Evaluating Science Communication
Evaluating Science Communication
net/publication/329214406
CITATIONS READS
72 941
1 author:
Baruch Fischhoff
Carnegie Mellon University
495 PUBLICATIONS 48,727 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Baruch Fischhoff on 10 December 2018.
Edited by Dietram A. Scheufele, University of Wisconsin–Madison, Madison, WI, and accepted by Editorial Board Member May R. Berenbaum October 4, 2018
(received for review June 18, 2018)
Effective science communication requires assembling scientists with this characterization, like Simon’s distinction (5, 6), is a heuristic
knowledge relevant to decision makers, translating that knowledge one. It captures some general features while breaking down in
into useful terms, establishing trusted two-way communication ways that reveal the communities’ need for one another.
channels, evaluating the process, and refining it as needed. Com- Scientists’ bounded rationality entails ignoring issues that they
municating Science Effectively: A Research Agenda [National cannot treat systematically, hoping to reach strong conclusions
Research Council (2017)] surveys the scientific foundations for within their discipline’s self-imposed constraints. Scientists from
accomplishing these tasks, the research agenda for improving them, different disciplines struggle to collaborate, because they bound
and the essential collaborative relations with decision makers and problems differently. Experimental researchers may be uncom-
communication professionals. Recognizing the complexity of the fortable with unruly field observations. Field researchers may
science, the decisions, and the communication processes, the report question the artificial conditions of experiments. Both may puzzle
calls for a systems approach. This perspective offers an approach to over computational models, while modelers may have little pa-
creating such systems by adapting scientific methods to the practical tience for the simplification of experiments or the qualitative ev-
constraints of science communication. It considers staffing (are the idence of field research. Scientists who study individuals may not
right people involved?), internal collaboration (are they talking to know what to do with the context provided by those who study
SUSTAINABILITY
one another?), and external collaboration (are they talking to other
groups or cultures, who may shake their heads at being ignored.
Each discipline owes its success to its tacit knowledge of how to
SCIENCE
stakeholders?). It focuses on contexts where the goal of science
work within its bounds. Those bounds can be so incommensu-
communication is helping people to make autonomous choices
rable that scientists from different disciplines struggle even to
rather than promoting specific behaviors (e.g., voter turnout, vacci-
agree about how to disagree (7, 8). Nonetheless, as argued by
nation rates, energy consumption). The approach is illustrated with
Communicating Science Effectively: A Research Agenda (1), the
research in two domains: decisions about preventing sexual assault
success of science communication depends on collaboration
and responding to pandemic disease.
across disciplines.
Practitioners’ satisficing entails paying attention to anything
science communication | evaluation | decision making | pandemics | that might be relevant and accepting imperfect solutions. Prac-
sexual assault titioners of different persuasions struggle to collaborate, because
they have different skills and norms. Those skills might include
2 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1805863115 Fischhoff
COLLOQUIUM
PAPER
evidence on the efficacy of self-defense measures, we concluded sensitivities, and collect feedback. Practitioners are also needed
that, if they knew research, women might differ in the strategies to manage the process, secure the relevant experts, and get them
that they chose. As a result, we needed to understand those talking with one another and external stakeholders. Without a
differences. As befits any communication project, we began by firm hand, normal group dynamics can lead to recruiting, re-
listening. Here, that entailed semistructured interviews with di- warding, and retaining people with similar backgrounds and
verse groups of women, men, and experts, eliciting their per- blind spots, who are overly comfortable talking to one another. A
spectives on both personal decisions (e.g., how to respond to an firm hand is also needed to let everyone offer opinions, while
assault) and societal ones (e.g., how to make assaults less likely). leaving ultimate authority to those most expert in a topic. That
These interviews revealed a rich decision space, with many will keep subject matter experts from editing for style rather than
possible options, outcomes, and uncertainties (37, 38). To struc- accuracy, social scientists from garbling the facts when trying to
ture that space, we created categories of options and outcomes, clarify them, and practitioners from spinning messages when the
seeking a level of granularity that would be useful to decision facts are needed.
makers. Within that structure, we summarized available research
on the effects of the options on the outcomes. Where the evidence Internal Consultation: Are They Talking Effectively with One Another?
