of
description that makes certain aspects of these technologies visible which
would remain invisible with a solely technical vocabulary.
By focusing on amateur videos of pre-empted accidents rather than
on the many videos of smooth autopilot driving also available, this paper
turns to examples in which the operational microtemporality of the car
not only replaces the driver’s activities. The pre-emption of crashes, as will
become clear, is itself a modus operandi. The accident, as a temporal event
in which computational agency unfolds, has become a learning situation
for the vehicle.3 Machine learning thus also implies that the machine has
to make mistakes in order to learn. Vehicles might become fully automated
at one point, but at the present moment it is important to keep in mind that
automation is gradual and extends mainly to computer-supported maneuver-
and routing-procedures. As recent accidents show, these technologies are
not yet as reliable as the manufacturers want the public to believe. The
most advanced cars available at the time of writing are rather sophisticated
software-based driving assistants with sensor-based emergency brake and
lane assistants on what the Society of Automotive Engineers has defined as
level 2, than autonomous self-driving cars that no longer rely upon a driver.
On this level, the automated system is authorized to take full control of the
car in specific situations (usually on a highway) while the driver can take her
or his hands off the wheel, but must be able to intervene at all times. While
the self-driving car without a driver is usually taken as the horizon of fully
automated autonomy (and currently available as a test-service by Waymo
in geofenced areas of California), the decisive challenges in understanding
these technologies amount to understanding their microtemporality. The
technologies operating on this level are already available.
Taking the fascination for a non-human and distributed capability of
decision-making as a starting point, this paper explores how the temporality
of micro-decisions is integrated into technical systems that interact with
their surroundings. With self-driving cars and other autonomous technol-

ogies from everyday applications of the internet of things to unmanned
drones, we observe a transition of digital technologies from the realm of
the virtual into the environment. The perspective presented here argues
that spatial embeddedness and situatedness into surroundings are only
possible because of a specific temporal operationability which is based
on micro-decisions—put bluntly, automated vehicles couldn’t effectively
interact with their environment if they weren’t able to decide in fragments
of seconds. In the words of engineers: “Driving a vehicle implies taking
decisions continuously based on the current awareness of the vehicles
situation and its likely evolution.”4
The Speed of Decisions
Micro-decisions are algorithmic processes of digital technologies charac-

terized by quantity, speed, and automation.5 Their temporality is an effect
of the relation between the sheer numbers of calculations and the velocity
of automated processing. They surmount in surpassing human capabilities.
Their number and speed can only be accomplished by computers: their
quantity is their quality. They are not instantaneous, though, because the
execution of every micro-decision takes time—the car never reacts in “real
time.” I suggest to use the term decision, as a choice between alternatives, to
delineate processual events which depend upon the openness of alternatives
and the contingency of input, in this case from the car’s surroundings. This
temporality is denoted by the micro: its intervals are too short for human
attention. Consequently, a description of autonomous cars as agents of
micro-decisions needs to account for the larger infrastructures that make
micro-decisions possible as enactments of automated agency in a temporality
accessible only to digital computers.
As elements of what Katherine Hayles calls infrastructural “cognitive
assemblages,”6 micro-decisions constitute a new mode of power that we are
yet in the process of understanding. Micro-decisions decide upon possibilities
that shape life in digital cultures: who can move to a specific location and
who can’t, who has access and who has no access, who is connected and who
is disconnected, probably also who is a potential victim of an accident and
who is not. Such decisions are not associated with individual decision-mak-
ers; rather, they are effective because they take place automatically—in
unfathomable numbers and as quickly as possible—according to fixed sets
of rules. They represent the smallest unit and the technical precondition
of a politics of autonomous, adaptive technologies—and of our potential

opposition to it. They question traditional concepts of agency, because the
act of deciding is separated from individual deciders. Automation depends
on the time-critical analysis of the environment, the calculation of the
probabilities of different futures and according reactions, and finally on
“temporal ‘resolution’ and speeds of decision.”7 In this sense, the techno-
logical autonomy of self-driving cars is an effect of a microtemporality that
bestows the capacity of environmental interaction on them. A heuristic of
autonomous technologies that covers the role of micro-decisions should
therefore be accompanied by a media archaeology of these temporalities.
To understand the consequences of autonomy based on micro-decisions,
it is necessary to establish new a vocabulary.8 In the description of the
accident presented so far, a series of concepts was invoked that this paper
attempts to rearrange in the following four chapters:
Computational Agency and Cognitive Assemblages
In order to understand how micro-decisions foster computational agency, it

is important to situate them within the complex interrelations of technical
and environmental elements that constitute a self-driving car as what
Katherine Hayles calls a “cognitive assemblage.”9 Its modes of operation
can be subdivided into three categories: first, regulation and control of
relevant mechanical properties, e.g. steering, braking, signaling etc.; sec-
ond, monitoring the environment (supported by a host of sensors, optical
cameras, GPS, radar, sonar, laser and Lidar) and processing this information
(identification, pattern recognition, categorization); and third access to
road maps and updated traffic information collected by vehicles and to
databases and upgrades maintained by manufacturers.
In their capability of solving cognitive tasks, autonomous cars, their drivers
and their environments constitute a cognitive assemblage: “As a whole, a
cognitive assemblage performs the functions identified with cognition in
general: flexibly responding to new situations, incorporating this knowledge
into adaptive strategies, and evolving through experience to create new
strategies and kinds of responses.”10 In the dynamic unfolding of human and
non-human, material and biological actors into distributed agency, which
Hayles describes with this term, “the spectrum of decision-makers”11 is widely
expanded. Following Hayles, micro-decisions can be regarded as crucial
elements of cognitive assemblages characterized by constant time-critical
processes of synchronization between their components. To refer to this
concept of distributed agency is relevant in this context, because micro-de-

cisions do not simply replace a central cognitive unit, but can be applied
on different levels with varying tasks. Their agency is distributed within
different layers of operationability, it is situated, embodied and interactive.
Autonomy does not mean that the car is independent of the other
components of the assemblage, but a component of an infrastructure that
endows specific actors with the capacity of decision-making and environ-
mental agency. The media theoretical challenge lies in understanding this
temporality not as the agency of a calculating device, but as an infrastruc-
tural effect of a “cognitive assemblage.” In this regard, recent technological
developments challenge traditional concepts of agency, autonomy, and
decision. An investigation of these concepts has to constantly re-adjust the
language of description. This approach may have another, unintended, but
not unwanted consequence: what a human decision is and what constitutes
the agency behind it may become more and more unclear.
The Temporality of Reaction Time
Temporality, so far, has referred to the intervals of intervention. But how

are microtemporal decisions included into intervals of intervention? In the
development of ADAS systems, the perception-response-time of human
drivers and computers has become a “critical design element for vehicular
safety systems, which interact with the driver.”12 In this context, a look at the
history of the concept of reaction time in experimental physiology reveals
that this temporality is conceived of as neither a unitary process nor a simple
reflex. Historical research on reaction time conceptualizes acts of decision
as variables of intervals of reactions between an input (signal subject to
decisions) and an output (pressing of a key). In the 1950s, this process was
formulated in terms of information theory. In the concept of reaction time
developed by cognitive sciences and employed by automotive engineers,
the interval of reaction turns out as the time it takes for the experimental
subject to decide between choosing and pressing a key or not pressing a key.
On the basis of this research, decisions, as temporal variables of information
processing, are integrated into the design of interfaces and control elements
for the interaction between vehicles and human drivers.13 Accordingly, the
automation of decisions as micro-decisions can be seen as an attempt to
implement automated reaction times.
Contemporary conceptions of reaction time are strongly influenced by
a seminal paper that experimental psychologist William Edmund Hick
published in 1952.14 His approach to cognitive research draws upon nine-

teenth century experiments and reframes them through information theory.
Influenced by Claude Shannon’s work, this approach defines reaction time
as the interval between input and output of stimulus and impulse, as the
sum of “cognition time” and “choice time.” Between cognition and choice, a
decision, as a choice between alternatives, takes place about which reaction
to a pre-defined stimulus is necessary.
Hick is interested in the rate of gain of information and, accordingly, the
“mathematical relation between reaction time and number of alternatives.”15
His experimental research determines the interval of a reaction depending
on the number of alternative choices. In the experiments conducted by Hick,
the decision consists in pressing one of up to ten keys if one of up to ten
corresponding lamps flashes. The experiment ultimately shows that if more
choices, that means a series of lamps, are added, or the frequency of flashing
is raised, choice time increases logarithmically in analogy to Shannon’s
definition of information as a logarithmic function of alternatives.16 As
a result, the duration of a decision appears as an effect of the number of
alternative choices. Consequently, the interval of a decision is bound to
the number of choice alternatives: the more alternatives, the greater the
amount of information to be processed. With more alternatives, the level of
uncertainty increases, which leads to a delay in reaction time. Autonomous
technologies, as information-processing machines, correspondingly operate
with probabilities of events that reduce the number of choices.
Hick’s law, as it was subsequently called, only applies to situations in
which decisions are simple and reactions correspondingly quick. Because it
introduces a conception of reaction as information processing, this approach
has become highly influential in cognitive sciences and in engineering
all kinds of interfaces in which several options are presented. Measuring
human reaction times on the basis of Hick’s Law had a huge influence on
the design of safety systems, road intersections and the car’s interfaces.17
As for example a recent project by BMW on a decision-making algorithm
for emergency brake assistants demonstrates,18 Hick’s law facilitates the
correlation of human and non-human reaction times and demonstrates
that the temporality of automation can be understood as the time of deci-
sions—be they based on human or algorithmic agency. The act of decision
takes place between input and output and as such it can be operationalized
as micro-decisions.
The time it takes a self-driving car to react is influenced by different
factors: the speed of calculation, depending upon the number of alternative
choices, which is determined by the clock rate of the car’s GPU (in the
most advanced version, Nvidia’s DRIVE PX 2 has a clock rate of 8 Teraflops

(Floating Point Operations Per Second), which makes this factor almost
negligible for reactions); the transmission times between the different
modules of the car; and the frequency with which it scans its surroundings.
The mid-range radar sensor mentioned in the beginning emits signals every
50 milliseconds (twenty times per second). The radar signal needs time
to be transmitted depending upon the distance of the scanned objects. If
the system decides to come to a stop, the Tesla iBooster brake system can
reach full braking in 150 milliseconds. The time it takes to stop depends
upon the vehicle’s speed and the road conditions. Adding these times to the
time needed for information processing results in an estimated reaction
time. In this constellation of different temporalities, the decision-making
process is the synchronizing element between the environment and the
reaction of the car. In the sequence of events, it is the orchestrating factor.
The Decidability of Alternative Futures
The biggest challenge for autonomous vehicles is not only to constantly

scan their environment and to map their surroundings, but to project
the results of different possible reactions and to prevent accidents before
they happen. Reacting in the time before an accident in a pre-emptive
mode of environmental embeddedness means to realize an alternative
future before the irrevocability of the accident precludes any alternatives.
If an accident is to be pre-empted, this already implies a future in which
the accident happens. Pre-emption operationalizes decisions, which rest
upon the alternativity of its options. The main operation of the car’s CPU
is to constantly transform different futures into probabilities that can be
calculated and require vast numbers of micro-decisions in extremely short
time intervals. This differentiality of potential futures is the basis of the
decidability.
To calculate the probabilities of different scenarios and possible re-
actions to them is not only a question of reaction, but also of prediction.
“Critical traffic situations can require a decision among several unfavorable
alternatives for action. Here again, virtual assessment can support the
development of transparent decision algorithms.”19 The car has to predict
the probabilities of possible futures and choose a reaction according to a
set of codes of behavior that can be either preprogrammed or be extracted
from the diversity of enacted situations by machine learning. These rules
can then be applied to new situations depending upon the pre-given goals
defined beforehand in what can be called macrodecisions—for example

that the safety of pedestrians is more valuable than the safety of passengers.
Pre-programmed, determined reactions never comprise the contingency of
the environment. The rules that the car (respectively the fleet) has learned
and that were coded into its algorithmic set-up are not a set of reactions
in which ethical assumptions are simply translated into algorithms, but
conditions of probabilities, as Lucy Suchman and Jutta Weber argue: “The
behavior is not pre-programmed, but rather the outcome of a kind of
systematized tinkering and situated experimenting of the system with its
environment.”20 Recognized patterns can be used as predictors for future
behavior, as Jack Stilgoe stresses in his analysis of machine learning for
self-driving cars: “Deep learning systems are seen by their creators as means
of engaging with an uncertain world that is impossible to capture with a set
of formal rules. However, in developing rules, such systems may create new
social uncertainties. In gaining the ability to recognize and make decisions
about unfamiliar information, they lose the ability to account for their
actions.”21 In other words: the car has to make decisions under conditions
of uncertainty and at the same time creates further uncertainty as an actor
in the environment.
In the example discussed so far, prediction and action conjoin to arrive at
a decision without a human decider and their agency. The resulting cognitive
assemblage of automated traffic, I want to argue, cannot be explained by
the operationability of a algorithms. Rather, it is important to call them
decisions with all the conceptual consequences this entails, instead of
trivializing them as predetermined executions of fixed rules. The act of
deciding, traditionally embedded in concepts of intentionality, but now
fulfilled by machines, has to be rethought in its microtemporality.22
The heuristics of calling these processes decisions follows the intuition
that as decisive acts they take place in the temporal interval of interruptions.
Interruptions, in the tact of digital processing, are the precondition for
decisions, because they interrupt in order to introduce an alternative.23
Micro-decisions are more than the processing of a determined sequence of
calculations—they always decide between possibilities in an interruption.
Decisions require time. Reactions require decisions. Interruptions free up
this time by adding durations of stasis to the temporality of transmissions.
The duration of calculation, not its moment is decisive.
To use the term decision takes into account their openness. In this sense,
a decision is never absolute—it always implies an alternative. A decision
can always be different. As Niklas Luhmann put it in a paper on decisions
in formal organizations: “It [the decision] constructs the alternativity of
its alternative regarding “what could be,” and it always constructs it in the
present.”24 Luhmann points out that decisions depend upon the potentiality
of alternatives and consequently also upon the difference between past and
future. For the observer, “the decision before the decision is different from the
decision after the decision.”25 While the openness of the decision before the
decisive act lies in the multiplicity of alternatives, that means the contingency
of the future, a decision after the decisive act appears as what Luhmann
calls “thickened contingency”:26 it could have been different, but it is fixed.
To speak of decisions instead of programs, rules, or algorithms opts to
unclose this potential of contingency. The basic fact of decidability—in
opposition to determination—is important here. Acts of decision—and this
is central for the scope of my argument—are decisions not only between two
or more alternative chains of reaction, but also between different futures.
In a technical and mathematical context, a de-cision (Ent-Scheidung) is
more than the execution of a predetermined protocol or a programmed
algorithm, more than the definition of a possible answer. A decision is always
an act that draws a distinction and requires an alternative. As predictions
of possible outcomes, they are bound to the future. In order to decide, a
self-driving car needs alternatives that are in turn products of algorithmic
prediction and the calculation of probabilities.
Alternativity is also the reason why micro-decisions should not be
mistaken for algorithms. Micro-decisions and algorithms are situated on
different conceptual levels. Micro-decisions are coded as algorithms, but
they only become operational in a quantity, speed and automation which
cannot be reduced to algorithms. Not all executions of algorithms are
micro-decisions—algorithms can be slow, analog and based on human labor.
Nonetheless, it is obvious that the importance of algorithms has esca-
lated since the rise of digital computing in the second half of the twentieth
century due to new velocities of calculation. But even though the speed of
execution might be important for the application of specific algorithms,
this temporality is not a necessary condition for their implementation. In
this sense, the term algorithm does not encompass temporality, which is the
key element of automated decision-making.27 Algorithms are temporally
nonspecific and can in principle be performed by a human or a machine at
any speed. While algorithms can be translated back into human work, the
speed and quantity of micro-decisions are not substitutable. The focus on
algorithms prevalent in recent media theoretical discussions conceals that
even though micro-decisions are programmed as executions of algorithms,
their effects lie in their temporality and automation which should not be
mistaken as effects of algorithms.
The Time of Autonomy
The dimension of autonomy presented so far challenges traditional concep-

tions of agency and accountability. In this regard, the automation of traffic
leads to a reversal of contingency of the so-called trolley problem,28 which
results in juridical and ethical grey areas. Though the ethical dimension of
automated, time-critical technologies cannot be discussed here extensively,
the following remarks on the trolley dilemma serve to demonstrate how
autonomy and microtemporality are intertwined.
The trolley dilemma was introduced by Philippa Foot in her seminal
paper The Problem of Abortion and the Doctrine of the Double Effect in 1967.
This scenario touches the roots of ethical considerations by negotiating
human interaction with technology. Foot describes a runaway tram driver
who has to decide between two tracks: five people are working on track
one and one person on track two. Anyone of them is bound to be killed
if the tram takes the track. With the rise of automated cars, the resulting
ethical dilemma needs to be reformulated: How should the car react in a
situation in which the driver cannot intervene? For example, a car may
have to decide if it rather takes the risk of destroying itself and killing
the driver or of crashing into a group of people certainly killing several
of them.29 The new challenge lies in the fact that decisions about life and
death and the accompanying responsibilities might be outsourced to
algorithms. In such a scenario, responsibility couldn’t be accredited to a
human actor—a problem even more urgent since the first deadly crash
with an automated car in May 2016, in which a Tesla wrongly identified a
blue truck for the sky.30
This discussion about agency and accountability has become very popular,
as its steady recurrence in newspapers and publications shows. So far, these
discussions have not registered the importance of temporality for the trolley
driver’s decision. Usually, descriptions of the dilemma do not account for
the time, respectively for the urgency of the decision that needs to be made.
In the scenario described by Foot, it is not mentioned that the driver of a
tram only has a few seconds to decide between their options of killing five
human beings or one. The dilemma, I want to argue, consists not only in
the fact that both decisions result in casualties, but is multiplied by the
fact the driver doesn’t have the time to reflect their options or even the
philosophical implications of their decision. The interval of intervention,
which is not covered by Foot’s argument, needs to be taken into account
in contemporary ethical discussions, because the microtemporality of
decisions—not accessible for the human driver—is operationalized in
automation. While a human driver might be unable to come to a rational

conclusion in a few seconds, the car’s micro-decisions are perfectly ratio-
nal—though not necessarily convincing—because they are calculated in
the speed of data processing.
In how far the ethical discussions have to be transformed by taking into
account microtemporalities is a question that has to be left to experts in this
field. In the context of this paper, it is important to note that automation
and decision-making are connected on the basis of microtemporality. In the
words of engineers again: “This challenge leads to relevant implications on
the design of the automated driving system: the automated system has to
make decisions in all situations and cannot rely on the driver to take back
responsibility instantly.”31 As this quotation shows, engineers deduce the
autonomy of a vehicle from its potential to decide in all situations, which
means at all times and in all temporalities. Micro-decisions thus can be
seen as a core element of autonomy.
At the current state of technology, though, the autonomy of decisions
is a messy business. Practically, the experience of driving a Tesla shows
the need to balance agencies of decision and to constantly revise them.
In a study on Tesla’s Model S, Mica Endsley has demonstrated that the
different systems of automation embedded in the car result in a series of
complex situations of decision-making on behalf of the driver. Agency is
constantly re-distributed. The technologies employed by a Model S at the
time of Endsleys research in late 2016 (version 7.2) included Adaptive Cruise
Control (ACC) to adjust its speed to traffic, Autosteer to hold the line and
Auto Lane Change. Anxieties arise for example because the ACC reacts if
the car in front brakes or stops, but not at traffic lights. Auto Lane Change
seems unreliable in detecting cars next to the Tesla. Endsley argues that
while driving on a highway is indeed automated to a large extent, all other
traffic situations need constant intervention and attention by the driver.32
In this regard, the automated execution of specific routines also creates
the possibility of distraction. In situations in which the machine is unable
to decide and the human driver has to intervene, distinct and intuitive
interfaces are required that address the driver’s situational awareness in
an unmistakable way. The comfort of automated driving, Endsley argues,
may lead to a slowdown of attention and, consequently, of reaction time.
The most important task for designers is the integration of different modes
of attention through human-computer-interaction and mental models
of the car. This, in fact, turns out to be an integration of different levels
of decision-making—and an integration of different reaction times as
intervals of decisions.33
In this respect, the concept of micro-decisions may prove to be relevant

for a discussion of the agency in question and finally for new concepts
of the interaction between drivers and cars. Autonomous technologies
can be understood as situated technologies capable of adaptation to their
surroundings. As Suchman and Weber have demonstrated in the context
of unmanned drones, “the project of machine intelligence is built upon,
and reiterates, traditional notions of agency as an inherent attribute and
autonomy as a property of individual actors.”34 As an alternative in the vein
of science and technology studies, they propose a conception of agency
as relational, embedded in human-machine-assemblages and constantly
re-configured. In addition to this perspective, the approach presented here
supposes to take the temporality of micro-decisions into account for the
constitution of computational agency and autonomy.
Conclusion
The uncanny that may creep upon us while watching the video with which
this paper started is a symptom for the necessity to rethink our relation to
such technologies. To speak about decisions as their defining characteristic
and to borrow this concept from social theory does not yet entail that there
is any social intention behind their operations. Machines do not (yet?) have
intentionality in the sense of conscious acts or reflexivity. But they can, as
philosopher Peter-Paul Verbeek suggests, achieve intentionality in the sense
of intendere: they can realign and orient something.35 The alternative be-
tween and mutual exclusion of intentional, reflexive human decision making
and deterministic, mechanical procedures is misleading. Machines that carry
out micro-decisions are of course produced and managed by human agents,
who also program their protocols and algorithms. The measures according
to which decisions are made are necessarily established in protracted in-
stitutional negotiations between various interest groups, manufacturers,
regulators, engineers and coders, obvious in ethical discussions about the
implementation of life-threatening reactions in autonomous vehicles and
the juridical problems this entails. The implementation of micro-decisions
is based on collective or individual macrodecisions. Yet, the great mass of
micro-decisions can only be executed by computers, and it is this mass that
underlies the technical definition of successful digital communication or
traffic on computer-supported networks. Micro-decisions have become as
effective as they are precisely because they circumvent the laborious human
act of decision-making on the basis of calculated probabilities.
Assuming the determinacy of micro-decisions by machines results

in a simplistic division between determinate computational agency and
free human agency. What is determined cannot be decided. In the words
of Niklas Luhmann, decisions imply a “minimum of unpredictability.”36
While it is obvious that human agency and human decision-making are
different from computational agency, that does not mean that we should
imagine agency only as a blueprint of human agency against which machinic
agency would seem to be deterministic. The concept of micro-decisions
wants to circumvent the shortcomings of decisional determinism by tak-
ing into account three aspects. First, conceptualizing micro-decisions as
predetermined acts of rendering a preprogrammed set of rules would also
mean that such decisions cannot be changed and are bound to past events
instead of anticipating and regulating the future. Second, if we def ine
micro-decisions as determined, we would not be able to understand the
technological developments that allow for example an autonomous car to
react to the complexity and contingency of its environment. Calculative
environmental technologies need to account for the unpredictable.37 Third,
defining decisions as deterministic would imply that they are immediate—
that they happen without taking time. The fact that micro-decisions—and
that no transmission and no medium—can ever be immediate, that they
always take time, leads to a non-deterministic, decisive understanding of
micro-decisions. This means that the politics of micro-decisions, including
potentials to transform them, are based upon the interruption of connection
and the fact of constant disconnection.
To speak about decisions also means to bear in mind that no decision is
ineluctable and that every decision can be reached in a different manner—
that it is possible to modify them for the better, but that they can also turn
out for the worse. Yet even a bad decision is better than no decision, which
leaves no room for improvement—for example on how autonomous cars
react to specific situations in which people are in danger. To make decisions
in advance, to determine them, or even to abolish the act of decision-making,
is, in every case, to reduce what is possible. Nonetheless, micro-decisions are
neither inherently good nor inherently bad. For the operations of a network
such as that of automated traffic, however, they are unavoidable. This means
that we need a conceptual framework that helps us to understand their mode
of power in its microtemporal dimension. To simply delegate decisions to
machines and then to conceive them as determined algorithms, as something
that necessarily happens as it happens, is at core a depoliticizing act. In this
regard, it is the task of critical humanities to politicize machines. This means
to take the alternatives of each decision, be it fast or slow, as a starting point.
Notes
1. Tesla (blog), Tesla, September 26, 2016, https://www.tesla.com/de_DE/blog/

upgrading-autopilot-seeing-world-radar. For a demonstration, see Eric Love-
day, “Tesla Model X Sees Two Vehicles Ahead In Automatic Braking Test—
Video,” Inside EVs, November 10, 2016, https://insideevs.com/tesla-model-
x-sees-two-vehicles-ahead-automatic-braking-test-video/. Autopilot here
means the combination of different Advanced Driving Assistance Systems:
collision warning, autosteer lane centering, self-parking, automatic lane
changes, and traffic-aware cruise control. In this regard, Tesla offers a good
example of the constraints, challenges, and potentials of driving assistants,
although it is certainly not the only competitor in the field.
2. Fred Lambert, “Tesla Autopilot’s New Radar Technology Predicts an Ac-
cident Caught on Dashcam a Second Later,” Electrek, December 27, 2016,
https://electrek.co/2016/12/27/tesla-autopilot-radar-technology-predict-ac-
cident-dashcam/.
3. See Paul M. Leonardi, Car Crashes without Cars: Lessons about Simulation
Technology and Organizational Change from Automotive Design (Cambridge,
MA: MIT Press, 2012).
4. Javier Ibanez-Guzman, Christian Laugier, John-David Yoder, and Sebastian
Thrun, “Autonomous Driving: Context and State-of-the-Art,” in Handbook of
Intelligent Vehicles, ed. Azim Eskandarian (Berlin: Springer, 2012), 1286.
5. See Florian Sprenger, Politics of Microdecisions: Edward Snowden, Net Neu-
trality and the Architecture of the Internet (Lüneburg: Meson Press, 2015).
6. See N. Katherine Hayles, Unthought: The Power of the Cognitive Noncon-
scious (Chicago: University of Chicago Press, 2017).
7. Axel Volmar, “Zeitkritische Medien im Kontext von Wahrnehmung, Kom-
munikation und Ästhetik,” in Zeitkritische Medien, ed. Axel Volmar (Berlin:
Kadmos, 2009), 9–26, 10.
8. In a similar direction, Beatrice Fazi has argued that “computational process-
es of decision-making and prediction might constitute a novel modality of
thought: not despite the deterministic, deductive, axiomatic, procedural
and operational character of these processes, but rather because of it.” M. B.
Fazi, “Can a Machine Think (Anything New)? Automation Beyond Simu-
lation,” AI & Society 34, no. 4 (2018), 10. In addition to Fazi’s argument, the
emphasis here rests on the temporality of these processes.
9. See Hayles, Unthought.
10. Hayles, Unthought, 118–19. Hayles prefers the term assemblage over
network, because it denotes “continuity in a fleshy sense,” while network
implies “a sense of sparse, clean materiality” (118).
11. Hayles, Unthought, 115.
12. Azim Eskandarian, “Fundamentals of Driver Assistance,” in Handbook of
Intelligent Vehicles, ed. Azim Eskandarian (Berlin: Springer, 2012), 491–535,
499.
13. See Katherine L. Plant and Neville A. Stanton, “Identifying the Importance
of Perceptual Cycle Concepts during Critical Decision making in the Cock-
pit,” Procedia Manufacturing 3 (2015).
14. For the importance of Hick, see Robert W. Proctor and Darryl W. Schnei-
der, “Hick’s Law for Choice Reaction Time: A Review,” Quarterly Journal
of Experimental Psychology 106, no. 16 (2018). On the history of physiolog-
ical research on reaction times see Henning Schmidgen, “The Donders
Machine: Matter, Signs, and Time in a Physiological Experiment, c. 1865,”
Configurations 13 (2005); and Jimena Canales, A Tenth of a Second: A History,
Pbk. ed. (Chicago: University of Chicago Press, 2011).
15. W. E. Hick, “On the Rate of Gain of Information,” Quarterly Journal of
Experimental Psychology 4, no. 1 (1952): 11. Hick’s law is the basis for a model
of intelligence in which the speed of information processing is an indicator
of intelligence. Though he doesn’t use the term decision and talks about
choice, the term has become common in the application of Hick’s law.
16. In a follow-up article published a year later, Ray Hyman underlines that
“the choice-reaction-time experiment can be looked upon as a model of a
communication system” (Ray Hyman, “Stimulus Information as a Deter-
minant of Reaction Time,” Quarterly Journal of Experimental Psychology
45, no. 3 (1953): 188) and determines the bits of information per stimulus
presentation.
17. See Tibor Petzoldt and Josef F. Krems, “How Does a Lower Predictability
of Lane Changes Affect Performance in the Lane Change Task?,” Applied
Ergonomics 45, no. 4 (2014). and Chia-Fen Chi and Ratna S. Dewi, “Matching
Performance of Vehicle Icons in Graphical and Textual Formats,” Applied
Ergonomics 45, no. 4 (2014).
18. See Philipp Reinisch, Peter Zahn, and Dieter Schramm, “Using a Reaction
Time Model for Determining a Collision Avoidance System’s Brake Timing,”
IFAC Proceedings Volumes 43, no. 7 (2010). For other examples, see Marc
Green, “‘How Long Does It Take to Stop?’: Methodological Analysis of Driver
Perception-Brake Times,” Transportation Human Factors 2, no. 3 (2000).
19. Thomas Helmer et al., “Safety Performance Assessment of Assisted and
Automated Driving in Traffic Simulation as Knowledge Synthesis,” in
Automated Driving, ed. Daniel Watzenig and Martin Horn (Berlin: Springer,
2017), 473–94, 492.
20. Jutta Weber and Lucy Suchman, “Human–Machine Autonomies,” in Auton-
omous Weapons Systems: Law, Ethics, Policy, ed. Nehal Bhuta et al. (Cam-
bridge: Cambridge University Press, 2016), 75–102, 88.
21. Jack Stilgoe, “Machine Learning, Social Learning and the Governance of
Self-Driving Cars,” Social Studies of Science 48, no. 1 (2017): p. 6.
22. The concept of decision has been discussed in many contexts, but so far not
systematically applied to technologies. See for example Derek P. McCor-
mack and Tim Schwanen, “Guest Editorial: The Space-Times of Decision
Making,” Environment and Planning A 43, no. 12 (2011); Peter Adey and
Ben Anderson, “Event and Anticipation: UK Civil Contingencies and the
Space-Times of Decision,” Environment and Planning A 43, no. 12 (2011),

doi:10.1068/a43576.; Florian Hoof, “Medien managerialer Entscheidung:
Decision-Making ‘At a Glance,’” Soziale Systeme 20, no. 1 (2016).
23. On the importance of interruptions for micro-decisions see Sprenger, Poli-
tics of Microdecisions, p. 86.
24. “Sie konstruiert die Alternativität ihrer Alternative unter dem Gesicht-
spunkt ‘was sein könnte;’ und sie konstruiert sie in ihrer Gegenwart.” Niklas
Luhmann, “Die Paradoxie des Entscheidens,” Verwaltungs-Archiv 84, no. 3
(1993): 291. All translations my own.
25. “Die Entscheidung vor der Entscheidung [ist] eine andere Entscheidung als
nach der Entscheidung.” Luhmann, “Die Paradoxie des Entscheidens,” 293.
26. In Luhmann’s original German, “verdichtete Kontingenz.” Luhmann, “Die
Paradoxie des Entscheidens,” 293.
27. That does not imply that algorithms are unimportant for automated deci-
sions, see Serjoscha Wiemer, “Von der Matrix zum Milieu: Zur Transforma-
tion des Entscheidungsbegriffs zwischen homo oeconomicus und evolu-
tionärer Auslese,” in Medien der Entscheidung, ed. Tobias Conradi, Florian
Hoof and Rolf F. Nohr (Münster: Lit, 2016), 23–46.
28. See Philippa Foot, “The Problem of Abortion and the Doctrine of the Dou-
ble Effect,” Oxford Review, no. 5 (1967).
29. See Jean-Francois S. A. Bonnefon and Iyad Rahwan, “The Social Dilemma
of Autonomous Vehicles,” Science 352, no. 6293 (2016), doi:10.1126/science.
aaf2654. For the difference between accountability and responsibility see
Maya I. Ganesh, “Entanglement: Machine Learning and Human Ethics in
Driver-Less Car,” Aprja 6, no. 1 (2017). For a critique of the applicability of
the trolley dilemma, see Tobias Matzner, “Autonome Trolleys und andere
Probleme: Konfigurationen Künstlicher Intelligenz in ethischen Debatten
über selbstfahrende Kraftfahrzeuge,” Zeitschrift für Medienwissenschaft 21
(2019), 46–55, doi:10.25969/mediarep/12632
30. US Department of Transportation, National Highway Traffic Safety Admin-
istration, ODI Resume, June 28, 2016, https://static.nhtsa.gov/odi/inv/2016/
INCLA-PE16007-7876.pdf. See Victoria A. Banks, Katherine L. Plant, and
Neville A. Stanton, “Driver Error or Designer Error: Using the Perceptual Cy-
cle Model to Explore the Circumstances Surrounding the Fatal Tesla Crash
on 7th May 2016,” Safety Science 108 (2018), doi:10.1016/j.ssci.2017.12.023.
31. Jan Becker, Maria-Belen Aranda Colas, and Nordbruch Stefan, “Bosch’s
Vision and Roadmap Toward Fully Autonomous Driving,” in Road vehicle
automation, ed. Gereon Meyer and Sven Beiker (Heidelberg: Springer, 2014),
49–60, p. 50.
32. From an engineering point-of-view see also Victoria A. Banks and Neville
A. Stanton, “Keep the Driver in Control: Automating Automobiles of the
Future,” Applied Ergonomics 53 (2016), doi:10.1016/j.apergo.2015.06.020.
33. Some engineers even argue that human agency needs to be included for
automation to become effective: Neville A. Stanton and Philip Marsden,
“From Fly-by-wire to Drive-by-wire: Safety Implications of Automation in
Vehicles,” Safety Science 24, no. 1 (1996). See also Sam Hind, “Digital naviga-
tion and the driving-machine:: Supervision, calculation, optimization, and
recognition,” Mobilities 20, no. 1 (2019).
34. Jutta Weber and Lucy Suchman, “Human–Machine Autonomies,” in Auton-
omous Weapons Systems: Law, Ethics, Policy, ed. Nehal Bhuta et al. (Cam-
bridge: Cambridge University Press, 2016), 75–102, 98.
35. Peter-Paul Verbeek, “Ambient Intelligence and Persuasive Technology: The
Blurring Boundaries Between Human and Technology,” NanoEthics 3, no. 3
(2009).
36. Luhmann, “Die Paradoxie des Entscheidens,” 287. In Luhmann’s original
German, “ein Mindestmaß an Unvorhersagbarkeit.”
37. While Luciana Parisi describes algorithmic decisions with Alfred North
Whitehead’s process ontology as modi of prehension, the focus here rests
on their technical implementation: “The computation of relations thus re-
quires that preplanned decisions become replaced by prehensive capacities
of decision making, which afford the parametric system the freedom to
establish unintended connections between parameters within the con-
strained conditions of sequential programming.” Luciana Parisi, Contagious
Architecture: Computation, Aesthetics, and Space (Cambridge: MIT Press,
2013), 138.
About the Author
Florian Sprenger is Professor for Virtual Humanities at the Institute for

