Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=bher20 Download by: [University of Arizona] Date: 29 February 2016, At: 14:36 Human and Ecological Risk Assessment: An International Journal ISSN: 1080-7039 (Print) 1549-7860 (Online) Journal homepage: http://www.tandfonline.com/loi/bher20 Methods for Communicating the Complexity and Uncertainty of Oil Spill Response Actions and Tradeoffs Ann Bostrom, Susan Joslyn, Robert Pavia, Ann Hayward Walker, Kate Starbird & Thomas M. Leschine To cite this article: Ann Bostrom, Susan Joslyn, Robert Pavia, Ann Hayward Walker, Kate Starbird & Thomas M. Leschine (2015) Methods for Communicating the Complexity and Uncertainty of Oil Spill Response Actions and Tradeoffs, Human and Ecological Risk Assessment: An International Journal, 21:3, 631-645, DOI: 10.1080/10807039.2014.947867 To link to this article: http://dx.doi.org/10.1080/10807039.2014.947867 Accepted author version posted online: 29 Jul 2014. Submit your article to this journal Article views: 176 View related articles View Crossmark data Citing articles: 2 View citing articles
Transcript
Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=bher20
Download by: [University of Arizona] Date: 29 February 2016, At: 14:36
Human and Ecological Risk Assessment: An InternationalJournal
Methods for Communicating the Complexity andUncertainty of Oil Spill Response Actions andTradeoffs
Ann Bostrom, Susan Joslyn, Robert Pavia, Ann Hayward Walker, KateStarbird & Thomas M. Leschine
To cite this article: Ann Bostrom, Susan Joslyn, Robert Pavia, Ann Hayward Walker, KateStarbird & Thomas M. Leschine (2015) Methods for Communicating the Complexityand Uncertainty of Oil Spill Response Actions and Tradeoffs, Human and Ecological RiskAssessment: An International Journal, 21:3, 631-645, DOI: 10.1080/10807039.2014.947867
To link to this article: http://dx.doi.org/10.1080/10807039.2014.947867
Accepted author version posted online: 29Jul 2014.
Methods for Communicating the Complexityand Uncertainty of Oil Spill Response Actionsand Tradeoffs
Ann Bostrom,1 Susan Joslyn,2 Robert Pavia,3 Ann Hayward Walker,4 Kate Starbird,5
and Thomas M. Leschine6
1Daniel J. Evans School of Public Affairs, University of Washington, Seattle, WA,USA; 2Department of Psychology, University of Washington, Seattle, WA, USA;3School of Marine and Environmental Affairs, University of Washington, Seattle,WA, USA; 4SEA Consulting Group, Cape Charles, VA, USA; 5Human CenteredDesign & Engineering, University of Washington, Seattle, WA, USA; 6HumanDimensions of the Environment and the School of Marine and EnvironmentalAffairs, University of Washington, Seattle, WA, USA
ABSTRACTComplexity and uncertainty influence opinions, beliefs, and decisions about
health, safety, and other kinds of risk, as demonstrated in research on health, climatechange, storm forecasts, pandemic disease, and in other domains. Drawing from thisresearch, this article summarizes insights into how people understand and processuncertain or complex information and explores key oil spill and oil spill response-relevant issues regarding the communication of complexity and uncertainty—fromthe presentation of uncertainties around forecast parameters to the deployment ofonline oil spill response simulation tools. Recommended practices from this arti-cle include (a) to continue to develop and evaluate interactive Web-based oil spillresponse simulations to help users explore tradeoffs in response decisions, (b) totake how people simplify information into account in designing communicationsprocesses and products (and evaluate), (c) to provide numbers along with verbalprobability descriptions, and (d) if using graphics, to communicate probability or un-certainty, using simple graphics and testing them, as effects may not be predictableand some kinds of graphics are easier to understand than others, depending oncontext, numeracy, and graphicacy.
Address communication to Ann Bostrom, Daniel J. Evans School of Public Affairs, Universityof Washington, Parrington Hall 327, P.O. Box 353055, Seattle, WA 98195-3055, USA. E-mail:[email protected] versions of one or more of the figures in the article can be found online at www.tandfonline.com/bher.
631
Dow
nloa
ded
by [
Uni
vers
ity o
f A
rizo
na]
at 1
4:36
29
Febr
uary
201
6
A. Bostrom et al.
INTRODUCTION
Making oil spill response plans and decisions requires communicating the com-plexity, uncertainty, and tradeoffs associated with response options, with stakehold-ers and concerned communities. This article describes what we know about strategiesfor these tasks, and concludes with recommendations to promote effective engage-ment of communities and individuals in discussions about oil spill issues, and toimprove response communications about dispersants and oil spills from and be-tween Unified Command (Federal and State On-Scene Coordinators and IncidentCommanders representing a spiller known as the Responsible Party), dispersantdecision-makers from coastal Regional Response Teams (RRTs), and other key stake-holders. Many of these key stakeholders are looked to by elected officials/politiciansand the public for leadership and assurance about oil spill response options.
Communicating about oil spills and oil spill responses involves conveying not onlythe logistics and politics of response decisions and actions, but also the science of oilspills and response options (Machlis and McNutt 2011, p 320). And like all science,the science of oil spills and spill response is inherently uncertain. The complex mixof incident-specific variables and unknown information amplifies these scientificuncertainties in spill situations, especially during the initial emergency phase. Tack-ling this as a risk communication task means acknowledging the uncertainties andcomplexities. Tackling this as a decision support task means providing actionableinformation. Accordingly, this article has two aims. The first is to describe currentand emerging approaches to conveying uncertainty in risk communications, withan eye toward how these approaches are and can be applied to oil spill responsesituations. The second is to explore current approaches to tackling the complexityof oil spill response and the sciences behind it, to assess what is currently done tocommunicate actionable information for oil spill response, and how to improve onthat.