was limited, as was usually the case, those limits were part of the Experts must combine their knowledge to realize its value. That
story. When uncertainty is great, advice in unproven. Unless those means jointly examining issues, connections between issues, and
limits are acknowledged, if things go badly, then decision makers the assumptions underlying those interpretations. Fig. 1 illus-
may bear the insult of blame and regret in addition to the injury trates a decision science tool for structuring such consultations
that occurred. Whatever they did, some “expert” had advised (44). A computable (i.e., nonnumeric) version of an influence
otherwise. diagram (45), it depicts actions as rectangles and uncertain var-
Given that uncertainty and the diversity of decision makers’ iables as ovals (gray if valued outcomes; white if intermediates).
circumstances, our project had no theory of change for encour- It was created to structure discussions at a meeting about the
aging specific behaviors among those concerned with sexual as- then-pending threat of H5N1 (avian flu). It has places for the
sault. However, we did have a theory of change for ourselves, science that could inform decisions faced by health officials,
structuring our efforts to inform those decisions. That engage- employers, parents, suppliers, and others, each wondering if and
SUSTAINABILITY
ment and subsequent ones have led to the theory of change that how to prepare for a possible pandemic. What should they ex-
SCIENCE
guides this proposal for implementing the recommendation of a pect regarding quarantine, home schooling, rationing, hospital
systems approach in Communicating Science Effectively: A closures, telecommuting, drug shortages, and social solidarity (or
Research Agenda (1). fracture)?
Translating science into such a decision-relevant form requires
Staffing: Are the Right People Involved? consultation on three levels. One is summarizing the science at
Effective science communications must contain the information each node (e.g., what quantities of antivirals will be available)
that recipients need in places that they can access and in a form and link (e.g., how effectively will vaccines reduce morbidity).
that they can use. Achieving those goals requires four kinds of The second is estimating interactions (e.g., how will morbidity
expertise (39). and mortality combine to affect social costs). The third is iden-
tifying contextual factors (sometimes called “index variables”)
Subject Matter Experts. The core of any science communication is that affect many model elements (e.g., is the society developed or
authoritative summaries of evidence relevant to decision makers’ developing) (40).
needs. That evidence may come from many sciences. For ex- Quantifying such models demands technical training and
ample, sexual assault decisions might be informed by results from material resources. However, sketching a model well enough to
psychology, sociology, criminology, and economics. Unless staff facilitate consultations only requires clear thinking and candid
have expertise in an issue or the capacity to absorb it (13), they conversation. To that end, before the H5N1 meeting, partici-
will have to ignore or guess at it (26, 40). pants completed a survey eliciting their beliefs about the issues in
Fig. 1 and several related models (46). The models were in-
Decision Scientists. Eager to share their knowledge, subject matter stantiated with scenarios to make their abstractions concrete.
experts may drown decision makers in facts that it would be nice The meeting and survey were anonymous so that participants
to know. Decision scientists can identify the facts that decision could work the problem without pressure to represent the firms
makers need to know. They can also characterize evidence or agencies in which many held senior positions. They were
quality, estimate decision sensitivity, and reveal hidden assump- drawn from public health, technology, and mass media; hence,
tions (16, 17, 30, 31). For example, a sensitivity analysis of the they could offer their views on the needs and responses of publics
decision faced by a young academic might conclude that “there that they might support in a pandemic, but typically know in
is no sure way to prevent a powerful figure in your field from more benign circumstances.
destroying your career.” A decision analysis of sexual assault These scientists and practitioners were brought together be-
advice might conclude that “it ignores restrictions on your cause interpreting such evidence requires actual conversation. It
freedom.” is not enough for members of one field to read the publications
of another. Publications reflect their authors’ perspectives and
Social, Behavioral, and Communication Scientists. Knowing what to not those of their entire field. They omit assumptions that go
say does not guarantee knowing how to say it. Coupled with the without saying when scientists or practitioners write for colleagues
normal human tendency to overestimate mutual understanding or clients. For scientists, those assumptions include bounds on
(41), scientists’ intuitions can be a poor guide to effective com- their discipline’s rationality. For practitioners, they include ac-
munication. Indeed, scientists’ success in the classroom may cepted limits to satisficing solutions.