Media Studies at the Ruhr-Universität Bochum. His research topics are
environmental technologies, transformations of traffic, and autonomous
technologies.
Part III
Lifetimes
9. Grounded Speed and the Soft
Temporality of Network Infrastructure
Nicole Starosielski
Abstract:
This chapter offers a descriptive view of cable temporality from the per-
spective of network operators and those involved in system maintenance
and repair. Examining the “soft temporalities” of these operators’ labor, the
chapter illustrates a tension at the heart of media’s infrastructures: in many
places, slowness, stability, embeddedness, and fixity of infrastructures are
what enable speed and acceleration. After describing the “grounded speed”
of the cable network, the chapter turns to the phenomemon of temporal
irruptions: moments when the assemblage of temporal processes that
enable network operation and network speed suddenly and radically
changes the network.
Keywords: undersea cable, infrastructure, slowness, telecommunications,

network operation
Outside the thick concrete walls of a Pacific Island undersea cable station, the
dense heat draws tourists to the beach and into the ocean. Many have come
here to break from the accelerated rhythms of hyper-stimulated lives—a
separation that some see as moving “out of time.”1 For others, “island time”
is a staged interruption in their acceleration, one that incorporates them
into longstanding colonial temporal regimes. Inside the station, a different
temporality is in operation. It is evident in the chilled air, where molecules
have been slowed to an optimal temperature for machine operation. And it
is evident in the building’s sparse population, which exists in stark contrast
to the crowded beaches. On the day that I visit the station, there are only
a few employees on the clock. As one of them gives me a tour, he points
to a laptop on top of a stack of servers. Here, he tells me, he could play
doi 10.5117/9789463727426_Ch09
178 Nicole Starosielski
video games with the least possible latency. His computer’s position in the
network’s geography gives him a slight temporal advantage in real-time
multiplayer online games. He is describing the spot—in a concrete building
intensely insulated from the rest of the island, an island that is visited
for its remoteness—where he is as close as possible to achieving network
instantaneity. It is the place where his signals would travel faster than
anyone else’s, even in the many cities the nearby visitors have come to
escape. Metaphorically and thermodynamically, this is both the hottest
and the coldest point on the island. The temporality of the cable station
inverts the rhythms of bodies beyond the walls: here, it is not that slowness
provides a break from acceleration, but rather, that slowness makes possible
extraordinary speed.
A century earlier, on another island in this same ocean, telegraph op-
erators shared a similar sense of spatial and temporal disjuncture as they
occupied the center and periphery at once. Sharing their stories in the cable
industry magazine, they reflected on “the monotony of our slumberous
existence,” challenged one another to prove that they lived in the loneliest,
most remote station, and looked forward even to the transient missionaries
that might drop by in the summer and provide the smallest diversion from
the banality of everyday life.2 The slowness of their world existed in stark
contrast to the intensity, speed, and pace of the network they operated.
This was often commented upon in stories and the company’s internal
communications. One operator remarked: “Though they have to wait three
months for letters to be forwarded from Honolulu … men on Midway can
obtain news from nearly all parts of the world in a few minutes.”3 For
these telecommunications workers, as for the people they connected, the
acceleration of communications could heighten the relative sense of the
slowness of their nonelectric surroundings, including their own bodies and
the labor that they performed.
These two examples, one from the early twentieth century and the other
from a hundred years later, illustrate a critical point about the temporality
of networked media: the speed of communication is often made possible by
incredible investments in stasis, stability, and slowness—molecular, bodily,
and architectural. While much interest in infrastructure’s temporalities and
microtemporalities has focused on acceleration, less research has examined
the temporal dilations and prolongations that make acceleration possible.
This, as Sarah Sharma points out in her foundational text In the Meantime,
is a critical blind spot of speed theory. 4 As she identifies in works from Paul
Virilio’s Speed and Politics to Jonathan Crary’s 24/7: Late Capitalism and the
Ends of Sleep, the focus on speed, time-space compression, and acceleration
Grounded Speed and the Sof t Tempor alit y of Ne t work Infr astruc ture 179
is often conducted from a position of temporal privilege, and often fails

to account for the multiplicity of lived times that scaffold acceleration.5
This chapter brings Sharma’s argument to bear on media infrastructures.
It offers a descriptive view of cable temporality from the perspective of
network operators and those involved in system maintenance and repair.
Following Gabriele Schabacher in this volume, it sees hardwiredness “as
a network effect with relative duration brought about by specific types of
labor.” If we consider the operations of these systems as solely a technical
process, rather than a performance of situated and geopolitically-specific
labor of care, we miss the complex, embodied temporalities of contemporary
digital infrastructure. And as a result, we fail to account for the decisions
involved in establishing, maintaining, and securing network infrastructure.
Like the taxi-cab drivers that transport global jetsetters in Sharma’s
study, there is a population whose everyday work and temporal coordination
makes possible the speed of the network—in the case of the cable system,
these include network operators, engineers, and suppliers, among many
others. The everyday work of cabling involves, for today’s operators, waking
up in the middle of the night, interrupted from sleep by a broken cable. It
includes the coordination of ship movements with weather patterns and
the active routing of internet traffic. One thing that distinguishes network
operators from many other laborers that maintain the acceleration of the
global elite is a co-presence of extraordinary temporal privilege (and a
concentrated personal investment in that speed) alongside a sense of an
extraordinary slowness, prolonged activities of care, and relative stability.
While this in itself is not unique—living in the midst of contradictory
temporalities and oppositional rhythms is characteristic of modern life—the
temporalities of these operators’ labor illustrate a tension at the heart of
media’s infrastructures: in many places, slowness, stability, embeddedness,
and fixity of infrastructures are what enable speed and acceleration. From
the operator’s standpoint, especially those that inhabit the cable station,
these two extremes are often brought together in poetic opposition, in-
nately intertwined, and perceptible in their bodies as an affect of remote
centrality. I call this dynamic grounded speed: the ongoing production of
accelerated rhythms through the consistency and regularity of network
bodies, architectures, practices, and environments.
In the first half of this chapter, I describe the grounded speed of the
undersea cable system. As a background for this analysis, I begin with a
narrative about the larger-scale temporal patternings of the cable network.
Some of the common descriptions of cable temporalities include: cables
erase time; cables accelerate; and network infrastructures are significant
because of their high speed and their microtemporal operations. In contrast

to these typical observations, the second section reveals some of the varied
decelerations, slowings, and stabilizings that make possible cable speed. First
and foremost is the slowness of human navigation through the ocean. The
slowness and sparseness of marine transport has produced relative safety
for cable systems: more boats would produce more cable breaks. Next to this
is the historical privileging of remote areas for centers of network traffic,
where the frequency of human and machinic movement is less likely to
disrupt or interfere with cable traffic. Lastly is the relative slowness of the
cable industry, an insular community where extended relationships between
operators emerge over years. This social world reproduces the cable network
as an intimate and familiar landscape. Tracking cable industry labor, past
and present, I show how the slowness of the network’s milieu grounds the
speed of interconnected global telecommunications.
Because network traffic is grounded—in the environments they extend
through, in the rhythms of the bodies those who operate them, and in the
patterned technics of their nonhuman milieus—the alteration of ground
rhythms alters the speed of the system itself. If these typically balanced
temporalities suddenly shift, the routinized temporality of an infrastructure
is suddenly inflected by changes in its temporal milieu. I call these moments
temporal irruptions: moments when the assemblage of temporal processes
that enable network operation and network speed suddenly and radically
changes the network. The fourth section of this essay charts how such
irruptions along cable routes produce disruptions in the patterned time of the
global network. Weaving through a set of breaks, disruptions, and irruptions
in the cable network, I show how the elongated rhythms that keep networks
safe can suddenly interrupt—and irrupt into—the cable system. I argue that
these irruptions direct our attention to the ongoing role of everyday, lived
and embodied temporalities and the rhythms of cabled sociality, and the
ways that infrastructures are “cyclical and repetitive processes of formation
and transformation.”6 As a complement to the hardwired temporalities that
occupy much of this book, I call these soft temporalities, as they are the
underbelly—a shifting and complex temporal substructure—of hardwired
media and infrastructural times.
Patterned Time
In his well-known essay “Technology and Ideology: The Case of the Tele-
graph,” James Carey demonstrates how the telegraph enabled the separation
of physical objects from communication for the first time and, in doing so,
was critical in the establishment of standard time, the transformation of
futures markets, and the evolution of time contracts.7 Cable infrastructure,
he showed, facilitated the production of modern, patterned time, and was
an essential part of the temporal grid that restructured the industrial
world. Undersea cables are the global links in this history, enabling the
standardization of time on a global scale. An early function of undersea
networks was to establish differences in longitude—which, as Richard
Stachurski documents, “were literally a matter of time.”8 One of the first
uses of the transatlantic cable was to send star-transit timings that would
help to determine longitude. The United States Coast Survey built tem-
porary observatories at both ends of the cable, and between October and
November 1866, clock signals were sent between the two locations on five
different nights, “yielding the first directly-measured longitude of the dome
of the U.S. Capitol west of the Greenwich Observatory: 5 hours 8 minutes
and 2.22 seconds.”9 The completion of the All-Red Line at the turn of the
twentieth century—a British network that encircled the earth, was an
infrastructure for the creation of Universal Standard Time.10 Today fiber
connectivity via undersea cables enables the global coordination of digital
time, and undergirds the precise operations of global positioning systems
(GPS) among other digital operations.
Undersea cable systems, as was the case for telegraph systems generally,
were perceived as annihilating time and space. This was true even within
the cable industry. For example, in the cablemen’s magazine, The Zodiac,
many issues featured the final lines of Rudyard Kipling’s “The Deep-Sea
Cables”: “They have wakened the timeless Things; they have killed their
father Time.” Although cable networks were understood to make time matter
less, in actuality they substituted one form of temporal difference (in which
temporal difference correlated with geographic distance) for another (a
temporal difference that correlated with distance from infrastructure). In
other words, the cable system produced a new set of temporal patternings in
which elite users located in privileged positions felt synced with other elite
users in privileged positions—they felt the distance between them less and
less as temporal delay. In turn, cables introduced a new temporal structure in
which those who did not have access to cabled systems or cabled information
would more frequently perceived themselves as behind, delayed, and distant.
In accelerating the production of a globalized temporality, whether through
finance, news, or standardized time, the submarine telegraph also intensified
a colonial temporal regime and was a means by which the Western empires
were able to dominate the nineteenth-century timescape.
Today’s undersea cable networks, which carry 99 % of transoceanic

internet traffic, continue to restructure media temporalities in ways that
parallel many of the telegraph’s original impacts. In the world of high-
speed trading, the cable system similarly both standardizes and accelerates
machinic temporalities. This is especially true given the recent turn toward
lower-latency cable systems. In network traffic engineering, latency is
the amount of time it takes for a signal to make a round-trip between two
terminals. Although the optimization of networks for speed has varied over
the decades of cable laying, recent networks have been designed specifically
as part of “low-latency” solutions. These have been developed to capture
the market of high-frequency traders as well as for the range of computa-
tional practices that are dependent on speed—from cloud computing to
the navigation of remote vehicles.
Take for example the 2015 Hibernia Express cable, which introduced a
new low-latency path between New York and London—clocking in at 58.95
milliseconds.11 In 2017, the Seabras-1 submarine cable supplied Seaborn
Networks with a “proprietary ultra-low latency (ULL) solution” (SeaSpeedTM)
between the BM&F Bovespa Stock Exchange in São Paulo, Brazil and Cart-
eret, New Jersey.12 Geography and the shortest route isn’t the only factor in
latency—other technical factors include the refractive index of the fiber,
the number of amplifiers, and the process of equalizing the fibers.13 Other
traffic-related factors include network congestion, firewalls, and over-used
routers. As one operator describes: “This also means that the integration
for the land and sea routes must be taken into account, every interface can
add latency.”14 Achieving speed in the cable network never occurs merely
through reduction of geographic distance, it often involves a consolidation
of technical interfaces and a reconfiguration of social practice. Just as cable
systems do not simply erase time, but re-pattern it differently for different
publics, the production of the cable’s temporal dynamics does not simply
involve acceleration, but are a complex temporal and social operation.
Grounded Speed
What is often missing from these descriptions of cable speed are embodied
practices that prop up acceleration—labor and social worlds assembled
into fixed rhythms. Pulling its workers into a set of patterned times, the
early telegraph system often sped up their movements and synced them
to a machinic temporality. Some operators were required not only to meet
a minimum speed, but also to ensure accuracy at this speed. As the poem,
“Cablemen,” describes cable work: “we work ‘em like clockwork … We must
work them like lightning,/ Must work ‘em pell mell,/ For it’s through the
thin cables/The Empire can tell/That all her Dominions and Children are
well.”15 H.F.B., the poem’s author, describes the work of the cablemen as both
“like clockwork” and “like lightening,” drawing connections to the machinic
and natural world to articulate the routinization and inhuman speed of
the operator’s body, all in service of the British Empire. The patterning of
the men’s bodies was structured by and in turn scaffolded British colonial
expansion. Another cableman laments the lack of recognition for the men
in the middle: “do you ever think how men/ Have worked the whole night
through,/ And done the best within their ken/ to rush that Press-work
through?”16 These descriptions of the operators’ work echo throughout
the cable magazines, with their details about late-nights and extended
hours, immediacy to global events, the repetitive sounds of clock and
telegraph, and the “dull routine” that propels the world forward.17 As Sarah
Sharma writes: “Temporalities are not times; like continually broken clocks,
they must be reset again and again. They are expected to recalibrate and
fit into a larger temporal order.”18 The temporality of the cable network
was anchored in the ongoing recalibration of cable workers’ embodied
temporalities.
This patterning was not only machinic—keyed by the telegraph and
modulated by cable. It was and continues to be social. In the telegraph
era, regulations governing the rhythms of everyday life were intended to
stabilize men’s bodies, and as a result, the network’s operation. One cable
operator recounted his time at the Southport Station in Australia, where for
probationers, “Church attendance once on Sundays was compulsory, and
there was a 10 p.m. curfew … The use of lamps in bedrooms was forbidden.”19
Here the use of modern technologies of light were forbidden (even as candles
were allowed) as part of the regulation of the men’s nighttime activities. This
ensured their adherence to a temporal pattern that would keep the cable
system operational. In the telephone cable era, more complex technological
systems would be deployed as a means of managing time and timing. For
example, aboard the cableship Neptun, an internal CCTV network enabled
“split-second timing” in cable laying operations—and for cable-laying
operators.20
The operators of the Eastern Telegraph System thus held together a
number of overlapping temporalities in their own bodies. On one hand, they
perceived their own role as an extension of the cable, annihilating temporal
distance between Britain and the colonies. They both upheld its speed, with
the actual speed of transmission reflecting the cumulative capabilities of
all operators on the route, and saw themselves as the very first receivers:
in 1926, one author remarked: “The work of a cable operator is a business of
thrills and chills, though many imagine it to be a humdrum, mechanical
occupation. You never know what is coming next. Cable operators have early
news of world-shaking events.”21 They occupied the subjective position, as
with many other inhabitants along the colonial lines, in which a colonial
temporal regime (the location of Britain as ahead) materialized in the
technological extension of the cable system. At the same time, stationed at
remote outposts around the world, as the anecdotes at the opening of this
chapter testify, the cablemen were keenly aware of the continued temporal
and geographic distance that positioned them at the fringes of empire.
As they waited for supply ships, for letters and visitors to arrive, and for
seasons to shift such that the environment would become passable, the
cablemen’s daily worlds shifted very slowly. The curfews, regulation of
activity, and patterning of social exchanges in the cable station provided a
slow and tedious rhythm of everyday life. Beyond this, a sense of slowness
was produced in their extended posts at colonial hubs: the cable staff at
any given location might remain in a given outpost for years. This strategic
choice of the system managers would in turn help to establish continuity
at the stations themselves.
Slowness, drag, and longer rhythms of movement did not inhibit the
system—they constituted it. For the cablemen this sense of delay and
stasis, perceived in their bodies and communicated through their internal
magazines, intensified the sense that the environment surrounding them
was “backward” or slow. At the same time they experienced an extraordinary
temporal privilege. This sociotechnical and affective dynamic is what I
describe as grounded speed. Today, grounded speed is the temporal substrate
of undersea digital infrastructure, and can be best perceived in the processes
of network construction, operation, and maintenance. The establishment
of new cable networks continues to take years in planning, financing, and
construction. Part of this process involves sending a survey ship to carefully
document the prospective ocean route. In installation, the cable is coiled
into a tank on the back of a cable ship, and dropped off the stern as the
vessel crosses the ocean. The cable rests on the seafloor for the duration of
its life, sometimes for decades without being disturbed. If the cable is broken,
even if the ship is deployed to the fault location immediately, it might still
take days or even weeks for the cable to be repaired, especially if the cable’s
owners are waiting to acquire permits for repair. All of these activities
take time, and the duration of this process is extended by the slowness of
marine transport, the need for careful operations, the tangled permitting
process that occurs in some territorial waters, and the fact that once on the
seafloor the cable cannot easily be retrieved. In other words, the difficulty of
negotiating with the aquatic and political environment ground the system,
making it relatively fixed once established, and discourage replacement.
In turn, the ocean itself and the lack of human contact protects the cable,
ensuring its continued speed.
Grounded speed is also present in the bodies of cable operators, though
they are no longer confined to cable stations in remote outposts. Instead,
many sit at screens in network operations centers (NOC), waiting for “alarms”
to tell them what parts of the network need to be checked or repaired. The
latest network operations technologies might release them from these
screens, enabling the delivery of alarms via text message or smart phone.
Whether in a NOC or not, the cable operators’ bodies stay at attention and
their movements are calibrated such that they can attend to the network at
a moment’s notice. Checking the network involves routine trips to a set of
locations, including cable stations and beach landings. It involves regular
maintenance and cleaning of these locations. While waiting for alarms
keeps cable operators holding focus, ready to react at all times, ongoing
maintenance is intended as a preventative measure, one that will forestall
disruption through routine practices. In both of these cases, grounded speed
exists in operation: signals keep flowing through the system at accelerated
rates because of cable operators’ repetitive motions, their familiarity with
a narrow set of routes, and their relative stasis.
Another way that grounded speed exists in cable operation is in the
relative stasis of the cable industry. Many people who work to construct
undersea cable systems, whether in supply, marine operations, management,
and even sales, have been in the industry for decades. In a recent presentation
at the SubOptic conference, the triennial event of the subsea cable industry,
analyst Kristian Nielsen described the problem of the “generation gap,”
and to prove his point, asked the members of the audience to keep their
hands raised if they had been in the cable industry for more than twenty
years. His audience, with many of their hands in the air, already understood
his point: they had worked with the same set of people since at least the
1990s. While Nielsen and others have identified this as a problem in the
industry—there are not very many younger members to take the place
of the existing cable labor force—the lack of turnover and the ongoing
stable presence of industry veterans grounds the system. Their continued
practice ensures continuity of operation, availability of knowledge, and a
vast archive of past negotiations with difficult social and environmental
actors. The system stays intact precisely because of the lack of change in
the cable world’s social fabric. Repetition (whether social or machinic),

prolongs interactions and slowness scaffolds speed.
Temporal Irruption
At the bottom of the Luzon Strait, relatively little changes. Sediment shifts
and marine creatures move, but the cables that extend through in this
narrow stretch between the Philippines and Taiwan stay in place. Cables
might shift slightly on the seafloor, but they rarely travel great distances.
The ocean is a stable and continuous context for cable systems.
In 2006, the ground suddenly shifted. The Hengchun earthquake triggered
a subsea landslide. It was not the sudden movement of the earth’s tectonic
plates that severed the cables, but the subsea landslide that sent waves of
sediment hurling down the seafloor. When accumulating slowly over time,
sediment itself does not disrupt a cable—many systems are buried at their
shore ends. But the shift in ecological temporality, the deviation from the
existing pattern of subsea currents and sediment motion, irrupted into the
temporal patterning of the cable system. Suddenly network traffic stopped
across the cable. Users at various locations suddenly could not load webpages
or send email. They were stalled, temporarily stopped by an irruption of
re-patterned deep-sea time.
Temporal irruptions are not simply the moments when a system fails. They
are indications that the ground that sustains infrastructure operation has
fallen out of sync. They are moments when the shifting temporality of other
phenomena—of ocean and atmosphere; of operators and inhabitants—alter
the rhythms of the cable network and the internet as a whole. While undersea
cables are largely secured from human interference in the ocean’s depths,
terrestrial fiber optic cables are regularly disrupted by ongoing projects of
ground disturbance, especially digging along the cable route. Local construc-
tion projects, which involve the shifting and redevelopment of architecture,
power lines, and water lines among others, often involve digging. This shift
in an otherwise stable temporal patterning likewise irrupts into the network,
causing a temporal disruption. In the ocean, storms have caused ships to put
down anchor, stabilizing themselves in an otherwise tumultuous sea, but in
turn severing cable systems. These are not mere moments of environmental
interference, they are moments of temporal disjuncture.
In moments of irruption, network time is out of sync. The slowness, the
cycles, and the patterns that had previously grounded the cable network
then prevent the network from resuming at speed. Take for example this
problem in the case of Arctic undersea cables. While cables always take
time to repair, in the Arctic frozen ice can inhibit a cable ship’s passage for a
significant part of the year. If a cable breaks during this period, one operator
reflects, “you need to sustain a fault for months sometimes.”22 The rhythm
of environmental phenomena, jarred out of sync with the network itself,
no longer serves as a continuous shelter, a ground for the system. Instead,
operators must find other ways—whether other cables or environments—to
sustain network traffic and temporal operations.
Conclusion
Over the past decades, media studies scholars have documented the temporal
regimes of network technologies, from the manipulations of infrastructural
optimization and efficiency, to the emergent microtemporalities of digital
systems, to planned social and technological obsolescences. This chapter
shows that the system of submarine cables is an infrastructure that helps to
prop up these temporal regimes, facilitating distinct and historically-spe-
cific patternings of time. It facilitates synced, standardized and universal
time, but it has also propped up varied colonial, capitalist, and globalized
temporal regimes. Looking beyond the hardware of cable systems, this
chapter’s study of network operations reveals the ways that all of these
regimes depend on the syncing up of numerous human and nonhuman
temporalities. Machinic time produced via cable systems is inevitably
interlaced with social and environmental time—the reconfiguration of
machinic time is both dependent on the organization of the temporality of
labor and seasons, and exists within its limits. Undersea cables are relatively
stable infrastructures compared to most digital systems, and this is in part
due to a multitude of extended and embodied rhythms: from the operators
standing by to the social fabric of cable landing stations and management
centers, where the goal is always stasis.
In these sites, the human–machine interface is always also a temporal
interface, where the patterned time of cable systems is enfolded in a mul-
tiplicity of other time-scales and temporal practices. If technical systems
hardwire media and social practice in a variety of ways, these systems are
nonetheless inevitably tethered to soft temporalities. Soft temporalities are
those that are variable, exist beyond the machinic, and emerge in embodied
practice. If hardwired temporalities encode temporal regimes and govern
activity, the soft temporalities of digital infrastructure comprise both its
base layer and its most vulnerable context.
Notes
1. Phillip Vannini, “In Time, Out of Time: Rhythmanalyzing Ferry Mobilities,”

Time & Society 21, no. 2 (2012): 241–269.
2. “The Gooney Clarion,” The Zodiac: The Journal of the Submarine Cable Ser-
vice, 6 (June 1912–May 1913): 61–72; “The Lonliest Station,” The Zodiac: The
Journal of the Submarine Cable Service 10 (June 1917–June 1918): 109.
3. “The Trail of Lonesome Midway,” The Zodiac: The Journal of the Submarine
Cable Service 9 (June 1916–June 1917): 116.
NC: Duke University Press, 2014).
5. Paul Virilio, Speed and Politics (Los Angeles: Semiotext(e), 2007); Jonathan
Crary, 24/7: Late Capitalism and the Ends of Sleep (London: Verso, 2013).
6. Gabriele Schabacher, “Time and Technology: The Temporalities of Care”
(this volume).
7. James W. Carey, “Technology and Ideology: The Case of the Telegraph,”
Prospects 8 (1983): 303–325.
8. Richard Stachurski, Longitude by Wire: Finding North America (Columbia:
University of South Carolina Press, 2009), 15.
9. Trudy Bell, “The ‘American Method’: The 19th-Century Telegraphic Revolu-
tion in Finding Longitude” (work-in-progress paper presented at the IEEE
Conference on the History of Telecommunications, St. John’s, NL, Canada,
Friday, July 27, 2001), 6.
10. Notably, Sandford Fleming, the Canadian who had proposed the adoption
of Universal Standard Time, was also the principal proponent for the All-
Red Line.
11. TeleGeography, “Trans-Atlantic Network Latency Reduced,” The Broadcast
Bridge, October 9, 2015, https://www.thebroadcastbridge.com/content/en-
try/3988/trans-atlantic-network-latency-reduced.
12. Seaborn, “Spread Networks and Seaborn Team Up to Provide SeaSpeedTM:
Brazil’s First Dedicated Ultra-Low Latency Subsea Route,” SeabornNetworks.
com, May 8, 2017, https://seabornnetworks.com/spread-networks-and-sea-
born-team-up-to-provide-seaspeed-brazils-first-dedicated-ultra-low-laten-
cy-subsea-route/.
13. Alasdair Wilkie, “Low Latency Cables” (paper presented at SubOptic
Conference, Paris, 2013), https://www.suboptic.org/wp-content/up-
loads/2014/10/NO02_Poster_105.pdf.
14. Wilkie, “Low Latency Cables.”
15. H. F. B, “Cablemen,” The Zodiac 9 (June 1916–June 1917): 211.
16. Crispian, “Cable (K)nights,” The Zodiac: The Cableman’s Paper 8, no. 85–96
The Zodiac: The Journal of the Submarine Cable Service, vol. 9 (June 1916–
June 1917): 44.
17. R. M. M., “The Man at the Key,” The Zodiac (March 1917): 227.
18. Sharma, In the Meantime, 7.
19. “Early Days at Southport,” State Archives, Southport Australia, Premiers

Papers Re: Pacific Cable, 6.
20. “Candid Camera … NEPTUN Version!” Underseas Cable World: A Bi-month-
ly report by U.S. Underseas Cable Corporation 1, no. 3 (Feb–March 1967).
21. George E. Clarke, “Cable Thrills,” The Zodiac: The Submarine Cable Service
Paper 17, no. 210 (1926): 316.
22. Hector Hernandez, “The Push for Polar Projects: Challenges and Remedies
Impacting the Design and Implementation of Arctic Submarine Cable
Systems” (paper presented at SubOptic Conference, New Orleans, April 11,
2019).
About the Author
Nicole Starosielski, Associate Professor of Media, Culture, and Communi-

cation at New York University, is author of The Undersea Network and Media
Hot and Cold, and co-editor of Signal Traffic: Critical Studies of Media Infra-
structure, Sustainable Media: Critical Approaches to Media and Environment,
Assembly Codes: The Logistics of Media and the “Elements” series (Duke
University Press).
10. Unruly Bodies of Code in Time
Marisa Leavitt Cohn
Abstract
This chapter examines how debates about the (im)materiality of software
comes to inhabit the practices of software engineering work who manage
the temporality of obsolescence and its entanglement with their own
careers, language prof iciencies, and expertise during the lifetimes of
systems they develop or maintain. It describes how bodies of code endure
materially in ways that exceed their formal understanding, revealing how
the hardwiring of temporality into digital systems takes place through
a moral economy of software work that devalues of code as it ages and
obsolesces. The habitus of the programmer is set within a disciplinary
regime that sustains the imaginary of software as immaterial, infinitely
flexible and malleable in spite of routine encounters with its material
recalcitrance.
Keywords: Maintenance, software work, temporality, legacy systems,

obsolescence
That which decays is not software.

This is a claim I encountered after giving a presentation of my work to
an audience of colleagues at my university that included designers, social
scientists, and engineers. I had spoken about the concept of “convivial
decay,” a term I have used to describe the achievement of engineers working
to maintain aging and obsolescing infrastructure.1 In the talk I had spoken
about the different forms of decay that engineers encounter—from instru-
ments shorting out, to programming languages reaching end of support,
to obsolescing standards for software change management. After the talk,
one of my colleagues, a software engineer who also provides IT support to
some of the university labs, explained to me that while intriguing, I had
made an error in claiming that software decays. “Software doesn’t decay,”
doi 10.5117/9789463727426_Ch10
192 Marisa Leavitt Cohn
he said, “everything but the software decays.” I understood his contention.

“Sure,” I replied, “but then, where is this software? Where does this thing
that does not decay exist?” He laughed at my question, dismissing it as a
philosophical concern best left for an afterwork conversation.
This disagreement reflects a debate between materialist and immaterialist
views on the nature of software that has animated much discussion among
scholars of computational media. We have, for example, the assertion
from Kittler in his well-known essay “There is no software,” that “all code
operations, despite their metaphoric faculties … come down to absolutely
local string manipulations and that is, I am afraid, to signifiers of voltage
differences.”2 On the other hand, we have claims that what is most essential
to the nature of software is its “operativity”3 or “executability,” an abstraction
that exists outside of the particular instantiations of hardware and software
that “coalesce” in order for the code to function. 4
Rather than stake out a position within this debate, my aim in this chapter
is to ask how this debate comes to inhabit the discourses and practices of
software engineering work. In my ethnographic fieldwork, programmers
and software engineers inevitably deal with material instantiations of
code, particularly in cases where they are working with long-lived systems
running on legacy software. Aging software systems require engineering
methods that are similarly forensic and archaeological as the methods of
media scholars unpacking historical or residual media. Yet at the same
time, these encounters with the materiality of code, while commonplace,
still seem to rupture with regularity a strongly held belief in software as
immaterial, as that which does not decay.
Materialist approaches to software are often adamantly reliant upon en-
gineering know-how. Kittler wishes to give engineering knowledge primacy
as the proving ground for theorization.5 Others in software studies have
debated whether you can theorize software without being a proficient coder.
And, as Dourish points out, when technologists work with code, the material
constraints of software systems are unavoidable. “Programmers understand
the ways in which digital structures can resist their will, every bit as much
as clay, wood, or stone. …Materiality—the nature of the substrates and the
properties that constrain and condition the designerly encounter—is at the
core of each experience.”6 If engineers know this materiality of software
so well, why does an immaterial view have such staying power? Why and
how is an immaterial view of software privileged and sustained despite the
obviousness of its materiality?
What I suggest here is there is an overlooked temporal dimension to this
tension between software as material/immaterial. As software ages, its
Unruly Bodies of Code in Time 193
materiality becomes more and more unruly, more recalcitrant. A new line
of code just written by a solitary coder/theorist might be easy to change,
but once embedded into the complexity of interoperating systems being
maintained by multiple generations of engineers, such changes become
the object of much organizational scrutiny and negotiation. Bodies of code
are increasingly bound up in the contingencies of historical organizational
decisions, material constraints of available technologies, as well as the
careers of those maintaining the code.
Through empirical material drawn from ethnographic fieldwork with a
large engineering organization, I argue that what is at stake in this tension
are the valuations of different forms of computational work over others.
The negotiations that take place in managing aging software are not only
a matter of securing computational systems from disastrous changes; they
are also a matter of how engineers manage the temporality of obsolescence
and the entanglement of their own careers, language proficiencies, and
expertise with the lifetimes of systems they develop or maintain.
If we were to track the biography of a software system over time, we might
observe an unfolding process of its materialization that mirrors in many
ways a process of decommodification described by Kopytoff.7 As software
ages it becomes increasingly “singular, unique, and unexchangeable” and
embedded into the social milieu of the organization. Those who want to
commoditize their expertise must then detach themselves from the concerns
of particular bodies of code and their accidental materialities, and align
with more universal, timeless ideals of code as immaterial. It is through this
marginalization of material concerns, I argue, that an immaterial view of
software is sustained within the moral economy of software work.
The following vignettes are taken from ethnographic fieldwork conducted
in 2010–2011 with the Cassini mission to Saturn at the Jet Propulsions Lab-
oratories. The case offers an extreme example of engineers working with
a long-lived infrastructure built to maintain and command a spacecraft
navigating in orbit around Saturn, that was built and launched in the 1990s.
While the hardware and software on-board the spacecraft is over a decade
old, so too is much of the software used to command it (known as the
“ground system”) because it must remain compatible with the spacecraft.
Nonetheless, there are plenty of pressures towards upgrade and adaptation
on the mission, due to programming languages reaching end of support,
new software management methodologies in vogue at the lab, as well as
down-sizing of personnel requiring the “leaning” of software work.
These vignettes are snapshots from within the lifetime of the mission
that capture the entangled biographies of engineering careers and bodies of
code.8 As Kopytoff notes, a cultural analysis of the biographies we tell about

objects can reveal a “moral economy that stands behind the objective.”9 We
can hear in these vignettes, evaluations of what constitutes a successful
career or a good software system. What is negotiated in the stories told to
me about the mission and its software systems are the attachments, both
economic and affective, that are formed with software over time. In the
discussion that follows the vignettes, I unpack further how time and code
are figured together within these stories, and how particular affective
attachments to code are pathologized and devalued.
Vignette 1: “Gripping the Casket”
William is a navigator who has joined the Cassini mission quite recently but
who loves the work, which he says is the most exciting work of navigation
going on at the Lab. No other mission has as much navigation work to be
done on an ongoing basis, and he associates much prestige with the fact that
mission still knows the location and trajectory of its spacecraft with such
precision and is still flying successfully after so many years. But William
also sees that many of the ways of doing navigation work at the mission
have become entrenched and is eager to get on board with new software
developments at the Lab—to harness the “power” of new software.
When I arrived and for full duration of my time in the field, the navigation
team was struggling to migrate to a new software system, and so were
running in what they called “parallel ops”—flying the spacecraft with
two software systems at the same time. These software tools are used to
determine and analyze the spacecraft’s current vector in space (i.e. where
it is and where it is currently headed) and to plan upcoming maneuvers
(for example “trimming” an orbit by firing thrusters).10 Cassini was a new
“customer” to a multi-mission software system called MONTE that had
long replaced the legacy system on all other missions at the lab. Cassini was
the only mission coming on board to the tool after launch and during the
long-term maintenance and operations phase of the work, after the team
has built up over a decade of experience with the spacecraft’s behavior and
the tools used to navigate it.
William explained that for many of the old-timers, navigators who had
worked on the Cassini mission since its launch, they could not see the
value of the new system. For one thing, he explained, there is the question
of why you would try to learn a new programming language when you are
“ten years deep” in another. He also discussed the risk of doing too much
work to adopt and adapt a new software tool. Working on software is “not
our job description” and does not have much visibility within the organi-
zation. But despite this William had decided to develop an interface with
the new software, through the creation of a new design mark-up language
(an Interface Description Language, IDL), that would accommodate the
mission’s need to run multiple analyses in parallel by adopting the same
naming conventions used to keep the various potential solutions distinct.
Over multiple meetings he tried pitch the new language to his own team
and to lab-wide meetings of navigators with the hopes that this would allow
his team to interface with the new software while maintaining older ways
of working.
William was particularly sensitive to the fact that navigation work can
change dramatically in nature depending on the software tools one uses.
He views working on software as a way of showing deep care for the mission
and the work that they do, but also a risk to one’s career. Aligning to software
work can make one indispensable if you are the only person who knows that
tool, which could be bad in the long run if some other more exciting mission
comes up. It also puts you at risk of becoming a software person and being
seen as someone to ask to write code rather than to design a mission tour.
“If anything it can be harmful for you. They hired me because I know
astronautics, they don’t care about my coding. I could become the code guy
and be downgraded to writing scripts for others. Instead of oh William is a
PhD rocket scientist, so have him design the maneuver, [it could be] oh no,
William is good for coding, put him there. So sometimes it is to your own
detriment to write good code.”
Yet for William, working with the new software was also a matter of
distancing himself from the legacy systems in operation at Cassini that are no
longer relevant. At a meeting with other navigators at the lab demonstrated
his new interface design language in a presentation on the unique challenges
faced by the Cassini mission which he outlined as a set of ten “pathologies”
within the legacy navigation database led to wasted effort.11 Through this
presentation, William works to demonstrate to his fellow navigators at the
lab that even within the context of a mission that is perceived as a “dinosaur”
he is capable of staying innovative and up to date in his ways of thinking
about navigation work. By positioning himself against the legacy software
to his broader professional network at the lab, William made sure that while
he was aligned to the mission system, he was not aligned to the software
code in the same way that his colleagues were.
His failures to get his team on board with the new design language
aside, his ability to narrate his own practice in the terms of the current
technoscientific paradigm by pathologizing the legacy software aligned him

to new disciplinary regimes, to such an extent that he was even called out
for “selling out” by one of his colleagues. Yet from his perspective, William
sees some of his colleagues as making the wrong kinds of alignments to
the legacy software, saying that they are “fearing the corpse but gripping
the casket.”
Vignette 2: Surfacing a bug
Another navigator, Chris, explained to me that he came to the Cassini

mission fresh out of his masters degree and found himself caught off guard
by the reality of his engineering work—being “thrown into the deep end of
[programming in] Unix.” Having had no introduction to it in school required
him to make a “big leap” to work in a space with so little of the program-
ming skills he needs everyday. He confessed to me that due to this lack of
experience with he “seem[s] to attract software bugs … or weird problems.”
He attributed this to the idiosyncrasies of his work, that he might have
configured something ever so slightly wrong or done something a slightly
different way. And then “all of a sudden,” he tells me, “I get a bug.” “[And all
the] old-timers are in my office staring at me wondering how could you do
that?!” He gives me an example of how one time, he was running a small test
on a file through a program in his own file space when suddenly he modified
an actual operations file that he should not have access to. “Technically I
wasn’t even supposed to have permission to edit [it]. And yet here I am doing
my own little work … And well people were starting to panic, wondering
… why is this input looking like this? how is he able to do it? Is the system
not secure? I ran a little thing on the side and suddenly I’m corrupting our
operations … and it ended up being … just a weird flaw in the legacy file.”
Chris seems simultaneously a bit abashed but also in awe of the fact that
by virtue of being new and knowing less and doing something “wrong” or
at least off the normal path he actually lead to a kind of discovery of an
anomaly hidden in the depths of the software system, something no one
knew was there. “I’m like trying to follow everything [they’re saying]. I
actually learned a lot that way. And so did everybody else.” He laughs and
explains that engineers much more experienced than him “the old timers”
were standing around his terminal saying “I didn’t know you could do this
or I didn’t know you could do that.” “And I just did it,” he adds.
In fact, everything about the way Chris explained this story came across
as a kind of confession, as if he were speaking in a way that he shouldn’t. His
voice lowered to a whisper and his tone implied both an apology for how
things work on such an old mission, and an appreciation for a system that
has been around for so long. He laughed uneasily as he told me this story
and added that when he hears about the legacy code “I hear people say ‘the
legacy code has been around for more than 40 years’” he thinks about how
it accrued over time as a kind of “patchwork” one that spans many people’s
careers, many hands having touched it over the years. And while he agreed
that the new ways of developing code, these newer programming languages
and platforms, do make code more maintainable, more evolvable, flexible
and extensible, at the same time this software being developed today, “I
wouldn’t say, I can’t say that this software will last another 40 years. Who
knows how long it is going to be relevant …” When I ask if after he surfaced
this bug in the system, were they able to fix it. “Not really, but [we could]
just be more aware of [it]. We call those ‘features.’ When it is something
that you can’t change and is just the way it is. It’s a feature. Like we have
features,” he says, as he gestures to his face.
Vignette 3: Return Tour of Duty
Reza is a software developer who began working at the lab in 1995 and
soon after began working on the Cassini mission in 1996 when they were
nearing their launch date. Like Chris, Reza took an apologetic tone when
explaining to me what he finds inspiring about working on this project.
Over the duration of his career, he had moved around a lot from mission to
mission, and it was only recently that he had been brought back in to work
on the Cassini mission. He had been tasked with jumping in to get a piece
of code working again that had broken when the person who wrote it had
left. I had asked him about what he was working on. Like Chris he could not
believe I would be interested in work he was doing with legacy code. But,
he explained, that is work he enjoys doing, picking up some code written
by other people. “Half of my career has been doing that actually, that is just
part of software development. You always enter a project. You’re seldom
aligned [so that] you’re there right at the beginning… . Or,” he continues,
laughing, “there’s always, even if you are right there right at the beginning,
some projects inherit code from other projects.”
I asked Reza what he likes about working with legacy code and he points to
a few things. First, he likes what he calls the “detective work.” “It is a matter
of going through the code and trying to decipher what was the intention,
what was he trying to [do]? What was the product he was trying to create
[and] why didn’t it get created. Because [there’s] the trail. You can see what
he was trying to do.” He also liked that working with old software held
nostalgia for him. Returning to a mission where he worked years back as a
young engineer, he encounters some of the same systems that were written
at those times. He described coming back to Cassini as a return tour of
duty, and it was in a tone of duty and care that he spoke about the mission
as one that provides a good working environment precisely because of the
care that has been put into it over so many years by so many engineers.
And it was also the “diversity” of a project where you might work one day in
something like a really “old X/Motif GUI that is not refreshing properly …
and from there I go to something like Drupal which I knew nothing about
yet… It’s just constantly changing.” In contrast, he said that working on
some of the newer missions made him feel like a programmer that is just
a replaceable cog in the system. He said the work is “cookie-cutter”—the
work broken down in a Fordist like manner making his work repetitive and
his own prior experience irrelevant.
Discussion
In these vignettes, we see how software work involves affective attachments,

as well as detachments, from ways of working with code. In William’s
story, he admonishes his colleagues for forming attachments to particular
ways of working with obsolescent software systems in ways that might be
detrimental to their careers. On the other hand, he is called out for being
too careless in how he cuts ties from old ways of working with software,
and too self-interested in how he capitalizes on a narration of the mission’s
software systems as backwards and pathological. Meanwhile, in Chris
and Reza’s accounts of the Cassini software, there is also an apology for
the forms of attachments one makes to legacy software. For Chris, as a
young engineer, he knows he is not supposed to find novelty and intrigue
in a system that is so old – a “well-oiled machine.” Like William, he hints
at a more proper and expected allegiance to the newer systems and ways
of working. He explains that the new systems are indeed better, parroting
the terms “extensible, maintainable, evolvable” in a voice that signals
ideological-speak.
Chris’s experience working with legacy systems has disrupted a particular
ideological attachment that he has encountered in the broader discipline
of software engineering, one that overvalues current systems and their
anticipated futurity. What does it mean, after all, for a system to be more
maintainable than one that has been maintained for over 40 years? Systems
are durable, not because of some attribute of the programming paradigm
in which they arise, but simply by virtue of people contributing to keep it
going. Taking this even further, Reza’s appreciation for legacy code (one
that he consistently apologized for) suggests that what is gained with newer
systems is not so much a cure-all against obsolescence, but a way to ensure
that histories are continuously truncated. If the new software is more “main-
tainable” it is in the sense that has become increasingly commoditized.
But in order for code to act as a commodity it must also be effaced of the
historicity that gives Reza pleasure in working with legacy code.
These stories reveal competing valuations of software, as well as of ways
of knowing and becoming affectively attached to software. One can care
too much for code in ways that are detrimental to career or one can be too
careerist in cutting ties from old code. A paradox presents itself in the idea
that one can be overly attached to the material concerns of software, yet
it is clear that such attachments are what keeps a system going, since the
code remaining operative relies on people who maintain proficiency and
understanding with legacy languages. It is also clear that while there are
competing valuations, there is one that dominates, inflecting the awe and
richness that both Chris and Reza see in the legacy code with apology and
irony. This speaks to the power of the disciplinary regime that governs
proper attitudes of the coding subject.
In thinking through these affective ties to old code and how they were dis-
avowed by Chris and Reza, I was reminded not only of Foucault’s discussion
of governmentality but also his essay on Technologies of the Self (Foucault
1988) in which he asks “through what operations we work upon our bodies
and souls, thoughts, and conduct … so as to transform themselves in order
to attain a certain state of … purity, wisdom, perfection, or immortality.”
While Foucault is talking about the techniques applied to the body, we can
easily extrapolate to techniques applied to the body of code, as I have done
so here by adapting his questions for my purpose:
– If one wants code to behave rationally and if we want to regulate code
according to certain principles of what makes code rational, what part
of code or ones work of coding should one renounce?
– How do individuals effect by their own means or with others a set of
operations on the body of code, on their conduct towards the code, so
as to transform the code into a state of perfection?
These questions apply quite readily to the context of software development

as seen in the vignettes above. They help to highlight two modes of knowing
and attaching to code. The mode of innovation in which one does not align
too closely to any aging or obsolescent system. And the mode of maintenance
and care in which one understands code materially through its longevity
and an attachment to the past as a source of insight. In the first code is
commodified, current, and promises longevity through a renunciation of
obsolescence. In the latter, the code itself is a rich tapestry or patchwork
that binds together many different careers that sustain the body of code’s
longevity – that is a way of knowing code that is rooted in continuity rather
than in a regime of anticipation and futurity.12
Others have pointed out that software is bound up with philosophical
commitments and morality.13 What I have highlighted here is how the moral
economy of software work also applies to its aging and obsolescence. As
John Durham Peters states, “obsolescence always raises moral questions
about the subjects and objects we neglect”14 . In the aging and obsolescence
of software, time and code are configured together. In long-lived systems,
particular temporalities of work must be maintained in order for the system
to remain vital, and likewise a system can “fail” for lack of those who know
how to program in older languages. At the same time, valuations of software
also shift as it ages. Old software resists commodification, displays too much
personality, bugs becoming features. In software engineering communities,
legacy is considered a derogatory word, referring to code that has stuck
around too long and become heavy. Old software is pathologized for being
mired in the past, and those who care too much for it are as well. Even the
terms used in software engineering to manage aging systems like rot and
grime15suggests a relation to the abject. At the same time, newer systems
and methods are adopted with a rhetorical promise of eternal youth, as the
solution that will never age.
A materialist approach to software contends that the “trope of imma-
teriality” is both analytically weak, smoothing over technical complexity,
and ideological in suggesting that digital systems liberate us from the
historical and material contingencies of other media.16 Drawing on Hayles,
Blanchette suggests that this trope is part of an ideological project that
has as its underlying “fundamental assumption, that informational pat-
terns (including human consciousness) are ontologically superior to their
(accidental) material instantiations (including the human body).”17 This
same ideology inhabits software programming work, in part because it is
a feature of obsolescence that an engineer can liberate himself from the
contingencies of the at-times arbitrary material conditions of the past by
choosing to write new code rather than maintain the old. It is thus also
the habitus of the programmer set within these disciplinary regimes that
forms part of the mundane apparatus sustaining the imaginary of software

as immaterial, infinitely flexible and malleable.
Conclusion
Ernst has claimed that “temporal transcendence of materiality is a faculty of