U.S. President Harry Truman famously joked about wanting a one-handedeconomist, since his economic advisor Edwin G. Nourse was always saying “on theone hand” and “on the other.” While considering both pros and cons is nearlyuniversally considered an essential element of thoughtful decision-making, advisingsomeone to consider the pros and cons of action in a crisis has the potential to leadto confusion rather than protective or risk reducing action. The conflict betweensimplicity on the one hand and information accuracy and sufficiency on the otheris a signature of crisis and emergency communications. Further, even in less criticalrisk situations, if the scientific and social bases for making a choice are difficult toconvey or understand, the resulting confusion can be exploited by parties preferringinaction (Freudenberg et al . 2008).
Any focus on actions requires an assessment of the context; what choices arethere and what do decision-makers need to know in order to act effectively? Oil spillresponders face technical decisions and occupational health challenges, in contrastto consumers who face economic and personal health choices. Even within suchcategories, levels of expertise and knowledge needs will range widely, with corre-sponding variation in prior knowledge, information processing capacity (Ericssonand Lehmann 1996), and ability to handle uncertainty and complexity in decision-making (Parker and Fischhoff 2005).
632 Hum. Ecol. Risk Assess. Vol. 21, No. 3, 2015
Dow
nloa
ded
by [
Uni
vers
ity o
f A
rizo
na]
at 1
4:36
29
Febr
uary
201
6
Communicating Oil Spill Complexity and Uncertainty
Much of the advice in vogue in oil spill response focuses on crisis and emer-gency situations; this guidance focuses on boiling information down to a handful ofspecifics (e.g., Covello’s advice, as cited in Reynolds and Seeger 2012, ch 1), getting“the facts right” (Reynolds and Seeger 2012, ch 3), and focusing on actions peopleshould take. Likely more important than the notion of boiling information down toa few small sound bites is recognizing the importance of providing enough specificsfor recipients to act on the information in order to effectively mitigate risk (Woodet al. 2012). Further, understanding the nature of the decisions faced is essential forassessing the quality of uncertainty communications. Expressing uncertainties thatare relevant for decision tasks and aligned with decision task demands improves theusefulness and comprehensibility of such communications (Joslyn et al. 2009).
COMMUNICATING UNCERTAINTY
Uncertainty encompasses a wide range of states, including lack of knowledge(epistemic or model uncertainty fall into this category), natural variability (alsocalled aleatory uncertainty) (Morgan and Henrion 1992; Eiser et al. 2012), ambiguity(lack of precision or clarity), and ignorance (Smithson 1989). While experts inthe field sometimes distinguish carefully among these (or types of uncertainties inother taxonomies, e.g., NRC 1994), responses to them share some common features.Aversion to ambiguity and uncertainty is a common finding (Camerer and Weber1992); people try to avoid it. A consequence of this is that people may prefer pointestimates even when they can be construed as misleading; for example, in the caseof providing worst estimates that have extremely low likelihoods. This exemplifieshow focusing on the uncertainty in a situation has the potential to deter appropriateprotective or mitigative actions; for example, if the focus increases the salience ofgains or losses (from a reference point such as the status quo), as people tend tobe risk averse for gains and risk seeking for losses (Kahneman and Tversky 1979;Tversky and Kahneman 1974, 1981; van Schie and van der Pligt 1995). However,suppressing known uncertainties may also be regarded as unethical, although notas unethical as distorting information (Smithson 2008).
While some sources advise acknowledging uncertainty (Reynolds and Seeger2012, p 156), proposals on how to communicate uncertainty in emergency planningand response are often general, focusing on recognizing that there are uncertainties.For example, the U.S. Centers for Disease Control and Prevention’s (CDC’s) Crisis& Emergency Risk Communication (CERC) principle of acknowledging uncertaintyand not overstating what you know is illustrated with this quote:
I want to acknowledge the importance of uncertainty. At the early stages of anoutbreak, there’s much uncertainty, and probably more than everyone wouldlike. Our guidelines and advice are likely to be interim and fluid, subject tochange as we learn more.
—Dr. Richard Besser, CDC Acting Director, H1N1 Press Conference,April 23, 2009. (Reynolds and Seeger 2012, p 14)
Relevant to this discussion is the evidence that “less is more” for information recip-ients who may face challenges understanding numbers, as distilled from empirical
Hum. Ecol. Risk Assess. Vol. 21, No. 3, 2015 633
Dow
nloa
ded
by [
Uni
vers
ity o
f A
rizo
na]
at 1
4:36
29
Febr
uary
201
6
A. Bostrom et al.
studies of the effects of communicating quantitative information and summarizedin the U.S. Food and Drug Administration’s (FDA’s) risk communication handbook(Fagerlin and Peters, ch 6 in Fischhoff et al . 2011). The advice stemming from thisresearch steers clear of communicating specific uncertainties around quantitativeestimates, focusing instead on the key message of keeping innumeracy and humaninformation processing limits in mind, generally by finding simple ways of sum-marizing point estimates. Up to a quarter of the U.S. public may be consideredinnumerate, and innumeracy is associated with greater susceptibility to questionwording and incidental framing, and contextual influences on judgments and deci-sions, but lower sensitivity to meaningful numerical differences (Gibson et al . 2013;Hart 2013; Kleber et al. 2013; Peters and Levin 2008).
