Hazardous Heuristics Heuristics and Biases: The Psychology of Intuitive Judgement by Thomas Gilovich; Dale W. Griffin; Daniel Kahneman Review by: Cass R. Sunstein The University of Chicago Law Review, Vol. 70, No. 2 (Spring, 2003), pp. 751-782 Published by: The University of Chicago Law Review Stable URL: http://www.jstor.org/stable/1600596 . Accessed: 28/05/2013 13:35 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. . The University of Chicago Law Review is collaborating with JSTOR to digitize, preserve and extend access to The University of Chicago Law Review. http://www.jstor.org This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions HazardousHeuristics Cass R. Sunsteint Thomas Judgement, ofIntuitive Heuristics andBiases:ThePsychology 2002. Dale W.Griffin, eds.Cambridge, & Daniel Kahneman, Gilovich, Pp xvi,857. In the early1970s,Daniel Kahnemanand Amos Tverskypropapersaboutdecisionsunderuncerduced a seriesof pathbreaking "peoplerelyon Theirclaimwas thatin assessingprobabilities, tainty.' principleswhichreducethe complex a limitednumberof heuristic valuesto simplerjudgand predicting tasksof assessingprobabilities didnotarguethatitis irKahnemanandTversky mentaloperations."2 they On thecontrary, rationalforpeopleto use therelevantheuristics. are quitevaluable.The claimedthatas a generalrule,the heuristics erproblemis thatin somecases,theylead "to severeand systematic theword"systematic." One ofthemost It is worthemphasizing rors."3 was thattheerrorswerenotranfeaturesof theirargument striking dom-theycouldbe describedandevenpredicted. in manyfields, have provedinfluential The resulting arguments to connect stemsfromtheeffort including law,4wheretheinfluence of ratherthanhypothetical, understanding legalanalysisto a realistic, The University of ofJurisprudence, ServiceProfessor t KarlN. Llewellyn Distinguished LiorStrahilevEricPosner, RichardPosner, Chicago.ThankstoReidHastie,DanielKahneman, research CarynCampbellprovidedexcellent itz,andAdrianVermeuleforvaluablecomments. assistance. Paul Slovic,andAmosTversky, eds, 1 Thekeypaperscan be foundin DanielKahneman, and Biases(Cambridge earlystudies underUncertainty: Heuristics 1982)(presenting Judgment literature shouldbe inthefaceofuncertainty). Theheuristics-and-biases abouthumanjudgment onprospect whichinvolves thenatureofpeople'sutility fromtheliterature theory, distinguished of uncertainty. See underconditions of risk,notmentalshortcuts underconditions functions andFrames(Cambridge andAmosTversky, 2001)(presenting DanielKahneman Choices, Values, andrelatedwork). recentstudiesaboutprospect theory 2 and BiunderUncertainty: Heuristics and Daniel Kahneman, AmosTversky Judgment underUncertainty 3, 3 (citedin note1) eds,Judgment Slovic,andTversky, ases,in Kahneman, forth thebasicfindings). (setting 3 Id. 4 See, forexample,Christine and RichardThaler,A Behavioral Jolls,Cass R. Sunstein, of 50 StanL Rev 1471,1518-19(1998)(exploring implications ApproachtoLaw andEconomics, biasesfordemandforenvironmental RogerG.Noll especially Superfund); regulation, judgment 19 JLegal forRiskRegulation, andJamesE. Krier,SomeImplications ofCognitive Psychology ofhealthand fortheregulation ofcognitive Stud747 (1990)(exploring psychology implications environmental risks). 751 This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 752 The University of ChicagoLaw Review [70:751 howhumanbeingsthinkandbehave.Ifhumanbeingsuse identifiable heuristics, andiftheymakesystematic errors, we mightbetterunderstandwhylaw is as it is,and we mightgeneratebetterstrategies for thatlaw actuallypromotessocialgoals.Mostprovocatively, ensuring an understanding ofheuristics andbiasesshouldimproveourunderstanding of thelegitimate roleofpaternalism in law.If peoplemake systematic errors,perhapsgovernment has,more oftenthanantipaternalists think, goodreasontooverride theirchoices. Theinfluence oftheheuristics-and-biases literature also stemsfromitsobviousconnectionwithparticular withwhichlawyers problems andpolicymakers are concerned. For example,thesystemof riskregulation has been said to showa combination of "paranoiaand neglect."5 An underofsystematic standing errorsmight helpshowhowandwhythisis so, andgivea senseofwhatmight be donebywayofresponse. KahnemanandTversky threegeneral-purpose emphasized heuristics: representativeness, and anchoring. availability, The availability heuristic has probablybecomethe mostwell-known in law.6When peopleuse thisheuristic, theyanswera questionofprobability byaskingwhether examplescomereadilytomind.7 How likelyis a flood,an airplanecrash,a traffic jam,a terrorist ora disaster attack, at a nuclear powerplant?Lackingstatistical knowledge, peopletryto thinkofillustrations. Thus,"a classwhoseinstances areeasilyretrieved willappearmorenumerous thana classofequal frequency whoseinstances arelessretrievable."8 Forpeoplewithout statistical knowledge, itis far fromirrational touse theavailability heuristic; theproblem is thatthis heuristic can lead to seriouserrorsof fact,in theformof excessive fearofsmallrisksandneglectoflargeones. KahnemanandTversky also suggested thatin thefaceofuncerestimatesare oftenmade froman initialvalue,or "anchor," tainty, whichis thenadjustedto producea finalanswer.'0 The initialvalue seemsto haveundueinfluence. Whatpercentage ofAfrican countries 5 JohnD. Graham,MakingSense of Risk:An Agendafor Congress,in RobertHahn, ed, and Lives Saved: GettingBetterResultsfromRegulation183,183 (Oxford1996) Risks,Benefits, (proposingmorebalanced approachestowardriskregulation). 