operative media” in the sense that there is such a “tight coupling” between
elements of hardware and the material knowledge of them, that they act as
a kind of time machine. When technologies are reactivated at a later time
they might require modification to get them working, but this is done in
such a way that “unleashes” the “latent presence” of their operations.18 What
he points to is the way that the “operation” itself endures in ways that the
material instantiations do not. This makes sense for understanding the
transience of software’s material instantiations and how older systems
become a repository of memory of older ways of working (as they did, for
example for Reza). However, this fails to make sense of how bodies of code
endure materially in ways that exceed their formal understanding.
As software ages, it can become no longer executable because of some
perturbation in the many interdependencies of a body of code and the
people who operate it. A person moves off the project, or a subsystem fails,
and the organizational memory is no longer there, if it ever was, to hold the
system together. This suggests that legacy software might become unable
to temporally transcend its materiality simply because the tight coupling
of hardware and knowledge has been lost. But it also, I think, betrays a
presentist bias in how we philosophize the nature of software that leaves
aside the problem of temporalities of software’s aging and obsolescence.
That is, even when we examine the historicity of computational media,
our analysis privileges how this historicity is made manifest in the present.
What about the latency of a bug in 50-year old software, as an aspect of its
(in)operativity, that is only revealed over the long lifetime of software. The
bug, which might also become a feature, is a constitutive part of software
as both a material and formal object.
As I have hoped to show, the hardwiring of temporality in digital systems,
is at least in part performed through the valuations of software as it ages.
Such valuations arise not only with regard to the performativity or execut-
ability of code, but also in regard to the multiple biographies of any body of
code, the many lives and careers it has touched or sustained. In a culture
where as code ages and becomes more mired in material concerns, the work
of maintaining it is devalued and even pathologized, privileges accrue to
the software yet to be written. This ethos shapes the attachments and moral
commitments of engineers to competing valuations of maintenance and
innovation. As in rubbish theory, legacy code is that which is not yet thrown
away but is durable despite its devaluation and troubles the moral economy
of software work.19 It is in this duration of unruly bodies of code in time
that the ideology of “immateriality” lives.
Notes
I would like to thank Axel Volmar and Kyle Stine for organizing the series of
workshops on Hardwired Temporalities as well as the participants in those
workshops for their feedback on this piece, particularly Gabriele Schabach-
er, Geoffrey Bowker, and Alexander Monea for their in-depth comments.
I also appreciate input from Nanna Thylstrup and Mace Ojala who signifi-
cantly advanced my thinking in later drafts.
1. See Marisa Leavitt Cohn, “Convivial Decay: Entangled Lifetimes in a Geri-
atric Infrastructure,” in Proceedings of the 19th ACM Conference on Comput-
er-Supported Cooperative Work and Social Computing (ACM, 2016), 1511–1523.
2. Friedrich Kittler, “There Is No Software,” Stanford Literature Review 9, no. 1
(Spring 1992): 84.
3. Wolfgang Ernst, “Micro-Dramaturgical Temporalities of Media Theatre: On the
Difference Between Performative and Operative Reenactment,” in Performing
Arts in Transition: Moving between Media (New York: Routledge, 2019), 55–68.
4. Rob Kitchin and Martin Dodge, Code/Space: Software and Everyday Life
5. Friedrich Kittler, “Real Time Analysis, Time Axis Manipulation,” Public
Culture 13, no. 1 (2017): 1–18.
6. Paul Dourish, The Stuff of Bits: An Essay on the Materialities of Information
7. Igor Kopytoff, “The Cultural Biography of Things: Commoditization as
Process,” The Social Life of Things: Commodities in Cultural Perspective 68
(1986): 64–91.
8. Steven J. Jackson, David Ribes, Ayse Buyuktur and Geoffrey C. Bowker,
“Collaborative Rhythm: Temporal Dissonance and Alignment in Collabora-
tive Scientific Work,” in Proceedings of the ACM 2011 Conference on Computer
Supported Cooperative Work (ACM 2011), 245–254.
9. Kopytoff, “Cultural Biography of Things,” 64.
10. There are many reasons that the navigation team found it difficult to
switch to the new software system. In contrast to the legacy code, which
was developed “in house” at the Cassini mission, adapted from earlier
missions to suit Cassini’s needs, the new software system was developed by
a new multi-mission software team at the lab that maintains a code base
and unit tests to keep the software working for multiple missions at once.
While other missions can provide requirements for the system and work
in advance of launch to test out the software, Cassini was the only mission
expected to change to the new software mid-flight. Not only was the new
software producing different results from the legacy software (literally plot-
ting different vectors for the spacecraft’s current trajectory), but each time
the new software was successfully tweaked to produce results for Cassini’s
navigation team, some new update to the software would break the team’s
in-house tools. Many of the initial issues discovered in transitioning to the
new software came down to the material instantiation of the code. For
example, the new software was written in Python, which stores variables
differently than the legacy code, in terms of how it handles truncation of
decimals. These types of fixes were relatively easy to deduce, but never fully
brought the tools into perfect congruence. The new software also ran very
slowly, something that is less of an issue for missions that design one set
of navigation maneuvers and parameters for a successful launch and entry
into a planet or stable orbit. The Saturn mission, however, had to consider
multiple maneuvers to execute on a weekly basis.
11. Many of these so-called pathologies came from inheritances from even ear-
lier missions that had done most of their navigation calculations by hand.
The legacy software draws on what was originally a digitized version of
paper files, which unlike a modern database did not have a clear separation
between form, syntax, and content. What used to be considered a “view” of
the database (a print out) is what is now treated as data.
12. Adrian Mackenzie, “Programming Subjects in the Regime of Anticipation:
Software Studies and Subjectivity,” Subjectivity 6, no. 4 (2013): 391–405.
13. Paul Dourish, Where the Action Is: The Foundations of Embodied Interac-
tion (Cambridge, MA: MIT Press, 2004); Christopher Kelty, “Geeks, Social
Imaginaries, and Recursive Publics,” Cultural Anthropology 20, no. 2 (2005):
185–214; E. Gabriella Coleman, Coding Freedom: The Ethics and Aesthetics of
Hacking (Princeton, NJ: Princeton University Press, 2012).
14. John Durham Peters, “Proliferation and Obsolescence of the Historical Re-
cord in the Digital Era,” in Cultures of Obsolescence. History, Materiality, and
the Digital Age, ed. Babette B. Tischleder and Sarah Wasserman (New York:
Palgrave Macmillan, 2015), 79–96.
15. Clemente Izurieta and James M. Bieman, “A Multiple Case Study of Design
Pattern Decay, Grime, and Rot in Evolving Software Systems,” Software Qual-
ity Journal 21, no. 2 (2013): 289–323.
16. Jean-Francois Blanchette, “A Material History of Bits,” Journal of the Amer-
ican Society for Information Science and Technology 62, no. 6 (2011): 1042–
1057, https://doi.org/10.1002/asi.21542.
17. Blanchette, “Material History of Bits,” 1044.
18. Ernst, “Micro-Dramaturgical Temporalities of Media Theatre.”
19. See Michael Thompson, “Time’s Square: Deriving Cultural Theory from
Rubbish Theory,” Innovation 16, no. 4 (2003): 319–330.
About the Author
Marisa Leavitt Cohn is an Associate Professor in the Technologies in Practice

group at the IT University of Copenhagen where she co-directs the ETHOS
Lab. She combines anthropological and design-oriented approaches to the
study of human computer interaction focusing on temporalities of work in
long-lived infrastructures and legacy system maintenance.
11. Screwed: Anxiety and the Digital
Ends of Anticipation
James J. Hodge
Abstract
This essay discusses anxiety as the paradigmatic malady of the present
and its correlation with the rise of always-on computing. Discussing
anxiety as “an expectation emotion,” the essay notes the ways in which
always-on computing has outsourced futurity to opaque computational
processes. This essay asserts that the latter fuels the former. However,
all is not lost. Through the analyses of three online texts—an artist’s
clock, a meme, and a YouTube video—the essay argues for a view of this
situation that recognizes the pleasurable and resolutely social dimensions
of anxiety. Discussion of these texts elucidates the dynamics of anxiety
in always-on computing and challenges the assumption of anxiety as an
individual problem.
Keywords: anxiety, new media art, social media, memes, death,

phenomenology
This essay is about anxiety and how we deal with it. It’s about anxiety as
a felt problem or condition specific to the historical present of post-indus-
trial cultures in the twenty-first century, which are characterized by the
saturation of lived experience by always-on computing, or the milieu of
smartphones, social media, and ubiquitous wireless networks. Anxiety, we
often hear, seems to be the distinctive malady of always-on computing.1 Of
course, anxiety is not new. Yet as the activist collective called the Institute
for Precarious Consciousness (IPC) theorize, anxiety takes on a newly
general character today arising from neoliberalism’s “obligation to be com-
municable.”2 To build on this formulation, always-on networks constitute
the infrastructural conditions of possibility for this “obligation,” where
doi 10.5117/9789463727426_Ch11
206 James J. Hodge
the obligation to be communicable involves a felt pressure and pleasure

of being available or “always on” to networks by virtue of the sheer fact of
carrying a smartphone and being enmeshed in networks. This essay, then,
seeks to elaborate the relation of anxiety as an “expectation emotion” to
always-on computing, especially for the ways it follows from the outsourcing
of future-oriented experience (prediction, anticipation, etc.) to machines
operating beyond lived experience.3
The essay also follows from a second issue identified by the IPC. Anxiety
is almost always viewed as a personal or individual problem. In seeking
to combat the deleterious and deflating effects of anxiety today, the IPC
recommends understanding anxiety as a social phenomenon. I heartily
agree with this prescription. To contribute to this project, I want to describe
and theorize not only the ways in which anxiety correlates with always-on
computing, but also the ways we are already managing anxiety as a collective
problem. I argue that the task of describing anxiety as a major dimension
of the lived experience of always-on computing requires attending to the
circulation of anxious objects, which generate a palpable sense of anxious
fun. Understood as networked aesthetic objects to be shared, they trouble
any strictly individual notion of anxiety and open up the social dimensions
of anxiety today. My discussion analyzes three anxiously fun objects: a clock
designed by an artist, a meme, and a viral video.
Start the clock.
Available on the online gift shop for Chicago’s Museum of Contemporary
Art at the discounted rate of $70.98 (marked down from $285!), designer
Brian Eaton’s Memento Mori Clock looks like a handsome, walnut-encased
alarm clock of modernist yesteryear. Its red and black digital display does
not tell the time, however, but rather how much time you have left. An
explanatory text reads,
Memento Mori serves as a reminder that all things must come to an

end. Upon turning the clock on, you are prompted to enter some basic
information about yourself. From there it pulls data from the World Health
Organization to establish a countdown, in minutes, to your inevitable
demise. At zero the clock will commemorate your life by playing a song, as
a final reminder of your time here on Earth. This song can only be heard
by those that make it to zero, as it’s impossible to cheat the countdown. 4
Is this a joke? While conceptually intriguing it’s difficult to say who would
be willing to participate in a countdown to her own death (maybe that’s
why it’s available for 75% off?). When I have shown people this item they
Screwed: Anxie t y and the Digital Ends of Anticipation 207
typically gasp, laugh, and exclaim along the lines of, “oh my god!” When I
half-seriously suggest that I am thinking of buying one, their tone quickly
turns serious. I must not buy one, they insist, sometimes even demanding
that I promise not to buy it. The very idea of this clock, they say, makes
them intensely anxious.
Why do my friends so discourage me from buying this clock? They say
the idea of the clock makes them anxious but I really think it’s the idea of
my having it. What difference would that make? Having or owning a clock
might do a few slightly uncomfortable things. Owning it might risk taking
on the anxiety the clock generates as my own. In effect, it might become an
extension and self-serving ratification of my anxiety, the anxiety that helps
to make me me and distinguish myself from others. I hope I don’t cheapen
the felt impact of anxiety when I note that anxiety today sometimes appears
as rather precious—distinguishing even as it distances. This certainly feels
like part of the problem. Relatedly, owning the clock may diminish the clock’s
power to produce a social bond instanced by the common exclamation of
“oh my god!” elicited precisely by sharing its image over text message, email,
social media, etc. Ownership risks taking it out of circulation as a way to
feel anxiety as a social feeling. I believe that is the real danger my friends
object to when they demand that I not buy it. If I owned this clock, I don’t
think I could really share it again, at least not in the same way as I do when
I don’t actually own it but just want to say, oh my god look at this. Put more
simply, the stakes I see this object opening up have to do with the nature
of togetherness set against the background of networked anxiety.
Let’s say we listen to our friends. Don’t buy the Memento Mori Clock.
Don’t plug it in next to the bed. The trouble is—at least in the semi-afflu-
ent portions of post-industrial Western culture—we’ve already by and
large bought the idea of this clock. The main issue is that the idea that
time—especially the time of anticipation, the time of the future—has
effectively been handed over from human perceivers to networked digital
infrastructures understood not merely as technology but rather as “the
living mediation of what organizes life.”5 Anticipation today is no longer the
work of human beings but rather it is the province of technical processes.
These processes largely operate beyond human experience even as they
have become normalized and ordinary. Following Anaïs Nony, we might say
that the Memento Mori Clock substitutes an operative logic of preemption
for one of anticipation. Nony writes, “Whereas anticipating is caring for
what could come next, preemption is the implementation of one single
possibility in the present, and the simultaneous reduction of a virtual
and potential future to a single line of interpretation.”6 Digital technics
208 James J. Hodge
transform the very nature of the future by substituting a machinic logic

of preemption for a human experience of care. It is difficult to argue with
her. Since World War II technology has been conceived as taking over the
human job of anticipation.7 Norbert Wiener’s insight that machines might
more effectively anticipate the flight patterns of enemy aircraft effectively
gave rise to cybernetics, the science of communication and control so vital
to the development of digital technologies.8 More recently, media theorists
such as Brian Massumi, Richard Grusin, and Mark B. N. Hansen have all
reckoned with the inheritance of this idea in the context of the pervasive
media environments of the twenty-first century.9 While these accounts focus
on the political and philosophical dimensions of technological futurity, I
want to emphasize the ordinariness of what Nony calls digital preemption.
From Google’s autofill feature and predictive text in text messaging appli-
cations to machine learning approaches to online discussion moderation
and Big Data-driven approaches to marketing (predictive analytics) and
law enforcement (predictive policing), a variety of digital technological
applications and approaches from everyday life demonstrate how widespread
this tendency has become.10
Over the last few decades the digital infrastructure of contemporary lived
experience has outsourced the human work of anticipation and care. During
the same period anxiety has arisen as the paradigmatic mental health issue
of the age. In the United States of America alone, the National Institute of
Mental Health estimates 40 million adults—or 18 percent—over 18 years
of age suffer from an anxiety disorder.11 This number is alarmingly high,
and it doesn’t even account for undiagnosed, undiagnosable, or just more
common or low-level forms of anxiety pervasive in post-industrial cultures.
As many have noted, anxiety runs rampant in contemporary culture and
has a marked if still-poorly understood relation to always-on computing.12
While I do not mean to offer a comprehensive explanation for this state
of affairs, I believe there is a deep connection between the technological
outsourcing of anticipation and the rise of anxiety. Following Ernst Bloch
as well as the DSM-V, anxiety is fundamentally a flawed relation to the
future; anxiety is anticipation gone wrong.13 Following Sigmund Freud’s
classic distinction, fear has something to fear but anxiety has no object.
Anxiety has nothing toward which to put itself in relation or to orient itself.
One overwhelming problem with our relation to the future today is that we
have increasingly outsourced futurity to digital machines. In Nony’s words
a society of preemption has replaced one of anticipation. Yet this does not
mean that can stop anticipating or seeking a relation to the future. The
objects orienting our relation to the future have, however, by and large
vanished into the opacity of technical processes subtending always-on

culture. Our relation to the future has fewer and fewer perceptible objects. It
is no wonder then that, as the IPC succinctly states, “we are all very anxious!”
In order to specify anxiety as well as the condition of anxiety in the
milieu of always-on computing, consider what Martin Heidegger calls the
nothing and nowhere of anxiety. For Heidegger, like Freud, anxiety has
no object. One fears something. But one is anxious in relation to a sort of
nothing or nowhere. To recall the Heideggerian tone of Nony’s argument,
anticipation involves “caring.” In Heidegger’s phenomenological imagination,
care [sorgen] captures the fundamental sense in which one orients oneself
in relation to the world, e.g. in intending, attending, taking up, seeing to,
etc. The challenge of anxiety is that it frustrates care as orientation or
comportment. Heidegger writes,
Accordingly, when something threatening brings itself close, anxiety

does not “see” any definite “here” or “yonder” from which it comes. That
in the face of which one has anxiety is characterized by the fact that what
threatens is nowhere. … that which threatens cannot bring itself close
from a definite direction within what is close by; it is already “there,” and
yet nowhere; it is so close that it is oppressive and stifles one’s breath,
and yet it is nowhere.
In that in the face of which one has anxiety, the ‘It is nothing and nowhere’
become manifest.14
For Heidegger, anxiety seems paradoxically proximate while yet threatening

from a distance. Anxiety has no precise location or even any embodied
specificity. He even puts quotes over the verb “see” to indicate how much
anxiety evades any specifically modal perceptual apprehension. Still, as
Heidegger emphasizes, the nowhere of anxiety takes effect in the mysterious
interface of body and world. It “stifles one’s breath, and yet it is nowhere.”
It is difficult to imagine a better description of the anxiogenic effects of
the internet!
It may seem paradoxical that a technological milieu so seemingly “ready
to hand” as always-on computing might generate so much anxiety, which
I’m specifying here an ambient frustration of care. Yet while devices such
as smartphones, wearables, and laptops may appear readily, the very thing
that defines their power—the network—retains its status as a kind of
nothing or nowhere. To be sure, scholars may direct our attention to undersea
cables, server farms, or other undeniably material forms of evidence for
the network’s existence as technological infrastructure.15 All the same, the
210 James J. Hodge
network remains something one simply can’t point at with satisfaction.

As I have written elsewhere, the network exists experientially as a “felt
relation of non-relation,” a sort of amodal pressure we feel in various ways
from phantom cell phone vibrations to the agony of the typing awareness
indicator and the progress bar.16
Always-on computing has no single experiential correlate, but anxiety
is certainly one of its most profound.17 The prominence of anxiety as a sort
of constantly felt “nothing” or “nowhere” certainly has much to do with the
experientially opaque nature of the infrastructure supporting always-on
computing in lived experience. However, the trouble is that we know very
well that this amodal “nothing” is constantly doing things out of sight that
nonetheless impact our lives. Privacy worries typically rule discussions
of this sort of below-the-radar-of-the-user activity in the form of NSA
dataveillance and data tracking by third-party software on smartphones.
These out-of-sight phenomena include the activity of algorithmic forces
with designs of their own targeting our dividual or data selves, i.e. our
smartphone-related behaviors from GPS data to shopping habits.18 As many
critics note, digital media do not address liberal human subjects, the sense
of personhood I have when I think about myself as “me.”19 This does not
mean, however, that the felt nothing of algorithmic addresses to my data
“self” does not have any effect on my liberal self. It certainly does. And the
most general result of this state of affairs is a general rise in anxiety against
the networked “nothing” and its aggressively opaque operations.
Anxiety names a felt non-relation to some sort of radically overdetermined
“nothing.” The name for this nothing today is the network. For Heidegger,
as for many other philosophers of anxiety, that nothing has do with death.
Importantly, death here is not an “object” toward which one feels anxiety
but rather what arouses anxiety. Death is instead a placeholder, an avatar,
or way of giving a name to something that can’t properly be named, a kind
of nothing or nowhere whose felt non-presence generates anxiety. It feels
only natural, then, that death would figure plainly in all three anxious
objects I discuss in this essay. Death also, of course, names a relation to
the future. That future relation is both radically overdetermined (everyone
dies) but also characteristically indeterminate (nobody knows what death
is). To depart gently from the philosophical seriousness of anxiety and
death, let us return to the web’s curious scrambling of these ideas in the
strange feeling of anxious pleasure occasioned by Memento Mori Clock,
which may be grasped as symptom of our uncertain relation to living newly
overdetermined lives governed by digital technics operating beyond the
scope of human experience.
Memento Mori Clock’s humor can’t be easily explained. On the one hand,
it’s funny. It must be. I mean, who takes seriously the ghoulishly over-serious
and foreboding declaration that “it’s impossible to cheat the countdown”?
On the other hand, maybe everyone? From Holbein to Hamlet to Game
of Thrones’ valar morghulis (“all men must die”), it is difficult to deny the
essential truth of memento mori (remember that you must die). All this,
however, seems too serious, too existential for an object resembling an
alarm clock (and for anything having to do with internet culture for that
matter). And yet, it feels newly revealing to note that the internet vernacular
in English for expressing laughter is “I’m dying” or simply “dead.” I can’t take
this seriously, can I? Put otherwise, I honestly can’t tell if this is a joke, or if
the joke is funny.20 The fact that I can’t quite tell certainly has something to
do with the internet and its ambivalent tone, which frequently renders humor
ambiguous.21 It also may have something to do with comedy’s disconcerting
ability to produce and dispel anxiety at the same time.22 Let’s hold onto
this impulse to not take things too seriously (even if we need also to take
things seriously!).
Memento Mori Clock certainly solicits anxiety but it also, at least briefly
and partially, sweeps anxiety away with a laugh. Such anxious ambiguity, in
turn, hinges on our relation to larger techno-cultural dynamics exemplified
by Memento Mori Clock, namely: the question of our faith in the capacity of
digital processes to determine future realities in advance. Will the number
displayed reliably count down to my death? I hope not, but also I hope so?
My hoping not is, of course, bound up with my desire not to die or even
to address such unpleasantness. But I hope so because I realize that as
much as I may not like it I actually do have a lot of faith in the power of
digital technologies to anticipate the future on my behalf. In some ways, we
simply must live our lives according to a certain faith in digital prediction.
Policy decisions based on climate change projections are based on just
such predictions, for example. In more everyday circumstances, many of
us depend on the predictions of traffic patterns calculated by GPS apps like
Google Maps or Waze. Yet something feels amiss when we put our faith too
blindly in the capacity of machines to anticipate the world for us on the basis
of something as impersonal as data-mining—especially when that world
comes into focus in something as singular and personal as the moment of
my own death. My death will probably not happen at the same time as the
timeline calculated by the clock drawing on data from the World Health
Organization—right? Although I can reasonably expect these two times to
diverge it’s unnerving to imagine the time of my estimated demise ticking
down. I know one thing—or probably—and I feel something very different.
212 James J. Hodge
I know and laugh ha! At the same time I also feel anxious and laugh haha.
At the root, then, the problem provoked by the ambiguous comedic nature
of Memento Mori Clock concerns not any single instance of preemption but
rather the problem of its massively general character grasped in relation
to the singularity of human life. We have become dependent on machine
thinking for a number of reasons. One of the consequences of this increasing
dependence is that it is becoming more and more jarring to see ourselves and
the very human narrative trajectories of our lives mapped onto algorithmic
predictions. We’re caught in a twenty-first century bind. We can’t tell what
to think because we can’t anticipate. We can’t anticipate because we have
outsourced anticipation to digital machines. This is where things get anxious.
But crucially, it’s also where things get weirdly fun.
The very word “anxiety” seems overwhelmingly negative, or even dis-
abling, as in the expression “crippling anxiety.” Yet always-on computing,
contemporary media by its very nature intermingles anxiety and pleasure.
By doing so it may not turn the “I can’t” of anxiety into a rousing “I can.” All
the same it meaningfully dislodges agency from the death-grip of, well, death.
The nothing of the network arouses anxiety but crucially it is not merely
anxiogenic. Networks also allow for new forms of sharing anxiety, or for
understanding anxiety as relational rather than as purely isolating. Because
sharing often simply feels good, this dynamic opens the door for re-artic-
ulating anxiety as simultaneously social and pleasurable. Memento Mori
Clock bears this out. It arouses anxiety in relation to the overdetermining
nature of always-on networks, but there’s also something about it driving
me to share it. Not because I want to spread my experience of anxiety, but
because encountering it together feels good. I don’t just want to tell friends
about it, I want to be there when they realize what it is and how it works. I
want to laugh again at this object, something I can’t do again by myself. I
want to experience it with someone else. I take pleasure in witnessing and
partially sharing the building no-no-no! realization I can read in the face
of my friend of just how dreadfully anxiety-inducing this terrible clock is.
Here it bears noting that the Memento Mori Clock is no outlier. Its dy-
namics recall other more widely known internet phenomena. Consider
the famous “This is Fine” meme. In that famous six-panel cartoon from
2013—reduced to two panels in the 2014 meme—a wall-eyed cartoon dog
in a bowler hat sits at a table with a cup of coffee in a room that is on fire.23
Smiling, certainly aware of his impending demise the dog states, “this is
fine.” Like Memento Mori Clock, “This is Fine” expresses a sense of anxious
fun. The world burns down around the dog, and he just smiles in the face of
it all. The meme depicts a paralyzed, dumb, knowing, and also cute version
Figure 11.1. Gymnastics scene from Final Destination 5 as seen in the YouTube video I Put Wii Music
Over a Final Destination Death Scene.
of contemporary experience today, where the world is often said to be

coming to an end or to resemble a dumpster fire. The meme represents the
anti-agentive anticipation of disaster in the face of forces outside individual
control. It presents an image of anxiety. Yet it also modulates or manages
that anxiety through laughter, or the fun of sharing the meme. Memento
Mori Clock, then, might be thought of as a more extreme version of “This
is Fine.” Like “This is Fine,” it expresses a sense of anxiety in the context
of digital networks. Both thrive on the possibility of constant circulation.
And both threaten to act like a declarative statement, the rough equivalent
of ending a text message with a period. But circulation keeps alive their
complex affective resonances of anxious fun while both continue as a kind
of network hot potato because each carries within itself an anxious image
of the end of circulation: conflagration and death.
In the last section of this essay I want to discuss one final example: a video
entitled I Put Wii Music Over a Final Destination Death Scene. Recommended
to me by YouTube’s algorithmic anticipation of my viewing pleasures and
posted on September 1, 2017 by a Belgian user named MATN, the video
currently has over 8 million views about one year later. This three-and-a-half-
minute video overlays music from the Nintendo Wii videogame platform’s
Mii Channel Plaza on top of a clip of a gymnastics routine gone horribly
wrong from the 2011 horror film Final Destination 5. Taking a close look at
this video will help to elaborate the peculiarly social pleasures of anxiety
in the milieu of always-on computing.
The lived experience of always-on computing involves a strangely par-
adoxical sense of freedom by design. The great achievement of this video’s
ordinary contribution to social media lies with the ingenious way in which it
formally dramatizes the dynamic I’ve been attempting to elucidate so far: the
214 James J. Hodge
solicitation of anxiety in always-on computing through the combination of

an indeterminate posture toward the world (being always on) encountering
the radically overdetermining processes and infrastructures governing that
make such “indeterminancy” possible in the first place. In this way, the
video represents a valuable text for recognizing the pleasurable and social
dimensions of anxiety. It does so in its mash-up of light-hearted videogame
aesthetics and horror cinema. The combination of music from the Nintendo
Wii with footage from Final Destination 5 makes unavoidable the algorithmic
nature of death in these movies.
Because this is a clip from a Final Destination movie we know that some-
one is going to die. The question of who will die, however, is less important
than the certain expectation of death. And sure enough, after running
through a number of gymnastics exercises—and after the camera teases
the viewer with a number of possibilities for how she will die—the gymnast
eventually meets a grisly end at the culmination of a Rube Goldberg-esque
chain of improbable events (another gymnast on a balance beam steps on a
fallen screw leading her to fall off and tip over a container of chalk, which a
fan then blows into the face of the main gymnast as she spins on the uneven
bars, blinding her during her dismount, and leading her to land so badly
that her spine snaps on landing as her body doubles back over her head).
In these remarkably cynical movies, Eugenie Brinkema observes, death is
not so much a figure as a function of design. Deaths in a Final Destination
movie do not add pathos to the psychologization of a character; instead
they fulfill a formal contract that might without exaggeration be termed
algorithmic. Deaths in a Final Destination are executions in the twin sense
of killing and “computational” in the sense of being executable from the
point of view of franchise logic. They are also “algorithmic” not merely in the
sense of being based in a set of rules reacting to variable conditions. Final
Destination deaths are algorithmic also in the sense that they outsource any
agential sense of anticipation to the rules themselves as processes governing
those same agents and their lifeworld. We don’t care about the characters’
fates. Because the characters are properly algorithmic we actually can’t.
Yet because the characters visibly appear to us they help us to grasp better
what it is we actually care about.
I wouldn’t call the film clip from Final Destination 5 by itself especially
anxiety-inducing. If you’re watching Final Destination, then you’ve signed
up to enjoy the elaborate execution of its characters. Suspenseful? Yes.
Anxiogenic? Probably not. The clip becomes an engine of anxiety, however,
when taken out of the context of the movie and overlaid with music from
the Wii platform. The light-hearted tone of the repeating melody contrasts
humorously with the gruesome mock-serious-but-also-somehow-genuine-

ly-serious tone of Final Destination 5. The contrast between the image and the
sound tracks makes the clip bearable in a very different way, and also, quite
differently enjoyable. In a way, the music opens the possibility for caring in
some way about the characters that Final Destination 5 on its own precludes.
Instead of just making the clip seem absurd and therefore funny, however,
the addition of Wii music in the context of YouTube draws out and makes
more literal and unavoidable the networked and “algorithmic” character of
Final Destination. As benign as the music sounds, it also evokes the digital
genre where one’s character almost always dies as a formal function of play.
The incorporation of a videogame imagination expands Final Destination’s
algorithmic aesthetic of overdetermined design by introducing some sense
of play in the sense of playability, or “room for play.”24 The exhibition of this
video on YouTube, moreover, further underscores the video as an ordinary
yet paradigmatic object of always-on computing, its status as an aspirational
meme, a text ripe for re-mixing and re-circulation. This next “algorithmic”
layer of signification underscores and clarifies the social pleasures it affords.
But still, what of anxiety?
When I show the video to friends it makes them squirm. I admit this
gives me pleasure. Their reported experience of anxiety, however, has very
little to do with death. The mashed-up specter of an especially algorithmic
form of death merely sets the scene for anxiety. Anxiety’s true emergence
as a socially pleasurable effect here has everything, however, to do with the
question of whether a gymnast will step on the fallen screw lying pointy
end up on the balance beam. Importantly, nobody gasps (at least not as
much) when the gymnast actually does step on the screw. Yet every time
the video features a close up of a heel arched over the screw my friends
audibly breathe in through clenched teeth. Two quotes from the top of the
comments section of this video seem especially relevant here: “a n x i e t y”
and “why did i only care about the nail that fell.” Indeed. Why care about
a small puncture wound when certain death is much worse? To follow
Heidegger, this makes total sense because one can’t actually can’t “care”
about death. One can, however, very much care about possible pain.
In describing anxiety Heidegger characterizes the failure of comporting
oneself toward its “nothing” and “nowhere” as stifled breath. Here I’m
guessing he has something like suffocation in mind, especially given the
way death arouses anxiety in his philosophical project. The idea of stifled
breath recalls for me my friends’ reaction to the flaunted possibility that
the gymnast might step on the screw. To be clear, their sharp intake of
breath is not only the most concrete evidence of the experience of anxiety
216 James J. Hodge
here, it is also both a little in the direction of sympathy but also in the not
wholly-contradictory valence of laughter. These reactions may seem like
opposites but they each suggest a glitchy frustration in our bodily inability
to synch up with or orient oneself in relation to the world. This intake of
breath strikes me as important, then, not only for the ways it recognizes
anxiety as an ambivalent social mode of aesthetic response generated by
this paradigm of ordinary anxiety in the context of always-on computing. It
matters that this experience of anxiety takes place, then, not just in reaction
to the anticipated possibility of a small puncture in the flesh, but also as a
reaction organized in the simultaneously indeterminate and overdetermined
thematics of Final Destination 5 and Wii music on YouTube.
Let us say just a bit more about death. Death is justly famous as a figure
that arouses anxiety. Try as we might we can’t do anything with death.
We die; there’s not much more to it than that. That futility only reinforces
death’s association with anxiety. The singularity of death, of my death,
further reinforces anxiety as an individualizing, isolating force. The amazing
thing about I Put Wii Music is that while the specter of death’s algorithmic
inevitability arouses anxiety, the felt anticipation of that screw sticking
suddenly into a naked foot effectively displaces death and gives a new
quasi-formless form of anticipation to anxiety in always-on computing. The
imagined ouch! of a screw in a foot gives a new and very different sense of
the possible than the specter of death. Namely, we can share this imagined
ouch in a way that we can’t share the specter of death. The possibility of this
event gets to conceptual core of anxiety. Renata Salecl emphasizes precisely
the way in which anxiety thrives on the experience of the possibility of
possibility. She writes,
When Kierkegaard analysed anxiety, he took it to be something that is

linked to possibility in existence. Here anxiety became specifically linked
to freedom, or as Kirkegaard says, it is linked to freedom’s actuality as
the possibility of possibility. … Anxiety is thus linked to the possibility of
being able, but, as such, it often appears as a feeling aroused by looking
down into a yawning abyss.25
As Salecl makes plain, anxiety concerns the possibility of being able.

Death—here in the trope of l’appel du vide—arouses anxiety because it
feels the grounding of possibility in its utter negation. Possibility wouldn’t
mean anything if it couldn’t be taken away. Death only arouses the possibility
of being able through its negation. By contrast, the screw solicits a vitally
distinct sense of anxiety as a sense of embodied ability, what Salecl calls “the
possibility of being able.” Let’s not forget that one of the most important and
distinctive social capacities of always-on life is the possibility of feeling hurt.
So how much can we actually do? How much are we really able? Are we
truly screwed? Being screwed sounds bad. But from a certain point of view
it also sounds pleasurable. This is what I Put Wii Music does to us. It tells us,
we’re screwed. We are all very anxious, and we have plenty of good reasons
for feeling this way. From irreversible effects of anthropogenic climate
change to terrorism to the resurgence of xenophobic nationalism—haven’t
you heard? The dark beauty of being screwed, however, is that it can be fun
even if it also feels bad. This does not mean we must simply accept our fate
by changing our perspective. Being screwed means dealing with it, just
as here it has meant recognizing some of the ways we are already doing
so. It means not just feeling anxious but also having fun, and having fun
means feeling something together. My job here has not been to show that
anxiety is somehow socially redemptive, only that that it might be social
at all. As a form of ambivalence, being screwed doesn’t so much allow us
to chart a course for change so much as it helps us to deal with what’s hard
or seemingly impossible. At the very least, it gives us something to feel and
to say: “I’m screwed!” The mere fact of voicing this condition opens up its
social character. In the context of always-on computing, as we’ve seen, the
obligation to be communicable makes things a little screwy. Haha. The fun
in anxious fun lies with getting worked over together. This may feel like the
end but it’s where we start today. Nobody gets screwed alone.
Notes
Thanks to Kris Cohen, Patrick Jagoda, and Kyle Stine for feedback.
1. See, for example, Alex Williams, “Prozac Nation is Now the United States of
Xanax,” New York Times, https://www.nytimes.com/2017/06/10/style/anxi-
ety-is-the-new-depression-xanax.html. June 10, 2017.
2. Institute for Precarious Consciousness, “We Are All Very Anxious: Six Theses
on Anxiety and Why It is Effectively Preventing Militancy, and One Possible
Strategy for Overcoming It,” Plan C (April 4, 2014): weareplanc.org.
3. Ernst Bloch, cited by Sianne Ngai, in Ugly Feelings (Cambridge: Harvard
4. https://www.mcachicagostore.org/shop/memento-mori-clock-52745.
5. Lauren Berlant, “The Commons: Infrastructures for Troubling Times,” Envi-
ronment and Planning D: Society and Space 34, no. 3 (2016): 393–419, 393.
6. Anaïs Nony, “Anxiety in the Society of Preemption: On Simondon and the
Noopolitics of the Milieu,” La Deleuziana 6 (2017): 102–110, 103.
218 James J. Hodge
7. Anticipation arguably only persists today in various apocalyptic imagina-

tions from American evangelistic Christianity to the doomsday scenarios of
popular fiction across media.
8. See Norbert Wiener, Cybernetics: or Control and Communication in the Ani-
mal and Machine, 2nd ed. (Cambridge: MIT Press, 1961); and Peter Galison,
“Ontology of the Enemy: Norbert Wiener and Cybernetic Vision,” Critical
Inquiry 21, no. 1 (Autumn, 1994): 228–266.
9. See Richard Grusin, Premediation: Affect and Mediality After 9/11 (New
York: Palgrave Macmillan, 2010); Mark B. N. Hansen, Feed-Forward: On the
Future of Twenty-First-Century Media (Chicago: University of Chicago Press,
2015); Brian Massumi, Ontopower: War, Powers, and the State of Perception
(Durham: Duke University Press, 2015); and Shane Denson, Discorrelated
Images (Durham: Duke University Press, 2020).
10. See Mark Andrejevic, Infoglut: How Too Much Information is Changing the
Way We Think and Know (New York: Routledge, 2013).
11. American Psychological Association, “Data on Behavioral Health in the
United States,” American Psychological Association. www.apa.org.
12. See, for example, Jean Twenge, iGen (New York: Atria Books, 2017).
13. As defined in the DSM-V, “anxiety is anticipation of a future threat.” It
overlaps with but is distinct from fear: “with fear more often associated with
surges of autonomic arousal necessary for fight or flight … and anxiety more
often associated with muscle tension and vigilance in preparation for future
danger and cautious or avoidant behaviors.” See Diagnostic and Statistical
Manual of Mental Disorders: DSM-5 (Arlington, VA: American Psychiatric
Publishing, 2013).
14. Martin Heidegger, Being and Time, trans. John Macquarrie and Edward
Robinson (New York: Harper Perennial, 1962), 231.
15. For example, see Nicole Starosielski, The Undersea Network (Durham: Duke
16. James J. Hodge, “Sociable Media: Phatic Connection in Digital Art,” Post-
modern Culture 26, no. 1 (September 2016): n.p.
17. Scott C. Richmond, “Vulgar Boredom, or What Andy Warhol Can Teach Us
About Candy Crush,” Journal of Visual Culture 14:1 (2015): 21–39.
18. See John Cheney-Lippold, We Are Data: Algorithms and the Making of Our
Digital Selves (New York: New York University Press, 2017).
19. See Kris Cohen, Never Alone, Except for Now: Art, Networks, Populations
(Durham: Duke University Press, 2017).
20. See Damon R. Young, „Ironies of Web 2.0,“ Post45 2, May 2, 2019: https://
post45.org/2019/05/ironies-of-web-2-0.
21. See Whitney Phillips and Ryan M. Milner, The Ambivalent Internet: Mischief,
Oddity, and Antagonism Online (Malden, MA: Polity, 2017).
22. Lauren Berlant and Sianne Ngai, “Comedy Has Issues,” Critical Inquiry 43,
no. 2 (Winter, 2017): 233–249, 233.
23. See “This is Fine,” Know Your Meme (2015–), knowyourmeme.com.
24. See Miriam Bratu Hansen, “Room-for-Play: Benjamin’s Gamble with Cine-
ma,” October 109 (Summer 2004): 3–45.
25. Renata Salecl, On Anxiety (New York: Routledge, 2004), 53; and Soren
Kierkegaard, The Concept of Anxiety, trans. Reidar Thomte (Princeton:
Princeton University Press, 1980).
About the Author
James J. Hodge is Associate Professor in the Department of English and

the Alice Kaplan Institute for the Humanities at Northwestern University.
He is the author of Sensations of History: Animation and New Media Art
(University of Minnesota Press, 2019). He is currently writing a book about
the aesthetics of always-on computing.
12. Beep: Listening to the Digital Watch
Sumanth Gopinath
Abstract
This essay considers the intertwined histories of the word “beep” and
the simple, single-oscillator tone by examining the digital watch of the
late 1970s–early 1980s. The watch and its beep marked a key economic
and technological development in which the US, as global hegemon and
economic powerhouse, was not the dominant agent. They also prefigured
the tinkling sound of ringtones on mobile phones and initiated the mass
mundanization of digital beeps, now shorn of the symbolic power of
recondite, expensive, and classified Cold War-era technologies. The watch
beep’s hardwired nature allows us to hear connections between that
historical moment’s different temporal scales, direct linkages between
which would soon be imperceptible thanks to the digital economy’s ever
increasing abstraction.
Keywords: beep, digital watch, sound studies, piezoelectricity, world-sys-

tems theory, integrated circuit (IC)
Beep. This word denotes several associations in English, including its noun
and verb forms, its evocative epizeuxis beep-beep, and its description of a
device, the beeper. The Oxford English Dictionary reveals that the beep is a
twentieth-century phenomenon, initially indexing the bustling modernity
of urban soundscapes and transit and expressed through the obtrusive car
horn.1 Plosives in the word’s beginning (voiced) and ending (unvoiced) frame
a close front unrounded vowel that accentuates higher frequency content,
particularly in American English. Sounding like the signal it describes, the
word’s onomatopoetic qualities grab your attention, quickly. The oldest
relevant entry in the dictionary is Edmund Wilson’s first novel I Thought
of Daisy (1929), as spoken by the titular character, Daisy Coleman, a chorus
girl who is the narrator’s love interest and who describes car horns thusly:
doi 10.5117/9789463727426_Ch12
222 Sumanth Gopinath
“‘It’s a shout. There’s a toot-toot, and a beep-beep—and an oorah—and a

