Stacked Networks Approach

International Journal of Actor-Network Theory and Technological Innovation, 4(1), 13-23, January-March 2012 13
Networks, Agents and Models:

Objections and Explorations
Fabian Muniesa, Mines ParisTech, France
Ivan Tchalakov, Plovdiv University, Bulgaria
ABSTRACT
Actor-Network Theory proves particularly inspiring in reconsidering the tenets of quantitative research
and computational methods in the social sciences. However, translating insights from this perspective into
operational models is problematic. The paper examines, in the form of a dialogue, critical problems of the
computational modelling of network topologies considered from the point of view of Actor-Network Theory.
In particular, the paper discusses the impetus of simulation and the inappropriateness of the distinction
between agents and links.
Keywords:
Actor-Network Theory, Agency, Modelling, Networks, Quantification, Simulation, Stochastic

Automata, Topology
INTRODUCTION
What a network is, what a model is or what a
quantity is are all questions worthwhile being
asked from a methodological viewpoint. At the
same time, all these things (networks, models,
quantities) are all particularly relevant in the
characterisation of contemporary cultures and
deserve also to be studied as such. Work in
Actor-Network Theory has certainly already
provided good occasions to engage into discussions on network topology, on modelling and
on quantification as both vehicles for scientific
investigation and as constitutive features of the
social world (e.g., Callon, 2006; Latour, 2010).
This paper elaborates on the discussion held at
a workshop hosted at the CSI (Centre de Soci-
ologie de lInnovation), in the Ecole des Mines

de Paris, on October 30th, 2008. The workshop
was part of the ATACD project (A Topological
Approach to Cultural Dynamics) and consisted
of an informal discussion on methods in network analysis, stochastic automata modelling
and Actor-Network Theory.1 The point of the
discussion is: how can Actor-Network Theory
still contribute to the renewal of network topology? The question is rendered here in the form
of a dialogue between two human agents, A and
B, freely inspired by the transcripts of the discussion: A is working on how to use stochastic
automata to model heterogeneous networks,
while B is mainly formulating objections. The
discussion starts with an argument about the
purpose of models and simulations.
DOI: 10.4018/jantti.2012010102
Copyright 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
14 International Journal of Actor-Network Theory and Technological Innovation, 4(1), 13-23, January-March 2012
OBJECTING TO SIMULATION
A: You dont seem very happy with the whole
idea of using mathematical models and
computer simulations to understand the
dynamics of networks. Cant you grant
them some empirical value? What do you
think a simulation is, anyway?
B: I can think of two major directions for
understanding the value of a simulation.
One would be a rather sociological one, a
direction in which you start with a dataset,
or a database. You dont care about things
such as whether or not you are confronting
a stable state what you primarily need
is to describe the dataset itself. You can
then imagine a computational model that
is useful or interesting in order to describe
the dataset, to find an analytical vocabulary
that allows making sense of the assemblage
of data. The other way of understanding
simulation would start with getting rid
of data and just playing the game of the
computational device. You might hook the
device onto a dummy database, a database
without data a kind of artificial situation
that does not correspond (does not need
to correspond and could actually hardly
correspond) to any empirical situation.
The second option is the fashionable tendency in computation models, in financial
computational modelling particularly. Just
build an artificial world and play the game
within describing, at best, the behaviour
of a world that you have provoked yourself.
A: Come on. That is a bit of an exaggeration. Model-based prediction of empirical
behaviour is a real possibility. What about
predictions based on the simulation of
real empirical data, such as predictions of
energy consumption which are based on
established trends and which allow for a
pretty accurate prospective modelling of
future behaviour?
B: Of course, prediction of behaviour. But we
are talking here rather about different cul-
tures of using models. Prediction is a very