6 A Lexis searchoflaw reviewsfoundwellover200 referencesto theavailabilityheuristic (as ofFebruarv2003). 7 See Tverskyand Kahneman,Judgment underUncertainty at 3, 11-14 (cited in note 2) (discussingtheavailabilityheuristic). 8 Idat11. 9 See TimurKuran and Cass R. Sunstein,AvailabilityCascades and Risk Regulation.51 Stan L Rev 683, 703-05 (1999) (exploringhow the availabilityheuristicand otherbiases "fuel mass delusionsthathave largeconsequencesforregulatorypolicy");Noll and Krier,19 J Legal Studat 755 (citedin note 4) (discussingcognitiveerrors). 10 See Tverskyand Kahneman,Judgment under Uncertainty at 14 (cited in note 2) ("In manysituations, people make estimatesby startingfroman initialvalue thatis adjustedto yield thefinalanswer."). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 753 KahnemanandTversky arein theUnitedNations?In one study, spun to obtaina numberbetween0 and 100,and asked a wheeloffortune thenumberthatemergedfromthewheelwas subjectsto saywhether It turnedout thatthe higheror lowerthantherelevantpercentage."l starting point,thoughclearlyrandom,greatlyaffectedpeople's answers.Ifthestarting pointwas65,themedianestimate was45 percent; ifthestarting pointwas 10,themedianestimatewas 25 percent. The processof anchoring-and-adjustment has an obviousapplicationto thesettingof damageawards,whereanmanylegalissues,including chorsplaya largerole.12 Whentherepresentativeness heuristic is involved, peopleanswer or causation-forexample,howlikelyis it a questionof probability thatobjectA belongsto class B?-by askingabout the extentto whichA resemblesB. Suppose,for example,that the questionis whethersome person,Nick,is a librarianor a farmer. If Nickis describedas shyandwithdrawn, and as havinga passionfordetail,most peoplewillthinkthathe is likelyto be a librarian-and to ignorethe thatis,thefactthatthepopulationhas farmorefarmers "base-rate," It shouldbe readilyapparentthattherepresentativethanlibrarians. willproduceproblemswheneverpeople are ignoring ness heuristic as theyare proneto do. In one study, base-rates, a majority ofjudges, in assessingprobabilities, fellpreyto therepresentativeness heuristic."3 Sincetheearlydiscussions ofheuristics andbiases,therehasbeen an explosionoffurther work,sometimes thebasicclaimsof contesting Kahnemanand Tversky,14 but usuallyoffering moreapplications, an of howtheheuristics improvedunderstanding work,and a discovery of many other heuristicsand biases. Heuristicsand Biases: The Psychologyof IntuitiveJudgmentoffersa massive,state-of-the-art treat- mentof theliterature, a similarbook publishedtwo supplementing The book is dividedintothreeparts.The first, decades ago."5 called Theoreticaland EmpiricalExtensions, elaborateson thethreemain heuristics and on severalrelatedheuristics and biases,including optimisticbias.The secondpart,called New TheoreticalDirections, discussestheroleofemotionsand affect, and alternative supporttheory, Thisdiscussionincludestheview,setforth on heuristics. perspectives mostprominently thatoutsidethe laboratory, by Gerd Gigerenzer, 11 Id. See ChrisGuthrie, andAndrew InsidetheJudicial Jeffrey Rachlinski, Wistrich, Mind,86 howjudges,likejuries,fallvictim CornellL Rev777(2001)(showing toanchoring). 13 See id. 14 See GerdGigerenzer, PeterM. Todd,andtheABC ResearchGroup,SimpleHeuristics ThatMake Us Smart27-28(Oxford1999)(contrasting theheuristics-and-biases notionofheuristics witha "fastandfrugal" notionbasedonboundedrationality andadaptation). 15 See Kahneman, underUncertainty Slovic,andTversky, eds,Judgment (citedinnote1). 12 This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 754 of ChicagoLaw Review The University [70:751 workverywell(p 559).Thethirdpart, heuristics our"fastandfrugal" offers a rangeofcasesinwhichintuiApplications, calledReal-World Thesejudgments includethosebyordinary go wrong. tivejudgments in the"hothand"phenomepeople,whofalselybelieve,forexample, whoseclinicaljudgnonin basketball(p 601),and thosebyexperts, are farless accuratethanactuarialjudgmentsof dangerousness 6 ments- a pointwithmanylegalapplications. forlawandpolicy. book,fullofimplications Thisis an impressive I will The collectionalso coversan extraordinary rangeofproblems. notbe able to comecloseto doingjusticeto ithere.InsteadI havea witha purpose:toconnectsomeoftherecentresearch muchnarrower andinparticular thoserelating toriskandlitigasetoflegalproblems, two aspectsof thebook deserve tionbehavior.In thatconnection, focusof Thefirst involvesa shiftfromthestrictly cognitive emphasis. to see howemotionsaffectdecisionand