produce unrealistic expectations for being able to communicate One practical method for describing these internal consultations
with general audiences, with no examinations providing feedback is social network analysis, created by asking members of a com-
on their success. Communicating Science Effectively: A Research munication team to describe their relationships (47, 48). A suc-
Agenda (1) identifies the diverse expertise available for un- cessful team will have the requisite connections among those
derstanding audiences, crafting communications, and evaluating associated with each link and node in the relevant models. Fig. 2
success (2–4, 35, 42, 43). shows such relationships as revealed in self-reports of “close and
collegial relations” in a study of six interdisciplinary research centers
Program Designers and Managers. Finally, science communication (49). For this center, the study concluded that “most. . .interactions
needs practitioners to create channels, recruit stakeholders, are concentrated in a small core of researchers. . .Disciplines from
disseminate messages, mind legal constraints, anticipate cultural the physical sciences dominate the core. . ., environmental
Health
care costs
Rate of
spread
Morbidity
Mortality
Non-health care
economic costs
Action
node
Chance
node
Social
Outcome costs
node
Fig. 1. Risk model for pharmacological interventions for a pandemic. Ovals indicate uncertain variables, which need to be predicted. Rectangles indicate
actions, which need to be planned and implemented. Reprinted by permission from ref. 29, Springer Nature: Journal of Risk and Uncertainty, copyright 2006.
scientists/social scientists dominate the periphery” (ref. 49, p. 57). Our sexual assault project adopted one imperfect approach
A theory of change for a communication team would specify which (36–38). It asked nonrepresentative samples of individuals
members must talk with one another, with a diagram measuring its recruited from diverse groups to complete open-ended surveys,
success. Of course, even parties who view their relationships as close allowing them to choose the issues and describe them in their
and collegial may not identify and correct all misunderstandings. own terms. These surveys were followed by confirmatory struc-
tured ones with similarly sampled individuals. Although we en-
External Consultation: Are They Talking Effectively with Other gaged diverse individuals who revealed a wide range of views, the
Stakeholders? Scientific communicators seek to be trusted part- consultation was indirect. Our avian flu project involved two days
ners of people making decisions where science matters. In this of intense direct consultation, building on a preparatory survey.
issue of PNAS, those include decisions about gene drives (50), However, it was with a highly select group experienced in sam-
autonomous vehicles (51), employment (52), and energy (53). pling public opinion but not authorized to represent it.
Earning that trust means providing the science most relevant to The medical world, with its traditions of informed consent and
decision makers’ valued outcomes in comprehensible form and shared decision making (22–24, 28), offers examples that might
be adapted to other settings. For example, to secure patient in-
accessible places. Existing research is the natural source of initial
put to its benefit–risk framework (30, 31), FDA created the
guidance for accomplishing those tasks (1–4, 41–43, 54). Voice of the Patient Initiative (56), with daylong exchanges on
However good the research, communicators must still consult critical issues (e.g., chronic fatigue syndrome, sickle cell disease)—
with their external stakeholders. It is unrealistic to expect them a model that some patient groups have adopted. The desire to
to know how people very different from themselves view their include patient experiences in clinical trials led to including self-
world or the communicators (8, 41, 55). Even if those consul- reported quality-of-life measures as outcomes. However, the result
tations only affirm what the research says, they are important as was a proliferation of measures with varying content and quality
“speech acts.” They show respect for the stakeholders as indi- that undermined the research effort. In response, NIH created an
viduals worthy of knowing and hearing. They need to occur inventory of psychometrically validated measures, freely available
throughout the process to maintain the human contact and so online, with adaptive testing for efficient administration (20, 21).
that communicators know what is on stakeholders’ minds and Recognizing the importance of evaluating communications,
stakeholders know what communicators are doing (1, 9, 35, 39). Communicating Risks and Benefits: An Evidence-Based User’s
However, making that happen can be challenging, especially with Guide published by FDA (43) ends each chapter with guidance
diverse, dispersed, and disinterested publics. on evaluation for no resources, modest resources, and resources
4 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1805863115 Fischhoff
COLLOQUIUM
PAPER
Discipline
= Astrophysics
= Environmental Chemistry
= Software Engineering
= Electrical Engineering
= Climate Change
= Meteorology
= (Bio)Geochemistry
= Paleoecology
= Enviro Sci Eng Policy
= Statistics
= Geographic Info Systems
= Enviro Soc Sci & Policy
= Resource Economics
Network Measures
Density = 63%
Cohesion = 1.4
Ave. Centrality = 11
SUSTAINABILITY
CENTER 1 CRN-T: shows all “close and collegial” connections by DISCIPLINE/FIELD based on responses to:
SCIENCE
“Please indicate the strength of your relationship with other center affiliates.”
