The sciences of science communication

Baruch Fischhoff1
Department of Engineering and Public Policy, and Department of Social and Decision Sciences, Carnegie Mellon University, Pittsburgh, PA 15213-3890

Edited by Dietram A. Scheufele, University of Wisconsin, Madison, WI, and accepted by the Editorial Board May 23, 2013 (received for review September
19, 2012)

The May 2012 Sackler Colloquium on “The Science of Science Com- The same mixed emotions may accompany phenomena that
munication” brought together scientists with research to communi- arise from the confluence of technology and society. How is it
cate and scientists whose research could facilitate that communication. that human actions can both discover antibiotics and encourage
The latter include decision scientists who can identify the scientific microbial resistance to them, create “smart” electricity grids and
results that an audience needs to know, from among all of the threaten them with cyberattacks, and produce a food supply
scientific results that it would be nice to know; behavioral scientists whose sheer abundance undermines the health of some people,
who can design ways to convey those results and then evaluate the while others still go hungry?
success of those attempts; and social scientists who can create the Without some grasp of the relevant science, it is hard to make
channels needed for trustworthy communications. This overview informed decisions about these issues. Those include private
offers an introduction to these communication sciences and their decisions, such as whether to choose fuel-efficient vehicles, robot-
roles in science-based communication programs. guided surgery, or dairy products labeled as “produced by animals
not treated with bovine growth hormone—not known to cause
decision making | environment | judgment | risk | uncertainty health effects.” And they include public decisions, such as whether
to support politicians who favor fuel efficiency, lobby for disclosing

W e need science for many things, the most profound of which

is supporting our sense of wonder about the world around
us. Everyone has had moments in which science has solved or
the risks of medical devices, or vote for referenda limiting agri-
cultural biotechnology.
Effective science communications inform people about the
created a mystery. For me, one of those moments was as an un- benefits, risks, and other costs of their decisions, thereby allowing
dergraduate in Introductory Geology at Wayne State University, them to make sound choices. However, even the most effective
when Prof. (Joseph) Mozola explained the origins of the Harris- communication cannot guarantee that people will agree about
burg surface, which gives central Pennsylvania its distinctive lattice what those choices should be. Reasonable people may disagree
pattern of rivers and gaps. At that moment, I realized that all about whether expecting X days less convalescence time is worth
landforms, even the flats of Detroit, were created by processes that expecting Y% more complications, whether the benefits of organic
science could decode. Another such moment was reading Joanna produce justify Z% higher prices, or whether their community has
Downloaded from https://www.pnas.org by 187.189.101.55 on March 21, 2023 from IP address 187.189.101.55.

Burger’s The Parrot Who Owns Me (1) and getting some insight been offered fair compensation for having a hazardous facility
into how our white-crested cockatoo, Big Bird, appeared to have sited in its midst. Reasonable people may also disagree about what
such human emotions, despite our species having diverged so long time frame to consider when making a decision (the next year? the
ago (even if he was born on Pittsburgh’s South Side). A third was next decade? the next century?), about whose welfare matters
following a proposal from our son Ilya, an evolutionary biologist, (their own? their family’s? their country’s? all humanity’s?), or
to take out our lawn and let our yard go feral with native Penn- about whether to consider the decision-making process as well as
sylvania flora—after which he taught us how to observe the sea- its outcomes (are they being treated respectfully? are they ceding
sons. Now, I can wonder why the first grackles arrived 10 days early future rights?).
this year, even though the main flock took over the yard on March The goal of science communication is not agreement, but fewer,
7th as usual. Was it a sign of climate change—or of my own im- better disagreements. If that communication affords people a
proved powers of observation? shared understanding of the facts, then they can focus on value
Science provides a sense of wonder not just from revealing the issues, such as how much weight to give to nature, the poor, future
world, but also from showing that the world can be revealed. For generations, or procedural issues in a specific decision. To realize
that reason, I tell my advisees, in our Decision Science major, that that potential, however, people need a venue for discussing value
they should join a research laboratory, any research laboratory, issues. Otherwise, those issues will percolate into discussions of
just to see how science is done. They may discover that they love science (2). For example, if health effects have legal standing in
the life, including the camaraderie of the graduate students and a facility-siting decision, but compensation does not, then local
postdocs who are the engines of much science. Or, they may find residents may challenge health effect studies, when their real
the constant fussing over details to be “mind-numbingly dull” (to concern is money. Science communication cannot succeed when
use the phrase that Calvin, of Calvin and Hobbes, applied to ar- people feel that attacking its message is the only way to get redress
chaeology) and then choose to get their science from places like for other concerns (3).
Nova, Frontline, the New York Times, and Wired. Either way, they Because science communication seeks to inform decision
will have seen the pursuit of uncertainty that distinguishes science making, it must begin by listening to its audience, to identify the
from other forms of knowing and the scientific turn of mind: trying decisions that its members face—and, therefore, the information
to get to the bottom of things, knowing that one never will.
Decisions That Require Understanding Science
This paper results from the Arthur M. Sackler Colloquium of the National Academy of Scien-
Although people can choose not to do science, they cannot choose ces, “The Science of Science Communication,” held May 21–22, 2012, at the National Academy
to ignore it. The products of science permeate their lives. Nuclear of Sciences in Washington, DC. The complete program and audio files of most presentations
power, genetically modified organisms, nanotechnology, geo- are available on the NAS Web site at www.nasonline.org/science-communication.
engineering, and xenotransplantation are just a few of today’s Author contributions: B.F. wrote the paper.
realities that would have been impossible, even inconceivable, The author declares no conflict of interest.
without scientific advances. Their very existence should be a This article is a PNAS Direct Submission. D.A.S. is a guest editor invited by the Editorial
source of wonder, even for people in whom they evoke a sense Board.
of terror (“we can do that?!”). 1
E-mail: baruch@cmu.edu.