fashionable issue indeed. What makes a
prediction possible is the certitude or belief
that you understand the initial state (and
those preceding it). This is not always the
case. So maybe understanding is more important. Understanding is definitely more
in the line of Actor-Network Theory not
to predict anything, but just to describe,
to understand what happens in a dataset.
I agree that as soon as you enter into the
business of computational models, there is
an urge to go into prediction mode. So
data from the past may be interesting, but
what is more important is the dynamics of
the model itself.
A: When you are trying to build a model,
there is a lot of trade, a lot of surplus
coming out of it, especially when you are
modelling small, local phenomena. First,
you are using not only certain variable
distributions to define the optimal states
and so forth, but there are also pretty good
statistical tools that enable you to establish
the main interdependencies between some
variables, the main types of interaction.
Then you can base your assumptions on
this specific model even if it refers to a
specific moment only it does not matter
so much whether this is a moment of crisis
or a stable situation. The logic is simple
enough: in the economy, for instance, if a
given company is on the market for long
enough, it necessarily accumulates some
experience that is reflected in the specific
values of the variables describing it. It
might be on the verge of collapse, like
Lehman Brothers, but you study it a few
months or weeks before it collapses. Yet,
even in this case your model will be viable
enough, i.e., more viable than those based
on pure computation, pure combinatory
logic. Second, there is no need to model
events in their full complexity. The idea
of stacked networks is very useful here,
right?
STAKED NETWORKS
B: Staked networks?
A: These are several layers of networks of
different kinds that intermingle together
through small ego-networks that cut across
them all.
B: Ok, I see. That is your idea of heterogeneous
networks as stacked networks. In the case
of, say, a transitional economy, you have
several kinds of networks, such as a network
of corporate interlocking directorates, and
then a network of capitalistic ownership,
and then a network of geographical proximity, and so forth, and then you have nodes
that allow you to make sense of how all
these networks are stacked together. All
right. But then you want to model that
(Figure 1).
A: One big challenge here is to model socalled nonhuman agents: technologies,
artefacts, money, and so forth. The question
is how to model their role, their place, their
behaviour, how to consider them as actors
described within the network. If one knows
how technology behaves and put some efforts to collect the relevant data imagine
for instance a study on competing telecom
protocols (X25, TCP/IP, OSI, etc.) that
would care for how each protocol functions
these nonhuman agents could perfectly be
modelled by a corresponding automata and
added to the compound of diverse human,
individual or corporate agents. Should we
use the same stochastic automaton, slightly
modified, that we used in order to model
human agents? Should we use a separate
automaton? It all depends on the identity
of the nonhuman agent.
B: But what do you mean by the identity of
a nonhuman agent?
A: Im not sure its about the singular identity
of one actor. I mean rather the different
identities of one and the same nonhuman
actor in each of the stacked networks. This
is a big challenge different faces of a
nonhuman agency that acts simultaneously
in different networks? Take money, for
instance, Viviana Zelizers money (e.g.,

Zelizer, 2001). The bankers money, the
farmers money, the farmers wifes money:
all partaking of different networks, but
connected in the ego-network of money.
So, you see, there are lots of challenges
coming from this modelling.
MODEL AS DEMONSTRATION
B: Yes, well, I see the empirical point. But we
need to clarify the purpose of modelling.
A: The most important thing is related to
demonstration. Take Claude Rosentals research on demonstrations used by software
developers (Rosental, 2005, 2008). We are
in a similar situation. If we manage to make
a piece of software sophisticated enough to
model the studied field, and then to model
other fields, you need at least two things.
First, you need data, which are specific for
the given field, allowing an outline of the
stable states, the distribution of variables,
interdependencies, and so forth. Then,
we need to introduce the corresponding
scripts in the model, to change the inputs
and modify the internal properties of the
automata. Then you could get the universality of the tool, not of the model but of
the tool. This is an ambitious goal that is
worth setting If we eventually manage
to build a compound of stacked networks
interconnected via ego-networks, and then
we build the model for this particular field,
and then we model another field, all major
questions about the configuring of stochastic automata will be solved. We will just
need to change the software a little bit. But
its skeleton, its basic properties would be
already available as result of the modelling of the initial field. We get new data, we
change some parameters in the construction
and we are ready to emulate the new field.
Inspiring, isnt it? Its about achieving the
universality of the tool. This is feasible and,
I believe, not very complicated: layered,
stacked networks, taken simultaneously
Figure 1. Different identities of nodes in stacked networks
not the Parsonian way, not considering