theearlyworkto an effort inseveralofthepaperson"dual Thesecondis theemphasis judgment. Accordingto these approcess"approachesto humanthinking. for decisions. One ofthese two making systems proaches, peoplehave retheotheris slower, butsometimes error-prone; is rapid,intuitive, andmorestatistical. One ofthepervasivethemesin thiscolflective, withtheintuitive andbiasescanbe connected lectionis thatheuristics systemmightbe able to systemand thattheslower,morereflective raisesthepossibility Thisemphasison correction makecorrections.'7 on whichseveralofthepapersalsofocus. of"debiasing," efThisReviewhas fiveparts.PartI discussessomereal-world PartII examinesone ofthemost and anchoring. fectsof availability and interesting papersin thebook,in whichDaniel Kahimportant of andgeneralization offera rethinking nemanand ShaneFrederick 16 See WilliamMeadow and Cass R. Sunstein, Not Experts,51 Duke L J629,629Statistics, 32 (2001) (arguingthatstatisticaldata is moreaccuratethanexpertreports). 17 The papers do not discussthe natureof the brain,but suggestiveresearchtendsto be supportiveof thedual-processidea. Some researchsuggeststhatthebrainhas special sectorsfor reactions,can be triggered emotions,and thatsome typesof emotions,includingsome fear-type beforethe more cognitivesectorsbecome involvedat all. See JosephLeDoux, The Emotional Brain:TheMysteriouis Underpinnings of EmotionalLife 157-69,172-73,283-96(Simon & Schusthe amygdala,appears to play a dister 1996).A small,almond-shapedregionof the forebrain, checkscomingfromthecerebralcortex.See tinctiverole in registering fear,withmorereflective id at 172-73 (suggestingthatstimulationof the amygdalaproduces"a sense of forebodingdanger,of fear,"and that"studiesofhumanswithamygdaladamage also suggestthatit playsa speof the cial role in fear").Indeed,some "emotionalresponsescan occurwithouttheinvolvement reasoning, higherprocessingsystemsof the brain,systemsbelieved to be involvedin thinking, and consciousness."Id at 161.The thalamicpathway,involvingtheamygdala,"cannotmake fine distinctions"but has a strongadvantagein speed. Id at 163. It "can providea fastsignal that warnsthatsomethingdangerousmaybe there.It is a quick and dirtyprocessingsystem."Id. An A patientwithamygdaladamage was asked to detectemotional finding: especiallyinteresting "mostclasses of expressions, exceptwhen expressionon faces,and she succeeded in identifying the facesshowedfear."Id at 173. This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 755 thewholeidea of heuristics.8PartIII explorestherole of emotions optimistic bias.PartV goesbeyondthe andaffect. PartIV investigates bookunderreviewto offersomespeculative aboutthepossiremarks -mental shortcutsthatgenerallywork bilityof "moralheuristics" and errorsin thinking aboutmorality well,butthatlead to systematic law. I. PARLOR GAMES? THE REAL WORLD OF AVAILABILITY AND ANCHORING and biases raiseda naturalset of The earlyworkon heuristics doubts: Are peoplelikelyto makesystematic errors?Are thesephein the real world?Are heuristics an artifactof nomenaimportant On one view,themistakes, oftenmade by unstrangeexperiments? by psydergraduate subjects,are a productof clevermanipulations or in markets, people do and in real-world environments, chologists, Theseissuesreceivehelpfulattention in theintroducmuchbetter.20 buttheyare notthebook'sexplicitfotion(pp 7-15) and elsewhere, to emphasizethatthegoal ofthe cus.To answerthem,itis important is emphatically literature notto showthatpeoheuristics-and-biases irrational. On the conple are fools,or thattheyare systematically KahnemanandTverskyemphasizedthattherelevantheuristics trary, workwell.But in thelaboratory, at least, are efficient and generally and it is theblunsometimes blunder, people who use theheuristics themostacademicattention. Consider,for deringthathas attracted example,thefactthatwhenaskedhowmanywords,on fourpagesofa novel,endin "ing,"peoplewillgivea largernumberthanwhenasked letter(p 21) a clear howmanywordshave"n" as theirsecond-to-last 18 See DanielKahneman andShanLi;rederick,Representativeness Revisited:Attribute Subtheheuristics andbiasesapproachto cognitive stitution processesofattrib(p 81) (generalizing toquestions aboutuncertain "notlimited utesubstitution, events"). 19 See Gigerenzer, at 27-28(cited Todd,andtheABC ResearchGroup,SimpleHeuristics areadaptive);RichardA. Posner, RationalChoice, Behavioral thatheuristics innote14) (arguing Economics,and theLaw, 50 Stan L Rev 1551,1551-53 (1998) (arguingagainstabandonmentof I do notdeal herewiththeissues therationality assumption). Exceptfora fewbriefnotations, I believe,in criticisms oftheheuristics-and-biases literature. numerous raisedbyGigerenzer's workwell(whichno one should criticisms thatGigerenzer's suggestthatsomeheuristics short, canbe designedso as toreduceoreveneliminate errors (whichno one deny)andthatproblems ofthelastpoint,see GerdGigerenzer, Calculated Risks: shoulddeny).Fora helpfuldiscussion How to Know WhenNumbersDeceive You 230 (Simon & Schuster2002) (introducinga three- forteaching individuals "howtoreckonwithrisk"). stepeducational program 20 Forevidencethatheuristics evenwhendollarsare andbiasesoperateintherealworld, Overreact? F.M.DebondtandRichardH. Thaler,Do Analysts see Werner (pp 678-85) involved, of security overreaction analysts);RobertShiller,IrrationalExuberance (finding systematic andoverconfidence in marketbehavior);Colin 136-47(Princeton anchoring 2000) (discussing A Reviewand Camerer and Robin Hogarth,The Effectsof Financial Incentivesin Experiments: thatfinancial 19JRisk& Uncertainty Framework, 7,7 (1999)(finding Capital-Labor-Production havenevereliminated anomaliesorpersistent incentives irrationalities). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 756 The University of ChicagoLaw Review [70:751 of theavailability heuristic. Severalof thepalaboratory illustration and demonstrate thatheuristics persgo wellbeyondthelaboratory intherealworld. produceerrors andRisk A. Availability andsafety. 1. Availability, health, abouthealth It is reasonableto expectthatpeople'sjudgments andBariskswouldbe affected andsafety bytheavailability heuristic, offerssome striking illustrations (p 730). Should ruch Fischhoff in casesofassault?In popularpubliresistance womenoffer physical advice(p 733).Thosewhoclaim cations,expertsoffercontradictory is a seriousmistakeconsistdisproportionately ofpeothatresistance sources,whomostlysee bad outcomesin ple fromlaw enforcement Hencepoliceofficers resistance. casesofphysical maywellbe victims at least"iftheyremembered whatthey of theavailability heuristic, ofwhattheywerenot hadseenandheard,butlackedan appreciation does notexplorethepossibility thatthose seeing"(p 733).(Fischhoff an producing whoresistandare hurtreceivespecialmediaattention, of Fischhoff discusseslay estimates additionalbias.)More generally, of forty-one causesof deathin theUnitedStates.He thefrequency are consistent withtheavailfindsthattheerrorsin theseestimates beforethedata abilityheuristic (and hencetheerrorswerepredicted wereseen) (p 737).Highlypublicizedcausesofdeath,suchas floods and tornadoes, are overestimated, whereasquieterkillers,such as are underestimated strokesanddiabetes, (p 738).Otherstudiesshow a similar pattern." is actualbehavioraffected bytheavailability Apartfromsurveys, heuristic? Thereis evidencethatit is. If floodshavenotoccurredin theimmediate past,peoplewholiveon floodplainsarefarlesslikely In theaftermath of an earthquake, to purchaseinsurance." thenumber of people seekinginsuranceforearthquakes risessharply-but thatnumberdeclinessteadilyfromthatpoint,as vividmemories re21 See W. Kip Viscusi,Jurors, and theMistreatment of RiskbytheCourts,30 JLegal Judges, A possiblecriticismof thesefindingsis thatthey Stud 107 (2001) (findingsimilarmisestimates). people typicallyare givena starting mightshow the effectof anchoring.In the relevantsurveys, number,such as the numberof deathsfrommotorvehicleaccidentseach year (around40,000). That startingnumberis necessaryto ensurethatnumbers, forimperfectly informedrespondents, willnotbe all over thelot.But thestartingnumber,as an anchor,mightalso compresstherange of answers,makinghighnumberslowerand low numbershigherthantheywouldotherwisebe. 22 For a vividdemonstration in the contextof catastrophes, and see JacobGersen,Strategy Cognition:RegulatoryCatastrophicRisk 57-60, 77 (unpublishedmanuscript2001) (presenting and analyzingevidencelinkingsubjectivebeliefsabout the likelihoodof floodsand theiractual occurrence).See also Paul Slovic,et al, CognitiveProcessesand Societal Risk Taking,in Paul Slovic,ed. The Perceptionof Risk 32. 39-40 (Earthscan2000) (discussinghow individuals'assessmentoffloodingprobabilityis strongly conditionedby theirrecentexperience). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 757 Noticethattheuse oftheavailability inthesecontexts cede.23 heuristic strongly suggeststhattheheuristics operateevenwhenthestakesare large.Anditis possiblethattheuse oftheavailability heuristic insuch contextsis fullyrationalforpeople who lack statistical knowledge. is thebestwayofminimizing Perhapsuse ofthatheuristic thesumof decisioncostsand errorcosts. But it seemsless usefulto debatethe rationality of the availability heuristic thansimplyto observethatit effect on actualbehavior. has a significant 2. Thesourcesofavailability. Whatin particular An interesting producesavailability? essayatofease ofimagery on perceived of temptsto testtheeffects judgments risk(p 98). The studyaskedsubjectsto read aboutan illness(Hyposcenia-B)that"wasbecoming increasingly prevalent" (p 99) on thelothesymptoms wereconcreteandeasyto cal campus.In one condition, imagine -involving muscleaches,low energy, and frequentsevere thesymptoms werevagueand hard headaches.In anothercondition, an inflamed to imagine, nervoussysinvolving liver,a malfunctioning tem,and a vaguesenseof disorientation. Subjectsin bothconditions wereaskedbothtoimaginea three-week periodinwhichtheyhadthe ofwhattheyimagined. Afdiseaseand to writea detaileddescription