www.pnas.org/cgi/doi/10.1073/pnas.1213273110 PNAS | August 20, 2013 | vol. 110 | suppl. 3 | 14033–14039

 that they need. In contrast, science education begins by listening decision may be irrelevant to another. Decision science formalizes
to scientists and learning the facts that they wish to convey (4). the relevance question in value-of-information analysis, which
Science education provides the foundation for science commu- asks, “When deciding what to do, how much difference would it
nication. The more that people know about a science (e.g., make to learn X, Y, or Z?” Messages should begin with the most
physics), the easier it will be to explain the facts that matter in valuable fact and then proceed as long as the benefits of learning
specific decisions (e.g., energy policy). The more that people more outweigh its costs—or recipients reach their absorptive
know about the scientific process, per se, the easier it will be for capacity (13). Although one can formalize such analyses (7, 11,
science communications to explain the uncertainties and con- 12), just asking the question should reduce the risk of assuming
troversies that science inevitably produces. that the facts that matter to scientists also matter to their audi-
Adapting terms coined by Baddeley (5), the sciences of science ences (2, 6, 8, 9, 11).
communication include both applied basic science (seeing how well For example, in the context of medical decisions (6, 7, 11), such
existing theories address practical problems) and basic applied analyses ask when it is enough to describe a procedure’s typical
science (pursuing new issues arising from those applications). In effects and when decision makers need to know the distribution of
Baddeley’s view, these pursuits are critical to the progress of basic outcomes (e.g., “most patients get mild relief; for some, though, it
science, by establishing the boundaries of its theories and identi- changes their lives”). Such analyses can also ask when uncertain-
fying future challenges. ties matter (e.g., “because this is a new drug, it will take time for
Science communications that fulfill this mission must perform problems to emerge”) and how precise to be (e.g., will patients
four interrelated tasks: make similar choices if told that “most people get good results”
and “researchers are 90% certain that between 30% and 60% of
Task 1: Identify the science most relevant to the decisions that patients who complete the treatment will be disease-free a year
people face. later”) (12, 14–16). Thus, such analyses, like other aspects of the
Task 2: Determine what people already know. communication process, must consider the range of recipients’
circumstances, paying particular attention to vulnerable pop-
Task 3: Design communications to fill the critical gaps (between ulations, both as an ethical duty and because they are most likely to
what people know and need to know). have unique information needs.
Task 4: Evaluate the adequacy of those communications. Because decisions are defined by decision makers’ goals, as well
as by their circumstances, different scientific facts may be relevant
Repeat as necessary. to different decision makers. For example, in research aimed at
helping women reduce their risk of sexual assault, we found that
The next four sections survey the science relevant to accom-
communications often ignore outcomes on some women’s minds,
plishing these tasks, followed by consideration of the institutional
such as legal hassles and risks to others (17). In research aimed at
support needed to accomplish them. Much more detail can be
helping young women reduce their risk of sexually transmitted
found in the rest of this special issue and at the Colloquium Web
infections, we found that communications often ignored out-
Downloaded from https://www.pnas.org by 187.189.101.55 on March 21, 2023 from IP address 187.189.101.55.

site (www.nasonline.org/programs/sackler-colloquia/completed_
comes important to many of them, such as relationships and
colloquia/science-communication.html).
reputations (18).
Task 1: Identify the Science Relevant to Decision Making As elsewhere, such applied basic science can reveal basic applied
science opportunities. For example, we found the need for addi-
Although advances in one science (e.g., molecular biology) may
tional research regarding the effectiveness of self-defense strategies
provide the impetus for a decision (e.g., whether to use a new drug), in reducing the risk of sexual assault, studies that are rare com-
informed choices typically require knowledge from many sciences. pared with the many studies on how victims can be treated and
For example, predicting a drug’s risks and benefits for any patient stigmatized. More generally, the methodological challenges of
requires behavioral science research, extrapolating from the con- asking people what matters to them, when forced to choose among
trolled world of clinical trials to the real world in which people risky options, have changed scientific thinking about the nature of
sometimes forget to take their meds and overlook the signs of side human values. Rather than assuming that people immediately
effects (6). Indeed, every science-related decision has a human know what they want, when faced with any possible choice, decision
element, arising from the people who design, manufacture, inspect, scientists now recognize that people must sometimes construct
deploy, monitor, maintain, and finance the technologies involved. their preferences, to determine the relevance of their basic values
As a result, predicting a technology’s costs, risks, and benefits for specific choices (19, 20). As a result, preferences may evolve,
requires social and behavioral science knowledge, just as it might as people come to understand the implications of novel choices
require knowledge from seismology, meteorology, metrology, (such as those posed by new technologies). When that happens,
physics, mechanical engineering, or computer science. the content of science communications must evolve, too, in order
No layperson could understand all of the relevant sciences to any to remain relevant.
depth. Indeed, neither could any scientist. Nor need they have such
vast knowledge. Rather, people need to know the facts that are Task 2: Determine What People Already Know
“material” to their choices (to use the legal term) (7). That Having identified the scientific facts that are worth knowing,
knowledge might include just summary estimates of expected communication researchers can then identify the subset of those
outcomes (e.g., monetary costs, health risks). Or, it might require facts that are worth communicating—namely, those facts that
enough knowledge about the underlying science to understand why people do not know already. Unless communications avoid re-
the experts make those estimates (8). Knowing the gist of that peating facts that go without saying, they can test their audience’s
science could not only increase trust in those claims, but also allow patience and lose its trust (“Why do they keep saying that? Don’t
members of the public to follow future developments, see why they think that I know anything?”). That can happen, for ex-
experts disagree, and have a warranted feeling of self-efficacy, from ample, when the same communication tries to meet both the
learning—and being trusted to learn—about the topic (9, 10). public’s right to know, which requires saying everything, and its
Thus, the first science of communication is analysis: identifying need to know, which requires saying what matters. Full disclo-
those few scientific results that people need to know among the sure is sometimes used in hopes of hiding inconvenient facts in
myriad scientific facts that it would be nice to know (11, 12). The plain sight, by burying them in irrelevant ones (6, 7, 10).
results of that analysis depend on the decisions that science Identifying existing beliefs begins with formative research, using
communications seek to inform. The scientific facts critical to one open-ended, one-on-one interviews, allowing people to reveal