social roles, not as attributes of the agent,
but with agents taken from the point of
view of Actor-Network Theory.
B: The universality of a tool is a very inspiring thing, indeed. I can understand the
objectives of mathematical modelling, this
notion of stacked networks and the aim of
constructing models, which have different
types of interactions, different patterns of
networks, but I think this is very complicated type of job you are dealing with. And
thinking that it will be possible to use this
model to describe the dynamics of real situations seems to me a bit too optimistic. If
you just consider mathematical questions,
it is probably very hard to succeed. If your
long-term objective is to use these models
to describe dynamics of whatever fields
you want to study, it looks much more
difficult. Im all for a far more modest
use of network analysis: using networks
to describe how some configurations could
be born out of the roles played by some
actors, in a situation that could be thought
of as a network. But thats a rather limited
perspective, of course, and with barely
any modelling. On the one hand, there is

the concern of describing, making sense
of some data, which is very difficult task.
On the other hand there are these computational tools, mathematical models that can
mimic this kind of behaviour, in almost a
game-theoretical way. These are two different directions. But maybe one interesting
compromise is to use simulations not as
research devices on their own, to produce
some data, but as a way to construct a vocabulary for understanding the database.
Im interested in what you said. You do not
have your stochastic automata yet, but the
very work you did in order to reflect about
the construction of this simulation has already produce a set of useful vocabularies
aiming at describing what is happening
in an heterogeneous dataset. You dont
have the stochastic automaton, but you
do have the personage, you do have now
the stochastic automaton as a conceptual
character that allows you to say: well, lets
consider what the actor is, lets consider
it as a kind of an ego-network trapped in
between a number of stacked networks, and
so on. And that is a key capacity in order
to describe a heterogeneous database! It is

fantastic to pursue the goal of building such
a computational device, but it is interesting to grasp what you have already done
for this computational venture in order to
describe, to make sense, to find the right
vocabulary to describe what is happening
in data.
HEURISTIC AUTOMATION
A: Well, we had to come up with these sorts
of concepts. But our key ambition is to
accomplish the automaton, really. We
want to spot its internal simplicities and its
abilities for generating interesting network
behaviour. An accomplished automaton
could be used to build many kinds of
research. Interesting research directions
would include applying the automaton to
simpler datasets. You do not need to have
fifteen variables Each sub-network could
be defined by a single variable, describing the basic (critical) relationship in the
network. It is difficult, of course, to decide
what this basic relationship is, but I think it
is worth thinking in that direction if not
one, we could combine two variables into a
single one, while the other variables could
be built-in as properties of the automaton
itself. So when such and such a script arrives
at the input, the automaton has to react in
a specific manner. The crucial point is to
clarify what you want, to clarify the initial
design that precedes building the automata
itself.
B: But you have first to define properties that
characterise your dataset. Not to try to
model your dataset, because your dataset
is too complex!
A: There is little sense in building purely
abstract models; we need to have datasets
in mind.
B: To build a machine that mimics the real
world, this is a hard task. But you can
have machines that are simple, and then
you say: well, it does not mimic a real
world dataset, this is a heuristic device,

an intellectual device actually. The tradition of complex systems in general starts
from some sort of mock complexity. You
have, for instance, mathematicians trying
to model financial markets with some
complexity in it, and then you notice that
what they are actually talking about is a
set of idiotic robots meeting in the CPU
of their computers. Nothing that would
look remotely like traders working in the
trading room of an investment bank! Its
totally all right to have these machines
working as intellectual devices in order to
foster some analytical categories, but not
as a depiction of practical reality.
A: One example: we have a hypothesis about
different speeds (or intensity) of interaction
within the sub-networks and between the
sub-networks. This is very important question, empirically. If you consider a case of
transitional economies, it has political implications. For example, you can consider
two types of speeds, two types of devices
with different speeds and try to study the
effects produced on the dynamics of each
sub-network. Using these results you could
have some insights as to how the real situation is configured.
B: Yes, I think this would be a more modest
approach, where the modelling should
be used more as a tool for studying very
limited sets of questions, even if models
are very complicated. Because very often
simple questions are very complicated. But
still I would be very reluctant to consider
a model as a straight simulation of what is
in your database. The model simulates, at
best, a research question. One example of
such a research question would be about
the different types of relations between
sub-networks, and the speed of interaction
within them. Another would be about the
shape of ego-networks or about the impact
of the introduction of a new actor into a
sub-network of firms.
A: So, for example, if we face the specific
problem of how a mediator transforms
into an actor, imagining a situation where