There is a rich body of empirical evidence regarding how to communicate un-certainties around numerical estimates that is applicable to oil spill response. Oilamounts, distances, transport rates—many numerical parameters are of interest tooil spill responders and publics concerned about potentially exposed ecosystems,fisheries, or human populations. In oil spills, communicating uncertainty about oilmovement, a topic in which the public has a great interest, is considered important(Beegle-Krause 2001). Experience has shown that it is difficult for even spill responseexperts to interpret uncertainty in the context of response decisions. It is complexenough that the U.S. National Oceanic and Atmospheric Administration (NOAA)publishes an extensive interpretation guide for graphical oil spill trajectories.1
Further, there has been considerable press around communication failures stem-ming from inadequate communications about uncertainties (e.g., oil flow rates fromthe wellhead in the BP Deepwater Horizon oil spill, oil quantities, and so forth,see, e.g ., Spotts 2010). Empirical research suggests that (a) verbal expressions of un-certainty lend themselves to diverse interpretations, depending on context, and soare readily subject to misinterpretation (Budescu et al . 2009; Wallsten and Budescu1995); (b) visual and graphical representations of uncertainty appear in many casesto be more interpretable than numbers, though numbers may be useful for taskswhere they can be applied directly (Shah and Freedman 2011; Shah et al. 2005),even when not understood perfectly (Savelli and Joslyn 2013), and visualizations canin some instances lead to more errors of interpretation than other representations(Savelli and Joslyn 2013); (c) not all visual and graphical representations of uncer-tainty are equal (Ancker et al. 2006; Cuite et al . 2008); some graphical presentationsof risk demonstrably influence risk preferences more than numerical presentations(e.g., bar charts showing relative frequencies of adverse outcomes, Chua et al. 2006;Stone et al. 1997); for example: (d) communicating uncertainty may (i) convey an im-pression of transparency or honesty that promotes good communication (Johnsonand Slovic 1995) and improve decision-making or trust in the information (Joslynand LeClerc 2012; Joslyn et al. 2013); (ii) convey an impression of incompetence(Johnson and Slovic 1995); or (iii) confuse people (Johnson and Slovic 1998); and(e) confirmatory information processing biases promote the tendency to notice andremember information that is consistent with one’s prior beliefs and may in someinstances lead people to select an upper or lower bound estimate that supports their
Communicating Oil Spill Complexity and Uncertainty
prior beliefs, in cases where they have access to a distribution or range of estimates(Viscusi et al . 1991, see also Russo et al. 1996; Savelli and Joslyn 2013). Finally, ev-idence suggests that people prefer to communicate uncertainty verbally, althoughthey prefer numbers when they are on the receiving end (Wallsten et al. 1993a,b).
To illustrate point (c) above, visual representations of uncertainty such as the coneof uncertainty the National Hurricane Center uses for possible hurricane tracks (i.e.,projected track plus uncertainties about the track) are subject to misinterpretationssuch as that the cone represents the area that might be affected (Broad et al. 2007).This type of misinterpretation of uncertainty information has also been called the“deterministic construal error” because it entails construing probabilistic informa-tion as deterministic information (Joslyn et al. 2009; Savelli and Joslyn 2013). Suchmisinterpretations can be explained in part by the concreteness principle: “Con-creteness represents the general notion that a judge or decision maker tends to useonly the information that is explicitly displayed in the stimulus object and will useit only in the form in which it is displayed. Information that has to be stored inmemory, inferred from the explicit display, or transformed tends to be discountedor ignored” (Slovic 1972, p 9). Further, Savelli and Johnson (2013) find that peopletend to ignore definitions when there is a visualization.
As noted above, improvements in decision-making can be achieved when expres-sions of uncertainty align with decision task demands (Joslyn et al . 2009; see alsoHegarty et al. 2010). To exemplify, for freeze warnings where freezing or below (thelow point) is critical to the outcome of the decision, expressing the chance of tem-peratures dropping below 32◦F is more useful and less confusing than expressingthe chance of temperatures being greater than 32◦F. Expressions of uncertainty likethis that align with task demands appear in laboratory experiments to reduce riskaversion (through increased withholding of precautionary action) in some contexts,or increase precautionary action when it is appropriate (i.e., reducing risk seeking)(Joslyn and LeClerc 2011; Savelli and Joslyn 2013). Notably, because graphics suchas confidence intervals can sometimes align with and reinforce misinterpretations,it is important to discover the likely misinterpretations empirically and address themdirectly in uncertainty expressions (Joslyn et al. 2009).
Stemming from these findings are several specific recommendations, in additionto the general recommendation to represent uncertainty:
(a) Include numbers with verbal probability descriptions, if verbal descriptions areused at all. Adding numbers (such as numerical confidence bounds for pointestimates) to verbal descriptions of uncertainty appears to improve meaning-fulness and accuracy, at least marginally, despite that adding numbers involvesincreasing the amount and complexity of the information (Budescu et al. 2012).
(b) Use simple graphics when possible and test their interpretation and effects ondecisions, bearing in mind that some kinds of graphical representations aremore interpretable than others, and that this depends on context, as well asindividual numeracy and graphicacy (Cuite et al. 2008; Savelli and Joslyn 2013;Shah and Freedman 2011; Shah et al. 2005). Some studies suggest that it isbetter to avoid graphics entirely unless those specific graphics have been testedand their interpretability and usefulness shown (Savelli and Joslyn 2013).
Hum. Ecol. Risk Assess. Vol. 21, No. 3, 2015 635
Dow
nloa
ded
by [
Uni
vers
ity o
f A
rizo
na]
at 1
4:36
29
Febr
uary
201
6
A. Bostrom et al.
(c) Be prepared for increased risk aversion and conflict when communicating un-certainties, as uncertainty can change responses to risk (i.e., risk avoidance orrisk-seeking behaviors); uncertainties give people more leeway to make choiceson the basis of their values, which may conflict. One of the benefits of com-municating uncertainty is the explicit acknowledgment of the roles of riskpreferences in decision-making under uncertainty.
(d) Evaluate communications of uncertainty, as effects may not be predictable (seeabove).
COMMUNICATING COMPLEXITY
Simulation and Simplification
Many environmental policy decisions are complex and characterized as persistent,with deep scientific uncertainties, conflicting values, and competing definitions.These characteristics typify oil spill policy and response decisions (Machlis andMcNutt 2011; Webler et al . 2011), fossil fuel transport failures as a class, and coastalregion management problems (Moser et al. 2012). Specific oil spill events exhibitmore uniformity of purpose (i.e., containing or cleaning up spills), but still entaildeep scientific uncertainties and complexity, in part because they involve oceanecosystems.