14034 | www.pnas.org/cgi/doi/10.1073/pnas.1213273110 Fischhoff

 whatever is on their minds in their normal ways of expressing edge. Nor can they fairly evaluate experts’ performance. For ex-
themselves (8, 21). Without such openness, communicators can ample, if public health experts seemed to be thinking 60% in 2005,
miss cases where people have correct beliefs, but formulate them then they might have appeared alarmist, once the pandemic failed
in nonscientific terms, meaning that tests of their “science literacy” to materialize—especially if they also seemed to be thinking 60%
may underestimate their knowledge. Conversely, communicators during the subsequent (2009) H1N1 scare. Thus, by communicating
could neglect lay misconceptions that never occurred to them (7, vaguely, experts make themselves needlessly vulnerable to second
8). Focus groups provide an open-ended forum for hearing many guessing. Perhaps they omit quantitative assessments because they
people at once. However, for reasons noted by their inventor, do not realize their practical importance (Task 1). Perhaps they
Robert Merton, focus groups cannot probe individuals’ views at exaggerate how well others could infer those probabilities from
depth, shield participants from irrelevant group pressure, or dis- everything else that they said. Perhaps they do not think that the
tinguish the beliefs that people bring to a group session from those public can handle such technical information. Whatever its source,
that they learn in it (22). As a result, focus groups are much more vagueness undermines the precision essential to all science.
common in commercial research than in scientific studies. Such communication lapses are not surprising. Behavioral re-
Having understood the range and language of lay beliefs, re- search has identified many ways in which people misread others’
searchers can then create structured surveys for estimating their minds (28, 29). For example, the common knowledge effect arises
prevalence. Such surveys must pose their questions precisely when people exaggerate how widely their beliefs are shared (30). It
enough to assess respondents’ mastery of the science. Thus, they can lead people to confuse others by failing to say things that are
should not give people credit for having a superficial grasp of a seemingly, but not actually, obvious. The false consensus effect
science’s gist nor for being sophisticated guessers. Nor should arises when people exaggerate how widely their attitudes are
they deduct credit for not using scientific jargon or not parsing shared (31). It can leave people surprised when others see the
clumsily written questions (10, 16, 23). same facts but make different choices, because they have different
Fig. 1 illustrates such a structured question, based on formative goals. The myth of panic arises when people misinterpret others’
research into how people think about one decision-relevant sci- intense responses to disasters as social disintegration, rather than
entific issue (24). In Fall 2005, as avian flu loomed, we had the op- as mobilization (e.g., the remarkable, if tragically incomplete,
portunity to survey the beliefs of both public health experts, who evacuation of the World Trade Center on 9/11) (32, 33). The myth
could assess the threat, and technology experts, who might create of adolescents’ unique sense of invulnerability arises when adults
options for keeping society going, should worst come to worst. assume that teens misperceive risks, without recognizing the fac-
When asked how likely the virus was to become an efficient hu- tors that can lead teens to act against their own better judgment
man-to-human transmitter in the next 3 years, most public health (e.g., social pressure, poor emotional regulation) (34, 35).
experts saw a probability around 10%, with a minority seeing Thus, because intuition cannot be trusted, communicators must
a much higher one. The technology experts all saw relatively high study what people are thinking. Otherwise, they choose to fly blind
probabilities (median = 60%). Perhaps they had heard the range when creating their messages (36). Given the stakes riding on ef-
fective communication, these studies should be conducted to a
Downloaded from https://www.pnas.org by 187.189.101.55 on March 21, 2023 from IP address 187.189.101.55.

of views among public health experts and then sided with the
more worried ones. More likely, though, they had heard the great publication standard, even if their topics lack the theoretical in-
concern among the experts and then assumed that the probability terest needed for actual publication. That standard includes test-
was high (“otherwise, why would they be so alarmed?”). Given that ing the validity of the elicitation procedures. One class of tests
the public health experts anticipated a 7% case-fatality rate (their examines the consistency of beliefs elicited in different ways.
median response to another question), a 15% chance of efficient For example, the question in Fig. 1 opened a survey whose final
transmission provides ample reason for great concern. However, question reversed the conditionality of the judgment. Instead of
such numeric judgments were nowhere to be found in the volu- asking about the probability of efficient transmission in 3 years, it
minous coverage of the threat. Even though the ambiguity of asked for the time until the probability reached several values
verbal quantifiers (e.g., “likely”) has long been known, experts are (10%, 50%, 90%). Finding consistent responses to the two for-
mulations suggested that respondents had succeeded in trans-
often reluctant to express themselves in clear numeric terms (14–
lating their beliefs into these numeric terms. A second class of tests
16, 25–27).
examines construct validity, asking whether responses to different
When decision makers are forced to read between the lines of
measures are correlated in predicted ways. For example, in a study
experts’ statements, they cannot make full use of scientific knowl-
of teens’ predictions for significant life events, we found that they
gave high probabilities of getting work, pregnant, and arrested if
they also reported higher rates of educational participation, sexual
activity, and gang activity, respectively—in response to questions
in other parts of the survey. That study also found evidence of
predictive validity, such that the events were more likely for teens
who assigned them higher probabilities. Whether these probabil-
ities were over- or underestimated seemed to depend on teens’
beliefs about the events, rather than on their ability to express
themselves numerically (34).