Fig. 2. Network diagram of self-reported close and collegial relationships among members of an interdisciplinary research program supported by the
National Academy of Sciences. Reprinted with permission from ref. 49.
commensurate with the personal, economic, and political stakes facilitate commissioning analyses from professionals when
riding on good communication. It simplest method is the think- circumstances warrant and resources allow (59).
aloud protocol, asking people to say whatever comes into their These methods all assume a world where, in the words of
minds as they read draft materials (57, 58). Communicating Science Effectively: A Research Agenda (1), “re-
Any communication that goes to a broader audience consti- searchers and practitioners. . .form partnerships” and “re-
tutes an indirect consultation, as recipients assess their rela- searchers in diverse disciplines. . .work together” (ref. 1, p. 9).
tionships with its source, based on what they infer about its There are precedents for creating boundary organizations
competence and trustworthiness. Fig. 3 shows section headings hospitable to such partnerships (14–16). The American Soldier
from an attempt to create a relationship that neither abandons project during World War II brought together social scientists
recipients nor provides unsupportable advice. It was written and practitioners (60). The Medical Research Council Applied
when editors of a journal (American Psychologist) refused to Psychology Unit did the same in the United Kingdom (61), as
publish a review that was critical of much customary advice have the Department of Homeland Security Centers of Ex-
without providing an alternative. It was my hope that it would be cellence (https://www.dhs.gov/science-and-technology/centers-
seen as respecting and empowering recipients. excellence). The latest US National Climate Assessment was
developed in consultation with stakeholders (e.g., in agriculture
Conclusion
Science communications succeed when recipients make better
decisions. Applying that standard means evaluating the opti-
mality of choices made with and without the communications.
With complex decisions and diverse decision makers, such
evaluation is typically infeasible. The alternative is asking how
well the communication process has followed a theory of change.
The present proposal offers a theory of change embracing a
systems approach as advocated by Communicating Science Ef-
fectively: A Research Agenda (1). It entails staffing with the right
people, internal consultation among them, and external consul-
tation with those whom they seek to serve. It embraces both the
bounded rationality of disciplinary scientists and the satisficing of
practitioners.
Its proposed procedures rely on simplified versions of scien-
tific methods adapted for use by organizations with limited re-
sources. They include think-aloud protocols, network analyses,
and qualitative formal analyses, precise enough to allow
quantitative analysis were data requirements met, but not
requiring it. The proposal assumes that anyone can create,
critique, and discuss a decision space with options and valued
outcomes; an influence diagram with the factors determining
those outcomes (Fig. 1); and a social network depicting work
relations (Fig. 2). Adopting science-like methods should also Fig. 3. Advice for reducing the risk of sexual assault (36).
1. National Research Council (2017) Communicating Science Effectively: A Research 34. Taplin DH, Clark H (2012) Theory of change basics (ActKnowledge, New York).
Agenda (National Academy Press, Washington, DC). Available at www.theoryofchange.org/wp-content/uploads/toco_library/pdf/
2. Fischhoff B, Scheufele D (2013) The science of science communication. Proc Natl Acad ToCBasics.pdf. Accessed June 8, 2018.
Sci USA 110(Suppl 3):14033–14039. 35. Dietz T, Stern P, eds (2008) Public Participation in Environmental Assessment and
3. Fischhoff B, Scheufele D (2014) The science of science communication II. Proc Natl Decision Making (National Academy Press, Washington, DC).