blah-blah, and a blurp-blurp.—Dad’s was a kind of a oorah—and it was a
humdinger, too!”2 The horn’s sound is not uniquely captured by “beep,” and
today a horn is typically described as honking. A search of The New York
Times reveals that the word doesn’t appear in the paper until the 1930s,
when describing a bicycle horn.3 In the postwar period, “beep” expanded
beyond vehicular horns to designate electronic signals, including those in
military technologies, touch-tone dialing, and other telephonic sounds;4
experimental electronic music and its uneven migration into pop/rock;
science fiction and space travel (partly via the Sputnik beep);5 and radio alert
and signal-scrambling systems like CONELRAD (Control of Electromagnetic
Radiation) and the Emergency Broadcast System (EBS).6,7 The expanded
use of the word “beep” suggests that beeps became central to the US Cold
War soundscape.
But if the initial conjoining of the beep and single-oscillator signal
occurred during the period spanning the late nineteenth century and
World War II, and a new crystallization began after the war, it would be the
long 1970s that made the beep newly ubiquitous by embedding them into
devices large and especially small: appliances, phones, games, computers,
and pagers. Moreover, due to the global growth of Japanese electronics
production and sales, the story was no longer predominantly a US-centric
one. Making that technological and sociocultural transformation possible
was the rise of semiconductor electronics, which enabled smaller and smaller
components to be tucked away silently into ever more ubiquitous devices.
As Ross Bassett puts it, the semiconductor transistor “is a base technology
of late-twentieth-century and early-twenty-first-century America. Through
it digital electronics have entered almost every area of American life, first
through the calculator, then through the digital watch, and finally through
the microprocessor.”8 Despite their pervasiveness in America and worldwide,
microelectronics are unseen and unheard, known only through their effects.
It is fitting that they have been accompanied by equally ubiquitous, beeping
small speakers.9
The widespread mundanization of the smartphone and hence portable
personal computer, with its sophisticated audio generation and playback
capabilities, have seemingly propelled the world into a newly intensified
dependence upon and engagement with integrated circuits (ICs), micro-
processors, and the complex sounds they enable. But we still live in the era
of the beep, which is unlikely to disappear anytime soon, given its utility
and simplicity. If the beep acts as a deep social layer composed of myriad
hardwired temporalities that continues to either interact with or ignore
Beep: Listening to the Digital Watch 223
sound reproduction,10 its formation merits further investigation. A crucial

development in the beep era is that of digital watch, in which the IC met
the beep in a mobile, mass-market context for the first time.
The history of the digital watch is complex and dramatic. It involves the
belated application of the nineteenth-century discovery and theorization
of piezoelectricity, international economic battles between the established
Swiss mechanical watch industry and newer, upstart watch industries based
in the United States and Japan, ethico-moral debates over the importance of
learning to read the analog clock face, ultimately leading to its coexistence
with the numerical watch display, and the inauguration, along with the
calculator, of a series of consumer fads for digital devices beginning in the
1970s. In what follows, I consider the digital watch at multiple temporal
scales. First, I recount the long history of the watch industry’s transna-
tional political economy, including its transition to digital timekeeping.
The industry’s temporality is interpreted here at the macro-temporal level
of Arrighi’s “long century,” in ways that dovetail with the century-long,
US-hegemonic history of the “beep” word and concept.11 I then explore the
micro-temporality of the digital watch, as produced by the piezoelectricity
of quartz crystals and microprocessor-calculated digital timekeeping. I
continue by examining the digital watch’s sounds and end by noticing their
presence in social spaces; these sounds seem to exist at a widely meso-tem-
poral scale from minutes to months and years. The patterned alignment of
these particular components and their corresponding social practices can
be interpreted as a form of hardwired temporality, but the simplicity of the
micro-temporal components in early mass digital timekeeping allowed a
direct and perceptible connection via the meso-temporal. To anticipate my
conclusion, listening to the digital watch allows us to further connect these
temporal scales, by articulating a beep (and its means of generation) with
the incipient decline of US economic hegemony. By becoming quotidian, the
beep—which prefigured the more versatile and abstracted mobile phone
and its sounds—lost its valences of power and potential.
The Political Economy of the Watch Industry
The digital watch upended the traditional global hierarchy of the watch in-
dustry, leading to the divide between luxury and economy exports stemming
primarily from Switzerland and East Asia, respectively. Before the digital
watch’s emergence, the industry was dominated by Swiss watch firms, as it
had been since the mid-nineteenth century.12 That dominance was arguably
a hard-won product of economic and technological contestation in Europe

over the course of the early modern period. Several contributing factors
include the emigration of French Hugenot watchmakers to Switzerland in
the sixteenth century, which brought crucial new talent to the country and
arguably crippled French watchmaking; the peculiar form of decentralized
social conservatism characteristic of Swiss industries, leading to the gradual
adoption of institutional changes and greater inclination toward honing
existing practices; and its comparatively high ratio of watchmakers to parts
manufacturers, encouraging the creation of a greater variety of products in
different price ranges.13 These developments led watchmakers in Geneva
to become increasingly competitive with the more established British
watchmaking industry in the eighteenth century. Then, with the rise of
yet another Swiss region for watch production in the Jura mountain valley
region, centered in Neuchâtel, in combination with generally lower wages in
Switzerland, the country’s products began to undercut British sellers at the
end of the eighteenth century and into the nineteenth century. Long being an
emulative industry up to the mid-eighteenth century, following innovations
from Britain or France, Swiss watchmakers increasingly took the lead in new
development, creating standardized tests, competitions, and watchmaking
schools, all with the strong support of the state and contributing to Swiss
preeminence in watch production by the mid-nineteenth century.
The primary contender for industrial dominance in the mid- to late
nineteenth century was the US. Driven by the demands of railroad
scheduling, the US watch market was temporarily closed to exports due to
wartime protectionism during the Civil War, giving a boost to postbellum
industrial growth in US watch manufacturing, which became increasingly
rationalized. US firms soon became strongly competitive with Swiss makers,
who were slow to adopt factory production methods. Industrialized US
watch firms, such as the Boston-based Waltham Watch Company, were of
crucial importance in establishing the quality of US watches, and the Elgin
Watch Company, located near Chicago, in turn drew on the expertise (and
sometimes personnel) of Waltham, leading to a powerful, two-company trust
at the forefront of the US watch industry in the late nineteenth century; the
Hamilton Watch Company, based in Lancaster, Pennsylvania, was another
important late entrant into the market that would assume an ever stronger
role among US firms. And yet, in 1885–1905, the Swiss watch industry up-
graded itself partly on the American model, reducing small-scale production
and moving toward vertical integration, and by 1910 Swiss watches again
dominated the global market. Crucial was the rise of the wristwatch (instead
of the older pocket watch), which Swiss makers produced more successfully
than US firms, who were hampered by “a uniform regulatory regime for

the production of railroad watches using machine-manufactured parts.”14
Swiss watch industry dominance continued through the interwar and
early postwar periods, with market share for low-end watches only facing
significant competition from US-based Timex in the 1950s and 1960s.
The rise of the Japanese watch in the 1960s and after was completely
unexpected by the Swiss and American watch industries. Although Japan
industrialized in the nineteenth century and sustained a sizable watch
industry led by Seiko, Citizen, and Ricoh in the early twentieth century,
its greatest advances in the watch industry occurred during the renewed
industrialization of the postwar period, after the devastating WWII bombing
of its factories and productive forces. Its credentials were boosted by a
key development. As Amy Glasmeier notes, “When Seiko was selected
as the timekeeper for the 1964 Winter Olympics, held in Sapporo, Japan,
its reputation was greatly enhanced. In full view of the entire world, the
Japanese watch company demonstrated what had heretofore been the
exclusive purview of the Swiss: to flawlessly keep, record, and display time
at exceptionally high levels of accuracy.”15 But the critical commodity for
the Japanese watch industry was the digital watch, which ultimately led
Japan to become the world’s largest watch producer, with annual production
figures growing from nearly 700,000 to over 86,000,000 watches from 1950
to 1980.16 US-American and Japanese scientists and companies collaborated
and competed in producing high-quality ICs (and later liquid crystal displays
or LCDs), first used in calculators and then watches: when they entered
some aspect of this market at all, US electronics firms often gave up on or
mistimed their investments in these technologies, to the benefit of Japanese
producers.17
Despite the relative weakness of the US watch industry after WWII, it
was competition between American and Japanese firms that would prod
the development of the digital watch, with the latter ultimately winning out.
After creating initial prototypes, Swiss firms stayed on the sidelines to their
detriment—until Swatch, founded in 1983, helped put an end to the “quartz
crisis” of that country’s watchmaking industry.18 In retrospect, the earlier
rise of American timekeeping—in contrast to the longer-established British
and Swiss industries—signaled the corresponding rise of US-American
hegemony in surprisingly direct ways (with the British watch industry’s
supersession and Swiss industry’s development arguably linking not only to
the early-modern-historical lineage mentioned above but also to the decline
of British hegemony and to central-European bids for global dominance in
the mid- to late nineteenth and early twentieth centuries, respectively).
This narrative suggests that the timekeeping business was fundamental

to the capitalist world-system more broadly—which might be expected,
given the essential role of clocks in the management of modern markets,
transportation, labor, and state administration. The Japanese development
of the digital watch likewise signaled a new possible capitalist hegemony in
East Asia—later furthered by China’s emergence as “the workshop of the
world.”19 Bound up with the long century of American hegemony, the beep
was absorbed and expanded dramatically in Japan and East Asia, alongside
digital devices more generally.20
Piezoelectricity and the Quartz Watch
The foundation of quartz timekeeping is the phenomenon of piezoelectricity.

Discovered by Jacques and Pierre Curie in 1880, it involves the generation of
an electrical polarization by applying mechanical stress to certain materials,
especially crystals, metals, and ceramics. This polarization creates the
possibility of generating an electrical current from mechanical pressure to
a piezoelectric material, but the reverse effect—generating a mechanical
movement from an electrical charge—was also predicted by Gabriel Lipp
mann one year after the Curies’ discovery. Although the basic mathematical
theorization of piezoelectricity was completed by Woldemar Voigt in 1890,
it was first employed in devices during WWI, when early efforts to detect
submarines via ultrasonic waves (later termed SONAR by the US military)
made use of quartz-based transducers. In 1921 Walter Cady created the first
quartz crystal oscillator, and in 1927 scientists at Bell Labs produced the
first quartz clock, using quartz as a stable oscillator.21
With the emergence of the integrated circuit in the 1960s, quartz clocks
came into widespread use; their basic principle can be understood as follows
(fig. 12.1). First, a battery (b) powers a transistor oscillator (not shown) that
activates the crystal resonator (a), causing it to vibrate and hence transform
electrical energy into elastic energy. The resonator smooths out the irregular-
ities of the oscillator, thanks to the very low power loss and high stability of
the quartz crystal. The resonator then reconverts the vibrations into a highly
stable electric current, sending it to the microprocessor (d), which counts
the number of pulses from the quartz (which is conventionally designed
to oscillate at 32,768 or 215 Hz). The microprocessor then repeatedly divides
the pulses by 2 in order to reach 1 Hz (or 1 cycle/second) for the purpose of
keeping time. The output information is then sent to the digital display (e)
or, if it is an “analog” quartz clock, to a watch movement mechanism that
Figure 12.1. Diagram of the quartz wristwatch structure, after David Penney, in David Landes,
Revolution in Time: Clocks and the Making of the Modern World (London: Viking, 2000), p. 376.
moves the second, minute, and hour hands of the clock face. (Hence, we
appreciate that just about all watches today, irrespective of their display
method, use digital timekeeping circuitry—even though “analog” quartz
clocks are typically described as “electronic” rather than “digital.” Mechanical
watches, presently a tiny minority of watches worldwide, are identified by
smooth, rather than pulsed, second-hand movement.) In a digital display,
a number of switches (c) are also used to control other, ancillary functions
like the alarm, stopwatch, and light, and thus are also mediated by the
microprocessor.22
In concerning the sonic aspects of the digital watch, we must further
examine and contextualize the series of switches (c) that drive the watch’s
ancillary functions. The earliest digital watches were sold as luxury goods,
in part as a result of the initial high cost of the timing circuits, and generally
competed in the high-end watch market. The first electronic (“analog”)
quartz watch on the market, the Astron SQ made by the Japanese maker
Seiko, was released in 1969 for 450,000 yen (or $1,250 US), on the heels
of several successful prototypes produced in Switzerland at the Centre
Electronique Horloger (CEH) for competition in the Neuchâtel time trials
in 1967. The American maker Hamilton, abandoning its unsuccessful efforts
to produce an analog quartz watch, sought instead to make a fully digital
watch, with a red light emitting diode or LED-based numerical readout
partly inspired by Hamilton’s digital timepiece for the film 2001: A Space
Odyssey (1968).23 Featured in the James Bond film Live and Let Die (1973),
the Pulsar cost $2,100 upon first release in 1972.24 However, the price of
microelectronics components soon dropped precipitously, inaugurating a
series of price wars within the industry. As Amy Glasmeier notes,
From 1974 to 1975, the price of a digital watch dropped from $125 to
$50. The next big step downward was forecast for 1977: some industry
watchers believed that prices would fall to $20. Then Texas instruments
stunned the market by introducing a plastic-encased digital watch for
$9.95! The actions of Texas Instruments were quickly followed by other
semiconductor makers who hoped to destabilize and thwart further
actions by watch assemblers who could not buy components cheaply
enough to compete at the $20 level … once Timex’s exclusive domain.25
The increasing overcapacity of the semiconductor industry signaled the

end of the integrated circuit’s “market-pull” phase, in which demand
for new semiconductors clearly outstripped production rates, to a new,
“technology-push” phase in which semiconductor manufacturers pushed
their way into new industries like watchmaking and undersold established
firms to meet the demands of expanding fixed capital sunk in production
equipment and facilities.26 A new, mass-oriented, spectacularized sales ethic
soon replaced the older, luxury-goods oriented marketing strategy, as seen
in a comparison between a mid-1970s watch ad by the Swiss-British luxury
brand Rotary (featuring an older, less efficient red LED display) and a 1978
Seiko Memory Bank Calendar Watch commercial (featuring the then-newer
LCD, science-fictional synthesizer music, and a tech-functional ethos).27
The micro-temporality of the digital watch was bound up with new
economic realities within the industry: it was cheaper and more accurate,
and its digital temporalities bypassed the previous mechanical temporalities
of the traditional watch movement. The digital watch’s temporality had new
visual indicators, chiefly the display that read out digits instead of hands
on a dial, but also its employment of a microprocessor and speaker system
that typically resounded its inner workings through a beep.
The Sounds of the Digital Watch
Within an increasingly competitive sales environment, companies in the

mid- to late-1970s began producing multifunctional watches, including
calculator and calendar watches, and incorporated sound into them. Some
early digital sound watches were alarm watches, and the alarm soon became
standard on digital watches, which often emitted short beeps when a button
was pressed.28 Among the earlier examples included the 1977 Seiko A039
Quartz LC Alarm Watch, which featured two different alarms with two
volume settings, an hourly signal, and a stopwatch function (likely including
start and stop beeps). By the turn of the new decade, novelty sonic watches
emerged, including some that played several different melodies, offered video
games, and even talked in the case of the OMNI Voicemaster. According to
one news article, this latter product “looks like a normal digital watch but,
when a small button next to the face is pushed, a male voice announces
the time—hour, minutes and a.m. or p.m. The timepiece also has a 24-hour
alarm—with the tune of Boccherini’s Minuet. When the set time arrives, the
watch states the time, then plays a melody for about 20 seconds. Unless the
wearer resets the alarm control within five minutes, the watch announces
‘Attention please,’ repeats the time and urges ‘Please hurry.’ The watch can
also be used as a talking stopwatch.”29
Perhaps the most important of these watches in terms of sales and
musical content were the Casio Melody Alarm watches first appearing at
trade shows in 1980 and released on the market the following year.30 These
watches featured twelve melodies, divided into two groups. The first group,
consistent across most of the watches, provided alarms on the hour or as
coordinated with the watch’s calendar function, such as playing “Happy
Birthday” on a designated birthday, “Jingle Bells” on Christmas Day, the
“Big Ben” chime for hourly alarms, and Mendelssohn’s “Wedding March”
for a wedding or anniversary. The second group involved daily alarms
that varied more frequently according to watch model; in each model,
however, the alarm was one of seven melodies coordinated with the days
of the week. The variety can be seen across two production batches of the
Casio H104 Melody Alarm Watch with different pre-programmed melody
modules (table 12.1). Module 82 features a weekly schedule beginning with
F. W. Meacham’s 1885 march “American Patrol” on Monday and a number
of European “folk songs” from Tuesday to Friday, before ending on the
weekend with the Japanese song “Sakura Sakura” on Saturday and the
main tune of Schubert’s Moment Musical No. 3 on Sunday. The melody
collection here seemingly trumpets Japan’s recent inclusion into the global
north and group of politically powerful nations, comprising the US and
most of the dominant European countries (France, Russia, Italy, Spain),
with Sunday being reserved for an Austrian classical composer.31 Module
142, in contrast, is more scattershot and less inclusive internationally.
Table 12.1. Casio H104 melody watch instructions, listing of melodies for two
separate sound-production “modules”
(Module No. 82)
Day of week Melody
Monday American Patrol (F. W. Meacham)
Tuesday Santa Lucia (Napoles folk song)
Wednesday Romanza de Amor (Spanish folk song)
Thursday Marche Royale (French folk song)
Friday Kalinka (Russian folk song)
Saturday Sakura Sakura (Japanese folk song)
Sunday Moments musicaux (Franz P. Schubert)
(Module No. 142)
Day of week Melody
Sunday Ungarische Tanz, Nr.5, F moll (Brahms)
Monday “Carmen.” Votre toast, je peux vous le rendre (G. Bizet)
Tuesday Little brown jug (American folk song)
Wednesday L’amour est bleu (Andre Popp)
Thursday A walk in the Black Forest (Horst Jankowski)
Friday Yellow Rose of Texas (American folk song)
Saturday Cassatio in G. (L. Mozart)
It is weighted more heavily toward classical music (Brahms, Bizet, and

Mozart), with some Europop (André Popp), European easy listening
(Horst Jankowski), and a couple of nineteenth-century US minstrel tunes
described, again, as “folk songs” (“Little Brown Jug” and “Yellow Rose of
Texas”).
With the exception of Popp, Jankowski, and “Happy Birthday,” these
melodies were no longer protected by copyright and could be adapted
freely. However, at least one version of the watch (perhaps the M-1230)
reproduced mostly English-language popular songs from the 1960s and
1970s, including “Never on Sunday” (a Greek and then English-language hit
song from the same-titled 1960 film), Simon and Garfunkel’s “The Sound
of Silence” and “Scarborough Fair (Parsley, Sage, Rosemary and Thyme),”
the Beatles’ “Ob-La-Di, Ob-La-Da,” “Rhythm of the Rain” (a 1962 hit for the
Cascades), and Billy Joel’s “The Stranger” (1977). These copyrighted tunes
were likely used without proper licensing or royalty payments. (This was
well before the US/UK music publishing industry had become extremely
driven to secure intellectual property rents in all possible venues, and
Japanese corporations’ geographical, linguistic, and legal distance surely
hampered enforcement.)
Finally, some Casio Melody Alarm watches included an LCD scale-display

showing notes corresponding to each melody’s pitches—specifically, their
white-key equivalents or counterparts represented within a C-diatonic
gamut, like a visualization of fixed-do solfège. Casio and other companies
eagerly demonstrated the utility of LCDs: some of their watches utilized
an LCD dial imitating the analog clock face, while others used the display
for portable video games.
By the 1980s, sounds on these watches were usually made by piezoelectric
speakers, which were developed over the course of the 1970s. Using the same
effect as the quartz timekeeping mechanism, a piezoelectric speaker is a
small transducer that transforms electrical signals into mechanical energy
without the mediation of a wire coil or magnet (as with standard speakers).
Scott Chou of Hattori Seiko Co. (Seiko’s name in 1983–1990), noted in 1989
that in most watches the piezoelectric element is a type of ceramic foam
connected to a thin silver contact, resonating at a frequency of about 4,000
Hz and taking advantage of the human ear’s great sensitivity at 4,000–6,000
Hz.32 The benefits of the piezo speaker were multiple. It required low power,
a primary concern during the 1970s (the red LED display fell out of favor
partly due to power needs). Like the watch mechanism itself it was highly
reliable. And, most importantly, it was remarkably compact—taking up
almost no space within the watch itself (present-day digital watches nestle
it snugly between the microprocessor and the battery).
Chou’s description of the alarm’s frequency seems correct: the stan-
dard digital watch beep is set at 4,096 Hz, the oscillating frequency of
the piezoelectric quartz crystal, 32,768 Hz, divided by 23 or undergoing
a half division three times. The watch alarm uses the original oscillating
frequency as the quartz watch timekeeping system to power the alarm,
almost certainly using the same microprocessor to compute the alarm’s
frequency. The pitch is a very high B7 or flat C8 (about 38 cents flat). Other
digital alarms and signals use the same principle, sounding at the same
frequency or an octave or two below—2,048 Hz (or high B6, flat C7) or 1,024
Hz (or high B5, flat C6). In the earliest watches, cost was likely a determining
factor in the selection of the alarm frequency; the alarm had to be solidly
within hearing range, and 8,912 Hz (32,768 Hz divided by 22) is a little bit
high to be effective as an alarm (and much less audible for elderly hearing
people33). The highest possible frequency would entail the smallest number
of half-divisions (which would not necessarily mean a cheaper circuit—that
would depend on the most efficient Boolean logical reduction of the alarm
clock circuit). Given the frequency’s commonness, it is possible that 4,096
Hz requires the smallest (and likely cheapest) piezoelectric sounder needed
to produce the sound. Moreover, since low frequencies generally require

more power to be heard and given this frequency’s placement within the
maximally perceptible frequency range of hearing, it probably uses the
least battery power.34
The Social Fallout of the (Sounding) Digital Watch
If the macro-temporality of the digital watch industry is its long centuries

and the micro-temporality of the digital watch is its quartz crystal vibra-
tions, it is the meso-temporality of the watch’s use, including its beeps,
that has shaped everyday life. Minutes, hours, and days structure that use,
through its orientation toward the working day; the watch’s alarms enact
both the production of labor discipline (such as daily waking for work) and
reproductive leisure (such as the late 1970s and early 1980s running boom,
to which its stopwatch function became fundamentally tied).35 Weeks,
months, and even years were marked by additional uses, from the calendrical
functions identifying the first two to the longer periodicities of leap-year
date corrections or replacing the watch’s battery, acts all accompanied by
ever-present beeps.
For a time, the digital watch seemed to be everywhere. Indeed, during the
1980s, a considerable ressentiment against it gathered steam, with much of the
energy trained upon the digital time display. As James Sterba wrote in 1982,
The anti-digit traditionalists have staged a comeback in the last few

years, partly by stealing a word from the computer business—analog—to
space-age the name of their old standard bearer. It is no longer just a watch.
It is an analog watch, which means it represents one quantity—time—by
another quantity, the motion of its hands on a dial. They argued that there
might be hidden danger to children, growing up on digits, who could read
time by simply knowing numbers but could not discern it analogously
from a watch or a clock with hands.36
Aiding the antagonists were the perceived sonic failings of the digital
watch; as the watch historian and partisan for the old mechanical watch
movement David Landes argued, “the tick of a good timepiece is a delight
to those who can appreciate a strong, regular beat,” whereas the British
designer Richard Porch complained that the digital watch is “a silent affair
that requires no attention.”37 Except, of course, when it beeps—especial-
ly as it often does on the hour, every hour, with a quasi-humanizing or
science-fictional chime likely intended to give some sonic character to the

otherwise silent device. Indeed, the collective annoyance with the digital
watch came to the fore in public art and performance spaces, generating
a mild “digital watch rage” recalled two decades later by ringtone rage
in similar circumstances. One journalist connected such experiences to
broader frustrations with the beep:
The contemptuous little “beep” never comes at the crescendo of a Wag-

ner symphony. It happens only during a dramatic pause in Hamlet’s
soliloquy, the turning point in a critical sales presentation, or at the
perfectly inappropriate moment of a torrid love scene. In the worst case,
a sudden, uncontrollable beep has the same social impact as a sudden,
uncontrollable belch. If digital watches were the only devices openly
mocking their owners, life in the Technobabble Jungle wouldn’t be so
frustrating.38
But a major difference between more recent cellphone interruptions and

digital watch intrusions is the regularity (and predictability) of the routine.
One editorial writer lamented,
To be in a movie theater now as the hour changes is to hear a beep chorus.

It sounds as though half the audience is wearing those digital watches
that emit one beep on the half-hour and a vigorous beep-beep on the
hour. The chorus would be tolerable if everybody’s watch chimed in
unison. But the movie theater beep-beeps start around 8:58:30, gather in a
crescendo, and then sputter along until 9:02 or so—a longish distraction,
and a puzzling one.39
The opponents of the fully digital watch and its beeps won out, relegating
the device to second-fiddle status within the industry and restoring the
analog dial to its once-primary position. 40 But, as historians Carlene
Stephens and Maggie Dennis put it, “Behind the dial of most new watches,
though, hummed an electronic heart.”41 Indeed, by 1997, mechanical watches
accounted for only 7.9 percent of total volume and 44 percent of sales value
globally. But the beeping digital watch had the last laugh: the same basic
principle of sound production made its way into the mobile phone, providing
the technological basis for the extremely annoying monophonic ringtone
during the 1990s and early 2000s. Synchronized to cellular networks, the
beeping phone became the new digital watch; in doing so, it bypassed the
problem of accurate, device-based timekeeping altogether.
Conclusion
Despite the appearance of critically praised, wrist-based devices such as

the Pebble Watch, Samsung’s Galaxy Gear, and the Sony SmartWatch and
the popular self-generated-data-tracking device FitBit, no device has yet
inspired a sufficient degree of excitement to propel wrist-based wearable
devices into the leading edge of popular technology consumption in the
new millennium. The closest exception might be the Apple Watch, first
released in 2015. ICT industry boosters hoped that the device would do
for the smart watch what the iPhone did for the smartphone and the iPad
for the tablet computer, and although its popularity is steadily rising, it is
difficult to predict its future.42 It initially targeted tech enthusiasts, fitness
watch users, Quantified Self movement participants, wealthy consumers,
and those nostalgic for outdated science-fiction protagonists, including
Knight Rider’s Michael Knight, Inspector Gadget’s niece Penny, and Dick
Tracy. For now the Apple Watch is still most functional when tethered to
an iPhone. 43 But it can make the same preset ring and alert sounds as the
iPhone, and, like many phones today, these functions are frequently silenced.
The smartwatch commodity could help to close the gap created by the
mobile telephone, which in the early to mid-2000s caused a precipitous
drop in wristwatch sales among younger consumers especially. 44 The
conflict between watch and phone—two technologies that offer what
Landes calls “cheap time”45— masks how they have influenced one another,
particularly in the era of the integrated circuit, which permits a degree of
repurposing remediation hitherto unimaginable. The digital watch, an
early wearable digital technology, was a precursor to the modern mobile
phone, particularly in the way it accumulated features, including some of
the same sonic ones. By listening to the digital watch, we learn something
about not only the relationship between two world-historical devices but
also the corresponding states of the capitalist world-system that made their
emergence possible.
The parallels between the respective sounds of the digital watch and
the mobile phone are easy to draw. 46 The digital watch’s beep became the
simple digital ringer of the mobile phone, and the melodic watch became
the monophonic ringtone. Ringtone rage in the public sphere and at per-
formances in the early 2000s was prefigured by the public disturbances of
the half-hourly chimes of the digital watch two decades earlier. Both the
digital watch and mobile phone used their sound production systems to
incorporate timers and handheld video games, and the digitized voices
of novelty watches even anticipated the voice assistants of Siri, Cortana,
and Alexa. And, the global political economies of both devices revealed a
tripolar dynamic, with competition between specialized locales in Europe
(Switzerland and Nokia’s Finland) and Japan and the US being relatively
shut out of the process while previously contributing technologies that were
transferred outward (integrated circuits and cellular telephony, single-os-
cillator tone-generators and FM synthesis). Moreover, the digital watch’s
trebly, tinkly beep presaged the monophonic ringtone’s aesthetics, which
are arguably a sonic corollary of global-regional cute (kawaii) culture—a
culture that articulates unevenly with Japanese state-economic power and
now exists worldwide, including in the US. 47 But the differences in device
purpose and design and the historical contexts of technological development
reveal clear dissimilarities: with the digital watch, there was no progression
in sonic fidelity, exhibited by the shift from monophonic, to polyphonic, and
to sound file as with the ringtone (and numerous precursor technologies); nor
did the digital watch make its tunes programmable or uploadable, leading
to a lucrative para-industry comparable to the ringtone industry. And the
flexibility of assigning distinct sounds to different contacts, functions, and
apps in the cellphone (and smartphone and smartwatch) bespeaks a world of
distributed, individualized labor-time management exploited by the digital
gig economy and only hinted at by older multifunctional digital watches. 48
Indeed, it is the very rigidity of the simple digital watch’s beep—unlike the
flexibility of and rapid changes in the mobile phone’s sonic production—that
allows one the rare possibility of hearing the micro-temporality of the
digital watch. It is the watch’s de facto employment of “scientific pitch,” or
C = 256 Hz (rather than the currently standard 261.63 Hz) that allows one
to perceive, via a specific tuning, the inner workings of the quartz crystal,
whose resonating frequency is just another flat C, out of hearing range, and
whose microprocessor divisions by two transpose it into audibility and then
back out of it, as it accurately calculates the length of a second. 49
In comparing the digital watch and the mobile phone, history would seem
to have repeated itself as tragedy and then farce, to invoke the now-clichéd
dictum from Marx’s The Eighteenth Brumaire. But if the mobile phone’s
ringtone was clearly the farcical repetition of a phenomenon past, what
was tragic about the digital watch’s beep? To revisit our initial discussion,
perhaps it lies partly in the epochal routinization of the ubiquitous and
now-residual “beep,” found in numerous household, personal, and industrial
devices—microwave ovens, washing machines, scanners, digital alarm
clocks, automobiles, cellular phones, home computers—and which since
the 1970s were predominantly made by integrated circuits, small speakers,
and simple, single-oscillator signals. The oscillator’s beep, at one time, held
a utopian promise, but the signal beeps from Sputnik now sound like an
alarm clock; the conquest of nature and space that they represent once
foretold of human betterment and today seems untenable. It is a state of
affairs worthy of a cold, clear-eyed tear or two.
Notes
Many thanks to the individuals who graciously gave me feedback on or

help with versions of this project, including Kate Altizer, Peter Burkholder,
Phil Ford, Anand Gopinath, Sudhir Gopinath, Beth Hartman, Karen Ho,
Marianne Kielian-Gilbert, Michael Klein, Ajay Limaye, Tom McAuley, Dan
Melamed, Ali Momeni, Alex Monae, Tina Muxfeldt, Ayana Smith, Jason
Stanyek, Jonathan Sterne, David Valentine, Christine Wisch, and the edi-
tors—with profuse apologies to those I’ve momentarily forgotten.
1. Oxford English Dictionary, s.v., “beep,” https://oed.com/.
2. Edmund Wilson, I Thought of Daisy, ed. Neale Reinitz (Iowa City: University
of Iowa Press, 2001 [1929]), 253–54.
3. A bicycle display ad for Gimbels department store lists “a small beep-beep
French horn” for 74¢. New York Times, June 5, 1933, 6.
4. See Axel Volmar, “Productive Sounds: Touch-Tone Dialing, the Rise of the
Call Center Industry and the Politics of Virtual Voice Assistants,” in The
Democratization of Artificial Intelligence: Net Politics in the Era of Learning
Algorithms, ed. Andreas Sudmann (Bielefeld: Transcript, 2019), 55–75.
5. In 1957 a retailer in Kansas City, MO had the Sputnik beep recorded and
pressed onto vinyl records, selling them for $2.95. See Mark Shanahan,
Eisenhower at the Dawn of the Space Age: Sputnik, Rockets, and Helping
Hands (Lanham, MD: Lexington Books, 2017), 82. Thanks to Dan Melamed
for mentioning Sputnik.
6. Peter E. Hunn, “CONELRAD,” in Christopher Sterling, ed., Encyclopedia of
Radio, vol. 1 (A–E) (New York: Fitzroy Dearborn, 2004), 605–607.
7. Bob Dylan mentions CONELRAD in the ninth verse of “Talkin’ World War
III Blues” (1963), and in the tenth verse he mentions dialing the operator
and hearing “When you hear the beep it will be three o’clock”; the song
juxtaposes two contemporaneously prominent beep contexts.
8. Ross Bassett, To the Digital Age: Research Labs, Start-up Companies, and the
Rise of MOS Technology (Baltimore, MD: Johns Hopkins University Press,
2002), 1.
9. See Jeffrey Zygmont, Microchip: An Idea, Its Genesis, and the Revolution It
Created (Cambridge, MA: Perseus Publishing, 2003), xviii and passim. For
example, early personal computer speakers producing simple beeps were
called “bippers” or “beepers”; see Karen Collins, Game Sound: An Introduc-
tion to the History, Theory, and Practice of Video Game Music and Sound
Design (Cambridge, MA: MIT Press, 2008), 29.
10. On the functional and historical differences between sound reproduction

and synthesized sound, see Jonathan Sterne, The Audible Past: Cultural
Origins of Sound Reproduction (Durham, NC: Duke University Press, 2003),
34–35.
11. Giovanni Arrighi, The Long Twentieth Century: Money, Power, and the Origins
of Our Times (London: Verso, 2010).
12. I draw heavily on Amy Glasmeier, Manufacturing Time: Global Competition
in the Watch Industry, 1795–2000 (New York: Guilford, 2000), 88–106 (on
Swiss watchmaking), 107–129 (on US watch production), 130–154 (on subse-
quent competition between the two), and 155–177 (on the Japanese watch
industry).
13. This is in comparison to Britain. See Glasmeier, Manufacturing Time, 92.
14. Glasmeier, Manufacturing Time, 128.
16. Glasmeier, Manufacturing Time, 177; figures from Table 8.4 (174–175).
17. On the Japanese and US-American scientists, entrepreneurs, and compa-
nies involved, see Bob Johnstone, We Were Burning: Japanese Entrepreneurs
and the Forging of the Electronic Age (Boulder, CO: Basic Books, 1999), 1–117.
On Japanese economic competitiveness vis-à-vis the US in relative capi-
tal and labor costs during the 1970s–80s and decreasing US advantages in
energy prices (due to the 1973 and 1979 energy crises), see Dale Jorgenson
and Masahiro Kuroda, “Productivity and International Competitiveness in
Japan and the United States, 1960–1985,” in Productivity Growth in Japan and
the United States, ed. Charles Hulten (Chicago: University of Chicago Press,
1990), 29–57, especially 38–41. The pre-OPEC oil crisis period (1968–1973)
witnessed intensive Japanese R&D investment in primary electrical and
electronics technologies, leading to a large accumulation of R&D stock
(including in semiconductor microchip manufacturing) that propelled the
digital watch market in the mid-1970s and after, continuing post-crisis and
scaling upward more effectively in Japan than the US. See M. Ishaq Nadiri
and Ingmar R. Prucha, “Comparison and Analysis of Productivity Growth
and R&D Investment in the Electrical Machinery Industries of the United
States and Japan,” in Productivity Growth in Japan and the United States,
109–133.
18. “Swatch,” Wikipedia, http://en.wikipedia.org/wiki/Swatch, accessed Octo-
ber 21, 2019.
19. Perry Anderson, “Jottings on the Conjuncture,” New Left Review II/48 (No-
vember–December 2007), 6.
20. See, for example, Mizuko Ito, et al., eds., Personal, Portable, Pedestrian: Mo-
bile Phones in Japanese Life (Cambridge, MA: MIT Press, 2005).
21. See Shaul Katzir, The Beginnings of Piezoelectricity: A Study in Mundane
Physics (Dordrecht: Springer, 2006); also, see Takuro Ikeda, Fundamentals
of Piezoelectricity (Oxford: Oxford University Press, 1996), 1–3; and “Quartz
Clock,” Wikipedia, http://en.wikipedia.org/wiki/Quartz_clock, accessed
May 21, 2009.
22. David Landes, Revolution in Time: Clocks and the Making of the Modern
World (Cambridge, Mass.: Belknap Press, 1983), 342–3 and esp. 377.
23. See “Watch,” Wikipedia, http://en.wikipedia.org/wiki/Wristwatch, accessed
May 21, 2009; and Carlene Stephens and Maggie Dennis, “Engineering Time:
Inventing the Electronic Wristwatch,” British Journal of the History of Science
33 (2000): 492–4.
24. See the Bond film clip at http://www.oldpulsars.com/ as of October 26, 2019.
26. Ernest Braun and Stuart MacDonald, Revolution in Miniature: The History
and Impact of Semiconductor Electronics, rev. 2nd ed. (Cambridge: Cam-
bridge University Press, 1982), 181–183.
27. See the Rotary ad here https://www.youtube.com/watch?v=G4Zfwq_2hLk
(accessed October 26, 2019) and the Seiko ad here https://www.youtube.
com/watch?v=IrP79jWm3Qc (accessed October 26, 2019).
28. See “Making the Digital Watch Serve the Customer,” in Tom Hyltin, The Digi-
tal Electronic Watch (New York: Van Nostrand Reinhold, 1978), 39–40.
29. Susan Kuzca, untitled, United Press International, 3 June 1981.
30. Pieter Doensen, Watch: History of the Modern Wristwatch: Design 1950–1983,
Electric 1950–1993 (Gent: Snoeck Ducaju & Zoon, 1994), 197; anonymous,
Casio press release, PR Newswire, 12 June 1980.
31. Japan joined the newly formed G6 in 1975 with France, Italy, West Germany,
the UK, and US.
32. “Q&A: How Alarm Watches Make Their Noise,” New York Times, 15 August 1989.
33. See Larry Humes, “What Is ‘Normal Hearing’ for Older Adults and Can
‘Normal-hearing Older Adults’ Benefit from Hearing Care Intervention?”,
The Hearing Review, July 14, 2020, https://www.hearingreview.com/in-
side-hearing/research/what-is-normal-hearing-for-older-adults, accessed
September 6, 2020.
34. Indeed, 4096 Hz is close to the frequency with the lowest decibel threshold;
see the chart in R. Murray Schafer, The Soundscape: Our Sonic Environment
and the Tuning of the World (Rochester, VT: Destiny Books, 1994), 116. Thanks
to Anand Gopinath, who suggested price determined the watch alarm
signal’s frequency and mode of amplification.
35. On the running boom, see Sumanth Gopinath and Jason Stanyek, “Tuning
the Human Race: Athletic Capitalism and the Nike+ Sport Kit,” in Music,
Sound, and Space: Transformations of Public and Private Experience, ed.
Georgina Born (Cambridge University Press, 2013), 128–148. Also see E. P.
Thompson, “Time, Work-Discipline, and Industrial Capitalism,” Past &
Present 38 (1967), 56–97.
36. James P. Sterba, “By 1990, Modern Quartz Will Have Nearly Silenced the
Tick,” New York Times, 9 February 1982.
37. Landes, Revolution in Time, 353; Richard Porch, “The Digital Watch: Tribal
Bracelet of the Consumer Society,” Design Issues 2/2 (Autumn 1985), 46.
38. Martin Smith (of the Orange County Register), “Technology’s Timebomb
Ticks,” The Oregonian, 3 January 1989.
39. Editorial, “Modern Times: Beeps,” New York Times, 18 January 1982. In

meetings during my father’s employment at Pullman in the late 1970s–ear-
ly 1980s, watches would start to beep around the hour, and the presiding
works manager would get annoyed and stop these meetings temporarily.
Digital watches were typically owned by men; this may still be true today.
(The gendering of the digital watch merits much greater study.) Sudhir
Gopinath, conversation on August 21, 2013.
40. Glasmeier, Manufacturing Time, 255. Moreover, as luxury goods, mechan-
ical watches made a comeback. See Joe Thompson, “Mechanical Watch-
es Almost Disappeared Forever. Here’s How They Didn’t,” Bloomberg,
4 January 2018, https://www.bloomberg.com/news/articles/2018-01-04/
how-mechanical-watches-survived-after-quartz-a-concise-history, accessed
September 1, 2020.
41. Carlene Stephens and Maggie Dennis, “Engineering Time: Inventing the Elec-
tronic Wristwatch,” British Journal for the History of Science 33 (2000), 497.
42. See Mike Murphy, “The Apple Watch is a Great Smart Watch and Finally a
Good Dumb One Too,” Quartz, 23 October 2019, https://qz.com/1719176/the-
apple-watch-series-5-is-the-best-one-yet/, accessed October 26, 2019.
43. Older smartwatches were peripherals, dependent on stronger devices.
Susan Ryan, Garments of Paradise: Wearable Discourse in the Digital Age
(Cambridge, Mass.: MIT Press, 2014), 235.
44. Annie Van Cleve, “A Time for Youth: Wristwatch Losing Ground in High-
Tech World,” The Capital Times (Madison, WI), December 1, 2006, http://
watch-talk.blogspot.com/2006/12/wristwatch-losing-ground-in-high-tech.
html, accessed 9 September 2014. Also see “Smartwatch,” Wikpedia.org,
http://en.wikipedia.org/wiki/Smartwatch, accessed October 22, 2019.
45. Landes, Revolution in Time, 365. After WWII, “cheap time” was purveyed
most by US-based budget watchmaker Timex.
46. See my The Ringtone Dialectic: Economy and Cultural Form (Cambridge, MA:
MIT Press, 2013).
47. See Larissa Hjorth, “Odours of Mobility: Mobile Phones and Japanese Cute
Culture in the Asia-Pacific,” Journal of Intercultural Studies 26 (2005), 39–55;
and Sianne Ngai, Our Aesthetic Categories: Zany, Cute, Interesting (Cam-
bridge, MA: Harvard University Press, 2012), 77–86.
48. See Melissa Gregg and Tamara Kneese, “Clock,” in Timon Beyes, et al., The
Oxford Handbook of Media, Technology, and Organization Studies (New York:
Oxford University Press, 2019), 95–105.
49. Scientific pitch, present in older digital casino games and slot machines
(also tuned to scientific C) as well as digital watches and alarms, is an
underappreciated facet of the auditory culture of an older digital moder-
nity. Thanks to Dan Melamed for this argument. Also see Bruce Haynes, A
History of Performing Pitch: The Story of ‘A’ (Lanham, MD: Scarecrow Press,
2002), 42, and “The Science of Music: What Is Beauty, and Why Is It Nec-
essary?”, The Schiller Institute, n.d., https://archive.schillerinstitute.com/
music/rev_verdituning.html, accessed October 22, 2019.
About the Author
Sumanth Gopinath is Associate Professor of Music Theory in the School of