the switch that propagates through the
network transforms itself into an actor,
then we need to change the properties of
the automaton. So it seems that while the
script (the switch) is just travelling
through the nodes, its properties change into
the property of a specific automaton and
it becomes a different agent. But then you
need specific data Anyway, I get the idea
of using the model as a tool for understanding some specific, peculiar situations. But
in any case this will not be a modelling of
rules of social interactions provided by
some sociological theory. Rules need to be
taken from some past situation, from data
describing some real phenomena.
B: If we tackle sociological theory, then maybe
we should talk a little about Actor-Network
Theory and the notion of agency.
AGENCY IN NETWORKS
A: Yes, the agency of nonhuman agents?
B: Well, I am not sure that the point of ActorNetwork Theory is to say that there are
some human agents, which have agency,
and that there are also some nonhuman
agents, which also have agency. Im not sure
that the problem is to develop some kind
of network analysis where every kind of
actor could be granted with agency. Is that
the most relevant point of Actor-Network
Theory? I would rather say that the point
is that no one and nothing have agency at
all either human agents or otherwise
unless those agents are networks. So in
some sense the challenge of Actor-Network
Theory is not to discuss the notion of agent,
but rather to reassess the notion of network.
A: Well, its true that the whole point is to
replace both the notion of actor and of
network with the notion of actor-network.
B: An interestingly misleading notion, by the

way. In French, an acteur-rseau refers
to an-actor-which-is-a-network but the
English version, actor-network, rather
suggests the idea of a-network-whichacts.
A: Was this difference in meaning created on
purpose?
B: Who knows? Perhaps! Anyway, the
problem with the vocabulary of network
analysis is that it spontaneously grants
agency to the agents, to the dots in the
graph, and not the edges between the dots.
This is not compatible with a theoretically
sound Actor-Network Theory stance! So
the idea would be to establish some theoretical vocabulary that would tackle this
issue, because as soon as we are heading
to the implementation of these models we
end up with a discussion about several kinds
of agents: we have human agents, persons,
then firms, then tools, etc. I think this is
a rather impoverished version, compared
with the potentials of Actor-Network
Theory, which is really to criticize completely this idea of agency as a property
of agents.
A: But maybe the idea of stacked networks
in which the linking agents are egonetworks can work as a possible solution,
right? Because the identities (or rather
hypostases)2 they obtain in each layer
are defined both by the interactions in
the corresponding sub-network, and the
interactions between these identities in the
ego-network a kind of looping, in which
one starts from nowhere or rather from the
entire space of ongoing interactions. I am
tempted to call that a topological space.
B: Yes, but in that case what you term a
sub-network shouldnt be taken in the
sociological sense of a field: the field
of politics, the field of business, etc. I
think one of the major proposals of ActorNetwork Theory is precisely to jeopardize
the notion of field and to say that there is

no such thing as a field in the sense of
Bourdieu, or a system in a Luhmannian
sense. Rather, it claims that everything is
completely heteronomous, that there is no
such a thing as an autonomous field.
DATABASES
A: Yes, right. Here we do not consider autonomous fields. But we consider, in a sense,
autonomous sets defined by the propagation
of specific mediators (or intermediaries, or
both). I mean datasets which correspond to
a heteronomous empirical reality but that
look autonomous because they correspond
to one kind of data source. The fields are
thus autonomous only as datasets because
if you are collecting information, there are
business databases, legal databases, scientific databases and so forth. I would like to
come back to the problem of a mediator
transforming itself into an actor: from the
point of view of ongoing interactions in the
topological space, databases are just mediators that stabilise and make predictable
these interactions. But all of a sudden they
become actors speaking or portraying
this space in a specific manner.
B: But that is actually the only reason why
you have such a thing as different subnetworks: because you construct them on
the basis of sources, which are instituted
as different, in a sense. A field is just the
performative result of the institutional
practice of a data provider.
A: Well, the idea that we shouldnt trust the
boundaries between fields is particularly
helpful in the study of transitional economies. The notion of recombination is
particularly illuminating here (Stark, 1996).
The existence of different sub-networks
should not derive from sociological assumptions but primarily from the properties of data sources. Consider the reality
of corporate mergers and acquisitions: it
is a profusion of political, kinship, social,
scientific networks. But if you want some

empirical databases, then you will need to
pick from several dedicated data sources.
You may say that they correspond to
stacked networks, but it is perhaps more
realistic to say that they are completely
mixed up. Any attempt at developing an
evolutionary approach should care about
the combinatorial evolution of networks
and about the evolution of data sources as
such.
OPTIMALITY AND BEHAVIOUR