Complexity in decision-making can refer to the number of decision attributes, thenumber of decision alternatives, or to other characteristics of the information avail-able about the decision, for example, including the social or political complexity ofthe decision context or processes. People prefer a moderate amount of complexity;too little is boring; too much overwhelms (Berlyne 1966). Preferences for complex-ity are context dependent. Fear, for example, can reduce preferences for complexity(Berlyne 1966).
In many circumstances our perceptual, judgment, and decision-making processeswork to reduce complexity (Slovic 1972). By selectively directing, enhancing, orinhibiting those processes, communications can shape their deployment and effectswithin limits, either unintentionally or by design. Examples of this kind of shapinginclude framing effects (Iyengar 1990; Levin et al. 1998; Peters et al. 2006) andanchoring effects (Tversky and Kahneman 1974; Epley and Gilovich 2006).
Strategies for making decisions under complexity include simplification, simula-tion, and partitioning or narrowing the scope of the problem. Simplification canbe done analytically, through modeling, but is also carried out though story telling(Kahneman 2011), or mental modeling (Morgan et al. 2002; Gentner and Stevens1983; Johnson-Laird 1983), which may mean using analogy (Bostrom 2008; Gentnerand Smith 2012), after determining the gist of the problem (on gist extraction, seeAdam and Reyna 2005). Mental models are our “inference engines” for simulatingevents (Bartlett 1932; Craik 1943; Gentner and Stevens 1983). For further discussionof mental models in the context of oil spill response see Bostrom et al . (2015).
Simplification and satisficing are routine in problem solving and decision-making.As put by Simon in 1959 (p 272): “The decision-maker’s model of the world encom-passes only a minute fraction of all the relevant characteristics of the real environ-ment, and his inferences extract only a minute fraction of all the information that
636 Hum. Ecol. Risk Assess. Vol. 21, No. 3, 2015
Dow
nloa
ded
by [
Uni
vers
ity o
f A
rizo
na]
at 1
4:36
29
Febr
uary
201
6
Communicating Oil Spill Complexity and Uncertainty
is present even in his model.” Simon goes on to describe how we quite actively—ifunwittingly—construct these fractions. While one might like to believe that this isintentional screening of situations to see what is relevant, many of these processesstem from how we process information cognitively. Whether they are adaptive pro-cesses depends on the match between the task and the process, which varies. In theensuing decades, research on behavioral decision-making has revealed some of thestructure of these active processes, including mental shortcuts (rules of thumb or“heuristics”), biases (such as the tendency to preferentially process information thatconfirms our prior beliefs), and how these vary as a function of individual differ-ences (e.g., numeracy, expertise or prior knowledge) or context (e.g., stress or timepressure) (Fischhoff 2012; Kahneman 2011; Kahneman and Frederick 2002; Payneet al. 1988, 2008; Peters et al. 2006; Simon 1955, 1956; Tversky and Kahneman 1974;Weber and Johnson 2009). An important manifestation of this is that effort oftentrades off against accuracy in decision-making. For example, when decisions arecomplex—with many attributes to consider and more than two or three alternativesto choose between—attributes may be considered one at a time and alternativeseliminated on that basis rather than according to an expected value calculation orother compensatory decision strategy (Payne 1976; Payne et al. 1992). On the otherhand, training and experience in the field (10,000 hours plus, or about 10 years ofexperience) create highly structured mental models that enable experts to recognizethe essential features of decision situations immediately, which may not be evidentto those with less experience (Chi et al. 1982, 1988; Ericsson and Lehmann 1996;March 1994). Awareness of these attributes of information processing and decision-making is a first step toward using them to improve the design of communicationsprocesses and products.
Story telling manifests itself in oil spill response both through scenario construc-tion and analysis (Leschine et al. 2015), and through narratives evident in socialmedia (Starbird et al. 2015) and responses to open-ended survey questions (Bostromet al. 2015). Stories have the advantage of engaging people more effectively thanstatistical evidence (Beach 2009; Kahneman 2011), and the disadvantage of beingspecific so that they tend to be relevant only metaphorically or by analogy, and eventhen are likely misleading (Kahneman 2011, ch 19).
While use of analogies and metaphors by oil spill response stakeholders remain tobe formally analyzed, analogies are prevalent in oil spill risk communications. Oneexample is an American Petroleum Institute poster that borrows a cake analogy(Figure 1) from Raffi Khatchadourian of the New Yorker (Khatchadourian 2011).Analogies can be extremely useful (Gentner et al. 2001), and even essential tolearning and discovery, but they can invite comparisons that evoke unanticipatedresponses from stakeholders (Johnson 2003; Roth et al. 1990; Slovic et al. 1990; forfurther discussion see Bostrom 2008). As with other forms of risk communication,evaluation is essential (Fischhoff et al. 2011; Fischhoff 2013; Walker et al. 2015).
Systems dynamics research examining mental models of dynamic systems showsthat intuitions often align with simpler, more linear processes and fail to accountadequately for accumulation and feedbacks (Booth et al. 2000, 2007; Moxnes andSaysel 2009; Sterman 2011). Systems dynamics researchers have had some success inimproving people’s intuitions about the performance of nonlinear systems, thoughit has been modest (Moxnes and Saysel 2009; Sterman 2010). Simple linear models
Hum. Ecol. Risk Assess. Vol. 21, No. 3, 2015 637
Dow
nloa
ded
by [
Uni
vers
ity o
f A
rizo
na]
at 1
4:36
29
Febr
uary
201
6
A. Bostrom et al.
Figure 1. Educational resource from Oil Spill Prevention + Response, a websiteof the American Petroleum Institute. Available at http://www.oilspillprevention.org/∼/media/Oil-Spill-Prevention/spillprevention/r-and-d/dispersants/the-role-of-dispersants-in-oil-spill-res.pdf. Reproducedcourtesy of the American Petroleum Institute from Role of Dispersants inOil Spill Response (2013).
generally predict behaviors—such as human performance—as well or better thanexperts do, if the models incorporate those variables identified as key by experts(Dawes 1979). However even analytic linear models may fail nevertheless to explainor predict well, especially for systems such as environmental or ocean ecologicalsystems that are complex and nonlinear (Lorenz 1963; Miles 2009).