Music at the University of Minnesota. His research interests include musical
minimalism, sound studies, new media, and experimental and popular
music. He is the author of The Ringtone Dialectic and co-edited volumes on
mobile music and Steve Reich.
Part IV
Futures
13. Captured Time: Eye Tracking and the
Attention Economy
Alexander Monea
Abstract
This chapter examines the history how eye tracking came to stand in
as a measurement for what people pay attention to. I argue that this
connection between eye movement and attention is problematic and may
lead to undesirable developments in the contemporary attention economy
as it is implemented across digital platforms and smartphone apps in
the near future—e.g., ads that pause playing when you look away from
the screen. The chapter traces the emergence of eye tracking technology
in the late nineteenth and early twentieth centuries and shows how
its implementation to measure attention in the mid twentieth century
cemented its purpose and has driven its development ever since.
Keywords: eye tracking, surveillance, attention, human-computer

interaction, media studies, history of technology
It is curious that today there are nearly as many popular authors arguing
that attention can be captured and sold reliably enough to form the bed-
rock of the contemporary economy as there are authors arguing that with
self-discipline we can harness our attention to either succeed in or escape
from that very same contemporary attention economy.1 The collective
wisdom across these texts is something like the following: when an indi-
vidual mechanism of attention capture works, it’s zombifying capacities
are near inescapable, but any individual mechanism of attention capture
will not be very efficient and will be highly exploitable, subvertable, or
even just ignorable. As such, the infrastructure of the attention economy
has been one of pure bombardment, of inefficient spamming, which when
collectively constant and immersive ensures that you are everywhere and
doi 10.5117/9789463727426_Ch13
244 Alex ander Monea
always ensnared by at least one or two of these mechanisms while you

ignore the rest. There is thus a fundamental ambiguity in the term. Our
attention has become both our greatest weakness and the source of our
salvation in the information age.
The reason that attention presents us with this paradox today is because
of its historical fusion with vision. It is this assumption—that the movement
of the eyes constitutes a sufficient stand-in for mental focus—that renders
the temporality of attention available for quantification and thus, in the
terms of Lorraine Daston and Peter Galison, “mechanically objective”
knowledge.2 As we’ll see, the slippage between the qualitative temporality
of everyday attention and the quantitative temporality of tracked eye
movements is one reason for this paradox—the second will be the radical
unpredictability of eye movements. Nonetheless, our culture is dominated
by an idea that the patterning of lived temporal phenomena can be put into
a circulatory relationship with what media theorists like Axel Volmar or
Wolfgang Ernst have described as the microtemporalities of time-critical
media like eye tracking sensors and the flow of electrons across the silicon
circuit boards of computational devices.3 Through the mediation of eye
tracking, the thoughts of mind and machine might be tethered across their
temporal registers. Mind and machine might be made to correspond with
and to pattern each other, which, in our most dystopic nightmare and the
wildest dreams of platform shareholders, might lead to a command and
control structure.
The first section examines the history of eye tracking technologies from
their conception in the nineteenth century to their stabilization in the
twentieth century. By mid-century, eye tracking technology was accurate,
available, and affordable—it was a hammer waiting for a nail, so to speak.
The second section shows how eye tracking came to be nearly synonymous
with tracking attention. Cyberneticists and psychologists looking to measure
attention turned to eye tracking as the only objective metric available
for attention. This cybernetic legacy continually inspires the rhetoric in
which eye tracking is positioned as a mechanism of command, control
and communication. While this representation has always been fraught,
eye tracking research often slips back into a reliance on science fiction
as it acknowledges but then quickly forgets its own limitations. The final
section argues that eye tracking technology currently constitutes the largest
gap in the attention economy and that contemporary technologies make it
increasingly feasible that attention merchants will try to capture that data.
I posit two dystopic futures, one soft and one hard. In the soft dystopia,
attention merchants are able to capture eye tracking data to, for instance,
Captured Time: Eye Tr acking and the Attention Economy 245
make sure we pay for freemium content with our attention to advertisements.
In the hard dystopia, the specificity of each individual’s eye movements
will become machine learnable and thus dynamic design processes will
be able to modulate our perception—and perhaps attention—in a much
more consistent way.
The future of the attention economy from digital platforms to smartphone
apps could be inflected by the advent of eye tracking technology being
placed by default in all our screen-based interfaces. The use of ubiquitous
eye tracking technologies to pattern the temporal flow of attention will likely
constitute a speed-up and convenience for some users while constituting
a slow-down and burden for others. It is easy to imagine a future where
affluent users access premium dynamic interfaces that increase their speed
and productivity while a large portion of less privileged users have their
attention trapped and monetized in exchange for access to apps, platforms,
or the internet writ large—think Zuckerberg’s “free” internet (via drones,
satellites, or balloons) on steroids.
The Mechanization of Eye Tracking, or Why the Eye Can’t be

Controlled
Throughout the nineteenth century many attempts were made to mea-

sure eye movement. Louis-Émile Javal, for instance, used afterimages to
determine that there was no vertical displacement of the eye as it glided
horizontally across text during reading. 4 This experiment is popularly
understood as the origin point of modern eye tracking. A number of
contemporaneous researchers were considering using feathers and bristles
attached to the eye to record its movements.5 By the turn of the century,
Edmund Burke Huey published a groundbreaking study on the psychology
of reading. Huey had successfully used plaster-of-Paris to build a mold that
fitted over the cornea that was connected by a light celloidin-covered glass
lever to a kymograph (fig. 13.1). The glass lever maneuvered the aluminum
pointer of the kymograph to record its movements on a smoked drum,
and later on smoked paper, as subjects read varying texts (fig. 13.2). Huey
writes, “During the reading, the reader was usually quite unconscious of
there being an attachment to his eye, and the reading proceeded as glibly
and easily as could be desired,” perhaps because he rendered participants’
eyes insensitive by the application of holocain or cocaine.6 Provided the
subject’s head was fastened between iron standards to keep it still, he could
successfully record their eye movements (fig. 13.1). Thus, by limiting the
Figure 13.1. Huey’s eye tracking apparatus
Figure 13.2. Scanpaths recorded by Huey

subject’s mobility in space, Huey could capture the foci of the subject’s
eye movements in time, thus allowing what would retrospectively be
enshrined as one of the f irst successful synchronizations of mind and
machine.
Despite Huey’s claims, the mechanical properties of his apparatus
limited its ability to objectively capture eye movement, in particular
because the inertia from the apparatus caused overshoots in the recorded
traces.7 In 1901, Raymond Dodge and Thomas Cline wrote, “Out of this
advance in the physiology of vision has developed a constantly growing
group of psychological questions which only a quantitative knowledge
of the eye movements can answer.”8 Dodge and Cline worked to develop
an apparatus capable of making exact measurements of eye movements,
with a particular focus on their angle velocity. A successful apparatus
would need to register eye movement without producing momentum
or inertia.9 The only possible solution was film, the perfect medium for
turning a sequence of captured spatial coordinates into a simulation of
phenomenological temporality. In a move that would determine many of
our contemporary eye tracking technologies, Dodge and Cline would catch
reflected light off the cornea on a sensitive film. In an apparatus jerry-built
with knitting needles, bicycle pumps, cardboard, and a modif ied 5 x 7
bellows camera, they would secure their subject’s head and photograph
their eye movements (fig. 13.3).
Their most important invention was a plate-holder inside the camera
that would move the f ilm vertically in a continuous and even motion
immediately behind a narrow horizontal slit through which the reflected
light off the cornea was received (fig. 13.4). In addition, they used a pendulum
within the plate-holder to create a time record by having its oscillations
periodically intercept the light allowed in by one side of the horizontal slit
in the plate-holder (fig. 13.5).
In later studies, Dodge would come to call these exposures ‘kinetograms’
(fig. 13.6).10 With these kinetograms, a mechanically objective representation
of the temporality of eye movements became available to researchers. The
reliability and cost effectiveness of this new medium far exceeded the
limited demands of eye tracking research to that date, still largely focused
on reading and visual impairments. It was, in a sense, a hammer awaiting
a nail that would not appear until eye movements became a stand-in for
human attention writ large. While this conjuncture would be instituted
largely by psychologists outside of opthamology laboratories, the stage was
set by researchers who moved from analyzing scanpaths during reading to
scanpaths during analysis of images.
Figure 13.3. Dodge and Cline’s eye tracking Figure 13.4. Dodge and Cline’s film box
apparatus
Figure 13.5. Dodge and Cline’s film recording of Figure 13.6. Dodge and Cline’s film recording of
eye movements eye movements
Figure 13.7. Buswell’s eye tracking apparatus
The use of film to capture eye movements became the norm in eye tracking
research for nearly 70 years and was quickly applied outside of the study of
reading. George Malcolm Stratton would use a similar photographic tech-
nique to examine eye movements when subjects viewed simple geometrical
patterns and line illusions. Stratton quickly found that he could reliably
predict the scanpaths that his subjects’ eyes would take when presented
with these patterns,11 and was particularly unnerved by how disharmonious
eye movements could produce the aesthetic experience of symmetry.12 As
Wade notes, “Stratton’s work is significant because it attempted to bridge the
gap between visual phenomena (illusions), cognition (aesthetic judgments),
and the underlying mechanisms (eye movements),” and further because
Stratton shifted the focus of scientific study from the movement of the
eyes to the locations selected for fixation.13 This focus on how we select
particular locations to fixate on would become a central research question
Figure 13.8. Buswell’s analysis of eye movements during viewing of painting
moving forward and would set the stage for the fusion of eye movement
and attention research.
The next major advancement came in 1935 when Guy Buswell published the
eye tracking records of nearly 2,000 individual scanpaths from 200 participants
each of whom viewed multiple pictures through an apparatus very similar
to Dodge and Cline’s (fig. 13.7).14 Rather than focusing on the eye movements
themselves, Buswell analyzed the fixation points, publishing density plots
from all of his recorded scanpaths (fig. 13.8). By reducing the spatial mobility
of the subject’s head, Buswell was able to use film to capture the temporal
dimensions of eye movements as sequences of fixations. This reduction then
allowed for a spatialization of these temporal dimensions (fig. 13.8).
From this innovative work, Buswell made some startling discoveries
about the uniqueness of his participants’ scanpaths. People tended to have
very different scanpaths when presented with the same picture, while only
the first handful of fixation points had even a semblance of homogeneity.
Additionally, the average duration of fixations for each picture varied widely
across observers. Buswell examined whether these differences correlated
well with other variables like artistic training, age, and race/ethnicity, but
found that within group variation always exceeded the average variation
between groups to such an extent that the correlations could not be consid-
ered significant.15 In essence, people’s phenomelogical temporalities were
radically individualized. While mechanically objective recording of eye
movements was now able to synchronize with the flow of phenomenological
vision in its past and present foci, it remained unable to predict future foci.
Alfred L. Yarbus would eventually extend this research by fusing the
film recordings with suction cup eye devices to investigate miniature eye
movements during fixation, detailed kinematics of individual saccades,
vergence, pursuit, and, perhaps most importantly, eye movements during
perception of complex objects.16 For this last emphasis Yarbus famously
analyzed the eye movements of subjects as they viewed Ilya Repin’s painting
The Unexpected Visitor. He found that while there was some homogeneity
in the scanpaths, especially in the earlier points of fixation, the scanpaths
were highly individualized. Additionally, when the same subject viewed
Figure 13.9. Yarbus’s analysis of scanpaths when viewing painting based on priming.
the same painting a number of separate times, each successive viewing’s

scanpath would be different, although the difference in scanpaths was
greater between different subjects than between successive viewings by
the same subject.17 In perhaps the most enduring illustration of early eye
tracking research, Yarbus extended Buswell’s research into priming’s effect
on scanpaths (fig. 13.9). As Yarbus notes, “Depending on the task in which a
person is engaged, i.e., depending on the character of the information which

he must obtain, the distribution of the points of fixation on an object will
vary correspondingly, because different items of information are usually
localized in different parts of an object.”18
The role of priming and other cognitive factors in determining scanpaths,
and in particular the sequence of selected points for fixation, is still central
to eye tracking research and difficult to predict with much accuracy.19
Yarbus’ study was recently reproduced and expanded to demonstrate that
what he found after priming a single subject upon multiple viewings of The
Unexpected Visitor was generalizable to multiple subjects.20
Cybernetic Analysis of Black Boxes, or Why People Try to Control

the Eye
It is in Paul Fitts’ research from the 1940s that eye tracking first was welded to
human attention. Here the foci of eye movements stand in for the foci of men-
tal attention and thus the scanpaths give an approximation of the sequence
and rhythm of human thought—particularly as a stream of consciousness.
Fitts was a pioneer in human factors, served as the sixth president of the
American Psychological Association, and developed important laws and
formulae for how humans use visual and proprioceptive feedback to situate
themselves in space and time. He explicitly positioned himself in a tradition
of interdisciplinary military research that was meant to better adapt humans
and machines for cooperative interaction in combat, with a particular focus
on aviation.21 The goal of much of this collaborative research between the
American Psychological Association and the military was to better select
and train pilots, better design aviation technology (esp. cockpits), and thus
to ensure optimal air force capacities. For Fitts, this was fundamentally an
informational problem about the visual and proprioceptive feedback loops
that allow a pilot to interact with an aircraft.22 This approach situates Fitts
in a transitional discursive tradition between behaviorism and cognitivism
that Paul Edwards has described as “cybernetic psychology.”23
Behaviorism understood the human mind to be a black box that could
only be studied through measuring its inputs (stimuli) and outputs (behavior)
with the ultimate goal of producing a stable theory of the functional relation
between the two such that behavior might be controlled through the engineer-
ing of stimuli. Cybernetics largely shifted the metaphor of the black box onto
electronics and used information theory to produce new methods of inferring
the contents of black boxes through analysis of random noise inputs, and then
used the analogy of mind and machine to argue that the same mathematical
analyses useful in understanding electronics could similarly be applied to
the human mind.24 The goal of cybernetic psychology was “to understand
the processes of perception, memory, and language in terms of formalizable
transformations of information and feedback circuits or control loops.”25
While some psychologists directly equated human attention with a
mechanical model of sensory inputs,26 Fitts was more open about the fact
that the connection between thought, attention, and vision was more a
matter of convenience. He writes, “If we know where a pilot is looking we
do not necessarily know what he is thinking, but we know something of
what he is thinking about. In servo terminology, we know what error signal
inputs he is operating on.”27 What the eye is focused on is only roughly
correlated to what the mind is paying attention to, which in turn is only
roughly corelated to what the mind is thinking about. Eye movements and
fixations were chosen as inputs to be studied because they were the only
variables that could be objectively observed and mechanically reproduced.
In short, eye movements became the stand-in for attention—and thus the
temporality of thought—because eye tracking technology was sophisticated,
affordable, and accurate enough to serve scientific and military purposes.
This is the point of fusion between a technology waiting for an application
and a theory of mind looking for a method. This entanglement of thought,
attention, and vision became essential to cybernetics as a whole.28 Take,
for example, the cybernetic understanding of teleology outlined by Arturo
Rosenblueth, Norbert Wiener, and Julian Bigelow.29 They understood hu-
mans and machines to pursue goals in the same way, by utilizing a cycle
of predicting future states, taking actions, and self-correcting their course
of action based on negative feedback received. Goals are “a final condition
in which the behaving object reaches a definite correlation in time or in
space with respect to another object or event.”30 The sensory feedback
received is primarily visual. Warren McCulloch went so far as to argue
that the mind was dominated by the eye. He writes, “The eye is not only
the most important of sense organs. It is the most complicated, being in
reality an invaginated evagination of the brain itself.”31 Here McCulloch is
arguing that in vision the brain is externalized and folded back in on itself;
it touches itself as it touches the world. This is no more nor less than the
cybernetic version of Maurice Merleau-Ponty’s insights about the chiasm—
the entanglement of sensor and sensation—whose anatomical definition
interestingly corresponds to the crossing optic tracts of the eyes that is so
fundamental to binocular vision.32 As Rudolf Arnheim has shown at great
length perception contains cognition, or as he puts it, “visual perception is
visual thinking.”33 This is central to cybernetic psychology: we must examine

sensory inputs—vision in particular—if we are to understand how the
mind works as a mechanism. In the context of aviation, as Fitts notes, “we
cannot study man’s motor system at the behavioral level in isolation from
its associated sensory mechanisms.”34
Eye tracking research continues to bear the imprint of this cybernetic
legacy today. The textbook Eye Tracking Methodology describes the visual
system repeatedly in terms of information processing, noise reduction,
command and control systems, and feedback circuits and loops—including
as an “attentional feedback loop.”35 While the author is careful to note that he
is speaking of visual attention and not attention writ large, the analogizing
of mind and machine leads him to habitually conflate the two in practice.
For instance, Duchowski refers to eye tracking data as “objective evidence
provided by users’ gaze and hence attention.”36 Wilson Geisler and Lawrence
Cormack similarly describe attention with a cybernetic vocabulary, and
Árni Kristjánsson goes so far as to argue that eye movements and attention
largely share the same neural resources.37 This is a discursive convention
that companies producing eye tracking technologies like Tobii are all too
happy to encourage. In their online marketing materials for Tobii Pro they
write, “Eye tracking is a unique method to objectively measure consumers’
attention and spontaneous responses to marketing messages.”38
As I’ve shown, this repeated analogizing of mind and machine was
embedded in eye tracking discourse through cybernetics in the mid-twen-
tieth century. The result of that legacy is that we are led to believe that
human “systems” might be just as amenable to command, control, and
communication as computational systems. However, as I’ve shown in the
previous section, our scanpaths are too susceptible to priming and other
idiosyncrasies for this analogy to ever hold true. Instead, just as Colin
Milburn has shown in the context of nanotechnology, eye tracking is as
much science fiction as science.39 By this I do not mean that there is no real
science behind the use of eye tracking. There is a great deal of innovative and
rigorous research being conducted through these means. What I do mean is
that the investment in and development of eye tracking technologies is as
much driven by science fiction stories about successful applications in the
future as it is by current applications. Regardless of the success or failure of
any of these anticipated future applications, the discourse on eye tracking
will continue producing science fiction so long as its cybernetic genealogy
remains intact. Eye tracking seems forever destined to fail to live up to its
hype because of the individual specificity of eye movements. Yet, like it is
our political party or our compulsively philandering spouse, we can’t help
but believe its perpetual promises that it has turned a corner, that this time
it will be different, that things are finally going to work out.
Computational Eye Tracking and Two Dystopic Futures
In 2016, Nvidia publicly announced ongoing research into the use of eye
tracking technology to make gains in computer processing power for virtual
reality. By taking advantage of foveal vision, Nvidia realized it could drasti-
cally reduce the processing power required to render VR graphics. In essence,
Nvidia argues that VR headsets can render the graphics outside a designated
range of the current eye fixation in lower resolution without the user noticing
any ill effects. This may prove essential for the future of VR. As Simon Parkin
writes, “When the player using the Nvidia system focuses on a new area of
the scene, eye-tracking software shifts the focus of the rendering in kind.
To render a full scene in VR at 90 frames per second, the lowest acceptable
frame rate in VR before users begin to report feelings of nausea, four million
pixels must be rendered at almost a hundred times a second. But by focusing
the rendering only on the player’s line of sight, huge computational savings
can be made.”40 While Nvidia will not produce VR headsets with eye tracking
technology, a number of companies are already moving in this direction.
For example, the Kickstarter-backed Fove headset is meant to be completely
controlled by the user’s eye movements,41 and companies like SensoMotoric
Instruments and Tobii Tech are already working to adapt their eye tracking
technologies to VR headsets. As people increasingly access visual information
through VR media, the natural inclusion of eye tracking technology promises
to offer a fresh new stream of user data.
As far back as 2011, Tobii was partnering with Lenovo to produce a laptop
that could be controlled with eye motions, but these technologies have
yet to sweep the consumer market. 42 This may in part be due to the high
cost, need to accommodate cumbersome hardware, and inaccuracy during
real-world use that such systems have demonstrated. Recently we have
seen rapid advances in software-based solutions that take advantage of
preexisting technologies, like webcams and forward-facing cameras on
tablets and smart phones, whose fixed position relative to the screen can be
exploited for eye tracking purposes.43 Machine learning has left eye tracking
technology largely unimproved, 44 and this is likely due to limitations in the
availability of large-scale datasets of captured eye movements—most have
around 50 subjects.45 By 2016, researchers were constructing “a mobile-based
eye tracking dataset containing almost 1500 subjects from a wide variety of
background, recorded under variable lighting conditions and unconstrained

head motion.”46 They fed this data into a convolutional neural network
and produced a system that could match state-of-the-art eye tracking
performance. Further, they were able to use dark knowledge approaches
to simplify their model enough for it to run in real-time on mobile devices.47
Provided the ubiquity of eye tracking technology is solely a software problem,
it is safe to assume that we are not far from eye tracking technology being
a standard feature of all smartphones, tablets, and laptops.
The soft dystopic vision of the future is one in which the current trend of
using click-through, typed input, scrolling, and time-spent-on-page data to
make these attention capture mechanisms more efficient is supplemented
with eye tracking data at web scale. Once our VR headsets, laptops, tablets,
and smartphones incorporate eye tracking technology, it is inevitable that
our eye tracking data will be collected as we navigate our digital worlds in
exchange for access to new digital goods and services. The collection of vast
stores of user scanpaths offers up a first solution to the individuality of our
eye movements: we then might collect enough data per user to statistically
model their individual eye movement behaviors in response to various
stimuli and priming. As user interfaces and designs become increasingly
modular and dynamic, it is not hard to envision a world in which this
collection of our eye movements can lead to the personalization of our
interfaces through ongoing dynamic iterations. For instance, every Google
account, or even every Google search for every Google account, would
generate a particular optimal placement for “sponsored content” so that it
might draw the most attention and produce the most click-throughs. This
will never be a world in which we become perfect ideological drones ready
for ongoing hypnotic commands from global capital, but it will be one in
which the effectiveness and efficiency of our current attention capturing
mechanisms is vastly increased.
Even should dynamic or personalized interface designs not prove possible
with eye tracking data at web scale, there are even easier to imagine dystopic
futures that can arise when eye tracking technology is embedded in all of
our screen technologies as a default. The business models of most of our
digital platforms and the bulk of our “freemium” smartphone apps are based
on advertising revenue. Imagine a world in which these companies could
monitor the consumptive labor of viewing advertisements that their users
are performing for them. Imagine a world in which advertisements pause
when you stop looking at them. Where you can’t return to primary content
until you’ve paid attention. Imagine how much more these companies could
charge advertisers if they could guarantee eyeballs on their adds, literally. We
aren’t far away from a Candy Crush app that tracks your eye movements to
make sure you are watching the advertisements that unlock daily boosters,
or from public WiFi hotspots that require you to focus on intermittent
advertisements to stay connected. The most likely dystopia is one in which
consumptive labor—i.e. watching advertisements—becomes a stable market
and attention gets standardized into currency that you pay with. Here the
phrase “paying attention” will finally take on its full meaning under capital.
In this first dystopic vision of the future one can already see the potential
emergence of even greater class and gender differentiation in the temporal
patterning of digital flows. As Axel Volmar and Kyle Stine have pointed
out in their introduction to this volume and Sarah Sharma has analyzed
in great detail, 48 there already exists a great class differentiation in our
contemporary lived temporalities, as the optimization and acceleration
of affluent people’s everyday lives is a privilege borne on the backs of the
underprivileged, whose lived temporalities are relationally overburdened,
disrupted, and thrown out of sync with the world. We already have an
instance akin to my imagined dystopia in the use of a platform like YouTube,
which serves more ads to people accessing the site via mobile phones than
on desktops and laptops—and keep in mind that the underprivileged people
of the globe are much more likely to only have internet access through their
mobile devices. One can easily envision the exacerbation of these divides
by imagining the future of affluent, white, cis-gendered men paying to use
optimized interfaces in the future to speed up their digital lives on platforms
subsidized by the attentional labor of, for instance, working class, black,
single mothers forced to pay attention to every second of the ads supporting
their access to digital platforms. This situation will only be worse in areas of
the globe without affordable broadband access. As platforms like Facebook
look to offer drone, satellite, and balloon-based mobile internet access to
their closed iterations of the “internet,” it is easy to imagine a future in
which these people’s entire internet access is facilitated by the extraction
of attentional capital through eye tracking technologies. In this future,
the digital temporality of large swaths of the globe might be patterned in
accordance to these attentional extraction mechanisms.
In the hard dystopic vision of the future is one in which this process
is much more successful than can be reasonably hoped for. Here we can
envision a process whereby the personalization afforded by eye tracking data
will lead to completely individualized interfaces and designs statistically
optimized to capture attention by structuring the sequence of points of
fixation in our scanpaths. This is a literal determination not only of what
but of how we see. In his book Visual Thinking, Rudolf Arnheim argues that
the “cognitive operations called thinking are not the privilege of mental
processes above and beyond perception but the essential ingredients
of perception itself,” and also that “visual perception … is not a passive
recording stimulus material but an active concern of the mind.”49 As such,
we can understand such a capacity for dynamic iterative design based on
personalized scanpath profiles to be a literal manipulation of mind. This
hard dystopia fits in well with a long tradition in film criticism ranging from
Horkheimer and Adorno’s woes over cinema relieving of us of any capacity
to imagine for ourselves to William Connolly’s more recent arguments that
the cinema can utilize affect to (re-)program its audience.50 The primary
difference here is that this will be a world in which the mechanism of
attention capture is individualized beyond all possibility of communal
experience, and thus short-circuits the capacity for building communities
through shared experience. In his three-volume work Technics and Time,
Bernard Stiegler worried that mass media have led to a world in which
everyone’s lived temporality is synchronized with broadcast media in what
he terms “the industrial temporalization of consciousness,” thus leading to
a cultural leveling and homogeneity. In this hard dystopia we can see the
opposite problem arising from network media and attentional modulation,
where every single person’s lived temporality is synchronized with a radically
individualized temporal patterning mechanism.51
The endpoint of eye tracking certainly does not need to be so grim. It
is extremely promising as a method for creating user interfaces that make
attention less intentional, so to speak, and thus less exhausting. This is
precisely the promise that Walter Benjamin saw in film: in opposition to
Horkheimer and Adorno, Benjamin argued that film was uniquely capable
of “distracted reception,” where audiences could develop a critical under-
standing of the film’s contents without having to expend the conscious
energy of intentionally directing their attention towards each nuance of
the film’s contents.52 Personalization based on scanpath models could do
something similar by making objects we need or ought to focus on easier to
pay attention to—more absorbing, interesting, or intuitive. It could be the
visual equivalent of Richard Thaler’s nudges.53 It might make learning or
communicating much easier or more efficient for many more people. The
problem is with the potential weaponization of this data, which seems all
but inevitable given the attentional infrastructure which such a technology
would get plugged into. In such an instance, what used to be our primary
defense against attention capture—namely, that eye movements and fix-
ations were highly individualized with no clustering of behavior around
demographic groupings—becomes our new primary weakness, which is
the personalization of attentional objects. Perfect echo chambers. Mass

solipsism. Catatonia. A horrific monadology. The true boob tube. A real-life
version of Ringu in which the husks of departed souls are found slouched
before screens that just read as static to second parties.
Notes
1. For evidence of the former, see Thomas H. Davenport and John C. Beck, The
Attention Economy (Cambridge, MA: Harvard Business Review Press, 2001);
Mara Einstein, Black Ops Advertising: Native Ads, Content Marketing and the
covert World of the Digital Sell (New York: OR Books, 2016); Nir Eyal, Hooked:
How to Build Habit-Forming Products, ed. Ryan Hoover (New York: Penguin,
2014); Ben Parr, Captivology: The Science of Capturing People’s Attention
(New York: HarperOne, 2016); Jonathan Taplin, Move Fast and Break Things:
How Facebook, Google, and Amazon Cornered Culture and Undermined
Democracy (New York: Hachette, 2018); James G. Webster, The Marketplace
of Attention: How Audiences Take Shape in a Digital Age (Cambridge, MA:
MIT Press, 2016); Tim Wu, The Attention Merchants: The Epic Scramble to Get
Inside Our Heads (New York: Vintage, 2017). For examples of the latter, see
Chris Bailey, The Productivity Project: Accomplishing More by Managing Your
Time, Attention, and Energy (New York: Crown Business, 2017); Matthew B.
Crawford, The World beyond Your Head: On Becoming an Individual in an Age
of Distraction (New York: Farrar, Straus and Giroux, 2016); Daniel Goleman,
Focus: The Hidden Driver of Excellence (New York: HarperCollins, 2015); Neen
James, Attention Pays: How to Drive Profitability, Productivity, and Account-
ability (Hoboken, NJ: Wiley, 2018).
2. Lorraine Daston and Peter Galison, Objectivity (Cambridge, MA: Zone
Books, 2007).
3. Axel Volmar, ed. Zeitkritische Medien (Berlin: Kadmos, 2009); Wolfgang
Ernst, Digital Memory and the Archive, ed. Jussi Parikka (Minneapolis: Uni-
versity of Minnesota Press, 2013).
4. Louis É. Javal, “Essai sur la phisiologie de la lecture,” Annales d’Oculistique
80 (1878): 240–274.
5. August Ahrens, Die Bewegung der Augen beim Schreiben (Rostock, Germany:
University of Rostock, 1891); Edmund B. Delabarre, “A Method of Recording
Eye Movements,” American Journal of Psychology 9, no. 4 (1898): 572–572;
Edmund B. Huey, “Preliminary Experiments in the Physiology and Psychol-
ogy of Reading,” American Journal of Psychology 9, no. 4 (1898): 575–586;
Edmund B. Huey, The Psychology and Pedagogy of Reading, with a Review
of the History of Reading and Writing and of Methods, Texts, and Hygiene in
Reading (New York: MacMillan, 1908); E. Rählmann, “Über den nystagmus
und seine ätiologie,” Archiv für Ophthalmologie 24 (1978): 237–242.
6. Huey, “Preliminary Experiments in the Physiology and Psychology of Read-

ing,” 25–26.
7. Raymond Dodge and Thomas S. Cline, “The Angle Velocity of Eye Move-
ments,” Psychological Review 8, no. 2 (1901): 147.
8. Dodge and Cline, “The Angle Velocity of Eye Movements,” 145.
9. Dodge and Cline, “The Angle Velocity of Eye Movements,” 147.
10. Raymond Dodge, “Five Types of Eye Movement in the Horizontal Meridian
Plane of the Field of Regard,” American Journal of Physiology 8, no. 4 (1903):
307–329; Raymond Dodge, “The Participation of Eye Movements in the
Visual Perception of Motion,” The Psychological Review 11, no. 1 (1904): 1–14.
11. George M. Stratton, “Eye-Movements and the Aesthetics of Visual Form,”
Philosophische Studien 20 (1902): 343.
12. Stratton, “Symmetry, Linear Illusions and the Movements of the Eye,” The
Psychological Review 13, no. 2 (1906): 95.
13. Nicholas J. Wade, “Pioneers of Eye Movement Research,” i-Perception 1, no. 2
(2010): 57.
14. Guy Buswell, How People Look at Pictures: A Study of the Psychology and
Perception in Art (Chicago: University of Chicago Press, 1935).
15. Buswell, How People Look at Pictures, 131.
16. Alfred L. Yarbus, Eye Movements and Vision, trans. Basil Haigh (New York:
Plenum Press, 1967).
17. Benjamin W. Tatler, Nicholas J. Wade, Hoi Kwan, John M. Findlay, and Boris
M. Velichkovsky, “Yarbus, Eye Movements, and vision,” i–Perception 1, no. 1
(2010): 7–27.
18. Yarbus, Eye Movements and Vision, 192. In reference to the figure, Yarbus
writes: “Seven records of eye movements by the same subject. Each record
lasted 3 minutes. The subject examined the reproduction with both eyes. 1)
Free examination of the picture. Before the subsequent recording sessions,
the subject was asked to: 2) estimate the material circumstances of the fam-
ily in the picture; 3) give the ages of the people; 4) surmise what the family
had been doing before the arrival of the “unexpected visitor”; 5) remember
the clothes worn by the people; 6) remember the position of the people and
objects in the room; 7) estimate how long the “unexpected visitor” had been
away from the family” (174).
19. Tatler et al., “Yarbus, Eye Movements, and Vision,” 24–25.
20. Marianne DeAngelus and Jeff B. Pelz, “Top-down Control of Eye Move-
ments: Yarbus Revisited,” Visual Cognition 17, no. 6–7 (2009): 790–811.
21. Paul M. Fitts, “German Applied Psychology during World War II,” American
Psychologist 1 (1947): 151–161.
22. Fitts, “The Information Capacity of the Human Motor System in Controlling
the Amplitude of Movement,” Journal of Experimental Psychology 47, no. 6
(1954): 381–382.
23. Paul N. Edwards, The Closed World: Computers and the Politics of Discourse in
Cold War America (Cambridge, MA: MIT Press, 1996).
24. Ross Ashby, An Introduction to Cybernetics (London: Chapman & Hall, 1956);
Norbert Wiener, Cybernetics: or Control and Communication in the Animal
and the Machine, 2nd ed. (Cambridge, MA: MIT Press, 1961).
25. Edwards, The Closed World, 179–180.
26. E.g. Donald E. Broadbent, “A Mechanical Model for Human Attention and
Immediate Memory,” Psychological Review 64, no. 3 (1957): 205–215.
27. Paul M. Fitts, Richard E. Jones, and John L. Milton, “Eye Movements of Air-
craft Pilots during Instrument-Landing Approaches,” Aeronautical Engineer-
ing Review 9, no. 2 (1950): 1–6.
28. Orit Halpern, Beautiful Data: A History of Vision and Reason since 1945
(Durham, NC: Duke University Press, 2015).
29. Arturo Rosenblueth, Norbert Wiener, and Julian Bigelow, “Behavior, Purpose
and Teleology,” Philosophy of Science 10, no. 1 (1943): 18–24.
30. Rosenblueth et al., “Behavior, Purpose and Teleology,” 18.
31. Warren S. McCulloch, “Information in the Head,” Synthese 9, no. 1 (1955): 235.
32. Maurice Merleau-Ponty, The Visible and the Invisible, trans. Alphonso Lingis,
ed. Claude Lefort (Evanston, IL: Northwestern University Press, 1968).
33. Rudolf Arnheim, Visual Thinking (Berkeley and Los Angeles: University of
California Press, 1969).
34. Fitts, “Information Capacity,” 281.
35. Andrew T. Duchowski, Eye Tracking Methodology: Theory and Practice, 3rd
ed. (Cham, Switzerland: Springer International, 2017). Duchowski even
describes the retina’s photoreceptors as “transducers” (18).
36. Duchowski, ix, 3–4.
37. Wilson S. Geisler and Lawrence K. Cormack, “Models of Overt Attention,”
in The Oxford Handbook of Eye Movements, eds. Simon P. Libersedge, Iain
Gilchrist, and Stefan Everling (New York: Oxford University Press, 2011),
439–454; Árni Kristjánsson, “The Intriguing Interactive Relationship
between Visual Attention and Saccadic Eye Movements,” in The Oxford
Handbook of Eye Movements, 455–470.
38. “Tobii Pro: Fields of Use,” Tobii, accessed May 12, 2018, https://www.tobiipro.
com/fields-of-use/.
39. Colin Milburn, Nanovision: Engineering the Future (Durham, NC: Duke
40. Simon Parkin, “Nvidia’s Eye-Tracking Tech Could Revolutionize Virtual Re-
ality,” MIT Technology Review, July 21, 2016, https://www.technologyreview.
com/s/601941/nvidias-eye-tracking-tech-could-revolutionize-virtual-reality/.
41. Parkin, “Point, Click, and Fire in Virtual Reality—with Just Your Eyes,”
MIT Technology Review, June 23, 2015, https://www.technologyreview.
com/s/538711/point-click-and-fire-in-virtual-reality-with-just-your-eyes/.
42. Tom Simonite, “A Laptop That Knows Where You’re Looking,” MIT Technol-
ogy Review, March 4, 2011, https://www.technologyreview.com/s/423198/a-
laptop-that-knows-where-youre-looking/. In fact, whole conventions have
since been held with prototypes demonstrating the future of commands in
screen-based media. See Simonite, “PC Makers Bet on Gaze, Gesture, Voice,
and Touch,” MIT Technology Review, January 10, 2013, https://www.technolo-

gyreview.com/s/509751/pc-makers-bet-on-gaze-gesture-voice-and-touch/.
43. Kyle Krafka, Aditya Khosla, Petr Kellnhofer, Harini Kannan, Suchendra
Bhandarkar, Wojciech Matusik, and Antonio Torralba, “Eye Tracking for
Everyone,” arXiv preprint arXiv:1606.05814 (2016): 2176–2184.
44. Xucong Zhang, Yusuke Sugano, Mario Fritz, and Andreas Bulling, “Appear-
ance-Based Gaze Estimation in the Wild,” in Proceedings of the IEEE Confer-
ence on Computer Vision and Pattern Recognition (2015), 4511–4520.
45. Qiong Huang, Ashok Veeraraghavan, and Ashutosh Sabharwal, “TabletGaze:
Unconstrained Appearance-Based Gaze Estimation in Mobile Tablets,”
Machine Vision and Applications 28, no. 5–6 (2017): 445–461; Yusuke Sugano,
Yasuyuki Matsushita, and Yoichi Sato, “Learning-by-Synthesis for Appear-
ance-Based 3D Gaze Estimation,” in Proceedings of the IEEE Conference on
Computer Vision and Pattern Recognition (2014): 1821–1828.
46. Krafka et al., “Eye Tracking for Everyone,” 2.
47. Geoffrey Hinton, Oriol Vinyals, and Jeff Dean, “Distilling the Knowledge in a
Neural Network,” arXiv preprint arXiv:1503.02531 (2015): 1–9.
NC: Duke University Press, 2014).
49. Arnheim, Visual Thinking, 13, 37.
50. Max Horkheimer and Theodor W. Adorno, Dialectic of Enlightenment, trans.
Edmund Jephcott, ed. Gunzelin S. Noerr (Stanford, CA: Stanford University
Press, 2007); William E. Connolly, Neuropolitics: Thinking, Culture, Speed
(Minneapolis, MN: University of Minnesota Press, 2002).
51. Bernard Stiegler, Technics and Time, 1: The Fault of Epimetheus, trans.
Richard Beardsworth and George Collins (Stanford, CA: Stanford University
Press, 1998); Bernard Stiegler, Technics and Time, 2: Disorientation, trans.
Stephen Barker (Stanford, CA: Stanford University Press, 2008); Bernard
Stiegler, Technics and Time, 3: Cinematic Time and the Question of Malaise,
trans. Stephen Barker (Stanford, CA: Stanford University Press, 2010).
52. Walter Benjamin, “On the Work of Art in the Age of Mechanical Reproduc-
tion,” trans. Harry Zohn, in Illuminations, ed. Hannah Arendt (New York:
Schocken, 1968), 217–253.
53. Richard H. Thaler and Cass R. Sunstein, Nudge: Improving Decisions about
Health, Wealth, and Happiness (New York: Penguin, 2009).
About the Author
Alexander Monea is Assistant Professor of English and Cultural Studies

at George Mason University (4400 University Drive MSN 3E4, Fairfax, VA
22030, United States). His research focuses on the history and cultural
impact of big data and machine learning. His upcoming book examines
heteronormative bias in internet content moderation.
14. Ahead of Time: The Infrastructure
of Amazon’s Anticipatory Shipping
Method
Eva-Maria Nyckel
Abstract
This text conducts a close reading of Amazon’s 2013 patent for a “Method
and System for Anticipatory Package Shipping” on three levels in order
to investigate the patent’s aspirations towards a potential hardwiring
of temporality. First, through the lens of media theory, the patent is
conceptualized as a medium for transporting knowledge over time itself.
On a second level, the patent is framed as a logistical medium both due
to its aspired effects in the logistical realm and its internal logic. Third,
the specific form of anticipation, prediction and prophecy is investigated
by leveraging Elena Esposito’s understanding of (digital) prophecy with
a particular focus on temporality.
Keywords: algorithms, Amazon, logistics, media theory, prediction,

temporality
What if effects preceded causes and answers preceded questions?