B: Talking about evolution: in stochastic
modelling there is also some sort of optimality unconscious that is worthwhile
historicising a bit.
A: You mean the pervasiveness of the point
of view of game theory?
B: Yes, the idea la von Neumann and Morgenstern according to which the point
of any kind of game would be gain. The
leitmotiv of the attainment of optimality:
if you do not play the optimality game you
are out, and this is the only possible game.
I object. Look at real life. Fortunately,
there are thousands of situations in which
people who do not like optimality, or who
do not like games altogether, are still in.
I would even say that there are thousands
of social games in which the rules are
set in a non-economic, although realistic
fashion. So I think model builders should
be trained to imagine worlds that are not
into the minimax business and be open to
other ways of theorizing behaviour other
than the game-theoretical, neoclassical
way. Come on, they always think in terms
of survival. The topic of survival of
so-called abnormal behaviour is always
a surprise for neoclassical game theory.
But in reality it is a rule. We are all the
living proofs of the survival of abnormal
behaviour.
A: I agree. The principles of game theory
are built into the properties of stochastic
automata implied by most existing simulations, in the very way a modelled system
arrives at stability. I think the real challenge is to envision a sort of stochastic
automata in which you do not make many
assumptions about the rules of behaviour
of the agents, and let them fold into the
agency of the networks. Well, for the sake
of computation you incidentally need to
start with the design of the properties of
the agents of the specific sub-network. One
automaton can simulate the behaviour of all
other agents. In fact there is one automaton
that changes its state after each iteration.
To model another sub-network you could
use the same automaton just changing
some of its properties according to the
corresponding dataset and so on, depending on the number of layers you have in
mind. Then you begin thinking about the
ego-network, and about how the very fact
of being together modifies the functioning of automata, which now become some
kind of modules of the larger automata
which is the ego-network. Yes, indeed its
the network that has agency, not the separate automata! But we have to care for the
stability of the sub-networks, which should
not start to fall apart. And this means that
we have to build into the initial automata
some additional properties, stemming not
from the functioning of the sub-networks,
but from the necessity of their mutual coexistence. And this obviously will cause
changes in the properties of the initially
designed automata, in each of the initial
sub-networks. This is a really complex
task.
FLAT NETWORKS
B: It is surely easier to study flat, heterogeneous networks than to model these stacked
networks. I see you want to try But do
not try to use data in the first place, then!
I think the first step is to demonstrate, to
show that this is a good idea to use the
A:
B:
A:
B:
notion of stacked networks in order to

understand the dynamics of networks I
do not understand how you will do that. Is
this another way of saying that there are
many relations between different categories
of entities?
Well, a very interesting and very strong
hypothesis is the one concerning a difference in speeds of interactions between
different sub-networks and in the egonetwork.
But it could be the same as following a
single network: some relations are stable
within this network, while there are others
that are rapidly evolving. That would be the
same assumption, but you hide this assumption behind this complicated machinery
of different networks evolving simultaneously. You could make exactly the same
assumption with a single network, saying
for example that some relations, some subregions of this network are stable or not,
taken into account the overall dynamics.
This is exactly the same assumption from
a formal point of view. Even from a mathematical point of view one could perfectly
as well say that one actor in a single flat
network has different roles or identities
depending on the type of relation in which
it is engaged. It has as many identities as
sub-networks in which it is engaged.
Talking about some segments of the network and about some sub-networks: it
is almost the same.
Yes, because of the assumption that fields
can be translated into network descriptions. If you have agents here and here
(Figure 2), well, this is a flat representation of a single network, which is exactly
the counterpart of the notion of stacked
networks. For example, when you start to
analyse the dynamics of this network, you
can make the assumption that the relations
between these different networks are very
weak. So this is just a particular case of
network shapes, and you do not need this
notion of stacked networks. Try to test
the hypothesis that these networks are
Figure 2. A flat representation of stacked networks
very specific configurations, in which the