Communicating Oil Spill Complexity and Uncertainty
knowledge into Arctic ERMA” (CRRC 2012, p 2). The workshop summary suggeststhat participants found Arctic ERMA promising but suggested several modifications,ranging from a low-bandwidth or stand-alone version (because of the unreliabilityof Web access), to having local people be “shepherds of the project” and includ-ing local and traditional knowledge and data (e.g., land ownership, Inupiaq placenames, BOEM and industry data including real-time high resolution ice observation,locations of staged response equipment, real-time currents and weather, and con-ceptual models, spill scenarios and restoration concept visualization, among othersuggestions) (CRRC 2012).
There are an increasing number of online simulators for complex systems, withwhich interested parties can play directly5 (Demski et al. 2013). In the domain ofwater management, for example, the Desert City Decision Center has developed Wa-terSim on the Web, an online simulation that is now being used in K–12 education.6
The more sophisticated simulations are similar to NOAA’s response simulationsin that they require downloading and installing software first (e.g., Climate Inter-active7 and the Decision Theater version of WaterSim). Recently, easily accessibleWeb-based simulators have proliferated, which suggests there may be a role foran easily accessible simple Web-interface simulator of spill response decisions thatwould illustrate response decision consequences and tradeoffs. One such exampleis the Response Operations Calculator (ROC), an online tool that allows evalua-tion of response cleanup methods.8 ROC allows users to compare combinations ofresponse methods, such as in situ burning, dispersants, and mechanical recovery,under simplified spill scenarios.
DISCUSSION
This article highlights six recommendations to complement those developed inother parts of this project (Bostrom et al. 2015; Leschine et al. 2015; Starbird et al.2015; Walker et al. 2015):
1. Include numbers with any verbal probability descriptions, for example, numericalconfidence bounds with verbal probability terms such as likely;
2. When possible, use simple graphics that have been empirically tested to commu-nicate probability and uncertainty;
3. Be prepared for value conflicts and differences in risk preferences when commu-nicating uncertainties;
4. Evaluate communications of uncertainty, as effects may not be predictable;5. Take how people simplify information into account in designing communications
(and evaluate); and6. Develop interactive Web-based oil spill response simulations that help users ex-
plore tradeoffs in oil spill response decisions.
5http://bit.ly/atmco2, http://bit.ly/stockflow, http://bit.ly/5c4pU3, or http://bit.ly/10Po0ah
These recommendations and the insights identified in this article are informedby behavioral decision research, a field that has only recently begun to achieverecognition in government operations. The Office of Science and Technology Policyin the United States began hiring in 2013 for a behavioral insights team, inspired bythe Behavioral Insights Team (BIT, or “nudge unit” after Thaler and Sunstein 2008)commissioned by UK Prime Minister David Cameron in 2010 (Oullier 2013). Theaim of these efforts is to harness the social and behavioral sciences to enable thedesign of better evidence-based public policies (OMB 2013). Thus, it is an opportunemoment for NOAA and other federal agencies involved in oil spill response tofamiliarize themselves with the empirical research strategies used in behavioraldecision research, and use them to achieve improvements in oil spill responsecommunications and management.
ACKNOWLEDGEMENTS
Gratefully acknowledged are the many insights and contributions of the partic-ipants at the July 24–25, 2013 Response Risk Communication Tools Peer-ReviewWorkshop held at the University of Washington, Seattle, including: Keeley Belva,NOAA Communications and External Affairs, National Ocean Service; DharmaDailey, Doctoral Student, Human Centered Design & Engineering, University ofWashington; Vicki Loe, Communications Coordinator, NOAA Office of Responseand Restoration; Amy Merten, Spatial Data Branch Chief, NOAA Office of Responseand Restoration; Debbie Payton, Chief Emergency Response Division, NOAA Officeof Response and Restoration; Bob Pond, Senior Oil Spill Advisor (retired), USCGHeadquarters; Debbie Scholz, Environmental Specialist, SEA Consulting Group;Richard Sheehe, U.S. Centers for Disease Control and Prevention/Sheehe Group;Emma Spiro, Assistant Professor, Information School, University of Washington;Tyler Scott, Doctoral Student, Daniel J. Evans School of Public Affairs, Universityof Washington; Jeannette Sutton, Senior Research Scientist, Trauma, Health andHazard Center, University of Colorado; Seth Tuler, Research Fellow, Social andEnvironmental Research Institute; Glen Watabayashi, Supervisory Scientist, NOAAOffice of Response and Restoration; Jeffrey Wickliffe, Assistant Professor, School ofPublic Health and Tropical Medicine, Department of Global Environmental HealthSciences, Tulane University. Appreciation also goes to Alicia Ahn, Melinda McPeek,and the Evans School Financial Services staff for their courteous and professionalcontributions to the project, and to the anonymous reviewers who provided addi-tional feedback on the articles.
FUNDING
Funding from the University of New Hampshire’s Coastal Response ResearchCenter (NOAA Grant Number: NA07NOS4630143, Contract: 13-003) is gratefullyacknowledged.