—John Durham Peters
In the industry of logistics, time is perhaps the most crucial resource not to be
wasted. A variety of reference parameters are dedicated to measuring time in
logistical processes, with names such as lead time, shipping time, and cycle
time, as well as time bucket, time fence, time to market, and time of circulation.
Reducing shipping latencies, providing real-time tracking of commodities,
and forecasting consumer demand are elementary to structuring temporal
doi 10.5117/9789463727426_Ch14
264 Eva-Maria Nyckel
relations between customers, carriers, and merchants. Yet even within this
context, Amazon’s 2013 patent for a “Method and System for Anticipatory
Package Shipping” marks a decisive shift in the industry’s thinking on time
saving in the way it extends logistics into the near-term future through a
practice of prediction-based package shipping ahead of incoming orders.1
The investigation of anticipatory package shipping, also referred to in the
patent text as a form of speculative shipping, is a speculative undertaking
in itself: At this point, we cannot know how far Amazon has actually
implemented the principles as described in the patent. The company has
neither confirmed nor denied the implementation of the patented meth-
ods.2 However, one could assume that Amazon’s logistical operations are
coordinated in a similar way and that the patent provides at least fractional
insights into how Amazon conducts logistical operations.3 On these grounds,
I approach the patent from three media-theoretical perspectives with an
eye to the system’s potential implementation, analyzing the medium of
the patent as a text genre with a specific temporality for the conveyance of
knowledge, conceptualizing the anticipatory shipping method as a new form
of logistical media for ordering time and space, and addressing the specific
form of prediction and corresponding temporality of the patented methods
in relation to Elena Esposito’s analysis of prophecies in the digital age.
The Patent as a Medium
Patents, as the name would suggest, are rich sources for scholarly investi-
gations in that they bring corporate research into the open. 4 In the case of
Amazon’s anticipatory shipping patent, however, the text does not entirely
deliver on this promise. Technical and methodical descriptions remain
rather vague, and the patent text does not clearly designate the nature of
the invention. As Albert Kümmel-Schnur has shown, it is not uncommon
for companies to submit and be awarded patents where the material or
process that is supposed to have been invented is unclear.5 This is the
case with Amazon’s patent for an anticipatory shipping process, which
represents a bundle of methods, not a concrete technology. A close reading
of the patent does not allow for an assessment of the concrete algorithms
that are to be used or the parameters crucial for the predictions. What is
described is rather a large number of different “embodiments,” or potential
material implementations, in an abstract way. The patent’s vagueness raises
a question of whether the company intentionally obfuscated corporate
knowledge to secure economic advantage over potential competitors. But
Ahead of Time 265
the language of patents in general might be a cause, being situated between

engineering-speak and legalese.
The medial form of the patent, as Christian Kassung argues, is one that
is necessarily situated between implicit and explicit knowledge. The act
of publishing a patent is connected to making knowledge explicit, at least
to some extent, that allows for the communication between the assign-
ee-to-be of the patent and the patent-granting institution.6 At the same
time, a patent gives room to discursive and diagrammatical operations
and aims to establish a new order of knowledge.7 The degree to which the
implicit knowledge of the inventor has to be made explicit is dependent on
the possibility of effective communication between the patent-granting
institution and the prospective assignee.8 Kassung argues, moreover, that
patents never describe a bare technology or a concrete apparatus by itself
but rather a collective of actants in Latour’s sense that must be connected
for the invention to be realized.9 Patents, in this understanding, project
a future network of actants. Thus, patents themselves have an inherent
temporal logic: they project a possible technological future that does yet
not exist at the time of the patent’s filing.10 Hence, patents are actors that
open, close, concentrate, and diffuse technical stories at the same time.11
In the case of Amazon’s anticipatory shipping patent, the company seeks
to extend conventional logistics into the future through a new network
of predictive actants. Central points or nodes of the logistical network of
conventional package shipping are called hubs. A hub in the context of the
patent can be a warehouse, fulfillment center, or sortation center. Fulfillment
centers are in essence the e-commerce variation of what has traditionally
been referred to as warehouses.12 Amazon’s fulfillment centers are modelled
on Walmart’s distribution centers, which at their introduction took over the
classical functions of warehouses. The change in name reflects the change
in function for these structures; whereas a warehouse would store goods, in
the ideal distribution center they are not stored at all. This is particularly
evident with so-called cross-docking distribution centers, where incoming
goods are immediately unpacked from trucks and sent through the center’s
conveyor belt system to the correct outgoing truck, headed for, say, a retail
store. Amazon’s fulf illment centers function somewhere in between a
distribution center and a classical warehouse.13
The patent provides a description of conventional order fulf illment
systems before elaborating the principles of anticipatory shipping, noting
that, traditionally, an item “is not shipped … until a customer places an
order for that item to be delivered to a specif ic delivery address.”14 In
conventional logistics, a customer order initiates the shipping process,
and the act of shipping triggers the payment. With anticipatory shipping,
however, the company spends money on the shipping process without
prior or simultaneous customer payment. Neither an order nor money is
the input.15 This is particularly important given that Amazon would pay
USPS and FedEx, or as of very recently, Amazon Delivery Service Partners,
for transporting the package.16
In the patent a form of speculative shipping is introduced, or the “shipment
of packages without completely specifying delivery addresses at the time of
shipping.”17 One important aspect of speculative shipping is the method of
late-select addressing, or “late addressing,” as it is also called in the patent,
which occurs as a package passes through one of the hubs. This method
allows for packages to be shipped without a specified delivery address and
for the delivery address to be added en route after the package has shipped.
A subform of speculative shipping is anticipatory shipping, which constitutes
the patent’s most interesting innovation:
In some embodiments, speculative shipping of a package may occur in

anticipation of a customer ordering items in that package, but before
such an order has actually occurred. In such embodiments, speculative
shipping may also be referred to as anticipatory shipping.18
In other words, the patented methods facilitate the shipping of packages

before they are even ordered, based on the ability “to forecast or predict
customer demand for a given item.”19 This forecasting model can ideally
draw on large amounts of data about customers, such as previous wish list
activities, shopping carts, product searches, and the duration of mouse
hovers over items.20 What is predicted by this mechanism, in essence, are
probable futures.
Of course, predictions can be wrong and orders might not be placed as
expected. To minimize financial risk, the patent also accounts for false
positives, or faulty predictions. What we might call a fallback mechanism
can be set up to keep costs related to an already-shipped package as low as
possible: redirecting the package to another geographical area where a future
purchase is likely, returning the package to its point of origin, incentivizing
a future purchase by offering a discount or other nudge toward a potential
purchase, or considering the costs of an already-shipped package as an
investment in customer’s “goodwill”:
In some instances, the package may be delivered to a potentially-interested

customer as a gift rather than incurring the cost of returning or redirecting
Ahead of Time 267
the package. For example, if a given customer is particularly valued (e.g.,

according to past ordering history, appealing demographic profile, etc.),
delivering the package to the given customer as a promotional gift may
be used to build goodwill.21
In this last option, false positives become a matter of building customer

relationships.22 In the case of offering a discount and nudging potential
customers toward a purchase, however, false positives from the antici-
patory shipping algorithm are countered instead using Amazon’s classic
recommendation algorithm for curating the display of products. If a package
were shipped but not yet met by a corresponding order, the product would
be offered via the e-commerce platform to customers in near proximity to
the current location of the shipped package at the given time. To make the
offer more attractive, a discount could be provided.
Subsequent to determining the potential cost of returning or redirecting

the near-proximity package, the package is offered to the potentially-in-
terested customer at a discounted price, where the discounted price
depends on the determined potential cost of return or redirection.23
Combining the anticipatory shipping and recommendation algorithms

allows for the fulfilling of prophecies. A sale is not only predicted but also
produced.24 In the following section, I discuss how this production occurs
through a specific ordering of time, or a hardwiring of temporality.
Anticipatory Shipping as New Form of Logistical Media
The etymology of logistics is often traced back to the Greek λογιστικός (logis-
ticos), which means “skillful at calculation” but also “pertaining to reason.”25
In her work on logistics, Gabriele Schabacher (2008) refers to the French
verb loger, originally meaning the temporary accommodation of soldiers or
guests in a housing space. It was only this latter of three meanings of logistics
that was taken up by researchers and experienced further differentiation in
military-historical contexts.26 Logistics, from a current point of view, can
be regarded as a discipline, a perspective, or a set of cultural techniques
engaged in (re)ordering time and space.27 Monika Dommann argues that
logistics, stemming from military strategies concerned with keeping fighting
forces supplied, designates an engineering science dealing with the symbolic
representation, monitoring, and control of flows of materials.
The declared goal of managerial logistical thinking, as Schabacher (2008)

describes it, is always optimization. Optimization is here understood in terms
of permanent improvement of logistics management and organizational
forms, aiming for the smooth flow of commodities, information, services,
and persons, preferably in a cost-saving way.28 The just-in-time (JIT) model,
also known as the Toyota Production System (TPS), concretely revolutionized
logistical processes: time was considered one of the most valuable resources
not to be wasted, with optimization being turned especially toward reducing
wasted time.29
In recent years, shipping processes have accelerated markedly, with
shipping latencies decreasing and JIT principles rendering commodities
constantly in motion, requiring them to be at the right place at the right time
and spending as little time as possible on shelves. The connection between
the logistical paradigm and time can be seen clearly in the corporate quest
for efficiency as timesaving. Deborah Cowen argues, with reference to
total cost analysis, that prior to the logistics revolution, different business
activities were seen as separate, with transportation being seen as following
the production process and not as a moment of it.30 With total cost analysis,
this changed, resulting in a new hardwiring of both time and space in the
realm of corporate logistics, as production facilities moved away from
metropolitan areas and distribution centers took their place. Such relocations
were counterintuitive within the old paradigm, but the higher transportation
bills made sense in terms of lowering overall cost.31
Alongside Walmart, Amazon is one of the largest organizations operating
in the realm of logistics and has built out a massive infrastructure for its
operations. Attention to the infrastructures of contemporary networks
of information and materials has been of great interest in recent media
studies. “Infrastructuralism,” as recently articulated by John Durham Peters,
“suggests a way of understanding the work of media as fundamentally
logistical.”32 For our purposes, the insight can also be reversed: the work of
logistics is fundamentally a question of media.
In his investigation of radar, Judd Ammon Case introduces logistical
media as “media of orientation” that “have to do with order and arrangement
first, and representation second, if at all.”33 This perspective on the power
to arrange seems indeed useful for understanding anticipatory shipping.
The arranging and re-arranging of relationships between different actors,
such as retailers, carriers, and customers as well as goods, databases, and
money, describes what Amazon does through different medial forms on a
daily basis. The patented methods connected to anticipatory shipping are a
way of rearranging the paths and in particular the temporal relationships of
Ahead of Time 269
these people, things, and institutions. A particular infrastructural ordering

of space and time produces a moment of exchange, a sale. It is only by having
the particular material arrangement of fulfillment centers, trucks, planes,
roads, sortation centers, and algorithms that the patent works.34
Peters describes logistical media in The Marvelous Clouds as having the
job of “ordering fundamental terms and units” and setting “the terms in
which everyone must operate.”35 As examples, he lists calendars, clocks, and
towers, as well as names, indexes, addresses, archives, and money. Taking
up the topic of new technologies later in the book, Peters adds that digital
media, such as those we see in Amazon’s order fulfillment systems, “serve
more as logistical devices of tracking and orientation” and “revive ancient
navigational functions: they point us in time and space, index our data,
and keep us on the grid.”36
Amazon’s system of anticipatory package shipping is governed by an
assemblage of digital media that produces an overlap of virtual and real
space: the “real” space, inhabited by potential customers, is made manageable
in terms of supply and calculation through a division into “geographical
regions” where space is organized according to grids.37 The grid, as an
instrument for partitioning, allocating, and relating entities, is a central
feature of logistical media. Peters writes: “The job of logistical media is to
organize and orient, to arrange people and property, often into grids.”38
The centrality of the grid to efforts at economizing time can be seen in
Alan E. Branch’s definition of logistics, from a business perspective, as
the “time-related positioning of resources ensuring that material, people,
operational capacity and information are in the right place at the right
time in the right quantity and at the right quality and cost.”39 What Branch
stresses, notably, is the “rightness” of place and time. Resources must be in
the right place to be on time, and the more minutely structured are time
and space, the more potential opportunity there is to be in a particular
place on time. The ability to coordinate objects in a grid of space in real
time, I argue, is made possible and provided through the digital—thus,
digital logistical media.
Real-time interconnection, as Ned Rossiter argues, is crucial to logistical
optimization: “Without real-time interconnection logistical operations
become exercises in inefficiency.”40 Rossiter, in this way, and in contrast
to Peters and Case, draws a line between prior forms of organization,
such as clocks and calendars, and contemporary digital technologies. He
describes logistical media as an “instrumentalization of location-aware
mobile technologies” that are “designed to exert control over the mobility
of labor, data, and commodities as they traverse urban, rural, atmospheric
and oceanic spaces and traffic through the circuits of databases, mobile
devices and algorithmic architectures.”41
Without diminishing the non-digital entities involved in fulfillment cen-
ters, trucks, schedules, sortation hubs, planes, roads, railways, and container
ports, I want to focus on the all-encompassing network of digital logistical
media governing these entities. Amazon’s patent describes a setting where
predictive algorithms, which are inherently digital technologies, govern
logistical processes. What is novel about the patent is a potential change in
algorithmic logic rather than in any material shipping infrastructure. Just as
total cost analysis comes to view business activities as interconnected, digital
systems compose these activities as literally and technically interconnected.
The alignment of data and goods described in Amazon’s patent is both
produced and ensured by digital devices of tracking and recording. In this
regard, the text stresses the importance of “machine-readable identifiers.”
That is, inventory items and packages must be automatically traceable
through “bar codes, magnetically or optically-readable characters, or oth-
er types of marking and scanning techniques,” such as “radio frequency
identifier (RFID) tags or other types of transponders.”42 The technological
identification of items allows for automatic control in the sense of both
monitoring and governing their direction, i.e., routing.
Digital logistical media, as the example of Amazon’s anticipatory shipping
method makes clear, represent a culmination of logistical techniques into a
potentially all-encompassing interconnective entity. In the contemporary
situation in which our world is ubiquitously populated by digital computers,
one could speak of a hardwiring of time (and space) through the networks
of algorithms and digital devices that govern the flow of information and
goods—thus setting the terms of a politics of digital time. The predictive
logic described in the patent is leveraged for not only logistical operations but
also customer behavior. The following section will therefore present a third
perspective on the patent by investigating the specific logic of prediction
inherent in the methods of anticipatory shipping.
Anticipation, Prediction, Prophecy
Predictive analytics have become increasingly important as processes of

decision-making in various sectors of contemporary societies. 43 The Latin
praedictiō has two meanings as “the act of mentioning in advance,” namely,
“prediction” and “prophecy.”44 Algorithmic predictions could be considered
to carry out this second meaning as the prophecies of our time, in the way
Ahead of Time 271
they structure the relation between present and future. Elena Esposito draws
a striking comparison between ancient divinatory practices and today’s
prediction practices performed by “web intelligence.” Prophecy takes on
different meanings depending on whether the future is considered open
or predetermined. In this regard, a key difference between divination and
algorithmic prediction lies in each society’s underlying understanding of the
future. While divination speaks to an already-decided future, algorithmic
prediction provides forecasts of probable futures. This difference becomes
particularly relevant and visible in the case of false positives. If a divinatory
prophecy did not come to pass, it was not the prophecy that was considered
false. Because the future was regarded as predetermined and already decided,
the interpretation was instead to blame: “one did not understand correctly,
but the oracle was right.”45 False positives were epistemologically impossible.
In a modern and contemporary understanding of time, the idea of a
predetermined future has, in the same way, become epistemologically
implausible. The future is rather considered as ever evolving, open, and
developing from human actions: “it cannot be known in advance because it
does not yet exist.”46 What is possible, though, is the algorithmic forecasting
of probable future behavior by looking at past behavior. When Esposito
describes the effort of economically-oriented systems to bring future events
and risks under control, she draws on Norbert Wiener’s cybernetic notion
of control. According to her reading of Wiener, “the past of a system is
never sufficient to determine its future. One can collect all the possible
information about the past, but it will not be enough to know the future.”47
Even if we cannot know the future, predictions can be produced through
algorithmic means. What we can get from algorithmic predictions, is a
“statistical future,” i.e., “the distribution of possible futures of the system
(the present future).”48 This idea of multiple possible futures corresponds
well to the patented methods of anticipatory shipping, which accounts
in advance for multiple futures. The most peculiar aspect of prediction
in anticipatory shipping is the way it accounts for false positives, what
we called its fallback mechanism. This fallback mechanism is, again, the
redirection of the package to another location, the return of the package to
where it came from, or the production of consumer demand for the product
by offering a discount and very short delivery time.
What remains unclear after a close reading of the patent, though, is what
exactly would cause the fallback mechanism to take effect. Basically, the
prediction would need to be declared wrong under specific circumstances.
In the case of anticipatory shipping, one cannot know if a predicted purchase
within a geographical area—after an indefinite amount of time—would
have occurred after all. 49 The validity of the prediction is temporally lim-
ited by, say, a “time to live”50 for the package – in order to set the fallback
mechanism into place to – in case of embodiment c) – induce the sale. Of
course, there is still no guarantee that the desired customer behavior will
occur, but at least there is a second chance. The form of temporal control
described in the patent follows the cybernetic meaning of control, as Esposito
describes it:
Control, in this case, is achieved not by comparing the input (the present)
with a goal (the future), but by comparing the input with memory (the
past), in order to draw an always renewed and always open projection of
possibilities. The future is rewritten again in each present. … This future
is open without being indeterminate, and uses the past to multiply the
available possibilities.51
In the third embodiment of anticipatory shipping’s fallback mechanism,

reality is readjusted to the prediction. This embodiment reveals itself as
the most radical articulation of rewiring temporalities because the future
order within a geographical area is not only predicted but produced in the
present. In other words, potentiality is sought to be turned into an actuality
through technological, specifically, algorithmic, means.
Conclusion
The algorithmic governing of shipping networks and fulfillment centers

towards anticipatory strategies represents an articulation of rewiring
temporalities. With Amazon’s predictive algorithms, it is not only time that
is brought under digital control but also, and particularly, the aggregated
purchase behavior of clients. As Wolfgang Ernst writes, “Algorithms do not
focus on the individual event but detect patterns, rhythms, and regularities
to be extrapolated into the future.”52 The operation of nudging potential
customers toward a purchase is basically an effort of aligning the “world” and
prediction, of increasing the possibility that the prediction will come true.
In many cases in the press and online, Amazon’s patenting of anticipatory
package shipping has brought forth articulations of uneasiness. One poten-
tial reason for these reservations might be found in the company’s lack of
transparency regarding its algorithms. Robert Seyfert and Jonathan Roberge
point out that users often expect Amazon’s conventional recommendation
algorithm to fail:
Ahead of Time 273
The astonishment often expressed when Amazon’s recommendation

algorithms correctly predict (or produce) our taste, and directly result
in a purchase, goes hand in hand with complaints of how wildly off the
mark they are. We have come to expect failing algorithmic systems and
we have indeed become accustomed to dealing with them.53
Indeed, following Seyfert and Roberge, at times the only response to Am-
azon’s miscalculations is to laugh. It is remarkably hard to peek into the
inner workings of algorithms and understand their potential mistakes,
particularly involving the anticipatory package shipping methods on the
table. To determine whether Amazon’s algorithms succeed or not in pre-
dicting our aggregated taste and shopping behavior is basically impossible
from the outside.
Not only do the predictive algorithms remain opaque, but the relations
between patented and actually implemented methods at Amazon also re-
main unclear. Amazon has not yet commented on the implementation status
of anticipatory shipping, as the company typically declines to comment on
patents.54 A quick search for patents assigned to Amazon Technologies Inc.
returns a list with 8,493 results.55 This might be an effect of the company’s
current policies, trying to patent everything that seems to be potentially
applicable and profitable in the future. The long list of patents, however,
makes it even more difficult to determine which patens have been or will
be implemented.
Irrespective of the actual implementation status of the methods described
in the anticipatory shipping patent, one can observe that Amazon seeks in its
business practices to collect ever more customer data and offer ever shorter
delivery times. Nevertheless, the patented methods can be used to think about
the rewiring of temporalities set out by predictive algorithms in ever faster
and more efficient logistical processes. In essence, the patent along with the
presented approach, allows us to think about a rewiring of temporality in
terms of effects preceding causes, or shipments preceding orders.
Notes
1. Joel R. Spiegel, Michael T. McKenna, Girish S. Lakshman, and Paul G. Nord-

strom, Method and System for Anticipatory Package Shipping, US Patent
No. 8,615,473, filed August 24, 2012, and issued December 24, 2013.
2. See Atle Mikkola Kjøsen, “Anticipating Realization: Value’s Logic of Move-
ment and Amazon’s Anticipatory Shipping,” Conference Paper, Historical
Materialism (UK), 2014, https://www.academia.edu/9288107/Anticipating_

Realization_Values_Logic_of_Movement_and_Amazons_Anticipatory_Ship-
ping1. After the patent on anticipatory shipping was published in late 2013,
it got some attention from the news and various technology-oriented blogs,
but there is only little scholarly work on this patent, notably that by Kjøsen.
At this point, I would also like to sincerely thank Kjøsen for a productive
email exchange and generous comments on the draft text. Furthermore,
this paper owes a lot to and builds on the collaboration with Nikolaus
Pöchhacker addressing the relation of anticipatory shipping to the produc-
tion of markets, see Nikolaus Pöchhacker and Eva-Maria Nyckel, “Logistics
of Probability: Anticipatory Shipping and the Production of Markets,” in
Explorations in Digital Cultures, ed. Marcus Burkhardt, Mary Shnayien and
Katja Grashöfer (Meson Press, 2020), https://explorations.meson.press/
chapters/logistics-of-probability/.
3. For a close look on the influence of patents on software development in the
US, see Gerardo Con Díaz, Software Rights: How Patent Law Transformed
Software Development in America (New Haven, CT: Yale University Press,
2019); “Intangible Inventions: A History of Software Patenting in the United
States, 1945–1985” Enterprise & Society 18, no. 4 (2017), 784–794. doi:10.1017/
eso.2017.36.
4. An alternative take on patents would be their conceptualization as “imag-
inary media” themselves or as documents describing imaginary media,
following Jussi Parikka’s elaborations on Eric Kluitenberg (ed.), Book of
Imaginary Media: Excavating the Dream of the Ultimate Communication
Medium (Amsterdam and Rotterdam: Debalie and NAi Publishers, 2006).
Given that patents are oriented toward future technology that may or may
not be invented, or we may or may not know if a patent has been imple-
mented as in the present case of Amazon’s anticipatory package shipping,
which could be described as “media imagined, non-existent, but worthy of
exploration in terms of how it can reinvigorate current media cultural de-
sign and debates.” Parikka, What Is Media Archaeology? (Cambridge: Polity
Press, 2012), 61.
5. Albert Kümmel-Schnur, “Patente als Agenten von Mediengeschichte,” in
Bildtelegraphie: Eine Mediengeschichte in Patenten (1840–1930), ed. Albert
Kümmel-Schnur and Christian Kassung (Bielefeld: Transcript, 2012), 16.
6. Christian Kassung, “Die Zukunft des Wissens und eine Geschichte der Pat-
ente,” in Nicht Fisch – nicht Fleisch. Ordnungssysteme und ihre Störfälle, ed.
Thomas Bäumler, Benjamin Bühler, and Stefan Rieger (Zürich: Diaphanes,
2011), 156. Although Kassung’s text is mainly concerned with patents for
concrete technical apparatuses, his diagnosis remains true for software
patents and patented methods.
7. Kassung, “Die Zukunft des Wissens,” 153. See Bowker (1992) as well as Pot-
tage and Sherman (2010) For further analyses of patents, see Geoffrey Bowk-
er, “What’s in a Patent?,” in Shaping Technology/Building Society: Studies in
Sociotechnical Change, ed. Wiebe E. Bijker and John Law (Cambridge, MA,
Ahead of Time 275
and London: The MIT Press, 1992), 53–74; Alain Pottage and Brad Sherman,
Figures of Invention: A History of Modern Patent Law (Oxford: Oxford Uni-
versity Press, 2010).
8. Kassung, “Die Zukunft des Wissens,” 156.
9. See for example Bruno Latour, Pandora’s Hope: Essays on the Reality of Sci-
ence Studies (Cambridge, MA: Harvard University Press, 1999).
10. Another reason why patents could be understood as documents describing
imaginary media (see Parikka, What Is Media Archaeology, 61; Kluitenberg,
Book of Imaginary Media).
11. Kümmel-Schnur, “Patente als Agenten von Mediengeschichte,” 27.
12. See Kjøsen, “Capital and Post-Cybernetic Control: On Amazon’s Patent for
Anticipatory Package Shipping.” Presentation on April 5, 2018, Faculty of In-
formation and Media Studies, Western University, London, Ontario; Oliver
Nachtwey and Philipp Staab, “Die Avantgarde des digitalen Kapitalismus,”
Mittelweg 36, no. 6 (2015), 67.
13. See Kjøsen, “Capital’s Media: The Physical Conditions of Circulation” (PhD
diss., University of Western Ontario, 2016), https://ir.lib.uwo.ca/etd/4156/.
14. Spiegel et al., 16.
15. Kjøsen, email message to author, July 11, 2018.
16. For more on Amazon Delivery Service Partner, see Louise Matsakis, “Why
Amazon Is Giving Employees $10,000 to Quit,” Wired, May 14, 2019, https://
www.wired.com/story/amazon-delivery-paying-employees-to-quit/.
18. Spiegel et al., 16. Emphasis added.
22. However, unasked-for gifts by corporations are not always the best idea in
terms of Customer Relationship Management, as they basically violate the
“rules” of gift-giving and can produce irritation, even anger, for customers.
This was the case e.g. with Apple’s “free gift” of automatically downloading
U2’s then-new album for all iTunes users in 2014. See Robert Booth, “U2’s
Bono Issues Apology for Automatic Apple iTunes Album Download,” The
Guardian, October 15, 2014, https://www.theguardian.com/music/2014/
oct/15/u2-bono-issues-apology-for-apple-itunes-album-download.
24. At this point—on the important difference between demand and sale—I
wish to quote Atle Kjøsen from an e-mail correspondence on July 11, 2018:
“Often “demand” is confused for “need” and for “purchase.” In the case of
the patent, “demand” is actually already proven by Amazon because the
consumer has added items to their wish list, shopping cart, looked at the
items x number of times and so on. There is clearly a demand there, but
that demand may not be backed up by cash. Hence, the whole point of
logistics is to not match supply with demand, but match supply with cash.
… What is predicted and produced is not demand (it was already there),
but a sale: the moment whereby commodities change hand for money. But
in that case, it is not merely the algorithms that produce a sale, but also the
physical infrastructure of FCs, hubs, trucks etc. (the infrastructure of the
order fulfillment system). After all, if the package is not in a geographical
area where demand has been detected, the sale cannot be produced.”
25. The Brill Dictionary of Ancient Greek, 2015 ed., s.v. “logistics”; see also
Deborah Cowen, The Deadly Life of Logistics: Mapping Violence in Global
Trade (Minneapolis: University of Minnesota Press, 2014), 26; and Gabriele
Schabacher, “Raum-Zeit-Regime: Logistikgeschichte als Wissenszirkulation
zwischen Medien, Verkehr, und Ökonomie,” Agenten und Agenturen: Archiv
für Mediengeschichte, ed. Lorenz Engell, Joseph Vogl, and Bernhard Siegert
(Weimar: Bauhaus University, 2008), 137.
26. Schabacher “Raum-Zeit-Regime,” 137.
27. For more on cultural techniques, see Cornelia Vismann, “Kulturtechnik und
Souveränität,” Zeitschrift für Medien- und Kulturforschung 1 (2010), 171–181;
Liam Young, “Cultural Techniques and Logistical Media: Tuning German
and Anglo-American Media Studies,” M/C Journal 18, no. 2 (2015), http://
journal.media-culture.org.au/index.php/mcjournal/article/viewArticle/961.
For more on the ordering of time and space, see Monika Dommann, “Han-
dling, Flowcharts, Logistik: Zur Wissensgeschichte und Materialkultur von
Warenflüssen,” Nach Feierabend 2011: Zirkulationen. Zürcher Jahrbuch für
Wissensgeschichte 7, ed. David Gugerli et al. (Zürich: Diaphanes, 2011), 75.
For a history of logistics from a media-theoretical perspective, see Scha-
bacher, “Raum-Zeit-Regime” and Dommann, “Handling, Flowcharts, Logis-
tik.” For an investigation on the political economy and potential violence of
global logistics processes, see Cowen, The Deadly Life of Logistics.
28. Schabacher “Raum-Zeit-Regime,” 144f.
29. For more on TPS, see Taiichi Ohno, “Toyota Production System” (1978), The
Roots of Logistics: A Reader of Classical Contributions to the History and Con-
ceptual Foundations of the Science of Logistics, ed. Peter Klaus and Stefanie
Müller (Heidelberg: Springer, 2012), 173–182.
30. Cowen, The Deadly Life of Logistics, 36–38. See also Kjøsen, Capital’s Media,
78. Kjøsen argues for qualifying the current capitalist mode of production as
“logistical … due to the increased centrality of logistics to business since the
1970s” (73).
31. Cowen, The Deadly Life of Logistics, 38.
32. John Durham Peters, The Marvelous Clouds: Toward a Philosophy of Elemen-
tal Media (Chicago: University of Chicago Press, 2015), 37.
33. Judd Ammon Case, “Geometry of Empire: Radar as Logistical Medium”
(PhD diss., University of Iowa, 2010), 1, http://ir.uiowa.edu/etd/474.
34. Kjøsen, email message to author, July 11, 2018.
35. Peters, The Marvelous Clouds, 37.
37. Spiegel et al., 15. This real space is also built up and ordered by the infra-
structure of fulfillment centers, sortation hubs and so on. These hubs are
Ahead of Time 277
not placed randomly. Cowen elaborates on how logistics is a reorganization

of space; before factories were located close to the market, but after the
logistics started calculations of total cost, factories were often located closer
to important transportation nodes.
39. Alan E. Branch, Global Supply Chain Management and International Logis-
tics (New York: Routledge, 2009), 1.
40. Ned Rossiter, “Coded Vanilla: Logistical Media and the Determination of
Action,” South Atlantic Quarterly 114, no. 1 (2015), 139.
41. Rossiter, Software, Infrastructure, Labor: A Media Theory of Logistical Night-
mares (New York: Routledge, 2016), 8.
42. Spiegel et al., 17. See also Christoph Rosol, RFID. Vom Ursprung einer (all)
gegenwärtigen Kulturtechnologie (Berlin: Kadmos, 2007).
43. See also Pöchhacker and Nyckel (2020), 1.
44. Oxford Latin Dictionary, 1994 ed., s.v., “praedictiō.”
45. Elena Esposito, “Digital Prophecies and Web Intelligence,” Privacy, Due Pro-
cess and the Computional Turn: The Philosophy of Law Meets the Philosophy
of Technology, ed. Mireille Hildebrandt and Katja de Vries (Oxon and New
York: Routledge, 2013), 134. The future was perceived as predetermined, and
therefore immutable. According to this belief, it could not be altered by hu-
man actions—“not even by those who try to oppose it” (134). As an example
of the future’s inevitability Esposito refers to the story of Oedipus.
46. Esposito, “Digital Prophecies and Web Intelligence,” 134.
47. Esposito, The Future of Futures: The Time of Money in Financing and Society
(Cheltenham: Edward Elgar, [2009] 2011), 194.
48. Esposito, The Future of Futures, 194; see also Wiener, Cybernetics: Or control
and Communication in the Animal and the Machine (Cambridge, MA: MIT
Press, 1948), 37, 92–93.
49. Alternatively or in addition, one could suspect that costs may play a decisive
role in this scenario, though time and cost are, of course, connected quite
closely. Thus, declaring a failed prediction could be something like “Possible
sale x has a y % chance in z hours/days.” Kjøsen, email message to author,
July 11, 2018.
50. “Time to live” (TTL) is an attributed number that enables a mechanism
to prevent the indefinite rerouting of packages; however, in this case,
data packets or datagrams in the Internet, enabled and structured by the
Internet Protocol: “This number designates the maximum number of hops
that that datagram is able to take before it is deleted. At each hop, the
time-to-live is decreased by one. If the time-to-live reaches zero, the routing
computer is obligated to delete the datagram. This ensures that datagrams
will not hop around the network indefinitely, creating excess congestion.”
Alexander R. Galloway Protocol: How Control Exists after Decentralization
(Cambridge, MA: MIT Press, 2004), 45. This description—respectively the
reappropriation of the package metaphor as providing a surplus meaning—
works surprisingly well for the temporal control of packages “hopping”
through the network of hubs, fulfillment centers, and delivery centers as

conceptualized in the anticipatory package shipping patent. For further dis-
cussion of this similarity between the routing of data packets and packages,
see Christoph Neubert and Gabriele Schabacher, “Verkehrsgeschichte an
der Schnittstelle von Technik, Kultur und Medien. Einleitung,” in Verkehrs-
geschichte und Kulturwissenschaft (Bielefeld: Transcript, 2012), 25.
51. Esposito, The Future of Futures, 194.
nological Media, trans. Anthony Enns (London and New York: Rowman &
Littlefield, 2016).
53. Robert Seyfert and Jonathan Roberge, “What Are Algorithmic Cultures?” in
Algorithmic Cultures: Essays on Meaning, Performance and New Technologies,
ed. Seyfert and Roberge (London and New York: Routledge, 2016), 14f.
54. Alan Boyle, Amazon Wins a Pair of Patents for Wireless Wristbands That
Track Warehouse Workers,” GeekWire, January 30, 2018, https://www.
geekwire.com/2018/amazon-wins-patents-wireless-wristbands-track-ware-
house-workers/; Celan Yeginsu, “If Workers Slack Off, the Wristband Will
Know (And Amazon Has a Patent for It),” New York Times, February 1, 2018,
https://www.nytimes.com/2018/02/01/technology/amazon-wristband-track-
ing-privacy.html.
55. Google Patents, searched June 14, 2019, https://patents.google.com.
About the Author
Eva-Maria Nyckel is a PhD candidate in media studies at the Humboldt

University Berlin. Her research interests are logistical media, enterprise
software and digital capitalism. In her PhD thesis she writes about the
history, theory and politics of “programmable labour.”
15. Artificial Neural Networks, Postdigital
Infrastructuresand the Politics of
Temporality
Andreas Sudmann
Abstract
This essay examines the infrastructures and temporalities of modern
AI technology based on artificial neural networks (ANN) and aims to
contribute to a more substantial understanding of its political challeng-
es. In order to unlock the different temporalities of ANN, a theoretical
framework for the relationship of media and infrastructures is suggested
that also might help to distinguish between the different levels of analysis
related to specific steps and aspects of the machine learning process (the
collection and production of learning data, the training of AI models
etc.). An important reference point for the following considerations is
ethnographic research conducted at TwentyBN,1 a Toronto and Berlin
based AI company specialized in ANN and computer vision that just
recently developed an app for the fitness market.
Key words: artif icial intelligence, machine learning, artif icial

neural networks, computer vision, science and technology studies,
media-ethnography
Since around 2012, machine learning based on artificial neural networks

(ANN) has become the dominant paradigm of artificial intelligence (AI)
research and increasingly determines the technological and infrastructural
conditions of contemporary computer culture. Whether new developments
in the field of text, speech, or image recognition, whether concrete use cases
such as self-driving cars, virtual avatars, or medical diagnostic systems, the
list of innovations that involves ANN-based machine learning is extremely
doi 10.5117/9789463727426_Ch15
280 Andreas Sudmann
long. Although some scholars emphasize the epistemological limits of ANN,

it is nevertheless surprising to see how, since very recently, they became such
a powerful technology of prediction, capable of calculating and mastering
the uncertainties and fuzziness of future events.2
The status of ANN as predictive systems already demonstrates the
fundamental importance of investigating this technology in terms of its
temporal dimensions. From the perspective of media studies, which in part
also sees itself as a decidedly critical discipline, the particular relevance
of this focus lies not only in exploring the epistemological potential of this
specific approach in AI. Rather, it is the central aim of this essay to show
how a deeper investigation of the infrastructures and temporalities of
modern ANN-driven AI technology also contributes to a more substantial
discussion of its political challenges.
Like basically all AI technologies since the 1950s, ANN have been nego-
tiated mostly in the mode of utopian discourse, from Frank Rosenblatt’s
Perceptron model3 to Google’s Neural Machine Translation. The term utopia,
in the sense of its etymological origin, refers to a non-place, in temporal terms
typically to the not-yet-existent of an unknown future. Indeed, the current
advances in that particular field of AI do once again fuel the speculative
notion of a so-called “technological singularity” as the very event after which
all future development of machine intelligence will no longer be predictable
or comprehensible to humans (while also signifying the very date when
machines will be superior to humans in most, if not all, core areas of human
expertise). 4 Michel Serres has argued that the concept of time in itself
articulates the dialectics of predictability and unpredictability.5 However,
in the case of ANN, we are confronted with a specific dialectic or paradox
of temporal orders that simultaneously resonates with the speculative
dystopian (rather than eutopian) character connected to current visions
of an artificial general intelligence: how the accelerated development of a
predictive technology might turn into something—at least for humans—
fundamentally unpredictable and inaccessible.
But these characteristics are not concerned merely with future scenarios
because ANN are already perceived fundamentally as black boxes not
understandable or at least only partially understandable by humans.6 Their
black-box status encompasses many aspects, not only the opacity of the
ANN itself or its specific medial learning operations, or the tremendous
speed of micro-temporalities connected to AI technologies and their re-
spective decision-making processes that we can understand as submedial
forms operating below the threshold of human perception,7 but also the
complex industrial practices and environments that generate and form
Artificial Neur al Ne t works, Postdigital Infr astruc tures 281
the technological processes of modern AI. In other words, the event of a

technical singularity is an escalation of what is already being discussed as
one of the core problems of ANN technology: its fundamental inaccessibility.
Obviously, such an event (if it ever becomes a reality) will result in enormous
distortions, not least in political terms. Against this very background, one
must take such an event as seriously as possible, even if, from today’s per-
spective, it seems a scenario of a still distant future or unlikely in principle.
Nevertheless, it is not very helpful to constantly speculate about the political
or social consequences of a technology’s future if we have just started to
understand the current situation of AI in general and ANN in particular,
both theoretically and historically. Even if the present and its temporality
is always exactly what is already over when we talk about it.8 Accordingly,
the aim of this essay is to explore the relation between infrastructures and
temporalities with regard to ANN-based machine learning, specifically to
provide a substantial basis for future political discussions of this technology.
A central aspect of ANN in political terms, particularly concerning its
present form, is that we are dealing with a powerful technology of opti-
mization and automation, which in numerous areas not only expands the
respective capabilities of humans, but also clearly outperforms them. And it
is worth highlighting that such ANN systems do not just master domain-spe-
cific tasks faster or more efficiently than humans. Instead, in many areas,
such as image recognition, they also outperform humans in qualitative
sense. However, as many scholars working in the transdisciplinary field of
science and technology studies have argued, it is problematic to limit the
political-epistemological relationship to the binary relationship of humans
and machines or humans and technology. While it is important to stress that
current machine learning technologies are still very much dependent on the
activities of human beings and human labor, we also have to acknowledge
that we are dealing with complex infrastructures and chains of operations,
where humans, things, practices, and media are inseparably connected.
Theoretical Approaches to Media and Infrastructures
But how do we make sense of these different layers of temporalities involved

in the industrial and scientific infrastructures of current AI? And how
deep do we have to explore all the elements that constitute and shape
these smart technologies? The answer to this question can only be given
if we are willing to investigate the specifics of ANN technologies and try
to understand the complexity of the technology in terms of its manifold
282 Andreas Sudmann
temporal dimensions, at least on a conceptual level. Obviously, there is

no lack of potentially useful approaches: Scholars working in the field of
media archaeology, science and technology studies, historical epistemology,
critical code studies, platforms studies, etc., have provided many accounts
for the critical understanding of our data-driven algorithmic culture, also
and especially with regard to the nexus of infrastructures, media, and
temporality.9 And yet we have to acknowledge that studies in this specific
area of research (as with media studies) have neglected machine learning
in general and ANN in particular as an important field of research for a
very long time. In the course of the current AI boom and especially since
the spectacular victory of DeepMind’s AlphaGo, the situation has changed
significantly: AI has already become the prevalent topic in every field of the
humanities and sciences. Nevertheless, it has to be said that the debate on
ANN or machine learning in fields such as media studies, and cultural studies
more generally, is still in its infancy. But what does it mean to approach the
infrastructures and temporalities of AI and ANN, especially from a media
studies perspective? And even more central in theoretical terms, what is
the difference between infrastructures and media (if there is any) as key
concepts for understanding AI technologies?
As one might imagine, there is not easy answer to this question. And yet,
while infrastructures and media share certain characteristics—for example,
that they refer to the cultural productivity of (partly) invisible operations and
structures10—I would claim that a key difference between the terms is that
“infrastructure” is a concept more useful for describing the arrangement or
organization of (different) media, while, in turn, media are the heterogenous
entities that constitute and contribute to shaping certain infrastructures in
the first place. This consideration is by no means at odds with the idea that
an infrastructural or logistic dimension is central to many diverse forms of
“old” and “new” media.11 Still, we need a term for the arrangement and orga-
nization of different media, especially with regard to systemic and rule-based
aspects of their stabilization, as well as with regard to the interoperability and
connectivity of and associated with media and their related practices. And it
seems to me that the term infrastructure is suitable for precisely this purpose.
In the following, I would like to demonstrate how the distinction between
media and infrastructure is also helpful for understanding the different
temporal logics and dimensions of AI technologies: For example, although
the differences between media as input data play an important role in choos-
ing and shaping a particular AI infrastructure for mastering an individual
learning task, we also have to take into account that media differences can
be decisive for developing advanced forms of AI. For this very reason, it is
important to deal with the specifics of the relationship between media and
infrastructures as they pertain to AI technologies.
However, as I outline in this chapter, ANN systems are characterized by a
rather general temporal characteristic that is of great relevance both trans
historically and across different applications and specific infrastructures,
and as such challenges the perception of ANN as a recent manifestation
of digital change.
An important reference point for the following observations and con-
siderations is my current ethnographic research at TwentyBN,12 a Toronto
and Berlin based AI company specialized in Deep Learning solutions and
computer vision that just recently developed an app for the fitness market.
Big Data, Machine Learning, Crowdworkers, and the Politics of