separate sub-networks are linked through
one single agent. You make the assumption that the speed of exchanges between
these different sub-networks, or intensity
of exchange, is lower than the intensity of
the exchange within the sub-networks.
A: You mean that the stacked networks are a
limit case of one of the types of flat networks.
B: This is definitely more understandable
from the perspective of Actor-Network
Theory. It is exactly a description of this
network structure (Figure 2), which is a
very specific case.
A: You mean it is better not to oppose flat
networks and stacked networks?
B: Yes, because otherwise you are implying
a different type of topology, I think. So
it seems to me it will be easier, from a
mathematical point of view, to weaken the
very strong assumptions of stackedness. It

would be easier to introduce, for example,
the possibility of some relations between
sub-networks which are not going through
ego-networks.
A: However, the idea of stacked networks
seems to me closer to the deeper assumptions of Actor-Network Theory we
mentioned before. When you take the
loosely related star-shaped regions of a
flat network and then twist them onto
each other, you arrive not just at a few
nodes that mediate, but at the possibility
for each node to be simultaneously immersed in several layers, and hence to be
considered as an ego-network.3 Otherwise
its true that, in transitional economies for
instance, such configurations of stacked
networks or of flat networks where the
separate sub-networks are linked through
one single or a few agents are exceptions
rather than a rule. You also should keep

in mind that after almost two decades of
transition in post-socialist European economies relationships are gradually settled in
a more ordered way and one could admit
that a kind of re-autonomisation did occur.
B: Yes. But this is an empirical matter. Not a
necessity. We should avoid the obsession
of considering that every situation looks
for equilibrium, or at least looks for stability. This is something that is completely
discussible! I insist: abnormal situations
are not monsters, but something that really
happens very often. That is why people are
making (or loosing) money, anyway: out
of disequilibrium. In a sense, there are no
gains, no profits, without disequilibrium.
More generally, disorder is something
quite productive. It is possible to tackle
disorder or dis-optimality as a situation
that is stable. It is funny that quants in
finance are now more into Mandelbrot
than into plain neoclassical economics
(e.g., Mandelbrot & Hudson, 2004).
CONCLUSION
A: Maybe we should refrain from assessing
stability and instability and limit ourselves
to descriptions of the state of networks, at
best of kaleidoscopically changing states.
And to attempting to trace and visualise the
continuity in their evolution as networks.
From a mathematical topological point of
view, the continuity between a circle and
a square is not a big deal.
B: In a sense, maybe Actor-Network Theory is
more topological than we thought: I mean
in the sense of mathematical topology, not
of network topology. It is precisely a point
of view for which there are no dimensions,
no coordinated space. Everything is selfcontained in the networks, including their
metrics.
REFERENCES
Callon, M. (2006). Can methods for analysing large
numbers organize a productive dialogue with the
actors they study? European Management Review,
3(1), 716. doi:10.1057/palgrave.emr.1500051
Kapriev, G., & Tchalakov, I. (2009). Actor-Network
Theory and Byzantine interpretation of Aristotles
theory of action: three points of possible dialogue.
In Bamm, A., Getzinger, G., & Weiser, B. (Eds.),
Yearbook 2008 of the Institute for Advanced Studies
on Science, Technologies and Society (pp. 207238).
Munich, Germany: Profil.
Latour, B. (2010). Tardes idea of quantification. In
Candea, M. (Ed.), The social after Gabriel Tarde:
Debates and assessments (pp. 145162). London,
UK: Routledge.
Mandelbrot, B., & Hudson, R. L. (2004). The (mis)
behavior of markets: a fractal view of risk, ruin, and
reward. New York, NY: Basic Books.
Rosental, C. (2005). Making science and technology
results public: a sociology of demos. In Latour, B.,
& Weibel, P. (Eds.), Making things public: atmospheres of democracy (pp. 346349). Cambridge,
MA: MIT Press.
Rosental, C. (2008). Weaving self-evidence: a sociology of logic. Princeton, NJ: Princeton University
Press.
Stark, D. (1996). Recombinant property in East European capitalism. American Journal of Sociology,
101(4), 9931027. doi:10.1086/230786
Tchalakov, I., & Burton, P. (Eds.). (2001). Project
TACTICS. Sofia, Bulgaria: LIK.
Tchalakov, I., & Kapriev, G. (2005). The limits of
causal action: Actor-Network Theory notion of translation and Aristotles notion of action. In Bamm, A.,
Getzinger, G., & Weiser, B. (Eds.), Yearbook 2005
of the Institute for Advanced Studies on Science,
Technologies and Society (pp. 389433). Munich,
Germany: Profil.
Zelizer, V. A. (2001). The social meanings of money:
pin money, paychecks, poor relief, and other currencies. Princeton, NJ: Princeton University Press.
ENDNOTES
1
Participants to the workshop were Fabian