Task 3: Design Communications to Fill the Critical Gaps

A century-plus of social, behavioral, and decision science research
has revealed many principles that can be used in communication
design. Table 1 shows some of the principles that govern judgment
(assessing states of the world) and choice (deciding what to do,
given those assessments). For example, the first judgment princi-
Fig. 1. Response by public health experts (black bars) and technology experts
ple is that people automatically keep track of how often they have
(gray bars) to the following question: “What is the probability that H5N1 will observed an event (e.g., auto accident, severe storm) (37), pro-
become an efficient human-to-human transmitter (capable of being propa- viding an estimate that they can use for predicting future events, by
gated through at least two epidemiological generations of affected humans) applying the availability heuristic (13). Although often useful, re-
sometime during the next 3 years?” Reproduced with permission from ref. 76. lying on such automatic observation can lead people astray when

Fischhoff PNAS | August 20, 2013 | vol. 110 | suppl. 3 | 14035

 Table 1. Partial list of behavioral principles
Judgment Choice

People are good at tracking what they see, but not at detecting People consider the return on their investment in making decisions.
sample bias. People dislike uncertainty, but can live with it.
People have limited ability to evaluate the extent of their People are insensitive to opportunity costs.
own knowledge. People are prisoners to sunk costs, hating to recognize losses.
People have difficulty imagining themselves in other visceral states. People may not know what they want, especially with novel questions.
People have difficulty projecting nonlinear trends.
People confuse ignorance and stupidity.

an event is disproportionately visible (e.g., because of media re- Even within a domain, knowledge may vary widely by topic. People
porting practices) and they cannot correct for the extent to which forced to learn about cancer may know little about diabetes.
appearances are deceiving. For example, people may think that People adept at reading corporate balance sheets may need help
(widely reported) homicides are more common than (seldom with their own tax forms. As a result, knowing individuals’ general
reported) suicides, even though these two causes of death occur level of knowledge (sometimes called their “literacy”) in a domain
about as often (38). The first choice principle (in Table 1) is that provides little guidance for communicating with them about
people consider the return on investing in decision-making pro- specific issues. For that, one needs to know their mental model for
cesses (39). Applying that principle could lead them to disregard that issue, reflecting what they have heard or seen on that topic
communications that seem irrelevant, incomprehensible, or un- and inferred from seemingly relevant general knowledge.
trustworthy (40). The study of mental models has a long history in cognitive
Each of these principles was a discovery in its time. Each is sup- psychology, in research that views people as active, if imperfect,
ported by theories, explicating its underlying processes. Nonethe- problem solvers (46–48). Studies conducted with communication
less, each is also inconclusive. Applying a principle requires in mind often find that people have generally adequate mental
additional (auxiliary) assumptions regarding its expression in models, undermined by a few critical gaps or misconceptions (8,
specific circumstances. For example, applying the first judgment 21). For example, in a study of first-time mothers’ views of child-
principle to predicting how people estimate an event’s frequency hood immunization (49), we found that most knew enough for
requires assumptions about what they observe, how they encode future communications to focus on a few missing links, such as the
those observations, how hard they work to retrieve relevant importance of herd immunity to people who cannot be immunized
memories, whether the threat of biased exposure occurs to them, and the effectiveness of postlicensing surveillance for vaccine side
and how they adjust for such bias. Similarly, applying the first effects. A study of beliefs about hypertension (50) found that many
Downloaded from https://www.pnas.org by 187.189.101.55 on March 21, 2023 from IP address 187.189.101.55.

choice principle to predicting how people estimate the expected people know that high blood pressure is bad, but expect it to have
return on their investment in reading a science communication perceptible symptoms, suggesting messages about a “silent killer.”
requires assumptions about how much they think that they know A classic study found that many people interpret “once in 100
already, how trustworthy the communication seems, and how ur- years” as a recurrence period, rather than as an annual probability,
gent such learning appears to be. showing the importance of expressing hydrologic forecasts in lay
Moreover, behavioral principles, like physical or biological terms (51).
ones, act in combination. Thus, although behavior may follow At other times, though, the science is so unintuitive that people
simple principles, there are many of them (as suggested by the have difficulty creating the needed mental models. In such cases,
incomplete lists in Table 1), interacting in complex ways. For ex- communications must provide answers, rather than asking people
ample, dual-process theories detail the complex interplay between to infer them. For example, risks that seem negligible in a single
automatic and deliberate responses to situations, as when people exposure (e.g., bike riding, car driving, protected sex) can become
integrate their feelings and knowledge about a risk (9, 10, 13, 35, major concerns through repeated exposure (52, 53). Invoking
38). Getting full value from behavioral science requires studying a long-term perspective might help some (“one of these days, not
not only how individual principles express themselves in specific wearing a helmet will catch up with you”) (54). However, when
settings, but also how they act in consort (41). Otherwise, com- people need more precise estimates of their cumulative risk, they
municators risk grasping at the straws that are offered by simplistic cannot be expected to do the mental arithmetic, any more than
single-principle, “magic bullet” solutions, framed as “All we need they can be expected to project exponential processes, such as the
to do is X, and then the public will understand”—where X might proliferation of invasive species (55), or interdependent nonlinear
be “appeal to their emotions,” “tell them stories,” “speak with ones, such as those involved in climate change (56). If people need
confidence,” “acknowledge uncertainties,” or “emphasize op- those estimates, then someone needs to run the numbers for
portunities, rather than threats” (42). Indeed, a comprehensive them. Similarly, people cannot be expected to understand the
approach to communication would consider not only principles of uncertainty in a field, unless scientists summarize it for them (6,
judgment and choice, but also behavioral principles identified in 11, 16, 27, 57, 58).
studies of emotion, which find that feelings can both aid com-
munication, by orienting recipients toward message content, and Task 4: Evaluate their adequacy and repeat as necessary
undermine it, as when anger increases optimism—and diminishes Poor communications cause immediate damage if they keep
the perceived value of additional learning (43, 44). A compre- people from using available science knowledge. They cause lasting
hensive approach would also consider the influences of social damage if they erode trust between scientists and the public. That
processes and culture on which information sources people trust happens when lay people see scientists as insensitive to their needs
and consult (45). and scientists see lay people as incapable of grasping seemingly
Whereas the principles governing how people think may be basic facts (2, 9, 24, 50). Such mistrust is fed by the natural ten-
quite general (albeit complex in their expression and interaction), dency to attribute others’ behavior to their essential properties
what people know naturally varies by domain, depending on their rather than to their circumstances (59). Thus, it may be more
desire and opportunity to learn. People who know a lot about natural for laypeople to think, “Scientists don’t care about the
health may know little about physics or finance, and vice versa. public” than “Scientists have little chance to get to know the public.”