Temporality
To unlock the different temporalities of modern ANN and specifically to

understand the political dimension of their media and infrastructures, we
first of all have to distinguish between different levels of analysis related to
specific steps and aspects of the machine learning process. One important
perspective in this regard is to shed light on the level of collecting and
producing the large amounts of data necessary for training a machine
to master a certain learning task. For example, in the case of so-called
supervised machine learning, it is not enough simply to have access to
large data sets; it is also necessary to have pairs of corresponding inputs
and outputs as training data. A profound challenge here is that in many
use cases, learning data in accurately labeled form is not available, and
producing it can be a very time-consuming and costly working process.
To address this problem, AI companies like TwentyBN specializing in
ANN technologies hire legions of crowdworkers through platforms such
as Amazon Mechanical Turk or CrowdFlower, responsible for labeling and
producing the learning data. A general and and economically productive
advantage of crowdsourcing in temporal terms is that instead of having a
few employees working on monotonous, repetitive tasks again and again,
such activities can be distributed more or less in parallel to hundreds or
thousands of workers, thus saving a massive amount of time.13
One of the downsides of this form of work in late capitalism is that
crowdworkers are typically poorly paid, which in turn regularly has a
negative impact on the work they are engaged for. For their labor to pay off
at least to some extent, they must complete their specific tasks as quickly
284 Andreas Sudmann
as possible. Hence, to fulfill the task of producing and labeling data most
efficiently, it happens quite often that they develop subversive strategies to
earn more money within a certain time frame. Unsurprisingly, such practices
of “cheating” do have serious consequences; in essence, it means that the
algorithms are trained with corrupted data. As a result, AI companies
like TwentyBN usually cannot use the data produced by crowdworkers
directly; instead, they are forced to monitor and evaluate its quality using
test algorithms or manual inspection. As the activities of crowdworkers
illustrate, ANN systems and their specific infrastructures do have profound
effects on the temporalities of lived experience, they not only exert pressures
of conformity or standardization but also lead to strategies for avoiding
the temporal regimes of a cognitive capitalism and its infrastructures now
increasingly shaped by data-driven machine learning.
The creation of appropriate learning data can take months or even years,
depending on the specif ic purpose of the learning data and AI model.
The training and testing of an AI model is also very time-consuming. For
high-end AI applications, such as advanced machine vision systems, training
involves not only thousands but typically even millions of such cycles or
epochs, while other less complex classification tasks, such as distinguishing
simple geometrical forms (as in Rosenblatt’s perceptron model), might
only demand a few hundred training cycles. Hence, the temporal span of a
training process can also vary a lot, from a few hours to several weeks. More
generally, how fast an ANN can be trained for a certain problem depends
on number of different parameters, including the quantity and quality of
the training data, the specific architecture of the ANN, and the hardware
resources available. For these very reasons, it is crucial to understand the
specific characteristics of AI infrastructures that make use of ANN and
other technologies of machine learning.
Real-Time Capability and the Temporal Challenges of Cognitive

Capitalism
Another important point to consider regarding the temporal dimensions

of AI infrastructures is that many intelligent systems must be capable of
operating in real-time.
When for example TwentyBN recently developed its fitness app, it was
particularly important to the company that their AI system is capable of
providing immediate feedback to the users whenever they perform a certain
exercise like jumping jacks or push-ups incorrectly.
And also especially in sensible application areas, such as self-driving cars,

ANN-based AI models involve strict latency requirements and “demand
lightning-fast deep learning inference, usually within tens of milseconds
for each sample.”14
Many modern AI technologies are therefore media configurations, which
should or must guarantee a certain degree of responsiveness (Rechtzeitigkeit),
which means they have to operate with regard to a varying ratio of processing
speeds and a time window to be adhered to by these processes.15
While forms of “inference acceleration” have significantly improved over
the last few years, also due to new hardware architectures like Google’s
TPUs, there is a further serious problem of temporality to consider here. For
example, ANN models (like any other machine learning model) are based
on the assumption that there are no “distributional shifts in the input and
output data over time.”16 Machine learning and in particular ANN systems
work so well, regardless of the specific prediction task, because the world,
i.e., the structure of data a statistical model of AI can learn, usually does not
change quickly or with hard transitions but usually slowly and smoothly.17
And yet, the important assumption of ANN and other machine learning
models that the distribution of input and output data remains more or less
stable over time does have its specific limits, especially if we think about
application areas like information security, “where fast-paced evolution of
the underlying data generating mechanism is a norm (in the case of security,
it is because both players, the defender and the adversary, are constantly
striving to outmatch his opponent by changing his own strategies, thus
exploiting the opponent’s unguarded vulnerabilities).”18
Nevertheless, the basic principle of the industrial use of modern technol-
ogy as a whole is to ensure that technical problems can be dealt with quickly
and flexibly, for example, in the field of AI with regard to the much-discussed
problems of bias structures. However, this does not always work as smoothly
as desired. When in 2016 Microsoft presented its chatbot Tay to the public,
the company had to shut down the system only after sixteen hours. Users
quickly hacked the system by training it for racist, anti-Semitic, and sexist
articulations, which unsurprisingly led to quite a large public controversy.
But even if the example of Tay illustrates some serious problems and limits
of current machine learning systems, we should not forget that at least with
regard to machines, problematic bias structures can be easily corrected, or
at least there is the option to simply shut down systems that do not work
properly. And this is obviously different with regard to human beings.
However, within current academic discourses of philosophy, social sci-
ence, and cultural studies, temporal aspects seem to play a less important
286 Andreas Sudmann
role when we discuss the specific potentials of AI. This may be because AI
itself is overdetermined by a rhetoric of progress emphasizing efficiency
and flexibility in accordance with the neoliberal logic of the temporal
regimes of late capitalism.
While this form of subjugation to neoliberal orders of time is typically
rightly questioned by scholars across different fields, in the humanities
and beyond, we should also keep an eye on the potentials of AI, even if the
technology and its temporal order as a whole or in part are never neutral,
serve neoliberal interests, or take on a questionable and teleological logic
of progress. To give just one of many examples, if traffic will become more
and more dominated by autonomous vehicles in the future, or if road traffic
will be regulated by modern AI systems, this may be seen as a problematic
affirmation of neoliberal values of optimization and efficiency because the
intention is not simply to reduce traffic jams but also to create the possibility
of coping with greater traffic loads, which in turn represents a serious
ecological problem. On the other hand, there is at least some evidence that
self-driving cars will significantly reduce the frequency of traffic accidents,
simply because they are faster and better able than humans to recognize
dangerous situations as such.19 A critique of AI, which in particular focuses
on aspects of temporality, should therefore take such ambivalences of
technology into account and discuss them in a differentiated way.
The Potential of ANN to Understand and Identify the “New”
In any case, it is interesting to note that the critique of AI often includes a

critique of the temporality of these systems. For example, the fact that machine
learning technology inevitably uses past data to predict the future has often
led to skepticism from scholars, claiming that data-driven machine learning
systems based on ANN are inherently conservative and incapable of producing
something genuinely new.20 Indeed, one could argue that recent discussions
about algorithmic biases underpin this critical perspective. Likewise, many
of us know from our everyday experience that it can be quite boring when,
for example, recommender systems simply suggest titles similar to the music
we already like. It might be also fair to say that we don’t necessarily have to
make the algorithms responsible for this: If we have the impression that an
AI-powered system is only suggesting music we already know or like, we
could also take this as an invitation to question the limitations of our taste in
music. But that’s not the point I’m primarily concerned with. Instead, I would
like to question the criticism I have just presented for two other reasons.
First, doubts seem justified as to whether ANN-based systems are really

incapable of producing something new in the emphatic sense. Isn’t the
example of DeepMind’s AlphaGo a powerful reminder that we can witness
an AI model operating with moves that are perceived as innovative and
surprising by the global community of Go players? Of course, prima facie
it seems legitimate to argue that, according to such a perspective, one does
not just take the rhetoric and PR strategies of the AI industry for granted,
but also carry out a questionable anthropomorphizing and mystification of
technical systems that in their core consists of statistical or computational
processes. Nevertheless, it might be advisable to distinguish between Genesis
and Geltung here. Similar to Turing’s argumentation in the 1950s regarding
the justification of his famous test, one could argue that what matters is
less whether an AI system really is creative or if it de facto has the potential
in itself to create something genuinely new. Rather, what counts is if the
system manages to appear to be creative or innovative.
Second, the question would be if the fetishization of AI, which has often
been criticized, does not provoke another form of fetishization, namely
the fetishization of an anthropocentric figure of the emphatically new.21
All “human” achievements, whether in art or science, are in fact based on
complex and preceding mediators, i.e., on already existing technologies,
media, infrastructures, communication processes, and forms of knowledge.
Obviously, we apply different standards to machines than to people. But
why? Do we have to face the fact that a post-anthropocentric view of AI is
so difficult to accept, not only because we are not willing but also because
we are simply unable to do so? I do not suggest we have to answer that
question in the affirmative. Rather, the point is that one must be aware
of the specific paradoxes when discussing anthropocentric concepts such
as learning, knowledge, or the “new” in relation to intelligent machines.
However, the key challenges that AI systems are currently confronted
with are indeed less about being able to identify or generate something
completely new than about being able to take account of the temporal
dimension as well as the specific context of data.
Recurrent Neural Networks and the Temporal Significance of

Media Differences
It is important to acknowledge that also in this very area ANN have made
signif icant progress in the last ten years. A crucial factor here was the
advanced development of so-called recurrent neural networks (RNN). A
288 Andreas Sudmann
key characteristic of RNN is their ability to identify patterns in sequences of

data, e.g., for learning tasks such as text understanding, speech recognition,
and even stock market developments. To do so, RNN need to fulfill the same
task for each of the sequence’s elements, while the output depends not only
on the current input but also on the previous computations. In other words,
RNN have a kind of memory function (internal state) that allows information
to persist. Theoretically, RNN can process data sequences of infinite length,
but in practice they are constrained so that they can look back only a few
steps. Although the gradients of RNN can quite easily be computed, they are
difficult to train due to their nonlinear iterative characteristics. According
to Ilya Sutskever,
A small change to an iterative process can compound and result in very

large effects many iterations later; this is known colloquially as “the
butterfly-effect.” The implication is that in an RNN, the derivative of the
loss function at one time can be exponentially large with respect to the
hidden activations at a much earlier time. Thus, the loss function is very
sensitive to small changes, so it becomes effectively discontinuous.22
However, these limitations and difficulties do not change the fact that the
leaps in performance in the area of speech recognition, such as with Alexa,
and Siri, or in the field of machine translation have been considerable since
around 2009.
Recurrent neural networks have also made signif icant progress in
the field of computer vision. In 2016, I had the honor of introducing what
might be considered as the most important ANN innovation since the
famous ImageNet-Paper by Krizeshvsky et al.23 in an article for the German
newspaper Frankfurter Allgemeine Zeitung,24 where I described how for the
first time in the history of information technology, TwentyBN developed
an ANN system to automatically recognize hand gestures and basic (and
yet complex) activities (even before Baidu, Facebook, and Google were able
to master this fundamental AI problem). And, indeed, the infrastructure
(or the media) that enabled this computer vision technology could hardly
be more relevant from a media studies perspective: For the first time and
in contrast to most approaches in this field of AI research up to this point
(which typically used photographs or still images for computer vision tasks),
TwentyBN’s model has been trained with hundreds of thousands of short
video clips, showing different kinds of (basic) activities, such as opening
objects, throwing and catching something, or stack things.25 The implications
of this approach for our thinking about infrastructures, media, temporality,
and their relations are very significant: What the system developed by
TwentyBN implicitly demonstrates is that already the sheer media difference
between time-based and non-time-based media seems to be an essential
precondition for the development of an advanced AI. Accordingly, since
2017, we are confronted with new epistemological condition of our global
computer culture, namely that we witness the emergence of an AI that is
already capable of understanding basic forms of common-sense knowledge,
i.e., perceiving and “understanding” the world in ways similar to human
beings.26
And—hardly less relevant—we also have to consider that the specific
technological potential of media difference to allow such advanced forms
of machine intelligence exists—at least in a certain sense—outside the
realm of history, culture, and ideology. On the one hand, the specif ic
content and meaning of the videos shown to the system is less decisive for
the learning operation than the basic fact that it is trained with moving
images at all. On the other hand, the “content” of the model matters
insofar as the system has to be trained with a great variety of gestures
and actions so that it is capable of generalizing well. Nevertheless, the
learning system also perceives and processes every input similarly, without
a deeper sense of meaning, the gesture of a Hitler salute (Hitlergruß) not
being different from the activity of turning a bottle, stacking books, or
doing jumping jacks. Hence, ANN-based computer vision takes place
within an infrastructural arrangement in which the broader cultural,
social, and historical contexts of the learning material play no significant
role in terms of their technological operations. Functionally decisive
is the diversity of the displayed material in itself, not its specif ic and
semantically charged composition.27
As the example outlined above shows, it is very important to focus on
the specific, inherent temporalities of learning algorithms, their Eigen-
zeitlichkeit, which at least partially retain autonomy and contingency over
those temporal regimes that characterize the commercial and scientific
infrastructures of machine learning technologies as a whole—such as the
inscription of history or the specific temporal logic of how machine learning
tasks are organized as industrial or scientific processes.
Against this background, in view of their principal indifference to content
and meaning, ANN systems are a media technology that is not fundamen-
tally different from the gramophone or photography as technologies of the
nineteenth century or from the digital computer as a technology of the
twentieth century. And yet ANN turn the time relations of information
technology upside down.
290 Andreas Sudmann
Massive Parallelism as a Postdigital Configuration of Future

Information Technology
An important prerequisite for the implementation of a model such as

that developed by TwentyBN, however, is the performance of the modern
computer hardware on which the networks are trained, namely graphics
processing units (GPUs).28 What distinguishes the temporality of ANN both
generally and transhistorically is their massive parallelism. At the end of
the 1980s, Friedrich Kittler wrote:
Today’s artificial intelligences run faster, more parallel, but not funda-
mentally different from those who “follow the principle of the Universal
Discrete Machine … With it, the media system is closed. Storage and
transmission media both merge into a principle circuit that can simulate
all other information machines simply because it stores, transmits and
calculates in each individual program loop.29
In contrast to Kittler’s perspective, I would like to argue that ANN systems,

especially because of the parallelism of their information processing, do
indeed represent a significant break with the logic of digital computers of the
von Neumann architecture and that they actually stand for the emergence
of a post-digital computer culture avant la letter.30 In fact, neural networks,
whether artificial or “natural,” are in at least two respects a counter-model
to the way digital computers function according to von Neumann’s serially
organized architecture. First, since a single artificial neuron is usually
either active or not, in this respect it usually functions according to binary
logic, such as the switching states of a digital computer. The weighting of
the activity between the neurons, i.e., the strength of their connections,
is however mostly represented by floating point numbers (positive and
negative) in neural networks. And this representation is so finely structured
that the corresponding values can be understood as quasi-analog. As media
of information transmission, ANN do not operate using binary units, such
as 0 and 1, but in an almost analog form (even if the analog values are still
based on a digital substrate).
Second, it must be emphasized that the masses of interconnected neurons,
activated by an input, fire together simultaneously or in parallel, thus ulti-
mately forming a complex emergent system that abolishes the discreetness
of the elements it consists of (the layers of neurons and their connection).
This extreme or massive parallelism of information processing can indeed
count as the essential characteristic of ANN, distinguishing it from the von
Neumann architecture of classical digital computers. Due to the described

properties, an ANN is therefore a blurred system (Unschärfesystem) with
probabilistic results, whose operations can be described rather as analog
than digital.31 Whether ANN is primarily implemented as software, as
it is at present, or increasingly as hardware, as it may be in the future, is
irrelevant for its characterization as a post-digital information technology.
Whatever future processor technologies modeled on neural networks will
look like concretely, they share the basic characteristics that already position
ANN as a “software medium.” Anyone who simply understands the current
developments of ANN, particularly in terms of its political dimensions, as
a further expression of the digital revolution or the digitization of culture
is in fact using the wrong category—at least in part.
Conclusion
In recent decades, the evolution of computers and the processes of digiti-

zation have been extensively investigated with regard to their temporal
conditions, implications, and effects. Again it was Friedrich Kittler who
argued that the computer, as a universal medium that can scan and simulate
all other media, is in principle the end of media history.32 Furthermore,
several scholars have conceptualized the temporal logics of digitization
as a transformation process of simultanization. The current renaissance
of ANN now once again draws attention to the history of sub-symbolic
information processing and its specif ic logics of temporality. With a
few notable exceptions, the field of media and cultural studies has until
very recently payed little attention to this. Paul Virilio, for example, has
written about AI-driven vision machines that replace human perception,
machines that look back at us and observe us, but despite its interest in
the dromological view on temporality, he did not realize the enormous
temporal implications of the connectionist AI paradigm that is now on
the rise again.33
But what exactly has changed? One could assume that thanks to the
massive parallelism of ANN-based machine learning we are confronted
with a new dimension or quality of technological acceleration and thus
have reason to believe that the event of a technological singularity might
indeed not be that far away. But this remains to be seen. Obviously, the era
of digital computers has not yet ended. Nevertheless, it is important to think
about the changed conditions of a sub-symbolic computer culture, which
we may no longer be able to adequately describe as digital.
292 Andreas Sudmann
In contrast to the world of cultural imaginations, where immortality

and death, series and events, infinity and finiteness, constitute a symbolic,
interwoven network and thus a mysterious world sui generis, the question
of AI infrastructures and their temporalities as a techno-political condition
of the present almost adheres to something soothingly down-to-earth: be it
as a view of the material preconditions of metaphysical speculations or as a
more precise determination of the organizational relationship of media in
relation to their environments. Perhaps today more than ever, to preserve
the utopian potential of AI technology, we must lead it out of the mode of
its speculative discourse.
Notes
1. “Media and Infrastructures of Artificial Intelligence: Computer Vision,

Transfer Learning, and Artificial Neural Networks as Black Box.”, funded by
the German Research Foundation since 2019.
2. See Ian Goodfellow, Yoshua Bengio, and Aaron Courville, Deep Learning
3. See Frank Rosenblatt, “The Perceptron: A Probabilistic Model for Informa-
tion Storage and Organization in the Brain,” Psychological Review 65, no. 6
(1958): 386–408. https://doi.org/10.1037/h0042519.
4. See Vernor Vinge, “The Coming Technological Singularity: How to Survive
in the Post-Human Era,” in NASA Lewis Research Center, Vision 21: Interdisci-
plinary Science and Engineering in the Era of Cyberspace, NASA Conference
Publication CP-10129 (San Diego, CA: NASA, 1993), 11–22; Nick Bostrom,
Superintelligence: Paths, Dangers, Strategies (Oxford: Oxford University
Press, 2014).
5. See Michel Serres and Bruno Latour, Conversations on Science, Culture, and
Time, trans. Roxanne Lapidus (Ann Arbor: University of Michigan Press,
1995).
6. Andreas Sudmann, “Deep Learning als dokumentarische Praxis,” Sprache
und Literatur 48, no. 2 (2017): 155–170.
7. Wolfgang Ernst, Digital Memory and the Archive, ed. Jussi Parikka (Minne-
apolis: University of Minnesota Press, 2013); Mark B. N. Hansen, Feed-for-
ward: On the Future of Twenty-First-Century Media (Chicago: University of
Chicago Press, 2015); and Shintaro Miyazaki, “Algorhythmics: Understand-
ing Micro-Temporality in Computational Cultures,” Computational Culture 2
(2012), http://computationalculture.net/algorhythmics-understanding-mi-
cro-temporality-in-computational-cultures/.
8. Jacques Derrida, Marges de la Philosophie (Paris: Minuit, 1972).
9. See Lisa Parks and Nicole Starosielski, eds., Signal Traffic: Critical Studies of
Media Infrastructures (Urbana: University of Illinois Press, 2015); Gabri-
ele Schabacher, “Medium Infrastruktur: Trajektorien soziotechnischer

Netzwerke in der ANT,” Zeitschrift für Medien- und Kulturforschung 2 (2013):
129–148; Axel Volmar, ed., Zeitkritische Medien (Berlin: Kadmos, 2009).
10. Susan Leigh Star, “The Ethnography of Infrastructure,” American Behavioral
Scientist 43, no. 3 (1999): 380.
11. See John Durham Peters, The Marvelous Clouds: Toward a Philosophy of
Elemental Media (Chicago: University of Chicago Press, 2015).
12. “Media and Infrastructures of Artificial Intelligence: Computer Vision,
Transfer Learning, and Artificial Neural Networks as Black Box,” funded by
the German Research Foundation since 2019.
13. Nevertheless, as I have learned during my ethnographic study so far, there
are many crowdworkers who take on these microtasks repeatedly, some-
times even for years.
14. Yanchen Wang, “Deep Learning in Real Time — Inference Acceleration
and Continuous Training.” Medium.com, 2017. https://medium.com/synce-
dreview/deep-learning-in-real-time-inference-acceleration-and-continu-
ous-training-17dac9438b0b.
15. Uwe Brinkschulte and Heinz Wörn, Echtzeitsysteme. Grundlagen, Funktions-
weisen, Anwendungen (Berlin: Springer, 2005); Volmar, Zeitkritische Medien,
10, 18.
16. Wang, “Deep Learning in Real Time.”
17. Ethem Alpaydin, Machine Learning. The New AI (Cambridge, MA: MIT Press,
2016).
18. Wang, “Deep Learning in Real Time.”
19. See Sprenger in this volume.
20. See, e.g., Matteo Pasquinelli, “Machines that Morph Logic: Neural Networks
and the Distorted Automation of Intelligence as Statistical Inference,” Glass
Bead Journal, Site 1, Logic Gate: The Politics of the Artifactual Mind (2017),
https://www.glass-bead.org/article/machines-that-morph-logic/?lang=en-
view.
21. See also Marvin Minsky’s remarks from the early 1980s on the creativity of
machines and how in comparison human creativity tends to be overrated
and mystified whenever people discuss the limits of AI. Marvin Minsky,
“Why People Think Computers Can’t,” AI Magazine 3, no. 6 (1982), 3–15.
22. Ilya Sutskever, “Training Recurrent Neural Networks.” PhD diss., University
of Toronto, 2003, 10–11.
23. Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton, “ImageNet Clas-
sification with Deep Convolutional Neural Networks,” Advances in Neural
Information Processing Systems 25 (2012): 1097–1105.
24. Andreas Sudmann, “Wenn die Maschinen mit der Sprache spielen,” Frank-
furter Allgemeine Zeitung, November 2, 2016. However, a first demo version
was not presented to the public until 2017. See https://www.youtube.com/
watch?v=7Fntm8cCbNM&t=45s.
25. Later, in early summer 2019, TwentyBN started to develop its AI trainer for
the fitness market. Accordingly, the company then focused on creating
294 Andreas Sudmann
videos, showing different fitness exercises in order to train the AI model.

The corresponding app called “Fitness Ally” was released exclusively on iOS
in May 2020.
26. At the same, we have to acknowledge that AI’s learning systems, even if they
most certainly become more and more embodied as well as situated in our
empirical world, nevertheless always remain an alien form of intelligence,
very much different from the embodied intelligence of human beings.
27. Of course, this erasure of context is again a problematic ideological
operation. In addition, the operations of the learning algorithms and in
particular the learning data itself can be biased in many different respects.
See Alexander Monea, “Race and Computer Vision,” The Democratization
of Artificial Intelligence: Net Politics in the Era of Learning Algorithms, ed.
Andreas Sudmann (Bielefeld: Transcript, 2019), 189–208.
28. Sudmann, “Wenn die Maschinen mit der Sprache spielen.”
29. Friedrich Kittler, “Die Künstliche Intelligenz des Weltkriegs: Alan Turing,” in
Friedrich Kittler and Georg-Christoph Tholen, ed., Arsenale der Seele. Litera-
tur- und Medienanalyse seit 1870 (Munich: Fink, 1989), 187–202, 195.
30. Andreas Sudmann, “Szenarien des Postdigitalen. Deep Learning als Medien-
Revolution,” in Machine Learning: Medien, Infrastrukturen und Technologien
der Künstlichen Intelligenz, ed. Christoph Engemann und Andreas Sud-
mann, (Bielefeld: Transcript, 2018), 55–73.
31. At the same time, it is still important to note that ANN based systems oper-
ate as deterministic machines. See Sudmann, “Szenarien des Postdigitalen.”
32. Friedrich Kittler, Grammophon, Film, Typewriter. (Berlin: Brinkmann und
Bose, 1986).
33. Paul Virilio, La Machine de Vision (Paris: Galilée, 1988).
About the Author
Andreas Sudmann is a media scholar and AI researcher at Ruhr-University

of Bochum, Germany. He is the author of several books, edited collections,
and essays in the field of media and digital culture studies, including The
Democratization of Artificial Intelligence. Net Politics in the Era of Learning Al-
gorithms (Columbia University Press/ Transcript 2019), “Games and Artificial
Intelligence” (Special Issue of Eludamos. Journal for Computer Game Culture
Vol. 10 (2019), and “Artificial Intelligences” (“Künstliche Intelligenzen”),
German Journal for Media Studies (Zeitschrift für Medienwissenschaft)
Vol. 21 (2019).
16. Technics of Time: Values in Future
Internet Development
Britt S. Paris
Abstract
To understand how time can be considered both a technical and sociocul-
tural design value, this chapter investigates named data networking (NDN),
a new networking protocol conceptualized to replace addressed-based
internet protocol and promises to increase both the speed and the effi-
ciency of the internet. Bernard Stiegler’s technics and time framework
guides the analysis of time-based values articulated by the NDN project
principals to demonstrate how collective temporality is built into technical
systems as engineers reconcile social and cultural concepts of temporality
with computational and architectural time and resource constraints in
network design. This chapter shows that although efficiency is a time-
based technical value driving NDN development, the sociocultural values
of information temporality is much less understood by project principals.
Keywords: technics, internet infrastructure, time, temporality, technology

design, Bernard Stiegler
The advent of 5G, Google Fiber, and other recent technologies that augment
internet speed has meant very little to most people. With the technologies
currently available to most users, this super high-speed internet is difficult
to access.1 At the same time, internet service providers (ISPs) underserve
large swathes of the population because they are averse to building the
new infrastructure that would bring internet services up to speed for these
groups. Internet speed is addressed in this paper as a fundamental dimension
of technical time present in technological engineering projects. To charac-
terize how time is considered both a technical and social design value in
developing internet infrastructure, I look to named data networking (NDN),
doi 10.5117/9789463727426_Ch16
296 Britt S. Paris
a new networking protocol conceptualized to replace addressed-based

internet protocol and, according to its project documents, dramatically
increase both the speed and the efficiency of the internet.
The internet is often described as a “network of networks” that transmits
information, or data, in the form of small, discretized packets, using a stan-
dardized set of instructions for communication procedures, called protocols.
Packets then circulate through the internet following the procedures of TCP/IP
(Transmission Control Protocol/Internet Protocol), HTTP (Hypertext Transfer
Protocol), and other common protocols that structure internet networks. Alex
Galloway’s book Protocol points to the ways these structuring procedures effect
patterns of control over human relation after decentralization.2 He states that
protocols promote information speed, so that networked computer-dissem-
inated information becomes a “natural extension of the user’s own body.”3
Picking up where Galloway leaves off, I elaborate in this chapter on a
specific example of how protocols interact with and extend users’ bodies to
promote information speed. Guiding this project is Bernard Stiegler’s notion
of technics as fundamental to the phenomenon of time. For Stiegler, technics
are co-constitutive of human temporality, where technics refers broadly to
material practices, techniques, and technologies that shape and are shaped
by human temporality and the time of social coordination. In the first volume
of Technics and Time, Stiegler focuses on subjective temporality, analyzing
how technologies of materialized, externalized time inform how individuals
experience and conceive of time.4 Extending this analysis in volumes two
and three, he addresses collective temporality through the modes by which
time has been made material and exterior to the human and, importantly for
this chapter, how this externalization of time allows for social and cultural
coordination.5 Applying Stiegler’s framework to the time-based values artic-
ulated by the NDN project principals, we can see how notions of collective
temporality are built into the technical systems surrounding protocols and
how engineers reconcile social concepts of time with computational and
architectural constraints in network design. We also see, however, the current
limits of these engineering promises and how, although efficiency is a time-
based technical value driving NDN development, the user-facing, social value
of information temporality is much less understood by project principals.
Internet Past, Present, and Future
Internet Protocol (IP) is an elaborate addressing system that determines

the origins and destinations of packets. Transmission Control Protocol
Technics of Time: Values in Future Interne t Development 297
(TCP) sets parameters for sending and confirming receipt of packets as they
are transported to and from their destinations. TCP requires end-to-end
communication, or that a sender and receiver at each end must confirm
they are both online for a packet to be sent. The TCP/IP layers connect
the lower-level hardware and the user-facing applications layers and are
often called the networking and control layers using the International
Organization for Standardization (ISO) framework.6 Once two networks
are connected, end-to-end communication becomes possible with TCP/
IP. In this scenario, any end node on the internet can communicate with
any other, regardless of their physical location or network affiliation. TCP/
IP’s end-to-end communication requires only that the end nodes “know”
traffic transmitted through the network, therefore keeping the network
architecture itself “dumb” or unaware of any details about the packets
being transferred. It is this end-to-end design that many proponents claim
has allowed the internet the openness to become a global communication
system.7
The widespread popularity of social media and mobile devices in the
mid-2000s precipitated a paradigm shift in the nature of internet traffic.
Mobile phones liberated people from their desktops and pushed for increased
internet connectivity to accommodate users who tapped into mobile streams
on the go. While the original premise of the Web was sharing existing
documents, social media encouraged a notion of a new mode of using the
Web that generates and transmits enormous amounts of data. The increased
demand for connectivity and application-based communication caused
internet researchers and funding agencies to worry that the new paradigm
shift would be the demise of TCP/IP and end-to-end communication.8 To this
end, the National Science Foundation’s (NSF) Computer and Information
Science and Engineering (CISE) program initiated the Future Internet
Design (FIND) program, the first-generation future internet projects funded
by the NSF.9 In 2009, the NSF hosted a Future Internet Summit to survey
the results and formulate a call for the next round of projects under the
banner of Future Internet Architectures (FIA). For this round of funding,
sociologists and policy experts were mobilized under the banner of the
Values in Design (VID) Council to help formulate the call and to work with
the FIA Architecture project winners as part of an anticipatory ethics
project. This project pushed FIA engineers to begin the design process
with socio-cultural values in mind.10 In 2010, CISE funded four projects at
$8 million each for three years; in 2014, Named Data Networking (NDN)
was one of three programs that were awarded a further three years of Next
Phase support through 2017.11
298 Britt S. Paris
Named Data Networking (NDN)
Fundamentally, NDN wants to increase internet efficiency by simplifying

networking to transmit data based on the name of the data, not its loca-
tion. Overall, the NDN documents show six architectural principles that
guide the NDN architecture. The first three, they say, are gleaned from
the successes of IP routing, and the latter three build from the failures
or challenges that IP routing has presented in recent years. Nearly every
document concerned with what NDN is and how it works describes the
hourglass architecture of the existing TCP/IP internet that “makes the
original IP design elegant and powerful” because it takes packaged data
and delivers it to applications and users. The hourglass architecture with
the narrow waist is represented in figure 16.1. It centers on a “universal
network layer (IP) implementing the minimal functionality necessary for
global interconnectivity.”12
The thin waist of named data is the focus of NDN design. This design
feature is opposed to the thin waist of IP architecture that transmits packets
according to IP addresses. This simple change at the thin waist results in
significant differences between IP and NDN in their function.
In IP networks, nodes and links may overload once content becomes
popular and is requested often, such as a video going viral. In NDN,
more requests also mean more nodes will have a copy of the popular
content in the cache. 13 NDN’s routing strategy then focuses on the IP
address and control signaling to transmit packets, but it instead focuses
on requests or “interests” calling for named data cached within the
network nodes. The probability that a node near the application or
a user on the path to the content generator has a cached copy of the
content increases by its popularity. Via the NDN caching mechanism,
copies of content are automatically distributed toward the parts of the
network where the request is made.14 One will note, however, that push
notif ications from producers of interests are not something that NDN is
built to accommodate because all data must be requested. This makes it
diff icult to build real-time applications over NDN, for reasons detailed
later in this chapter.
Moreover, the popular press promotes NDN as an architecture that
is “faster” or “swifter,” or that “allows smoother content streaming” than
today’s TCP/IP-based internet.15 Given that these various terms for improved
efficiency are lauded in the project documents as advantages and that they
form the public-facing rationale for the projects, it makes sense to examine
what these time-laden terms mean in practice.
Figure 16.1. Comparing IP and NDN at the narrow waist16
Efficiency as Technical Value
Stiegler’s first volume of his Technics and Time series focuses on how all
technologies, broadly defined, are both material and temporal, capable of
physical movement across space and memory transmission across time.
In keeping with Stiegler’s notion of technologies as material and, as such,
temporal artifacts, the builders of the NDN technical system make decisions
about time and how to conceive of it.
In the design of technical systems, the broad notion of efficiency is often
cited as the primary goal. It is also a clear juncture at which the material
dimensions of technical time become bound with notions of collective
temporality. The importance of time as one of the first and most fundamental
computational resources was highlighted by Ada Lovelace in 1843 in her
description of the technical requirements for Charles Babbage’s Difference
Engine:
In almost every computation a great variety of arrangements for the

succession of the processes is possible, and various considerations must
influence the selections amongst them for the purposes of a calculating
engine. One essential object is to choose that arrangement which shall
tend to reduce to a minimum the time necessary for completing the
calculation.17
Efficiency is bound with time and is articulated as a primary goal of net-

working research. This sentiment was reflected in a focus on efficiency in
many of the NDN documents. Shilton found that with NDN, efficiency was
300 Britt S. Paris
often articulated as dynamism, in comparative statements such as, “Content

transfer via [NDN] is always secure, yet the results show that it matches
the performance of unsecured HTTP and substantially outperforms secure
HTTPS.”18
Lixia Zhang, NDN’s primary networking engineer, suggested that
efficiency drives internet technology and that user demand determines
how engineers balance efficiency with functionality, but she noted that
user-facing designs only move forward as technology allows them to do
so.19 Indeed, the most frequently expressed values in the NDN founding
documents were those responding to technical pressures, specif ically
methods for ensuring improved efficiency within the material constraints of
the network architecture. And, moreover, many of the values emphasized by
the NSF’s 2007 request for proposals were technical values such as scalability
and reliability.20 This is not surprising in a research setting where technical
innovation is the primary motivator and marker of success. However, when
taken together, these expressions of values suggest that the sociocultural
values that the VID Council hoped to impart in the anticipatory ethics stage
of the project were understood as less important than the technical values.
Zhang’s remarks also encapsulate the model by which NDN network
engineers think about values in design. User demand is understood super-
ficially in terms of the technological functionality at the application layer,
instead of any complex reckoning of what social values, hierarchies, and
other ethical considerations might actually affect user activities, broadly
defined. Other NDN respondents articulated notions of promoting efficiency
by balancing computational resources like bandwidth and storage; the
simplest computational resources are those of computation time (the number
of steps necessary to solve a problem) and memory space (the amount of
storage needed while solving the problem).
NDN’s lone application developer noted that when producing applications
that can generate and handle real-time audiovisual streams, there are issues
of timestamps, which are fundamental to the coordination of nearly all
networks. Routing strategies, such as the best-route strategy used in the
TCP/IP internet requires data to be timestamped to be made into packets
and sent. Similarly, NDN’s routing strategy mentioned above does the same.
However, because data must be called by name in real-time flows, each
piece of data must have a timestamp or the order number available in the
data’s namespace. As each segment is produced, the consumer must call for
the data using the proper name, order, or timestamp. Then the application
must order the data incoming to the consumer in the right way and quickly
enough so that it appears to be a real-time stream on the consumer side.
At the level of applications, Gusev describes that the overhead for the
packet header for the NDN named data alone is like 30% of its size, which is
really inefficient.”21 Gusev reported that when developing applications: “We
work on algorithms first and make sure they are configured properly. Once
that is sorted, we are very concerned with how well the algorithms work in
a particular scenario. Generally, we aren’t as interested in optimization as
we are in just getting the applications to work.”22
This highlights two important issues that must be mentioned in terms
of NDN and new internet infrastructure engineering in general: First,
infrastructure rarely faces users, so it perhaps makes sense that they aren’t
thinking as much about speed or user-facing temporality, even though this
is lauded in public-facing documents. Second, building out infrastructure
is slow, diff icult work. Gusev, working alone, struggles with designing
even basic applications because the NDN namespace, a core component
of the NDN architectural design, is not configured to facilitate real-time
applications.
While managing efficiency is overwhelmingly considered to be a bal-
ancing act between time and the materiality of the technical system, and
users are not considered in the equation, there are technical junctures at
which the materiality of the system is subject to the temporalities of social
coordination. Gusev noted that NDN uses C++ code for all application
development because that language has low barriers to entry and allows new
collaborators to focus on the important issues of technical design instead
of learning a new language. In this case, using C++ saves everyone time and
assumes that the optimization of the work will happen at some later date.
Similarly, protocols by their very common definition entail sociocultural
coordination. NDN testbed manager John DeHart maintained that Net-
work Time Protocol (NTP) is instrumental in running the testbed.23 NTP
coincides with IP and is an enduring standard that has been established by
standards governance bodies. NTP syncs devices within networks according
to time-stamped data to ensure smooth end-to-end communication over the
internet networks. The in-project decisions to use a common programming
language and run the testbed in NTP suggest an aim toward easier in-project
coordination, so that the work is also interoperable with other standardized
bits of infrastructure.
Thinking through how notions of efficiency are leveraged and put into
practice in the NDN project shows how time is considered a material
thing—a computational resource that can be broken into many subsets,
assigned timestamps and organized accordingly in technical practice. This
technoscientific truth is alluded to in volumes two and three of Stiegler’s
302 Britt S. Paris
Technics and Time, with reference to how networks transmit temporal

objects quickly and widely on the basis of informatics:
Informatics, as Daumas understood it, is the industrial exploitation

of information’s value made possible by the development of electronic
technologies as elements of the mastery of speed through the stabilizing,
processing, and transmission of signals that are recordable and storable
in electronic memory, making possible the control of information’s
circulation though the establishing of networks.24
For Stiegler, the result of this new programming culture industry amounts
to the “pauperization” of the social and cultural realms; however, he leaves
open to further analysis how this happens in practice. Indeed, he seems
to regard informatics as monolithic and incapable of being questioned or
held accountable, and treats humans as mere agents within the informatics
industry who do the work of pushing technology and demanding higher
speeds of humans. What becomes necessary is to bracket out the part of
Stiegler’s project that focuses on the cultural implications of informatics and
investigate how the technological project in question, NDN, is constructed
by people to push toward interface speed and consider how NDN engineers
think of notions of speed in their work.
Speed and User-Facing Temporality
The notion of speed permeates Stiegler’s Technics and Time. Following its
definitions throughout the series, one can get lost in a number of complex
anthropological examples that point to the human possibilities of memory
and information transmission across space and how contemporary technol-
ogies accelerate society and culture to the degree that the present becomes
ungraspable and the future nearly incomprehensible. The definition I use for
this project is that of the acceleration of technology, which Stiegler claims
urges humans to move ever faster and faster to keep up with increasingly
“real-time” technical systems.25 He locates Adorno and Horkheimer’s analysis
of the culture industry within the longer arc of information technology,
seeing a connection between today’s data practices and the way cinema
historically encouraged a loss of social and communal interaction in favor of
industrial prerogatives for sustaining increases in production and economic
profit through regulating consumption.26 He says the informatics culture
industry demands higher speeds but says little of the values animating the
industry other than, in naming it a culture industry, tying it to capitalism’s