terdoingso,subjectswereaskedto assess,on a ten-point scale,their thedisease.The basicfinding likelihoodofcontracting was thatlikeliin thetwoconditions, wereverydifferent hoodjudgments witheasilyimaginedsymptoms makingpeople farmoreinclinedto believethat theywerelikelyto getthedisease. forpolicyand law.The publicdeThereare severalimplications mandforlaw shouldbe muchhigherifpeoplecan easilyimaginethe harmin question;in suchcases,thelaw mightwellreflecta kindof to imagine, we mightwellsee a But iftheharmis difficult hysteria.25 26 patternofneglect.We wouldtherefore predictthateasilyimaginable 23 Paul Slovic,et al, Preference for InsuringAgainstProbable Small Losses: InsuranceImplications,in Slovic,ed, The Perceptionof Risk at 51 (cited in note 22) (discussingstudyof one- inearthquake-prone inflood-prone homeowners areasandtwo-thousand homeowners thousand heuristic affected individual decisions toinsure). areasthatfoundthatbiasfromtheavailability 24 It is reasonable, as suggesting to readKahnemanandTversky thattheheurishowever, inthisway-thatsomeofthetime,at least,theheuristics be defended ticscannotentirely operof"ing"as wouldimprove Considerthediscussion ate eventhougha littlethought judgments. discussed opposedto "n" as thenext-to-last letter, above,and consideralso theLindaproblem, below. 25 I am emphasizing theavailability heuristic here,butwhenpeopleimaginebad events, in whichtheyfailto considerthelikelito "probability theymightalso be vulnerable neglect," hood of harm.See Cass R. Sunstein,ProbabilityNeglect:Emotions,WorstCases, and Law, 112 noton itslikelihowpeople"focuson theadverseoutcome, Yale L J61,62 (2002)(illustrating hood"anddiscussing possiblepolicyimplications). 26 Comparethefinding afterdrinking, thatteens'ratesofriskbehaviors-smoking, driving This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 758 The University of ChicagoLaw Review [70:751 harmswouldlead to relatively and relagreaterprivateprecautions privategroups concern.Well-organized tivelygreatergovernmental to publishould,and do,takeadvantageofthisheuristic, attempting cizevisibleexamplesofharmsto whichtheyseekto drawattention.28 camThe pointalso offersimplications aboutpublicinformational wantspeople to take protective paigns.If the government steps,it ina vividrather shouldprovideinformation aboutsymptoms thanstaon examplesthatcan laterbe brought to tisticalway(p 102),relying mind.(Terrorists showa goodintuitive oftheavailabilunderstanding terrorist actislikelyto ityheuristic, simply becausea highly publicized consider createfarmorefearthanthebarestatistics warrant; the*Sepinthefall tember11attackandtheactsoftheWashington, DC snipers of2002.29) intherealButthereis an interesting puzzleforthoseinterested In manycontexts, worldusesof thisheuristic: multiple imagesarelitIt is nothard erally"available."Considertheproblemofgunviolence. to findcases in whichthepresenceof gunsled to manydeaths,and also casesin whichthepresenceofgunsallowedlaw-abiding citizens to protectthemselves In thefaceofconflicting inagainstcriminals.> whichcases are especiallyavailable?Whyshouldone or anstances, otherkindof case be available?The behaviorof themedia,and of relevantinterest here.Manypergroups,is undoubtedly important are in reality no suchthing, ceived"epidemics" butinsteada product unsafesex-can be reducedbyaddressing andbiases,inpartbyexplaining heuristics thatthe leadsteensto overestimate heuristic theriskbehaviorof theirpeers.See Baruch availability andBiasesinApplication thatbettereducation Heuristics aboutbi(p 747) (arguing Fischhoff, asesleadsteenstoreduceundesirably risky behaviors). 27 See NollandKrier, 19JLegalStudat767-71(citedinnote4) (analyzing thecorrelation ofavailability withthepolitical demands forriskpolicy). 28 See KuranandSunstein, 51 StanL Revat715-35(citedinnote9) (discussing availabilitycampaigns). 29 Although extensive andevenextraordinary onlytenpeoplewerekilledbythesnipers, weretaken.Forexample, OctoberSAT testing atseveralareaschoolswascanceled. precautions See Erik Brady,WeekendPlans Fall Victimto SniperFears,USA TodayA01 (Oct 14,2002) (re- on theextensive takenin thewakeoftheWashington and porting precautions DC, Virginia, Thesizeofthepanicis noteasytoexplainonthenumbers alone.The Maryland sniperattacks). statistical riskofbeingkilledbythesnipers was,at everystage,quitelow.I believethattheextremefearwasproduced oftheavailability heuristic andprobability bya combination neglect. a two-week Comparethefactthat,on average, twenty-five peopledie in autoaccidents during timespanin an area thesizeofWashington causea panic.See Mark DC. Suchdeathshardly But So Are the Odds, USA Today A06 (Oct 18, 2002) Memmott,Fear May Be Overwhelming, severalstatistical withthesniperattacks inordertohighlight thepublic's (drawing comparisons overreaction). 30 See Donald Braman and Dan M. Kahan, More Statistics, Less Persuasion:A Cultural Theoryof Gun-RiskPerceptions 26-27, 26 n 63, workingpaper (2002), online at the http://www.law.uchicago.edu/academics/circulation5.pdf (visitedFeb 18,2003) (discussing cultural determinants ofgunattitudes ofcompeting casesofoffengiventhereadyavailability siveordefensive usesofguns). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 759 incidents.3' But this of mediacoverageof gripping, unrepresentative does not providethe whole picture.Beliefsand orientations are a to be sure;butwhatis availableis also a prodproductofavailability, beliefsand orientations. In otherwords,availability uctofantecedent maybe endogenousto individual predispositions. Social processesare quiteimportant here,forapparently repreanecdotesandgripping examplescanmoverapidlyfromone sentative Once severalpeople startto takean exampleas personto another.32 bytheiropinion, probative, manypeoplemaycometo be influenced In thedomainofrisks, "availability casgivingriseto cascadeeffects.33 cades" help to accountformanysocialbeliefs,and herelocal variawithdifferent tionsare likely, examplesbecomingsalientin different lead likecommunities. Indeed,processesof deliberationtypically mindedpeople to accept a moreextremeversionof theiroriginal ofcertainavailableexamples views,34 makingit likelythattheeffects The problem groupdiscussion. willbecomegreatlyamplified through Bemightwellbe aggravated bycertainmediaandnewtechnologies. incidents can be madewidelyavailablealmost cause of theinternet, in whichmisleading or instantly, creating "availability cybercascades" is rapidlyspread.35 With the same technologies, false information And groupdiscussion can amplify theseprocessesoffear-mongering.36 different culturalorientations playa largerolein deterundoubtedly 31 See Michelle 31,2002),onlineat Cottle,Summer Scare,TheNewRepublicOnline(July (visitedNov 15,2002)("[T]akethese http://www.tnr.com/docprint.mhtml?i=life&s=cottleO73102 mediasince addinthehalfdozenothercasesmentioned bythenational much-hyped abductions, oftheyear... [it]stilldoesn'tqualify as a newcrimewave."). thefirst 32 See ChipHeath,ChrisBell,and EmilySternberg, in Memes:The EmotionalSelection 81 JPersonality & Soc Psych1028,1028(2001)(discussing threestudies CaseofUrbanLegends, a morereadywillingness topassalongstoriesthatevokedrelatively emotional stronger finding responses,especiallydisgust);Chip Heath, Do People Preferto Pass Along Good or Bad News? Valence and Relevance as Predictorsof TransmissionPropensity,68 Org Beh & Human Dec thatwhether newsis good,bad,or neither, will Processes79, 89 (1996) (finding individuals a congruent domainofsocialaction). itrapidly so longas itremains within propagate 33 See Shiller, Irrational Exuberance at 148-68(citedin note20) (discussing socialpres51 StanL Rev at 715-36(citedin suresleadingto information cascades);Kuranand Sunstein, of beliefs);Sushil formation note 9) (proposinga mechanismfor rapid,self-reinforcing Fads,and Informational Biikhchandani,et al, LearningfromtheBehaviorof Others:Conformity, ... implies cascadestheory Cascades,12JEconPersp151,168(Summer 1998)("[I]nformational by a small pervasivebut fragileherdbehavior.This occursbecausecascade~are triggered amountofinformation."). 34 See Cass R. Sunstein, andDissent8,University ofChicagoLaw & EconomConformity icsOlinWorking (visitedFeb18, PaperNo 164(2002),onlineathttp://papers.ssrn.com/id=341880 howindividual members' ofmoremodestviews andexplaining suppression 2003) (illustrating membiastowardextremism, forgroupsoflike-minded especially producesgroupdeliberative bers); Cass R. Sunstein,DeliberativeTrouble?WhyGroups Go To Extremes,110 Yale L J71, 88- howdeliberation are 94 (2000) (explaining maypolarizetheviewsofgroupsiftheirmembers like-minded composedofprimarily individuals). 35 See Cass R. Sunstein, 80-84(Princeton 2001). Republic.com 36 See id at 64-69. This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 760 The University of ChicagoLaw Review [70:751 A greatdeal ofwork,normawhatturnsoutto be available.37 mining remains tobe doneon thistopic. tiveandempirical, B. AnchorsandDamages werememoraThe originalstudiesofanchoring-and-adjustment thatwhen ble in partbecausetheywereso amusing. Theysuggested information about an are lack people appropriate value,they highly evenbyapparently irrelevant numbers.But theoriginal suggestible, studiesleftopenmanyquestionsaboutthenecessary for conditions andalso abouttheroleofanchoring outsideofthelaboraanchoring, tory. Gretchen offer a greatdealofhelpin ChapmanandEricJohnson thesequestions(p 120).Chapmanand Johnson showthat answering evenwhenpeoplethinkthatitis not(id); an anchoris oftenoperating thatanchorshaveeffects evenwhenpeoplebelieve,and saytheybelieve,thattheanchoris uninformative (id); and thatmakingpeople does notreduceanchoring awareof an anchor'seffect (id). Veryextremeor ludicrously anchorsalso seemto havean effect: implausible Estimatesof the yearthatAlbertEinsteinfirstvisitedthe United Statesare greatlyaffected byaskingpeopleto beginbyconsidering anchorsof1215or 1992(p 124).ChapmanandJohnson alsoshowthat economicincentives do noteliminate theeffects of anchors(p 125); is nota resultof casualnessabouttheunderlying task.It