14036 | www.pnas.org/cgi/doi/10.1073/pnas.1213273110 Fischhoff

 And it may be more natural for scientists to think, “The public choices made by people who receive a message with those of fully
can’t understand our science” than “The public has little chance informed individuals (6, 71). It can find that an otherwise poor
to learn our science.” Indeed, the very idea of science literacy communication is adequate, because recipients know its content
emphasizes the public’s duty to know over the experts’ duty already, and that an otherwise sound communication is in-
to inform. adequate, because recipients’ mental models are too fragmentary
To avoid such misattributions, communication programs must to absorb its content. Applying the standard is conceptually
measure the adequacy of their performance. A task analysis sug- straightforward with persuasive communications (72), designed to
gests the following test: a communication is adequate if it (i) encourage desired behaviors (e.g., not smoking, evacuating dan-
contains the information that recipients need, (ii) in places that gerous places): see whether recipients behave that way. However,
they can access, and (iii) in a form that they can comprehend. That with nonpersuasive communications, designed to help people
test can be specified more fully in the following three standards. make personally relevant choices (e.g., among careers, public
(i) A materiality standard for communication content. A com- policies, medical procedures), the test is more complicated. It
munication is adequate if it contains any information that might requires taking an “inside view,” identifying recipients’ personal
affect a significant fraction of users’ choices. Malpractice laws in goals and the communication’s contribution to achieving them (11,
about half of the United States require physicians to provide any 13, 17). That view may lead to creating options that facilitate ef-
information that is material to their patients’ decisions (60). (The fective decision making, such as subsidizing (or even requiring)
other states have a professional standard, requiring physicians sustained commitment to protective behaviors (e.g., flood in-
to say whatever is customary in their specialty.) The materiality surance), signaling its importance and encouraging long-term per-
standard can be formalized in value-of-information analysis (7, 11, spectives, as an antidote to myopic thinking. That view may also
12), asking how much knowing an item of information would affect require removing barriers, such as the lack of affordable options (73).
recipients’ ability to choose wisely. That analysis might find that Each of these three standards includes verbal quantifiers whose
a few items provide all that most recipients need, or that target resolution requires value judgments. The materiality standard
audiences have such diverse information needs that they require recognizes that users might face such varied decisions that the
different brief messages, or that there are too many critical facts best single communication omits information that a fraction need;
for any brief message to suffice. Without analysis, though, one someone must then decide whether that fraction is significant. The
cannot know what content is adequate (61). accessibility standard recognizes that not all users can (or should)
(ii) A proximity standard for information accessibility. A com- be reached directly; someone must then decide whether most
munication is adequate if it puts most users within X degrees of users are close enough to someone with the information. The
separation from the needed information, given their normal comprehensibility standard recognizes that not everyone will (or
search patterns. Whatever its content, a communication has little must) understand everything; someone must then decide whether
value unless it reaches its audience, either directly (e.g., through most people have extracted enough meaning. Although scientists
warning labels, financial disclosures, or texted alerts) or indirectly can estimate these impacts, they alone cannot decide whether
(e.g., through friends, physicians, or journalists). If there are too communicators’ efforts are good enough.
Downloaded from https://www.pnas.org by 187.189.101.55 on March 21, 2023 from IP address 187.189.101.55.

many bridging links, then messages are unlikely to connect with

their targets. If there are too few links, then people will drown in Organizing for Communication
messages, expecting them to become experts in everything, while As seen in the Colloquium and this special issue, scientifically
ignoring the social networks that they trust to select and interpret sound science communication demands expertise from multiple
information for them (9, 45, 62, 63). As a result, the proximity disciplines, including (a) subject matter scientists, to get the facts
standard evaluates the distribution of messages relative to recipi- right; (b) decision scientists, to identify the right facts, so that
ents’ normal search patterns. they are not missed or buried; (c) social and behavioral scientists,
As elsewhere, the need for such research is demonstrated by the to formulate and evaluate communications, and (d) communi-
surprising results that it sometimes produces, such as the seeming cation practitioners, to create trusted channels among the parties.
“contagion” of obesity and divorce (64) or the ability of messages Recognizing these needs, some large organizations have moved
to reach people through multiple channels (65, 66). Casman et al. to have such expertise on staff or on call. For example, the Food
(67) found that traditional, trusted distribution systems for “boil and Drug Administration has a Strategic Plan for Risk Communi-
water” notices could not possibly reach some consumers in time to cation and a Risk Communication Advisory Committee. The
protect them from some contaminants. Winterstein and Kimberlin Centers for Disease Control and Prevention has behavioral sci-
(68) found that the system for distributing consumer medication entists in its Emergency Operations Center. Communications are
information sheets puts something in the hands of most patients central to the Federal Emergency Management Agency’s Strategic
who pick up their own prescriptions. Downs et al. (49) found that Forecast Initiative and the National Oceanic and Atmospheric
first-time mothers intuitively choose search terms that lead to Web Administration’s Weather-Ready Nation program. The National
sites maintained by vaccine skeptics, rather than by proponents. In Academy of Sciences’ Science and Entertainment Exchange
each case, one would not know who gets the messages without answers and encourages media requests for subject matter expertise.
such study. For the most part, though, individual scientists are on their own,
Once users connect with a communication, they must find what forced to make guesses about how to meet their audiences’ in-
they need in it. Research has identified design principles for formation needs. Reading about the relevant social, behavioral,
directing attention, establishing authority, and evoking urgency and decision sciences might help them some. Informal discussion
(41, 69). The adequacy of any specific communication in revealing with nonscientists might help some more, as might asking mem-
its content can be studied by direct observation or with models bers of their audience to think aloud as they read draft commu-
using empirically based assumptions. Using such a model, Riley nications (41). However, such ad hoc solutions are no substitute
et al. (70) found that users of several methylene chloride-based for expertise in the sciences of communication and evidence on
paint stripper products who relied on product labels would not how to communicate specific topics (74, 75).
find safety instructions if they followed several common search The scientific community owes individual scientists the support
strategies (e.g., read the first five items, read everything on the needed for scientifically sound communication. One might envi-
front, read only warnings). sion a “Communication Science Central,” where scientists can
(iii) A comprehensibility standard for user understanding. A take their problems and get authoritative advice, such as: “here’s
communication is adequate if most users can extract enough how to convey how small risks mount up through repeated expo-
meaning to make sound choices. This standard compares the sure”; “in situations like the one that you’re facing, it is essential to