push to render all of human life as a commodity. Although technological
design practices that value material eff iciency account for part of the
equation of technical drivers of human time, they cannot complete it. It
becomes necessary to address human factors, in this case, by unraveling
these concepts of speed and how they manifest—or don’t—in the articulated
culture and political views of the NDN project principals.
At face value, NDN’s engineers envision their specific protocols not so
much as affecting the user-facing speed of applications but as offering, each
in its own way, a new palette of possibilities at the application layer because
of how protocol decreases network latency or enables higher performance
at the network edges. In the project with NDN, the principals alluded to,
but never clearly articulated, the idea that speed relates to user-facing
temporal experience. Moreover, the ways the principals articulated speed
are particularly evocative of how they view the advantages of their respec-
tive projects. However, engineers have a hard time effectively building for
protocol and interface speed in their work.
Jeff Burke noted that, while NDN is just a networking solution that seeks
to improve efficiency in the transmission of data over the internet, it is
possible that some notion of faster user experience features might be possible:
NDN is not going to change the speed of light. It’s not going to change
the typical behavior of networks. Is it going to be the same for the end
user? I’m not sure. For example, maybe the idea of scrubbing video that’s
streaming, because of the way that it happens on NDN versus how it
typically happens on IP—something like random access into video may
actually perform better. While you’re not talking about fundamentally
the network, I’d say the application-level behavior might be different. …
If we’re interested in applications that do more real-time selections
of—everything from the perspective on a scene from an immersive
camera or light-f ield camera—anything that involves making quick
decisions about what’s being delivered over the network.27
Increasing efficiency is again a persistent topic in engineers’ articulations of

time as a resource or user-facing temporality in these projects. User-facing
speed, on the other hand, is more of a countervailing concept that forces
technologists to think about temporality as the social value of their work.
Speed, or, more broadly and colloquially, user-facing temporality, is a
social dimension of efficiency, a crucial concept in NDN engineering that
has many potential valences. In technical practice, smaller-scale latency
304 Britt S. Paris
management in system components makes user-facing performance or

speed possible. Furthermore, locating speed as a subcategory of efficiency
highlights how dogged adherence to the technical design value of efficiency
seeks to make the design appear frictionless to users and, as a byproduct,
the technical underpinnings of the system inscrutable.
Materiality and Cultural Coordination of Technical Time
Recalling Alex Galloway’s claim that protocols promote information speed,

so that the information purveyed through computational devices become a
“natural extension of the user’s own body,” the moments of NDN design pre-
sented in this chapter support the idea that networking protocols materialize
time into information in a way that is beyond the perceptual capacities of
humans.28 While the principals seem to be unclear or even neglectful of how
this new protocol schema would change user-facing temporal experience
at the interface, there is evidence that the NDN technical system would
conceivably allow different types of temporal experiences. At the same
time, NDN invokes human temporal perception and experience as a partial
justification for its work.
Time is materialized into information as packets and data are named,
time stamped, and transported in order as they move through the NDN
network, as, for example, in Gusev’s discussion of the way timestamps are
allocated and used in the development of applications that can handle
real-time audiovisual streams. Through these temporal codes, time is made
into a thing and used to determine the most efficient path for the data to
take. This most-efficient path not only entails the lowest resource cost to
the network but also, as a result of its low overhead, theoretically causes
the network to function faster.
In-project standards, such as the use of C++ code, facilitate the devel-
opment of applications because of their low overhead and lowered use of
resources in the network and in any user’s CPU. Preexisting protocols such
as NTP also impart material requirements (the need for time-stamped
data) in the process of design and as such represent fundamental obstacles
or technical concerns that must be attended to in the design processes of
ordering functions in technical systems.
The technologies that Stiegler and, before him, Lewis Mumford wrote
about are defined as material, whether the physical objects that mediated
time in Stiegler’s case or those that performed social or cultural mediations
in Mumford’s case. In the same way, new networking protocols are not only
grounded in but also concerned with materiality. The engineers’ technical

goals—primarily, but not exclusively, efficiency—structured every choice
they made. The analysis presented here indicates that the technical design
value of efficiency is chiefly a material concern, that is, one of maximizing
use of material computational resources and keeping latency low, so that
the system can perform faster.
Finally, in NDN, we see that technical materiality has directly caused
problems with the social coordination of these protocol projects. While
efficiency and a faster internet were NDN’s partial stated goals, building
out the network and troubleshooting has taken more time than they had
originally imagined. Principals from NDN articulated how the namespace
load carried with each piece of data or content project is still quite bulky,
as it takes up more physical space and resources to process, and contributes
to the material load of these technologies. Thus, the social and presumably
economic good of the speed touted by NDN that would come from optimi-
zation has not yet come to pass.
Much work still must be done to get the NDNs’ functional efficiency to be
on par with the conceptual efficiency touted by project principals. It remains
to be seen whether this imagined efficiency or any combination of the related
concepts discussed in this chapter will come to fruition for NDN. In any
case, we see how efficiency and its subset of speed are used as marketing
tools to make an argument about NDN’s viability for the uninitiated. The
notion of speed is bound with concepts of the user-facing temporality that
would be made possible by using NDN instead of IP. However, this short
meditation has shown that NDN engineers are not thinking very much
about users, nor social or cultural values; they are instead more concerned
with the materiality of time as it is measured and distributed through the
NDN technical system. At present, NDN is built with regard to speed in
name only. I suggest that a deeper engagement with values would require
the principals to think seriously at all stages in the process to think about
how future interfaces will be built on top of the new networking protocol,
how those interfaces will affect users, their subjectivity, their bodies, and
their relationship to collectivity.
306 Britt S. Paris
Notes
Some language, analysis, and interview data presented here are also found
in the author’s previous work: Britt Paris, “Time Constructs, The Origins of
A Future Internet” (University of California, Los Angeles, 2018).
1. Farhad Manjoo, “What Do You Do With the World’s Fastest Internet
Service?” Slate, March 12, 2013, http://www.slate.com/articles/technology/
technology/2013/03/google_fiber_review_nobody_knows_what_to_do_
with_the_world_s_fastest_internet.html; and “Today, Kansas City. Tomor-
row, Oklahoma City!” Slate, March 13, 2013, http://www.slate.com/articles/
technology/technology/2013/03/google_fiber_internet_service_after_kan-
sas_city_where_will_the_gigabit_network.html.
2. Alexander R. Galloway, Protocol: How Control Exists after Decentralization
3. Galloway, Protocol, 67.
4. Bernard Stiegler, Technics and Time, 1: The Fault of Epimetheus, trans.
Richard Beardsworth and George Collins (Stanford, CA: Stanford University
Press, 1998).
5. Stiegler, Technics and Time, 2: Disorientation, trans. Stephen Barker (Stan-
ford, CA: Stanford University Press, 2009); Technics and Time, 3: Cinematic
Time and the Question of Malaise, trans. Stephen Barker (Palo Alto, CA:
Stanford University Press).
6. Hubert Zimmerman, “OSI Reference Model-The ISO Model of Architecture
for Open Systems Interconnection,” IEEE Transactions on Communications
28, no. 4 (1980): 425–32, https://ieeexplore.ieee.org/stamp/stamp.jsp?ar-
number=1094702.
7. David D. Clark, “The Design Philosophy of the DARPA Internet Protocols,”
Proceedings of SIGCOMM ‘88, Computer Communication Review 18, no. 4,
106–114, https://doi.org/10.1145/52324.52336; David D. Clark, John Wroclawski,
Karen R. Sollins, and Robert Braden, “Tussle in Cyberspace: Defining Tomor-
row’s Internet,” IEEE/ACM Transactions on Networking 13, no. 3 (June 2005),
462–475, https://doi.org/10.1109/TNET.2005.850224; J. H. Saltzer, D. P. Reed,
and D. D. Clark, “End-to-End Arguments in System Design,” ACM Transac-
tions on Computer Systems 2 no. 4 (November 1984), 277–288. https://doi.
org/10.1145/357401.357402; Tarleton Gillespie, “Engineering a Principle: ‘End-
to-End’ in the Design of the Internet,” Social Studies of Science 36, no. 3 (2006):
427–57, https://doi.org/10.1177/0306312706056047. Mark A. Lemley and Law-
rence Lessig, “The End of End-to-End: Preserving the Architecture of the In-
ternet in the Broadband Era,” SSRN Scholarly Paper ID 247737 (Rochester, NY:
Social Science Research Network), http://papers.ssrn.com/abstract=247737.
8. Clark et al., “Tussle in Cyberspace”; Gillespie, “Engineering a Principle.”
9. Darleen Fisher, “US National Science Foundation and the Future Internet
Design,” in ACM SIGCOMM Computer Communication Review 37 (2007):
85–87, http://www.nets-find.net/InvitationToJoin.php; “NSF Future Internet
Architecture Project,” National Science Foundation Future Internet Archi-

tecture Project (n.d.), http://www.nets-fia.net/.
10. Helen Nissenbaum, Luke Stark, and Katie Zeiwitz, “Values in Design
Council: An End of Project Report NSF Eager: Values in Design in the Future
Internet Architecture,” 2013, http://www.nyu.edu/projects/nissenbaum/vid/
pdfs/VIDCouncilReportv2.pdf.
11. National Science Foundation, “NSF Announces Future Internet Archi-
tecture Awards,” August 2010, https://www.nsf.gov/news/news_summ.
jsp?cntn_id=117611&org=OLPA&from=news.
12. Lixia Zhang, Deborah Estrin, Jeffery Burke, Van Jacobson, Beichuan Zhang,
James D. Thornton, Diana K. Smetters, et al. “Named Data Networking
(NDN) Project,” Technical Report NDN-0001, October 31, 2010, https://
named-data.net/publications/techreports/tr001ndn-proj/.
13. Zhang et al., “NDN Project,” 7; J. Takemasa, K. Taniguchi, Y. Koizumi, and T.
Hasegawa, “Identifying Highly Popular Content Improves Forwarding Speed
of NDN Software Router” (conference paper, 2016 IEEE Globecom Work-
shops), 1–6, https://doi.org/10.1109/GLOCOMW.2016.7848926.
14. Zhang et al., “NDN Project.”
15. Bob Brown, “IP Was Middle School, Named Data Networking Is College,”
Network World, October 8, 2015, http://www.networkworld.com/arti-
cle/2990834/network-management/ip-was-middle-school-named-data-
networking-is-college.html; Minerva Bauman, “NMSU Professor Working on
New Wireless Networks,” Las Cruces Sun-News, September 30, 2017, http://
www.lcsun-news.com/story/news/education/nmsu/2017/09/30/nmsu-pro-
fessor-working-new-wireless-networks/718803001/; Rutgers University,
“Designing the Future Internet,” ScienceDaily, October 20, 2016, https://
www.sciencedaily.com/releases/2016/10/161020115310.htm; and David Talbot,
“Your Smartphone and Tablet Are Breaking the Internet,” MIT Technology
Review, January 9, 2013, https://www.technologyreview.com/s/509721/your-
gadgets-are-slowly-breaking-the-internet/.
16. Van Jacobson et al., “Named Data Networking (NDN) Annual 2012-13
Report,” 2013, p. 2, http://www.caida.org/publications/papers/2013/named_
data_networking_2012-2013/named_data_networking_2012-2013.pdf.
17. Augusta Ada King Lovelace, Annotated Sketch of the Analytical Engine In-
vented by Charles Babbage, Extracted from the “Scientific Memoirs,” ed. Luigo
Federico (R. & J. E. Taylor, 1843), 710.
18. Van Jacobson, Deborah Estrin, Lixia Zhang, Beichuan Zhang, kc claffy,
Dimitri Krioukov, Dan Massey, et al., Named Data Networking (NDN) Annual
2012–13 Report (2013), 10, http://www.caida.org/publications/papers/2013/
named_data_networking_2012-2013/named_data_networking_2012-2013.pdf;
in Katie Shilton, “Anticipatory Ethics for a Future Internet: Analyzing Values
During the Design of an Internet Infrastructure,” Science and Engineering
Ethics 21, no. 1 (2015): 9, https://doi.org/10.1007/s11948-013-9510-z.
19. Zhang, personal communication, December 10, 2017.
20. Fisher, “US National Science Foundation and the Future Internet Design.”
308 Britt S. Paris
21. P. Gusev, personal communication, August 2, 2017.

22. P. Gusev, personal communication, August 2, 2017.
23. J. DeHart, personal communication, November 23, 2018.
24. Stiegler, Technics and Time, 2, 102.
25. For Stiegler’s discussion of the complications inherent in “real time,” see
Technics and Time, 2, 62–63.
26. Stiegler, Technics and Time, 3, 37.
27. J. Burke, personal communication, September 12, 2016.
28. Galloway, Protocol, 67.
About the Author
Britt S. Paris is an Assistant Professor in the Department of Library and

Information Science at Rutgers University, New Brunswick. Her work
is broadly focused on how values and ethics directives are rolled out in
technological projects and information systems and how these activities
affect political action.
Index
acceleration 9, 10, 21, 26–27, 30, 32, 86, 128, calendars 11, 18–21, 36n, 111, 114, 133, 228–229,
138, 144, 177–179, 182, 245, 257, 285, 291, 302 269
of culture and life 9–10, 32 n. 2, 36 n. 36, Canales, Jimena 25, 37 n. 66, 53, 172 n. 14
42, 138, 144, 177–179, 182, 257 capitalism 10, 12, 23–25, 28, 67, 133, 137, 187,
of technology 9–10, 30, 128–131, 144, 285, 226, 234, 276n, 283–286, 303
291, 293n, 302 capitalist time 18–19, 23–25
and modernity 9, 19, 32 n. 2 Carey, James 17, 35 n. 35, 180–181, 188 n. 7
see also speedup cellphone 233–235; see also mobile devices
accelerationism 10, 32 n. 2, 83, 86 and smartphone
aging (of technology and infrastructures) 14, chronic time (Benveniste) 20–21, 36 n. 44; see
31, 62, 66, 191–193, 200–201; see also also Benveniste, Émile
obsolescence Chun, Wendy 28, 34 n. 16, 69 n. 2, 135, 142 n. 72
algorithms 13–14, 17, 25, 28, 30, 80–82, 86, clocks 16, 18–25, 31, 37 n. 61
89–103, 126–142, 143–156, 157–174 clocking (microprocessors) 90–91, 92, 94
algorithmic bias 13, 33 n. 13, 286 quartz clock 94, 226
anticipatory shipping 263–278 atomic clock 95, 114–116
stock trading 28, 90, 127–129, 139 n. 19, 140 mechanical clock 21, 90
n. 25, 182 clock time 19–23, 47, 96, 108, 114–115,
machine learning 17, 82, 127, 142 n. 73, 119–120
148–150 stopwatch 18, 227, 229, 232
recommendation 80–81, 213–216, 267, 273 wristwatch 31, 95, 102, 113, 224–225, 227,
self-driving cars 157–174 234, 238 n. 23 and n. 30, 239 n. 41–44; see
animation 143–156 also wearables
anonymous time 24–25, 90–91 cloud computing 126, 153 n. 6, 182
anticipation 31, 80, 91, 112, 158, 172 n. 22, 200, computer time 14, 117, 128–129, 133, 137–138,
203 n. 12, 205–219, 263, 266, 270–272 143
anxiety 31, 110, 145–46, 152, 205–219 control (cybernetics) 67–68, 94, 208, 244, 271
artificial intelligence (AI) 83, 127, 134, 139, Coordinated Universal Time (UTC) 94–95,
143–156, 279–294 114–117
artificial neural networks (ANNs) 143, coordination
146–148, 150, 256, 279–294 of time technologies 11–15, 19–21, 28, 114,
Ashby, W. Ross 135–136, 142 n.81, 261 n. 24 116–119, 121–122, 181, 229, 247, 300–301,
atomic clock 95, 114–116 304
Augustine of Hippo 41, 77–78 of social practices 20–21, 23–25, 114, 179,
305
Babbage, Charles 30, 125–126, 129, 134, 138 n. of human and nonhuman actors 27,
4–5, 139 n. 8, 299, 307 n. 17 94–95, 114, 296, 305
Barthes, Roland 63–64, 72 n. 47 cosmic time 20–21
being-toward-death 79, 83, 91; see also Crary, Jonathan 33 n. 3, 178, 188 n. 5
Heidegger, Martin cultural techniques 15–16, 35 n. 29, 54 n. 17,
Benjamin, Ruha 13, 33 n. 13 55–57, 68, 73 n. 62, 140 n. 29, 267, 276 n. 27
Benjamin, Walter 91, 93, 101 n. 6, 258, 262 n. 52 cybernetics 14, 47, 54 n. 14, 67–68, 94, 135,
Bennett, Jane 12, 33 n. 9 147, 152, 208, 218 n. 8, 252–254, 261 n. 24,
Benveniste, Émile 20–21, 36 n. 41 271–272, 275 n. 12, 277 n. 48
Bergson, Henri 47, 78, 92, 95
Berlant‚ Lauren 14, 34 n. 22, 217 n. 5, 218 n. 22 data
big data 17, 28, 67, 80, 143–144, 208, 262, 283 tracking and surveillance 210, 234, 244,
bitcoin 130, 132; see also blockchain 255–258, 269–270
Bloch, Ernst 101, 208, 217 n. 3 transmission 17, 32, 51, 94–96, 99–100, 120,
blockchain 130, 133, 135, 140n, 141n; see also 144–145, 296–298, 302–303
bitcoin and Ethereum big data 17, 28, 67, 80, 143–144, 208, 262,
body (human) 10, 31, 44, 46, 49–50, 57, 145, 283
178–180, 183–185, 199–200, 209, 214, 296, processing 25, 28, 81, 90, 92, 94, 97, 99,
304–05; see also embodied and lived 164–165, 168, 172, 254–255, 291
temporalities training 144, 279, 283–285
310
storage 28–29, 50, 92, 96, 144–145, 300 hardwired temporalities 11–12, 16–18, 21, 46,
packets 95–97, 277 n. 50, 296–300, 304 57–58, 79, 82, 89–90, 92, 99–100, 111, 180, 187,
database 28, 147, 161, 195, 203, 255, 268, 270 222–223
data mining 211 Hartog, François 125, 135, 138, 142 n. 75
death 31, 43, 46, 51, 62, 79, 91, 136, 167, 206, Hayles, Katherine 127, 160–61, 171 n. 6, 200
210–216, 292 Hegel, Georg Wilhelm Friedrich 45, 51, 83
decision-making 29, 157, 159, 162–164, 168–170, Heidegger, Martin 25, 29, 37 n. 65, 66, 74 n.
171 n. 8, 172 n. 22, 270, 280 65–66, 79, 83, 85, 86 n. 4, 87 n. 15, 90–92, 98,
delay 92, 95–96, 125, 128–129, 133, 163, 181, 184 101, 209–210, 215, 218 n. 14
Deleuze, Gille 34 n. 17, 81, 87 n. 8 Helmholtz, Hermann von 26, 91
Derrida, Jacques 80, 86 n. 86, 292 n. 8 high-frequency trading 127–128, 140 n. 25, 182;
digital see also algorithms, stock trading
infrastructures 55, 64, 112, 120–121, 179, history of science and technology 21, 56, 79,
184, 187, 207–208 82, 134, 243
media 77, 113, 116, 143–145, 152–153, 210, Husserl, Edmund 29, 80, 86, 93–94, 97, 101n
269
networks 14, 18, 31, 111, 113, 121, 153, 213 information
technology 11, 78, 82–83, 86, 107, 144–145, society and age 90, 97, 244
152, 160, 211, 234 theory 162–163, 252
time 11–12, 28, 110, 112, 120–121, 124, 181, processing 97, 162–164, 172 n. 15, 254,
232, 257, 269–270 290–291; see also data processing
diurnal cycle 112–113, 133 infrastructure
Doane, Mary Ann 34 n. 17, 144, 153 n. 1 aging 14, 31, 62, 66, 191–192, 193, 200–201;
see also obsolescence
Edwards, Paul N. 14, 35 n. 24, 63, 72 n. 45, 252, as material 9, 14, 18, 19
260 n. 23 as social form 14
Eigenzeitlichkeit 68, 289 and modernity 14, 19, 23, 101, 181
epoché 98, 101 care of 55–58
Ernst, Wolfgang 13, 29, 122 n. 2, 128, 144, 153 digital 55, 64, 112, 120–121, 179, 184, 187,
n. 3, 154 n. 7 and n. 9, 202 n. 3, 244, 259 n. 3, 207–208
272, 278 n. 52, 292 n. 7 network 145, 179
Ethereum 133 temporality of 14, 55–58, 60–61
Esposito, Elena 263–264, 271–272, 277 n. 45 of temporality 9–12, 15, 18, 21, 23, 25, 27–28
evolution of technology 78–79, 126–127, 291 repair 14, 29, 35n, 56–57, 60–61, 64, 67–68,
179, 184–185, 187
fiber optics 17, 102 n. 13, 181–182, 295 maintenance 10, 29, 36, 55–57, 59–60,
film 13, 27, 34 n. 17, 44, 47–48, 154 n. 13, 65–68, 177, 179, 184–185, 194, 200, 202
213–214, 228, 230, 238 n. 24, 247–250, 258 Innis, Harold 20, 36 n. 40, 45
Foucault, Michel 46, 49, 69 n. 5, 99, 120, 123 innovation 13, 19, 35 n. 25, 142 n. 79, 200, 202,
n. 23, 199 224, 266, 279, 288, 300
future 17, 19–20, 29–30, 32, 46, 53, 57, instantaneity 10, 43, 91, 93, 127, 160, 178
66–67, 77–87, 135–136, 166, 206–211, 261 integrated circuit 15, 17–18, 35 n. 34, 36 n. 37,
n. 39, 265–266, 272–273, 280–281; see also 93–94, 99–100, 126, 131, 222–223, 226–228,
prediction and anticipation 231, 235, 244
Future Internet Design program 297 central processing unit (CPU) 164, 304
graphics processing unit (GPU) 163, 290
Galloway, Alexander R. 116, 122 n. 3, 123 n. 27, as semiconductors 99, 222, 228, 236 n. 9. 237
277 n. 50, 296, 304, 306 n. 2 n. 17, 238 n. 26
gender 56, 239, 257 interface (temporal) 19, 27, 187
Global Positioning System (GPS) 95, 161, 181, interface (computer) 108, 112, 122 n. 3, 143, 146,
210–211 149–153, 155 n. 28, 162–163, 168, 182, 187, 195,
Google 30, 117, 123 n. 31, 127–128, 143–156, 208, 245, 256–258, 302–305
211, 256, 259 n. 1, 278 n. 55, 280, 285, 288, 295 internet 28–29, 33, 90, 92, 99, 109–110, 113,
Grusin, Richard 33 n. 8 208, 218 n. 9 116–117, 133, 135, 136, 179, 182, 186, 209,
211–212, 245, 257, 295–305
Hansen, Mark B. N. 34 n. 16, 144–145, 153 n. 4, Internet time 95–97, 109, 117, 124 n. 33
208, 218 n. 9, 292 n. 7 iPhone 61, 234; see also smartphone and
hardware 28, 62, 95, 108, 112, 121, 187, 192–193, mobile devices
201, 255, 284–285, 290–291, 297 irreversibility 21, 29, 50, 52, 217
Index 311
Jackson, Steven 35 n. 25, 57, 61, 64, 69 n. 1, microtemporality 9, 11–12, 18–19, 25–30, 94,
202 n. 8 99–100, 108, 128–129, 142 n. 74, 146, 149, 157,
159, 161–162, 167–168, 170, 180, 223, 228, 232,
Kittler, Friedrich 29, 32, 34 n. 14, 41, 48–49, 51, 235, 280, 292 n. 7 see also microtime
54, 123 n. 16, 129, 132, 139 n. 9, 140 n. 33, 141 n. mobile devices 64, 69 n. 2, 98, 103, 221,
53, 192, 202 n. 2, 290–291, 294 n. 29–32 223, 233–235, 239 n. 47, 256–257, 262 n.
Krämer, Sybille 36 n. 39, 50, 54 n. 17, 129, 134, 45, 269–270, 297; see also cellphone and
140 n. 29 smartphone
modern time 90
labor 15, 19, 21, 29, 56, 166, 177–182, 185, 187, Mumford, Lewis 21–23, 36 n. 46, 304
226, 232, 237, 256–257, 269, 281, 283
labor-time 15, 19, 21, 23–25, 35 n. 27, 232, 235 Named Data Networking (NDN) 32, 295–308
latency 178, 182, 188 n. 11–13, 201, 285, 303, 305 neoliberal economy 12, 111, 120, 205, 286
Latour, Bruno 15, 33 n. 8, 35 n. 26 and 28, 56, network
64, 70 n. 10–11, 71 n. 19 and 23, 145, 154 n. 7, cable 17, 30, 60, 92, 99, 177–189, 209
265, 275 n. 9, 292 n. 5 digital 14, 18, 31, 111, 118, 121, 153, 213
leap second 29–30, 102, 107–109, 114–122, 128 telegraph 17, 35 n. 35, 92–93, 178, 180–183
legacy systems 191–203; see also obsolescence telephone 99, 127, 183
Leibniz, Gottfried Wilhelm 78, 83 undersea 177–189
leisure time 12, 232 time 112, 140 n. 36, 186
linear time 50, 131–132 wireless 98, 205
logistical media 24, 263, 264, 267–270, 276 n. Network Time Protocol (NTP) 95, 116–117, 137
27 and 33, 277 n. 40 new materialism 2, 12, 33 n. 9
logistics 11, 29, 33 n. 7, 68, 263–265, 267–269, Newton, Isaac 47–48, 132
275, 276 n. 25 and 37 Noble, Safiya Umoja 13, 33 n. 11
longue durée 9, 35 n. 30, 138 nonhuman 12, 15, 19, 27–28, 33 n. 8–9, 56,
89–90, 96, 107–109, 111, 116, 120, 146, 149,
Marx, Karl 15, 35 n. 27, 235 151–153, 180, 187
Massumi, Brian 208, 218 n. 9
McLuhan, Marshall 90, 96–98, 102 n. 22 obsolescence 15, 17, 31, 59–62, 67, 69 n. 2, 187,
McIlwain, Charlton 13, 33 n. 12 191, 193, 198–201, 203 n. 14; see also legacy
media systems
archaeology of 13, 34 n. 15, 89–90, 92, 97,
100, 103, 154 n. 7 and 13, 161, 274 n. 4, 282 pendulum 91, 93, 247
culture 10, 94, 100 Peters, John Durham 16, 20, 22, 24, 29, 34 n. 21,
digital 77, 113, 116, 143–145, 152–153, 210, 37 n. 61, 200, 203, 263, 268–269, 276–277
269 phenomenology 29, 80, 86, 97, 101 n. 7, 205
definition of practice turn 13, 34 n. 20, 155 n. 30
logistical 24, 263, 264, 267–270, 276 n. 27 preemption 31, 77, 81, 86, 159, 164, 207–208,
and 33, 277 n. 40 212, 217 n. 6
mass 258 prediction 16–17, 20, 29–30, 32, 80, 94, 134,
as time manipulation 9, 34 n. 14, 47–48, 136, 145, 158, 164–166, 170–171, 206, 208,
50–51, 54 n. 17, 129 211–212, 233, 249–253, 263–267, 270–273,
theory of 9, 45, 48, 51, 96, 124, 138, 153 n. 2, 280, 285–286
263, 277 n. 41 predictive policing 13, 208
mass 258 processing see data processing
technology 9, 13–14, 50, 69 n. 1, 92, 289 prophecy 263, 270–271
time-critical (zeitkritische Medien) 13, protocol 11, 18, 123 n. 27 and 29, 137, 140 n. 36,
28–29, 34 n. 16, 41, 95–97, 99–100, 102, 166, 169, 295–296, 301, 303–305; see also
108–109, 116, 128, 161–162, 167, 244 Transmission Control Protocol/Internet
memory (computer) 28, 34 n. 16, 96, 131–132, Protocol, Network Time Protocol, and
135, 139 n. 24, 142 n. 74, 272, 288, 300, 302 Named Data Networking
memory (human and social) 36 n. 50, 52, 80,
84, 90–91, 96, 102 n. 16, 201, 253, 261 n. 26, radio 27, 91–93, 95, 98–100, 118, 205, 270
299 real-time 10, 13, 28, 33 n. 10, 50–52, 91, 94, 97,
microprocessor see integrated circuit 113, 123 n. 16, 129, 133, 138, 140 n. 30, 160, 178,
microchip see integrated circuit 256, 263, 269, 284, 293 n. 14, 298, 300–304,
microtime 9, 11–12, 25–30, 18–19, 25–30, 94, 308 n. 25
99; see also microtemporality repair studies 56
312
reversibility 23, 29, 47–48, 50, 52 streaming (video) 10, 50, 96, 297–298, 300,
rhythm 303–304; see also video
bodily 31, 177–189 synchronization 9, 13–15, 18, 23, 35, 82, 84,
circadian 20–21, 27, 130; also as cosmic 89–90, 92–95, 97, 99, 111–113, 116–117, 123,
and celestial 126, 129, 133, 136, 140 n. 36, 144, 161, 164, 233,
seasonal 20, 36 n. 50, 133 247, 250, 258
environmental 16, 24, 31
technological 14, 23, 57, 94, 98–99, 102 n. techné 85, 100
24, 103 n. 25, 149, 272 technics 36 n. 46, 85, 86 n. 6, 87 n. 14, 144, 153,
work 24, 57, 202 n. 8, 252 180, 207, 210, 258, 262, 295–296, 302, 306 n.
Ricoeur, Paul 20, 24, 34 n. 17, 36 n. 42 4–5
routine 55–57, 60, 168, 183, 185, 191, 213, 233 technological determinism 86, 100
technological singularity 82–83, 135, 139, 141
Schabacher, Gabriele 29, 31, 69 n. 1, 71 n. 19, n. 62, 280–281, 291, 292 n. 4
72 n. 36, 73 n. 51 and 57, 74 n. 67, 102 n. 17, telecommunications 95, 117–118, 177–178
179, 188 n. 6, 267–268, 276 n. 25, 277 n. 50, television 10, 27, 95, 129, 183
292 n. 9 temporal regimes 11–12, 16–28, 57–60, 63,
science and technology studies 11–12, 15, 56, 69–70, 89, 95, 107, 109–122, 128, 181, 184, 187,
169, 281–282 191, 196, 199–200, 203, 284, 289
semiconductors 99, 222, 228, 236 n. 9. 237 n. colonial 133, 137–138, 177, 181, 183–184, 187
17, 238 n. 26; see also integrated circuits capitalist 10, 12, 18–19, 23–25, 28, 33 n. 3,
Serres, Michel 15–16, 35 n. 28, 280, 292 n. 5 36 n. 36, 37 n. 56, 133, 137, 187, 226, 234,
Shannon, Claude 99, 103 n. 31, 163 238 n. 35, 284–286; see also capitalism
Sharma, Sarah 10, 33 n. 5, 178–179, 183, 257 and clocks: clock time
Simondon, Gilbert 114, 123 n. 19 and 22, 146 calendar 18, 20–21; see also calendar
smartphone 10, 18, 31, 185, 205–206, 209–210, monastic 21, 90
221–223, 233–235, 245, 256–257, 297; see also temporality
cellphone and mobile devices etymology of 16
slowness 10, 49, 177–180, 184, 186 hardwired 11–12, 16–18, 21, 46, 57–58,
social media 33 n. 10, 80, 87 n. 11, 90, 102, 107, 79, 82, 89–90, 92, 99–100, 111, 180, 187,
123 n. 16, 127, 205, 207, 213, 257, 288, 297 222–223
software soft 31, 100, 177, 180, 187
and adaptability 16 embodied or lived 70 n. 14, 162, 179–180,
as applications 147, 151, 159, 210, 255–256, 182–183, 187
291 Thompson, E. P. 24, 37 n. 56, 133, 144, 238 n. 35
and bias 13, 33n time
care for 30–31, 56, 62, 64, 69 n. 2, 97, anonymous 24–25, 90–91
191–203 capitalist 18–19, 23–25
development 111–113, 147, 274 chronic (Benveniste) 20–21, 36 n. 44; see
and hardware 31, 62, 95, 108, 121, 191–203 also Benveniste, Émile
materiality of 31, 191–203 clock 19–23, 47, 96, 108, 114–115, 119–120
and time 123 n. 29, 133 191–203 computer 14, 117, 128–129, 133, 137–138, 143
software studies 192, 203 n. 12 Coordinated Universal Time (UTC) 94–
speed 95, 114–117
concept of 303, 305 cosmic 20–21
as efficiency 303–304 digital 11–12, 28, 110, 112, 120–121, 124, 181,
of light 91, 98, 127, 130, 303 232, 257, 270
beyond perception internet 95–97, 109, 117, 124 n. 33
processing 17, 26, 121, 131, 163, 168, 172 n. 15, historical 16, 77, 82, 132–133, 138
285 human perception of 26–27, 29–30,
theory of 178, 188 n. 5 42–45, 48, 50, 53 n. 2, 92, 95–97, 108,
speedup 9, 12, 26, 28, 32, 245; see also 112, 118, 127–130, 132, 137, 144–146, 162,
acceleration 250, 304
speed runs 150–151 labor-time 15, 19, 21, 23–25, 35 n. 27, 232,
Star, Susan Leigh 13, 27, 34 n. 22, 37 n. 60 and 235
72, 70 n. 7, 122, 293 n. 10 leisure 12, 232
Stiegler, Bernard 29, 32, 77, 80, 86 n. 6, 258, linear 50, 131–132
262, 295–296, 299, 301–302, 304, 306 n. 4–5, measurement of 19–20, 30, 36 n. 48, 108,
308 n. 25 112–120
Index 313
media 20, 25 time-criticality (Das Zeitkritische) 13, 28–29,

micro 9, 11–12, 18–19, 25–30, 94, 99–100, 34 n. 16, 41, 95–97, 99–100, 102, 108–109, 116,
128, 142 n. 74, 146, 149, 157, 159, 161–162, 128, 161–162, 167, 244
167–168, 170, 180, 223, 292 n. 7 timekeeping 18, 20, 22–23, 93, 95, 114, 223,
modern 90 225–227, 231, 233
network 112, 140 n. 36, 186 Transmission Control Protocol/Internet
nonhuman 15, 27, 89–90, 96, 187 Protocol (TCP/IP) 32, 97, 296–298, 300
order of 116, 118–121
reaction 15, 26, 158, 162–164, 168, 172 n. 14 video games 143–44, 150, 178, 229, 231, 234
and 16 Virilio, Paul 32 n. 2, 144, 178, 188 n. 5, 291, 294
real-time 10, 13, 28, 33 n. 10, 50–52, 91, 94, n. 33
97, 113, 123 n. 16, 129, 133, 138, 140 n. 30, von Baer, Karl Ernst 26, 37 n. 69, 41–44,
160, 178, 256, 263, 269, 284, 293 n. 14, 298, 46–47, 49
300–304, 308 n. 25 von Neumann, John 46, 290
religious 12
standardization of 11, 19, 23, 36 n. 50, 56, waiting 10, 58, 66–69, 73–74, 80, 178, 184–185,
114, 120, 181–182, 187, 224, 284; see also 244, 253
Coordinated Universal Time wearable 209, 234, 239 n. 43; see also clock:
time servers 95, 124 n. 33 wristwatch
time zones 95, 109, 111–112 Whitehead, Alfred North 119–120, 124 n.
universal 24, 94, 113–114, 116, 120, 187 34–35, 174 n. 37
time-axis manipulation 34 n. 14, 47–48, Wiener, Norbert 46–48, 54 n. 14, 100, 208, 218
50–51, 54 n. 17, 129 n. 8, 253, 261 n. 24 and 29, 271
In a crucial sense, all machines are time machines. The essays in
Media Infrastructures and the Politics of Digital Time develop the
central concept of hardwired temporalities to consider how technical
networks and infrastructural practices hardwire and rewire patterns
of time. Digital media introduce new temporal patterns in their
features of instant communication, synchronous collaboration,
intricate time management, and continually improved speed. They
construct temporal infrastructures that affect the rhythms of lived
experience and shape social relations and practices of cooperation.
Interdisciplinary in method and international in scope, the volume
draws together insights from media and communication studies,
cultural studies, and science and technology studies while staging
an important encounter between two distinct approaches to the
temporal patterning of media infrastructures, a North American
strain emphasizing the social and cultural experiences of lived time
and a European tradition, prominent especially in Germany, focusing
on technological time and time-critical processes.
AXEL VOLMAR is a Postdoctoral Fellow at the Collaborative Research

Center “Media of Cooperation” at the University of Siegen. He is
co-editor of Format Matters: Standards, Practices, and Politics in
Media Cultures and Interrogating Datafication: Towards a Praxeology
of Data.
KYLE STINE teaches Film and Media Studies at Johns Hopkins
University. His writings on cinema and technology have appeared in
Critical Inquiry, Discourse, Grey Room, and the Journal of Cinema
and Media Studies.
AUP.nl

電子書 Media

Uploaded by

Copyright:

Available Formats

電子書 Media

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

電子書 Media

Uploaded by

Copyright:

Available Formats

What are the main themes explored in the book series?

What are the main themes explored in the book series?

What are some of the approaches to temporal patterning discussed in the volume?

What are some of the approaches to temporal patterning discussed in the volume?

Media

Ams te rdam AXEL VOLMAR

Amsterdam University Press

Cover illustration: Pete Linforth (Pixabay), Pavlofox (Pixabay).

Cover design: Suzan Beijer

isbn 978 94 6372 742 6

Creative Commons License CC BY NC ND

All authors / Amsterdam University Press B.V., Amsterdam 2021

Infrastructures of Time: An Introduction to Hardwired Temporalities 9

Part I Media Philosophies of Time Patterning

1. The Suspension of Irreversibility: The Fundamental (and

2. Time and Technology: The Temporalities of Care 55

3. Problems of Temporality in the Digital Epoch 77

4. Suspending the “Time Domain”: Technological Tempor(e)alities

Part II Microtimes

5. Infrastructuring Leap Seconds: The Regime of Temporal

6. Life at the Femtosecond 125

7. Artificial Intelligence and the Temporality of Machine Images 143

8. Intervals of Intervention: Micro-Decisions and the Temporal

9. Grounded Speed and the Soft Temporality of Network

10. Unruly Bodies of Code in Time 191

11. Screwed: Anxiety and the Digital Ends of Anticipation 205

12. Beep: Listening to the Digital Watch 221

Part IV Futures

14. Ahead of Time: The Infrastructure of Amazon’s Anticipatory

15. Artificial Neural Networks, Postdigital Infrastructuresand the

16. Technics of Time: Values in Future Internet Development 295

This collection stems from two conversations started at McGill University in

Keywords: media theory, digital time, infrastructures, materiality,

All machines, whether mechanical, electronic, or symbolic, are in a crucial

emphasized this in terms of a cultural doctrine of accelerationism.2 However,

system or device is value neutral. Any divergence between the temporal

Bowker’s notion of “infrastructuring” as an active, ongoing process and

Technologies as Consolidated Temporalities

A guiding thread in this regard involves the processes by which technical

infrastructures. Scholars of the social construction of science and technology

Consider a late-model car. It is a disparate aggregate of scientific and tech-

Comprising parts of different vintage, technologies also operate on multiple

above and below: above, in the flexible software systems designed to

John Durham Peters elaborates these etymological connections in broad

in 1965, comparing the two generations of computers, wrote: “Another trend

Our project thus seeks to reassess material infrastructures as stabilized

Temporality, as seen throughout history but especially in an age of global

grid that locates personal experience within cosmic rhythms. 44 Out of

In his foundational work in the history of technology, Lewis Mumford (1934)

breathing of the lungs,” Mumford cites practices of agrarian subsistence: “The

In a similar way, modern technical systems and infrastructures should be

retirement should be managed. These struggles are largely crystallized in

chronoscopes, myographs, and photographic cameras, which operated

film, radio, and television, have constructed their own microtemporal

Layout of the Volume

media reveal “the fundamental relativity of each regularity of time.” Taking

2002), 151–162; Lauren Berlant, “The Commons: Infrastructures for Trou-

Museum of Vertebrate Zoology, 1907–39,” Social Studies of Science 19, no. 3

About the Authors

Infrastructures of Time: An Introduction to Hardwired Temporalities 9

2. Time and Technology: The Temporalities of Care 55

3. Problems of Temporality in the Digital Epoch 77

6. Life at the Femtosecond 125

7. Artificial Intelligence and the Temporality of Machine Images 143

10. Unruly Bodies of Code in Time 191

11. Screwed: Anxiety and the Digital Ends of Anticipation 205

12. Beep: Listening to the Digital Watch 221

16. Technics of Time: Values in Future Internet Development 295