anchoring followsthat"debiasing" is verydifficult inthiscontext. Anchorshave majoreffectson legal outcomes. The plaintiff's in termsof bothliability demandinfluences juryverdicts, judgments lowandimplausibly and amountsawarded(p 137).Evenimplausibly in negotiation highdemandsoperateas anchors(id). Openingoffers on settlements. An ingenious havea significant influence studyfinds thatanchorsaffect wereaskedtocomeup withapjudges,too.Judges propriate awardsin a personalinjurycase.40The studyinvolvedtwo The"no anchor"condition conditions. of involveda simplestatement thefacts.The "anchor"condition was thesameas thefirst, butwith one critical difference: Thedefendant filedan obviously momeritless tionto dismissthecase on thegroundthatthe$75,000jurisdictional 37 See Bramanand Kahan. More Statistics, Less Persuasionat 1, 18 (cited in note 30) (setofheuristicsaboutgunownership). tingforththeculturaldeterminants 38 See Tverskyand Kahneman,Judgment underUncertainty at 14-16 (cited in note2) (discussingstudiesshowinganchoringeffectpreventsproperscalingadjustment). 39 For a discussionof marketbehaviorand anchoring, see Shiller,IrrationalExuberanceat 135-42 (cited in note 20) (discussingquantitativeanchorsand moral anchorsin the contextof thestockmarket). 40 See Guthrie, 6 CornellL Rev at 790-91 (citedin note 12) (preRachlinski,and Wistrich, sentingjudges witha descriptionof a serious personal injurysuit withor withouta damage awardanchor). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 761 minimum had notbeen met.Almostall of thejudgesdeniedthemoon ulservedas an anchor,withlargeeffects tion,whichnonetheless In theno-anchor theaverageawardwas condition, timatejudgments. $1.24million, whileitwas$882,000intheanchorcondition.41 an invaluation"studies, Anchorsalso playa rolein "contingent goods,suchas increasedsafety methodofvaluingregulatory fluential and protection (p 137). Perhapsthemoststriking, andenvironmental evidenceto thiseffectcomesfroma studyofwillin a wayhilarious, ingnessto payto reduceannualrisksofdeathandinjuryinmotorveand minimum to elicitbothmaximum The authorsattempted hicles.43 to pay forsafetyimprovements. People werepresented willingness theyweredefiwitha riskand an initialamount,and askedwhether to pay thatamountto eliminate unwilling nitelywillingor definitely theamountdiswilling, therisk,or "notsure."If theyweredefinitely unwillplayedwas increaseduntiltheysaid thattheyweredefinitely ing;if theywereunsure,thenumberwas movedup and downuntil andmaximum. theminimum peoplecouldidentify ofanchors. On totesttheeffects The authorswerenotattempting the contrary, theywere alertto anchoringonlybecause they"had in which been warned"of a possibleproblemwiththeirprocedure, amountofmoneythat be undulyinfluenced bythefirst people"might theysaw displayed.""To solve thatproblem,the authorsallocated one withan initialdisplayof?25, to twosubsamples, peoplerandomly theotherwithan initialdisplayof?75.The authorshopedthattheanconsequencesfor choringeffectwouldbe small,withno significant values.Buttheirhopewasdashed.Forevery minimum and maximum to paywas higherwiththe?75 levelofrisk,theminimum willingness topaywiththe?25 startthe maximum than willingness starting point, intheannualriskofdeathby4 in Forexample,a reduction ingpoint!45 to payof?149 withthe?25 100,000produceda maximumwillingness 41 Id at 791. There is a possible responseto the authors'claim to have shownthe effectsof anchoring:Perhaps the motionto dismisssuggestedthatthe injurywas less seriousthanwas apparent.Whywould a lawyerfilea totallyfrivolousmotionto dismiss?But theabundantevidence of effectsfromanchorssuggeststhatthisis unlikelyto explain all or even muchof the authors' finding.See also W. Kip Viscusi,CorporateRiskAnalysis:ARecklessAct?,52Stan L Rev 547,558 (2000) (findingan anchoringeffectfrommonetaryvalue of lifeon juryawards,so muchso that companiesthatplaced a highvalue on humanlifeended up payinghigherpunitiveawards). 42 See, for example, Ian J.Bateman and Kenneth G. Willis,eds, ValuingEnvironmental ValuationMethodin theUS, EU, and DevelopPreferences: Theoryand Practiceof theContingent ing Countries(Oxford 1999) (presentingseveral studiesof contingentvaluation). 43 See Michael Jones-Leeand Graham Loomes, PrivateValues and Public Policy,in Elke U. Weber,et al, eds, Conflictand Tradeoffsin Decision Making 205, 210-12 (Cambridge 2000) (presentingdata forstrongstartingpoint effectson individual'swillingnessto pay forreducing auto accidentinjuriesand deaths). 44 Id at 210. 45 Id at 211. This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 762 TheUniversity ofChicagoLaw Review [70:751 to payof?232,withthe?75 starting value,buta minimum willingness inthatcase,of?350).