Fischhoff PNAS | August 20, 2013 | vol. 110 | suppl. 3 | 14037

 the distant parties. The Ocean Health Index (79, 80) uses natural
and decision science to identify the impacts that various stake-
holders value and then synthesize the evidence relevant to them.
The Blue Ribbon Commission on America’s Nuclear Future
adapted its ambitious plans for public consultation in response to
social science input about how to establish trusted communica-
tion channels (81, 82). Its recommendations called for sustained
communications, to ensure that proposals for nuclear power and
waste disposal receive fair hearings (83).

Conclusion
Communications are adequate if they reach people with the in-
formation that they need in a form that they can use. Meeting that
goal requires collaboration between scientists with subject matter
knowledge to communicate and scientists with expertise in com-
munication processes—along with practitioners able to manage
the process. The Sackler Colloquium on the “Science of Science
Communication” brought together experts in these fields, to in-
teract with one another and make their work accessible to the
general scientific community. Such collaboration affords the sci-
ences the best chance to tell their stories.
It also allows diagnosing what went wrong when communica-
tions fail. Did they get the science wrong, and lose credibility?
Did they get the wrong science, and prove irrelevant? Did they
lack clarity and comprehensibility, frustrating their audiences?
Did they travel through noisy channels, and not reach their
audiences? Did they seem begrudging, rather than forthcoming?
Did they fail to listen, as well as to speak? Did they try to per-
suade audiences that wanted to be informed, or vice versa?
Fig. 2. A process for integrating communication and decision making. Correct diagnoses provide opportunities to learn. Misdiagnoses
Reproduced with permission from ref. 24.
can compound problems if they lead to blaming communication
partners, rather than communication processes, for failures (46).
Downloaded from https://www.pnas.org by 187.189.101.55 on March 21, 2023 from IP address 187.189.101.55.

listen to local residents’ concerns before saying anything”; “we’ll For example, scientists may lose patience with a public that cannot
get you estimates on the economic and social impacts of that seem to understand basic facts, not seeing what little chance the
problem”; “you might be able to explain ocean acidification by public has had to learn them. Conversely, laypeople may lose faith
building on your audience’s mental models of acids, although we in scientists who seem distant and uncaring, not appreciating the
don’t have direct evidence”; or “that’s a new topic for us; we’ll challenges that scientists face in reaching diverse audiences. When
collect some data and tell you what we learn.” By helping individual a science communication is successful, members of its audience
scientists, such a center would serve the scientific community as should agree on what the science says relevant to their decisions.
a whole, by protecting the commons of public goodwill that each They need not agree about what decisions to make if they value
communication either strengthens or weakens. different things. They may not be reachable if issues are so po-
Realizing the value of science communication requires in- larized that the facts matter less than loyalty to a cause that
tegrating it with the decision-making processes that it is meant to interprets them in a specific way (45). However, in such cases, it is
inform. Fig. 2 (76) shows one version of the advice coming from social, rather than communication, processes that have failed.
many consultative bodies (2, 77, 78). In the center is a conven- Thus, by increasing the chances of success and aiding the di-
tional management process, going from initiation to imple- agnosis of failure, the sciences of communication can protect
mentation and monitoring. At the left is a commitment to two-way scientists from costly mistakes, such as assuming that the public
communication at each stage. Thus, a process begins by reporting can’t handle the truth, and then denying it needed information, or
preliminary plans and hearing the public’s thoughts early enough becoming advocates, and then losing the trust that is naturally
to incorporate its input. Sustained contact, throughout the pro- theirs (9, 24, 72). To these ends, we need the full range of social,
cess, reduces the chance of any party blindsiding the other with behavioral, and decision sciences presented at this Colloquium,
unheard views or unexpected actions. In this view, the process coupled with the best available subject matter expertise. When
itself represents a communication act, expressing a belief in the science communication succeeds, science will give society the
public’s right to know and ability to understand. greatest practical return on its investment—along with the sense of
The specifics of such consultations will depend on the decisions wonder that exists in its own right (84).
involved. Siting a factory, wind farm, or halfway house may require
direct conversation with those affected by it. Informing a national ACKNOWLEDGMENTS. Preparation of this paper was supported by the
debate may need social scientists to act as intermediaries between National Science Foundation Grant SES-0949710.