Acad Sci USA 111(Suppl 4):13583–13584. 36. Fischhoff B (1992) Giving advice. Decision theory perspectives on sexual assault. Am
4. Fischhoff B, Scheufele D (2018) The science of science communication III. Proc Natl Psychol 47:577–588.
Acad Sci USA, in press. 37. Fischhoff B, Furby L, Morgan M (1987) Rape prevention: A typology and list of
5. Simon HA (1947) Administrative Behavior (Macmillan, New York). strategies. J Interpers Violence 2:292–308.
6. Simon HA (1956) Rational choice and the structure of the environment. Psychol Rev 38. Furby L, Fischhoff B, Morgan M (1991) Rape prevention and self-defense: At what
63:129–138. price? Womens Stud Int Forum 14:49–62.
7. Kahneman D, Klein G (2009) Conditions for intuitive expertise: A failure to disagree. 39. Fischhoff B (2013) The sciences of science communication. Proc Natl Acad Sci USA 110:
Am Psychol 64:515–526. 14033–14039.
8. Medin D, Ojalehto B, Marin A, Bang M (2017) Systems of (non-)diversity. Nat Hum 40. Morgan MG (2017) Theory and Practice in Policy Analysis (Cambridge Univ Press, New
Behav 1:0088. York).
9. Dietz T (2013) Bringing values and deliberation to science communication. Proc Natl 41. Nickerson RA (1999) How we know—And sometimes misjudge—What others know:
Acad Sci USA 110:14081–14087. Imputing our own knowledge to others. Psychol Bull 125:737–759.
10. Scheufele DA (2013) Communicating science in social settings. Proc Natl Acad Sci USA 42. Breakwell GM (2014) The Psychology of Risk (Cambridge Univ Press, Cambridge, UK),
110:14040–14047. 2nd Ed.
11. Scheufele DA (2014) Science communication as political communication. Proc Natl 43. Fischhoff B, Brewer N, Downs JS, eds (2011) Communicating Risks and Benefits: An
Acad Sci USA 111:13585–13592. Evidence-Based User’s Guide (Food and Drug Administration, Washington DC).
12. Ashby WR (1956) An Introduction to Cybernetics (Chapman & Hall, London). 44. Bruine de Bruin WB, Güvenç U, Fischhoff B, Armstrong CM, Caruso D (2009) Com-
13. Cohen WA, Levinthal DA (1990) Absorptive capacity: A new perspective on science municating about xenotransplantation: Models and scenarios. Risk Anal 29:
and innovation. Adm Sci Q 35:128–152. 1105–1115.
14. Bidwell D, Dietz T, Scavia D (2013) Fostering knowledge networks for climate adap- 45. Burns WJ, Clemen RT (1993) Covariance structure models and influence diagrams.
tation. Nat Clim Change 3:610–611. Manage Sci 39:816–834.
15. Guston DH (2001) Boundary organizations in environmental policy and science: An 46. Bruine De Bruin W, Fischhoff B, Brilliant L, Caruso D (2006) Expert judgments of
introduction. Sci Technol Human Values 26:399–408. pandemic influenza risks. Glob Public Health 1:178–193.
16. Parker JN, Crona BI (2012) On being all things to all people: Boundary organizations 47. Carley K, Prietula MJ (1994) Computational Organization Theory (Lawrence Erlbaum,
and the contemporary research university. Soc Stud Sci 42:262–289. Hillsdale, NJ).
17. Thaler R, Sunstein C (2009) Nudge: Improving Decisions About Health, Wealth and 48. Moreno JL (1951) Sociometry, Experimental Method and the Science of Society
Happiness (Yale Univ Press, New Haven, CT). (Beacon House, Boston).
18. McCartney M (2017) Margaret McCartney: When organ donation isn’t a donation. 49. Hybrid Vigor Institute (2003) A multi-method analysis of the social and technical
BMJ 356:j1028. conditions for interdisciplinary collaboration (Hybrid Vigor Institute, San Francisco).
19. Schwartz A, Bergus G (2008) Medical Decision Making (Cambridge Univ Press, New Available at hybridvigor.net/interdis/pubs/hv_pub_interdis-2003.09.29.pdf. Accessed
York). May 10, 2018.
20. Cella D, et al.; PROMIS Cooperative Group (2007) The patient-reported outcomes 50. Brossard D, Belluck P, Gould F, Wirz CD (2018) Promises and perils of gene drives:
measurement information system (PROMIS): Progress of an NIH Roadmap cooperative Navigating the communication of complex, post-normal science. Proc Natl Acad Sci
group during its first two years. Med Care 45(Suppl 1):S3–S11. USA in press.