640 Hum. Ecol. Risk Assess. Vol. 21, No. 3, 2015
Dow
nloa
ded
by [
Uni
vers
ity o
f A
rizo
na]
at 1
4:36
29
Febr
uary
201
6
Communicating Oil Spill Complexity and Uncertainty
REFERENCES
Adam MB and Reyna VF. 2005. Coherence and correspondence criteria for rationality: Ex-perts’ estimation of risks of sexually transmitted infections. J Behav Decis Making 18:169–86
Ancker JS, Senathirajah Y, Kukafka R, et al. 2006. Design features of graphs in health riskcommunication: A systematic review. J Am Med Inform Assn 13:608–18
Bartlett FC. 1932. A theory of remembering. In: Remembering: A Study in Experimental andSocial Psychology, pp 197–214. Cambridge University Press, London, UK
Beach LR. 2009. Narrative Thinking and Decision Making: How the Stories We Tell OurselvesShape our Decisions, and Vice Versa. Available at www.LeeRoyBeach.com (accessed July20, 2013)
Beegle-Krause J. 2001. General NOAA oil modeling environment (GNOME): A new spilltrajectory model. In: Proceedings of the 2001 International Oil Spill Conference, pp865–71. Available at http://dx.doi.org/10.7901/2169-3358-2001-2-865. Tampa, FL, USA.March 26–29
Berlyne DE. 1966. Curiosity and exploration. Science, New Series 153:25–33Booth Sweeney L and Sterman J. 2000. Bathtub dynamics: Initial results of a systems thinking
inventory. Syst Dynam Rev 16:249–94Booth Sweeney L and Sterman J. 2007. Thinking about systems: Students’ and their teachers’
conceptions of natural and social systems. Syst Dynam Rev 23:285–312Bostrom A. 2008. Lead is like mercury: Risk comparisons, analogies and mental models. J
Risk Res 11:99–117Bostrom, A, Walker AH, Scott T, et al. 2015. Oil spill response risk judgments, decisions and
mental models: Findings from surveying U.S. stakeholders and coastal residents. HumEcol Risk Assess 21:581–604
Broad K, Leiserowitz T, Weinkle J, et al. 2007. Misinterpretations of the “cone of uncertainty”in Florida during the 2004 hurricane season. B Am Meteorol Soc 88:651–67
Budescu D, Broomell S, and Por HH. 2009. Improving communication of uncertaintyin the reports of the intergovernmental panel on climate change. Psychol Sci 20:299–308
Camerer C and Weber M. 1992. Recent developments in modeling preferences—Uncertaintyand ambiguity. J Risk Uncertainty 5:325–70
Chi MTH, Glaser R, and Rees E. 1982. Expertise in problem solving. In: Sternberg R (ed),Advances in the Psychology of Human Intelligence, pp 17–76. Erlbaum, Hillsdale, NJ, USA
Chi MTH, Glaser R, and Farr M. (eds). 1988. The Nature of Expertise. Erlbaum, Hillsdale,NJ, USA
Chua HF, Yates JF, and Shah P. 2006. Risk avoidance: Graphs versus numbers. Mem Cognition34:399–410
Craik KJW. 1943. The Nature of Explanation. Cambridge University Press, Cambridge, UKCRRC (Coastal Response Research Center). 2012. North Slope Borough Oil Spill Work-
Cuite CL, Weinstein ND, Emmons K, et al. 2008. A test of numeric formats for communicatingrisk probabilities. Med Decis Making 28:377–84
Dawes RM. 1979. The robust beauty of improper linear models in decision making. AmPsychol 34:571–82
Hum. Ecol. Risk Assess. Vol. 21, No. 3, 2015 641
Dow
nloa
ded
by [
Uni
vers
ity o
f A
rizo
na]
at 1
4:36
29
Febr
uary
201
6
A. Bostrom et al.
Demski C, Spence A, and Pidgeon N. 2013. Summary findings of a survey conducted in August2012– Transforming the UK Energy System: Public Values, Attitudes and Acceptability.Working Paper. UKERC, London, UK
Eiser JR, Bostrom A, Burton I, et al. 2012. Risk interpretation and action: A conceptualframework for responses to natural hazards. Internat J Disaster Risk Reduction 1:5–16
Epley N and Gilovich T. 2006. The anchoring-and-adjustment heuristic: Why the adjustmentsare insufficient. Psych Sci 17:311–18
Ericsson KA and Lehmann AC. 1996. Expert and Exceptional Performance: Evidence ofmaximal adaptation to task constraints. Annu Rev Psychol 47:273–305
Fagerlin A and Peters E. 2011. Quantitative Information. In: Fischhoff B, Brewer NT, andDowns JS (eds), Communicating Risks and Benefits: An Evidence-Based User’s Guide, Ch. 7. USDept. of Health and Human Services, Food and Drug Administration, Silver Spring, MD,USA
Fischhoff B. 2012. Judgment and Decision Making. Earthscan, New York, NY, USAFischhoff B. 2013. The sciences of science communication. Proc Natl Acad Sci
USA110(Supplement 3):14033–39Fischhoff B, Brewer NT, and Downs JS (eds). 2011. Communicating Risks and Benefits: An
Evidence-Based User’s Guide. US Dept. of Health and Human Services, Food and DrugAdministration, Silver Spring, MD, USA. Available at http://www.fda.gov/downloads/AboutFDA/ReportsManualsForms/Reports/UCM268069.pdf (accessed July 10, 2013)
Freudenberg WR, Gramling R, and Davidson DJ. 2008. Scientific certainty argumentationmethods (SCAMs): Science and the politics of doubt. Sociol Inq 78:2–38
Gentner D and Smith L. 2012. Analogical reasoning. In: Ramachandran VS (ed), Encyclope-dia of Human Behavior, 2nd edit, pp 130–6. Elsevier, Oxford, UK
Gentner D and Stevens AL (eds). 1983. Mental Models. Lawrence Erlbaum Associates, Hills-dale, NJ, USA
Gentner D, Holyoak KJ, and Kokinov BN (eds). 2001. The Analogical Mind: Perspectivesfrom Cognitive Science. MIT Press, Cambridge, MA, USA