1. Burger J (2002) The Parrot Who Owns Me (Random House, New York). 5. Baddeley AD (1978) Applied cognitive and cognitive applied research. Perspectives on
2. Dietz T (2013) Bringing values and deliberation to science communication. Proc Natl Memory Research, ed Nilsson LG (Erlbaum, Hillsdale, NJ).
Acad Sci USA 110:14081–14087. 6. Schwartz LM, Woloshin S (2013) The Drug Facts Box: Improving the communication of
3. Löfstedt R, Fischhoff B, Fischhoff I (2002) Precautionary principles: General definitions prescription drug information. Proc Natl Acad Sci USA 110:14069–14074.
and specific applications to genetically modified organisms (GMOs). J Policy Anal 7. Merz J, Fischhoff B, Mazur DJ, Fischbeck PS (1993) Decision-analytic approach to de-
Manage 21(3):381–407. veloping standards of disclosure for medical informed consent. J Toxics Liability 15(1):
4. Klahr D (2013) What do we mean? On the importance of not abandoning sci- 191–215.
entific rigor when talking about science education. Proc Natl Acad Sci USA 110: 8. Bruine de Bruine W, Bostrom A (2013) Assessing what to address in science commu-
14075–14080. nication. Proc Natl Acad Sci USA 110:14062–14068.

14038 | www.pnas.org/cgi/doi/10.1073/pnas.1213273110 Fischhoff

 9. Lupia A (2013) Communicating science in politicized environments. Proc Natl Acad Sci 48. Johnson-Laird P (1983) Mental Models (Cambridge Univ Press, New York).
USA 110:14048–14054. 49. Downs JS, Bruine de Bruin W, Fischhoff B (2008) Parents’ vaccination comprehension
10. Reyna VF (2012) A new intuitionism: Meaning, memory, and development in Fuzzy- and decisions. Vaccine 26(12):1595–1607.
Trace Theory. Judgm Decis Mak 7(3):332–339. 50. Meyer D, Leventhal H, Gutmann M (1985) Common-sense models of illness: The ex-
11. von Winterfeldt D (2013) Bridging the gap between science and decision making. ample of hypertension. Health Psychol 4(2):115–135.
Proc Natl Acad Sci USA 110:14055–14061. 51. Kates RW (1962.) Hazard and Choice Perception in Flood Plain Management (Univ of
12. Raiffa H (1968) Decision Analysis (Addison-Wesley, Reading, MA). Chicago Press, Chicago), Department of Geography Research Paper No. 78.
13. Kahneman D (2011) Thinking, Fast and Slow (Farrar Giroux & Strauss, New York). 52. Cohen J, Chesnick EI, Haran D (1971) Evaluation of compound probabilities in se-
14. Morgan MG, Henrion M (1990) Uncertainty (Cambridge University Press, New York). quential choice. Nature 232(5310):414–416.
15. O’Hagan A, et al. (2006) Uncertain Judgements: Eliciting Expert Probabilities (Wiley, 53. Slovic P, Fischhoff B, Lichtenstein S (1978) Accident probabilities and seat belt usage:
Chichester, UK). A psychological perspective. Accid Anal Prev 10:281–285.
16. Campbell P (2011) Understanding the receivers and the reception of science’s un- 54. Weinstein ND, Kolb K, Goldstein BD (1996) Using time intervals between expected
certain messages. Philos Trans R Soc 369:4891–4912. events to communicate risk magnitudes. Risk Anal 16(3):305–308.
17. Fischhoff B (1992) Giving advice: Decision theory perspectives on sexual assault. Am 55. Wagenaar W, Sagaria S (1975) Misperception of exponential growth. Percept Psy-
Psychol 47(4):577–588. chophys 18(6):416–422.
18. Downs JS, et al. (2004) Interactive video behavioral intervention to reduce adolescent 56. Sterman JD, Sweeney LB (2002) Cloudy skies: Assessing public understanding of global
females’ STD risk: A randomized controlled trial. Soc Sci Med 59(8):1561–1572. warming. Syst Dyn Rev 18(2):207–240.
19. Fischhoff B (1991) Value elicitation: Is there anything in there? Am Psychol 46(8): 57. Budescu DV, Broomell S, Por HH (2009) Improving communication of uncertainty in
835–847. the reports of the intergovernmental panel on climate change. Psychol Sci 20(3):
20. Lichtenstein S, Slovic P, eds (2006) The Construction of Preferences (Cambridge Univ 299–308.
Press, New York). 58. Politi MC, Han PKJ, Col NF (2007) Communicating the uncertainty of harms and
21. Morgan MG, Fischhoff B, Bostrom A, Atman C (2001) Risk Communication: The benefits of medical interventions. Med Decis Making 27(5):681–695.
Mental Models Approach (Cambridge Univ Press, New York). 59. Nisbett RE, Ross L (1980) Human Inference (Prentice-Hall, Englewood Cliffs, NJ).
22. Merton RF (1987) The focussed interview and focus groups. Public Opin Q 51(4): 60. Merz J (1991) An empirical analysis of the medical informed consent doctrine. Risk
550–566. 2(1):27–76.
23. Bruine de Bruin W, Downs JS, Fischhoff B, Palmgren C (2007) Development and 61. Feldman-Stewart D, et al. (2007) A systematic review of information in decision aids.
evaluation of an HIV/AIDS knowledge measure for adolescents focusing on mis-
Health Expect 10(1):46–61.
conceptions. J HIV AIDS Prev Child Youth 8(1):35–57.
62. Scheufele DA (2013) Communicating science in social settings. Proc Natl Acad Sci USA
24. Bruine De Bruin W, Fischhoff B, Brilliant L, Caruso D (2006) Expert judgments of
110:14040–14047.
pandemic influenza risks. Glob Public Health 1(2):178–193.
63. Eveland WP, Cooper KE (2013) An integrated model of communication influence on
25. Kent S (1994). Words of estimative probability. Sherman Kent and the Board of Na-
beliefs. Proc Natl Acad Sci USA 110:14088–14095.
tional Estimates: Collected Essays, ed Steury DP (Central Intelligence Agency, Wash-
64. Christakis NA, Fowler J (2009) Connected (Little, Brown, New York).
ington, DC), pp 151–166. Available at https://www.cia.gov/library/center-for-the-
65. Brossard D (2013) New media landscapes and the science information consumer. Proc
study-of-intelligence/csi-publications/books-and-monographs/sherman-kent-and-the-
Natl Acad Sci USA 110:14096–14101.
board-of-national-estimates-collected-essays/6words.html. Accessed July 26, 2013.
66. Nisbet MC, Kotcher J (2009) A two step flow of influence? Opinion-leader campaigns
26. Wallsten T, Budescu D, Rapoport A, Zwick R, Forsyth B (1986) Measuring the vague
on climate change. Science Communication 30:328–358.
meanings of probability terms. J Exp Psychol Gen 115:348–365.
67. Casman EA, Fischhoff B, Palmgren C, Small MJ, Wu F (2000) An integrated risk model
27. Fischhoff B (2012) Communicating uncertainty: Fulfilling the duty to inform. Issues Sci
of a drinking-water-borne cryptosporidiosis outbreak. Risk Anal 20(4):495–511.
Technol 28(4):63–70.
68. Winterstein AG, Kimberlin CL (2010) Usefulness of consumer medication information
28. Keysar B, Lin S, Barr DJ (2003) Limits on theory of mind use in adults. Cognition 89(1):
dispensed in retail pharmacies. Arch Intern Med 170:1314–1324.
25–41.
Downloaded from https://www.pnas.org by 187.189.101.55 on March 21, 2023 from IP address 187.189.101.55.