Muniesa, Ivan Tchalakov, Michel Callon, Liliana Doganova, Donka Keskinova and Petar
Kopanov. ATACD (A Topological Approach
to Cultural Dynamics) is a NEST Pathfinder
project supported by funding under the Sixth
Framework Programme of the European
Union (contract no. 043415). The workshop
also built upon earlier collaborations, focused on the TACTICS Project (Telematics
and Communications Technology Industrial
Comparative Study), an INCO FP4 RTD
Cooperation Project, developed from 1998
to 2000. For an overview, see Tchalakov and
Burton (2001). Support for these earlier collaborations was provided by the ACE/Phare
Programme in 1994 and the NATO Science

Fellowships Programme (France) in 2001. We
thank Celia Lury for her help, comments and
encouragement.
The concept of hypostasis was coined by
Byzantine orthodox philosophers to solve the
problem of relationships between the essential
properties of an entity and its existence. For
a possible reinterpretation of the Latin static
notions of being into more dynamic ones, see
Tchalakov and Kapriev (2005) and Kapriev
and Tchalakov (2009).
Maybe the very term of ego is misleading
here, since there is no center in the egonetwork it is rather heterarchical and not
hierarchical, like the dialectical notion of
self-develop by George H. Mead (as composed
by I and Me, none of which is superior).
Fabian Muniesa is a senior researcher at the Centre de Sociologie de lInnovation (CSI) in

the Ecole des Mines de Paris (now Mines ParisTech), France. His research contributions are
mostly located in the fields of science and technology studies and economic sociology. His
work primarily aims at developing a pragmatist, materialist approach to the study of calculation, valuation and organization. His past and current research topics include the automation
of financial markets, the practice of economics, the implementation of performance indicators,
and the pedagogy of business.
Ivan Tchalakov is associate professor at University of Plovdiv, Department of Institutional and
Applied Sociology and senior research fellow at Technology Studies Group at Institute for Studying
Societies and Knowledge, Bulgarian Academy of Sciences. He is working in the field of Science
and Technology Studies focusing on ethnographic studies of laboratory and engineering practice,
studies of innovations and entrepreneurship, as well as on science and technology development
in South-Eastern Europe after WW II. He is interested in exploration of new qualitative and
quantitative methods in studying and modelling the (actor) networks dynamics.

Stacked Networks Approach

Uploaded by

Copyright:

Available Formats

Stacked Networks Approach

Uploaded by

Document Information

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

Stacked Networks Approach

Uploaded by

Copyright:

Available Formats

International Journal of Actor-Network Theory and Technological Innovation, 4(1), 13-23, January-March 2012 13

Networks, Agents and Models:

Actor-Network Theory, Agency, Modelling, Networks, Quantification, Simulation, Stochastic

ologie de lInnovation), in the Ecole des Mines

tures of using models. Prediction is a very

instance, Viviana Zelizers money (e.g.,

Figure 1. Different identities of nodes in stacked networks

not the Parsonian way, not considering

any modelling. On the one hand, there is

to describe a heterogeneous database! It is

world dataset, this is a heuristic device,

into an actor, imagining a situation where

B: An interestingly misleading notion, by the

the notion of field and to say that there is

scientific networks. But if you want some

OPTIMALITY AND BEHAVIOUR

notion of stacked networks in order to

Figure 2. A flat representation of stacked networks

very specific configurations, in which the

very strong assumptions of stackedness. It

rather than a rule. You also should keep

Participants to the workshop were Fabian

Programme in 1994 and the NATO Science

Fabian Muniesa is a senior researcher at the Centre de Sociologie de lInnovation (CSI) in

You might also like