Gibson JM, Rowe A, Stone ER, et al. 2013. Communicating quantitative information aboutunexploded ordnance risks to the public. Environ Sci Technol 47:4004–13
Hart PS. 2013. The role of numeracy in moderating the influence of statistics in climatechange messages. Public Underst Sci 22:785–98
Hegarty M, Canham MS, and Fabrikant SI. 2010. Thinking about the weather: How displaysalience and knowledge affect performance in a graphic inference task. J Exp PsycholLearn 36:37–53
Iyengar S. 1990. Framing responsibility for political issues: The case of poverty. Polit Behav12 Cognition and Political Action:19–40
Johnson BB. 2003. Are some risk comparisons more effective under conflict?: A replicationand extension of Roth et al. Risk Anal 23:767–80
Johnson BB and Slovic P. 1995. Presenting uncertainty in health risk assessment: Initial studiesof its effects on risk perception and trust. Risk Anal 15:485–94
Johnson BB and Slovic P. 1998. Lay views on uncertainty in environmental health risk assess-ment. J Risk Res 1:261–79
Johnson-Laird PN. 1983. Mental Models. Cambridge University Press, Cambridge, UKJoslyn S and LeClerc J. 2012. Uncertainty forecasts improve weather related decisions and
attenuate the effects of forecast error. J Exp. Psychol-Appl 18:126–40Joslyn S, Nadav-Greenberg L, Taing MU, et al. 2009. The effects of wording on the understand-
ing and use of uncertainty information in a threshold forecasting decision. Appl CognitivePsych 23:55–72
642 Hum. Ecol. Risk Assess. Vol. 21, No. 3, 2015
Dow
nloa
ded
by [
Uni
vers
ity o
f A
rizo
na]
at 1
4:36
29
Febr
uary
201
6
Communicating Oil Spill Complexity and Uncertainty
Joslyn S, Nemec L, and Savelli S. 2013. The benefits and challenges of predictive intervalforecasts and verification graphics for end users. Weather, Climate, and Society 5:133–47
Kahneman D. 2011. Thinking, Fast and Slow. Farrar, Straus, and Giroux, New York, NY, USAKahneman D and Frederick S. 2002. Representativeness revisited: Attribute substitution in
intuitive judgment. In: Gilovich T, Griffin D, and Kahneman D (eds), Heuristics andBiases: The Psychology of Intuitive Judgment, pp 49–81. Cambridge University Press, NewYork, NY, USA
Kahneman D and Tversky A. 1979. Prospect theory: An analysis of decision under risk.Econometrica 47:263–91
Khatchadourian R. 2011. A reporter at large: The Gulf War—Were there any heros in the BPoil disaster? The New Yorker, March 14, 2011. Available at http://www.newyorker.com/reporting/2011/03/14/110314fa fact khatchadourian (accessed July 10, 2013)
Kleber J, Dickert S, Peters E, et al. 2013. Same numbers, different meanings: How numer-acy influences the importance of numbers for pro-social behavior. J Exp Soc Psychol49:699–705
Leschine T, Pavia R, Walker AH, et al. 2015. What-if scenario modeling to support oil spillpreparedness and response decision making. Hum Ecol Risk Assess 21:646–66
Levin IP, Schneider SL, and Gaeth GJ. 1998. All frames are not created equal: A typology andcritical analysis of framing effects. Organ Behav Hum Dec 76:149–88
Lorenz E. 1963. Deterministic nonperiodic flow. J Atmos Sci 20:130–41Machlis GE and McNutt MK. 2011. Ocean policy: Black swans, wicked problems, and science
during crises. Oceanography 24:318–20March JG. 1994. A Primer on Decision Making: How Decisions Happen. The Free Press, New
York, NY, USAMiles EL. 2009. On the increasing vulnerability of the world ocean to multiple stresses. Annu
Rev Env Resour 24:17–41Morgan MG and Henrion M. 1992. Uncertainty: A Guide To Dealing with Uncertainty in
Quantitative Risk and Policy Analysis. Cambridge University Press, New York, NY, USAMorgan MG, Fischhoff B, Bostrom A, et al. 2002. Risk Communication: A Mental Models
Approach. Cambridge University Press, New York, NY, USAMoser SC, Jeffress Williams S, and Boesch DF. 2012. Wicked challenges at land’s end: Man-
aging coastal vulnerability under climate change. Annu Rev Env Resour 37:51–78Moxnes E and Saysel AK. 2009. Misperceptions of global climate change: Information policies.
Climatic Change 93:15–37NRC (National Research Council). 1994. Science and Judgment in Risk Assessment. The
National Academies Press, Washington, DC, USAOMB (Executive Office of the President, Office of Management and Budget). 2013. Memo-
randum M-13-17 to Heads of Departments and Agencies from Sylvia M. Burwell, CeciliaMunoz, John Holder and Alan Krueger on next steps in the evidence and innovationagenda. Washington, DC, USA. Dated July 26, 2013. Available at http://www.whitehouse.gov/sites/default/files/omb/memoranda/2013/m-13-17.pdf
Oullier O. 2013. Behavioural insights are vital to policy-making. Nature 501:463Parker AM and Fischhoff B. 2005. Decision making competence: External validation through
an individual-differences approach. J Behav Decis Making 18:1–27Payne JW. 1976. Task complexity and contingent processing in decision making: An informa-
tion search and protocol analysis. Organ Behav Hum Perf 16:366–87Payne JW, Bettman JR, and Johnson EJ. 1988. Adaptive strategy selection in decision making.