21. Health Measures (2018) Comprehensive measurement systems. Available at www. 51. Hancock PA, Nourbakhsh I, Stewart J (2018) On the future of transportation in an
healthmeasures.net/index.php/. Accessed October 24, 2018. era of automated and autonomous vehicles. Proc Natl Acad Sci USA, 10.1073/
22. The Ottawa Hospital Research Institute (2017) Patient decision aids. Available at pnas.1805770115.
https://decisionaid.ohri.ca/. Accessed October 24, 2018. 52. Davis GF (2018) How to communicate large-scale social challenges: The problem of
23. Basu A, Meltzer D (2007) Value of information on preference heterogeneity and in- the disappearing American corporation. Proc Natl Acad Sci USA, 10.1073/
dividualized care. Med Decis Making 27:112–127. pnas.1805867115.
24. Dewitt B, Davis A, Fischhoff B, Hanmer J (2017) An approach to reconciling competing 53. Bruine de Bruin W, Morgan MG (2018) Reflections on an interdisciplinary collabora-
ethical principles in aggregating heterogeneous health preferences. Med Decis tion to inform public understanding of climate change, mitigation, and impacts. Proc
Making 37:647–656. Natl Acad Sci USA, 10.1073/pnas.1803726115.
25. von Winterfeldt D (2013) Bridging the gap between science and decision making. 54. Kahneman D (2011) Thinking, Fast and Slow (Farrar Giroux and Strauss, New York).
Proc Natl Acad Sci USA 110:14055–14061. 55. Fiske ST, Dupree C (2014) Gaining trust as well as respect in communicating to mo-
26. Fischhoff B (2015) The realities of risk-cost-benefit analysis. Science 350:aaa6516. tivated audiences about science topics. Proc Natl Acad Sci USA 111:13593–13597.
27. Versteegh MM, Brouwer WBF (2016) Patient and general public preferences for 56. Food and Drug Administration (2018) The voice of the patient: A series of reports
health states: A call to reconsider current guidelines. Soc Sci Med 165:66–74. from FDA’s Patient-Focused Drug Development initiative. Available at https://
28. Barnato AE (2017) Challenges in understanding and respecting patients’ preferences. www.fda.gov/ForIndustry/UserFees/PrescriptionDrugUserFee/ucm368342.htm. Ac-
Health Aff (Millwood) 36:1252–1257. cessed October 24, 2018.
29. Fischhoff B, Bruine de Bruin W, Guvenc U, Caruso D, Brilliant L (2006) Analyzing di- 57. Ericsson A, Simon HA (1990) Verbal Reports as Data (MIT Press, Cambridge, MA).
saster risks and plans: An avian flu example. J Risk Uncertain 33:131–149. 58. Merton RK (1987) The focussed interview and the focus group. Public Opin Q 51:
30. Food and Drug Administration (2013) Structured Approach to Benefit-Risk Assess- 550–566.
ment for Drug Regulatory Decision Making. Draft PDUFA V Implementation Plan. 59. National Research Council (2011) Intelligence Analysis for Tomorrow (National
FY2013–2017 (Food and Drug Administration, Washington, DC). Academy Press, Washington, DC).
31. Food and Drug Administration (2018) Benefit-Risk Assessment in Drug Regulatory 60. Lazarsfeld PF (1949) The American Soldier: An expository review. Public Opin Q 13:
Decision Making. Draft PDUFA VI Implementation Plan. FY2018–2022 (Food and Drug 377–404.
Administration, Washington, DC). 61. Reynolds LA, Tansey EM, eds (2003) The MRC Applied Psychology Unit. Wellcome
32. Fischhoff B (2017) Breaking ground for psychological science: The U.S. Food and Drug Witnesses to Twentieth Century Medicine (Wellcome Trust Centre for the History of
Administration. Am Psychol 72:118–125. Medicine at UCL, London), Vol 16.
33. Community Tool Box (2018) Learn a skill: Table of contents. Available at https://ctb.ku. 62. Lupia A (2013) Communicating science in politicized environments. Proc Natl Acad Sci
edu/en/table-of-contents. Accessed October 24, 2018. USA 110:14048–14054.
6 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1805863115 Fischhoff