69. Wogalter M, ed (2006) The Handbook of Warnings (Lawrence Erlbaum Associates,

29. Pronin E (2008) How we see ourselves and how we see others. Science 320(5880):
Hillsdale, NJ).
1177–1180.
70. Riley DM, Fischhoff B, Small MJ, Fischbeck P (2001) Evaluating the effectiveness of
30. Nickerson RA (1999) How we know – and sometimes misjudge – what others know.
risk-reduction strategies for consumer chemical products. Risk Anal 21(2):357–369.
Psychol Bull 125(6):737–759.
71. Eggers SL, Fischhoff B (2004) A defensible claim? Behaviorally realistic evaluation
31. Dawes RM, Mulford M (1996) The false consensus effect and overconfidence. Organ
Behav Hum Decis Process 65(3):201–211. standards. J Public Policy Mark 23:14–27.
32. Wessely S (2005) Don’t panic! Short and long-term psychological reactions to the new 72. Fischhoff B (2007) Nonpersuasive communication about matters of greatest urgency:
terrorism. J Ment Health 14(1):1–6. Climate change. Environ Sci Technol 41(21):7204–7208.
33. Silver RC, ed (2011) Special Issue: 9/11: Ten Years Later. Am Psychol 66(6). 73. Kunreuther H, Michel-Kerjan E (2011) People get ready: Disaster preparedness. Issues
34. Fischhoff B (2008) Assessing adolescent decision-making competence. Dev Rev 28(1): Sci Technol 28(1):1–7.
12–28. 74. Peters HP (2013) Gap between science and media revisited: Scientists as public com-
35. Reyna V, Farley F (2006) Risk and rationality in adolescent decision making: Im- municators. Proc Natl Acad Sci USA 110:14102–14109.
plications for theory, practice, and public policy. Psychol Public Interest 7(1):1–44. 75. Yocco V, Jones EC, Storksdieck M (2012) Factors contributing to amateur astron-
36. Fischhoff B (2011) Communicating about the risks of terrorism (or anything else). Am omers’ involvement in education and public outreach. Astronomy Education Review
Psychol 66(6):520–531. 11(1):010109.
37. Hasher L, Zacks RT (1984) Automatic processing of fundamental information: The case 76. Canadian Standards Association (1997) Risk Management: Guidelines for Decision
of frequency of occurrence. Am Psychol 39(12):1372–1388. Makers (Q850) (Canadian Standards Association, Ottawa, Canada).
38. Slovic P, ed (2001) Perception of Risk (Earthscan, London). 77. National Research Council (1996) Understanding Risk (National Academy Press,
39. Kahneman D, Tversky A, eds (2000) Choice, Values, and Frames (Cambridge Univ Washington, DC).
Press, New York). 78. Presidential/Congressional Commission on Risk Assessment and Risk Management
40. Löfstedt R (2005) Risk Management in Post-Trust Societies (Palgrave/MacMillan, (1998) Risk Management (Environmental Protection Agency, Washington, D.C.).
London). 79. Ocean Health Index. Available at www.oceanhealthindex.org/. Accessed July 26, 2013.
41. Fischhoff B, Brewer N, Downs JS, eds (2011) Communicating Risks and Benefits: An 80. Halpern BS, et al. (2013) Elicited preferences for components of ocean health in the
Evidence-Based User’s Guide (Food and Drug Administration, Washington, DC). California Current. Mar Policy 42(1):68–73.
42. Fischhoff B (1995) Risk perception and communication unplugged: Twenty years of 81. Dietz T, Stern P, eds (2008) Public Participation in Environmental Assessment and
process. Risk Anal 15(2):137–145. Decision Making (National Academy Press, Washington, DC).
43. Lerner JS, Keltner D (2001) Fear, anger, and risk. J Pers Soc Psychol 81(1):146–159. 82. Rosa EA, et al. (2010) Nuclear waste: Knowledge waste? Science 329(5993):762–763.
44. Slovic P, ed (2010) The Feeling of Risk (Earthscan, London). 83. Blue Ribbon Commission on America’s Nuclear Future (2012) Report to the Secretary
45. Haidt JH (2012) The Righteous Mind (Pantheon, New York). of Energy (Department of Energy, Washington, DC).
46. Bartlett RC (1932) Remembering (Cambridge Univ Press, Cambridge). 84. Fischhoff B, Kadvany J (2011) Risk: A Very Short Introduction (Oxford Univ Press,
47. Gentner D, Stevens AL, eds (1983) Mental Models (Erlbaum, Hillsdale, NJ). Oxford).

Fischhoff PNAS | August 20, 2013 | vol. 110 | suppl. 3 | 14039