J Exp Psychol Learn 14:534–52Payne JW, Bettman JR, Coupey E, et al. 1992. A constructive process view of decision making:
Multiple strategies in judgment and choice. Acta Psychol 80:107–41
Hum. Ecol. Risk Assess. Vol. 21, No. 3, 2015 643
Dow
nloa
ded
by [
Uni
vers
ity o
f A
rizo
na]
at 1
4:36
29
Febr
uary
201
6
A. Bostrom et al.
Payne JW, Samper A, Bettman JR, et al. 2008. Boundary conditions on unconscious thoughtin complex decision making. Psychol Sc 19:1118–23
Peters E and Levin IP. 2008. Dissecting the risky-choice framing effect: Numeracy as anindividual-difference factor in weighting risky and riskless options. Judgm Decis Mak3:435–48
Peters E, Vastfjall D, Slovic P, et al. 2006. Numeracy and decision making. Psychol Sci 17:407–13Reynolds B and Seeger M. 2012. Crisis and Emergency Risk Communications. Centers for
Disease Control and Prevention (CDC), Atlanta, GA, USARoth E, Morgan MG, Fischhoff B, et al. 1990. What do we know about making risk comparisons?
Risk Anal 10:375–87Russo JE, Medvec VH, and Meloy MG. 1996. The distortion of information during decisions.
Organ Behav Hum Dec 66:102–10Savelli S and Joslyn S. 2013. The advantages of predictive interval forecasts for non-expert
users and the impact of visualizations. Appl Cognitive Psych 27:527–41Shah P and Freedman EG. 2011. Bar and line graph comprehension: An interaction of
top-down and bottom-up processes. Top Cogn Sci 3:560–78Shah P, Freedman EG, and Vekiri I. 2005. The comprehension of quantitative information
in graphical displays. In: Shah P and Miyake A (eds), The Cambridge Handbook ofVisuospatial Thinking, pp 426–76. Cambridge University Press, New York, NY, USA
Simon HA. 1955. A behavioral model of rational choice. Q J Econ 59:99–118Simon HA. 1956. Rational choice and the structure of the environment. Psychol Rev 63:129–38Simon HA. 1959. Theories of decision-making in economics and behavioral science. American
Econ Rev 49:253–283Slovic P. 1972. Oregon Research Institute from Shakespeare to Simon: Speculations—and
some evidence—about man’s ability to process information. Research Bulletin 12(2).Available at http://www.decisionresearch.org/pdf/dr36.pdf (accessed January 2, 2014)
Slovic P, Kraus N, and Covello VT. 1990. What should we know about making risk compar-isons? Risk Anal 10:389–92
Smithson M. 1989. Ignorance and Uncertainty: Emerging Paradigms. Springer Verlag, NewYork, NY, USA
Smithson M. 2008. The many faces and masks of uncertainty. In: Bammer G and Smithson M(eds), Uncertainty and Risk: Multidisciplinary Perspectives, pp 13–25. Earthscan, London,UK
Spotts P. 2010. In finding BP oil spill flow rate, lab science meets real world. Christian ScienceMonitor, June 11
Starbird K, Dailey D, Walker AH, et al. 2015. Social media, public participation and the 2010BP Deepwater Horizon oil spill. Hum Ecol Risk Assess 21:605–30
Sterman J. 2010. Does formal system dynamics training improve people’s understanding ofaccumulation? Syst Dynam Rev 26:316–34
Sterman J. 2011. Communicating climate change risks in a skeptical world. Climatic Change108:811–26
Stone ER, Yates JF, and Parker AM. 1997. Effects of numerical and graphical displays onprofessed risk-taking behavior. J Exp Psychol-Appl 3:243–56
Thaler RH and Sunstein CR. 2008. Nudge: Improving Decisions about Health, Wealth, andHappiness. Yale University Press, New Haven, CT, USA
Tversky A and Kahneman D. 1974. Judgment under uncertainty: Heuristics and biases. Science185:1124–30
Tversky A and Kahneman D. 1981. The framing of decisions and the psychology of choice.Science 211:453–8
644 Hum. Ecol. Risk Assess. Vol. 21, No. 3, 2015
Dow
nloa
ded
by [
Uni
vers
ity o
f A
rizo
na]
at 1
4:36
29
Febr
uary
201
6
Communicating Oil Spill Complexity and Uncertainty
U.S. Department of the Interior. 1963. Federal Water Pollution Control Administration Policyon the Use of Chemicals to Treat Floating Oils. July 5. Washington, DC, USA
U.S. Environmental Protection Agency. 1994. 40 CFR Parts 9 ad 300. National Oil and Haz-ardous Substances Pollution Contingency Plan; Final Rule. September 15. Washington,DC, USA
Van Schie ECM and van der Pligt J. 1995. Influencing risk preference in decision making:The effects of framing and salience. Organ Behav Hum Dec 63:264–75
Viscusi WK, Magat WA, and Huber J. 1991. Communication of ambiguous risk information.Theor Decis 31:159–73
Walker AH, Pavia R, Bostrom A, et al. 2015. Communication practices for oil spills: Stakeholderengagement during preparedness and response. Hum Ecol Risk Assess 21:667–90
Wallsten TS and Budescu DV. 1995. A review of human linguistic probability processing:General principles and empirical evidence. The Knowledge Engineering Rev 10:43–62
Wallsten TS, Budescu DV, and Zwick R. 1993a. Comparing the calibration and coherence ofnumerical and verbal probability judgments. Manage Sci 39:176–90
Wallsten TS, Budescu DV, Zwick R, et al. 1993b. Preferences and reasons for communicatingprobabilistic information in verbal or numerical terms. Bull Psychonomic Soc 31:135–8
Weber EU and Johnson EJ. 2009. Mindful judgment and decision making. Annu Rev Psychol60:53–85
Webler T, Tuler S, and Dietz T. 2011. Modellers’ and outreach professionals’ views on therole of models in watershed management. Environ Policy Gov 21:472–86
Wood MM, Mileti DS, Kano M, et al. 2012. Communicating actionable risk for terrorism andother hazards. Risk Anal 32:601–15