`6 starting value(anda maximum, is thatwhenever The mostsensibleconclusion peopleare uncerand tainaboutappropriate effect, values,anchorshave a significant and sometimesa startlingly lawyers, largeone. Clevernegotiators, sometimes even shouldbe able to exploitthoseeffects, policymakers or apparently Thereis a anchor. an outlandish irrelevant byproviding forlegalreform realopportunity here,inpartbecauseanchorsmight wellproduceresultsthatare noteasyto defend, and in partbecause different anchorswillensurethatsimilarly situatedpeople are not to inform treatedsimilarly. Perhapslawyersshouldnotbe permitted suchas theannualprofits ofthe jurorsofpotentially effective anchors, punitive firm oreventheplaintiffs demand, at leastincasesinvolving awardsor hard-to-monetize awards.Or perhapsjudges compensatory shouldbe askedto reviewjuryawardscarefully and byreference to ofarbitrary anchors. comparison cases,so as toweakentheeffect This pointraisesa relatedone: Are groupsable to avoid the Theevidenceis mixed.In genjudgment errorsmadebyindividuals? eral,groupstendto polarize:Theytendto endup in a moreextreme At thesame positionin linewiththeirpredeliberation tendencies.48 time,groupshavebeenfoundto makebetterdecisionsthanindividuals withrespectto certainstatistical Thereis some eviproblems.49 dencethatgroupsareslightly betterat avoidingtheproblems created On theotherhand,someevidence heuristic.50 byuse oftheavailability heuristic is actually suggeststhatthe use of the representativeness ingroups.5" It seemsclearthatgroupprocessesdo notelimiamplified natetheuseofheuristics, anditremains tobe seenwhether andwhen errors. theyreduceorincreasetheresulting 46 Id. See NorbertL. Kerr,Robert J.MacCoun, and GeoffreyP. Kramer,Bias in Judgment: ComparingIndividualsand Groups,103 PsychRev 687,688-93 (1996) (reviewingseveralstudies of different typesofgroupand individualjudgmenterrorsand findingno clearpattem). 48 See Sunstein,110Yale L Jat 88-94 (2000) (citedin note34) (finding thathomogeneity, a ofgroups,reinforces commoncharacteristic groupmembers'tendenciesto go to extremes). 49 See Alan S. Blinderand JohnMorgan,Are TwoHeads BetterThan One?:An Experimental Analysisof Group vs.IndividualDecisionmaking1,NBER WorkingPaper No 7909 (2000), online at http://www.nber.org/papers/w7909.pdf (visitedFeb 18, 2003) (findingthatgroupsrethanindividualsto a statisticalurnproblemand a monespondedas rapidlyand moreeffectively tarypolicyexperiment). 50 Kerr,MacCoun, awd Kramer,103 PsychRev at 692 (cited in note 47), citingM.F. Stasson,et al, Group ConsensusProcesseson CognitiveBias Tasks:A Social Decision SchemeApproach,30 JapanesePsychRsrch68 (1988). 51 See id,citingG. Stasserand W.Titus,Effects Load and PercentageShared of Information on theDisseminationof UnsharedInformation duringDiscussion,53 JPersonality& Information Soc Psych81 (1987). 47 This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 763 II. Two SYSTEMS What,exactly, is a heuristic? Whenwilla heuristic be overridden bycognitive processesthatproducea moreaccurateunderstanding of theproblemin question?In a highlight ofthiscollection, Daniel Kahnemanand Shane Frederickmakereal progresson thesequestions (p 49).Theirdiscussion is packedwithnewmaterial, and I touchhere onlyon thepointsofparticular relevanceforpolicyandlaw. A. Dual Processing andAttribute Substitution Muchof theirargument turnson drawinga connection between heuristics anddual-process theories.52 Recallthatthosetheoriesdistinguishbetweentwofamilies ofcognitive operations, sometimes labeled I System andSystemII. SystemI is intuitive; itis rapid,automatic, and effortless. SystemII, bycontrast, is reflective; it is slower,self-aware, and deductive.Kahnemanand Frederickare carefulto disclaimthe view thatthe two systemsoperateas "autonomoushomunculi"; in theirview,theyrepresent"collectionsof processesthatare distinguishedbytheirspeed,controllability, andthecontents on whichthey operate"(p 51).TheysuggestthatSystemI proposesquickanswersto problemsofjudgment, and thatSystemII operatesas a monitor, conor overriding firming thosejudgments. Consider,forexample,someone who is flying fromChicagoto New Yorkin themonthafteran airplanecrash.This personmightmake a rapid,barelyconscious rootedin SystemI, thattheflight judgment, is quiterisky, butthere mightwell be a SystemII override, a morerealisticassessbringing mentto bear.In makinga distinction betweenSystemI andSystemII, Kahnemanand Frederickannouncea themethatplaysa significant roleinthisbook. KahnemanandFrederick also offera generalclaimaboutthenatureofheuristics: a processofattribute Theyoperatethrough substitution(p 53). In thisprocess,peopleare interested in assessinga "target and theydo so bysubstituting attribute," a "heuristic attribute" ofthe object,whichis easierto handle.Considerthequestionwhether more peopledie fromsuicidesor homicides. Lackingstatistical information, peoplemightrespondbyaskingwhether it is easierto recallcases in 52 See generallyShelly Chaiken and Yaacov Trope, eds, Dual-Process Theoriesin Social Psychology(Guilford1999) (reviewingrole of dual-processtheoriesin social cognitionand individualjudgment). 53 See, forexample,StevenSloman, TwoSystemsofReasoning(p 379) (reviewingevidence foran "associativesystem"and a "rule-basedsystem"ofcomputation);Paul Slovic,et al, TheAffectHeuristic(p 416) (advancing the distinctionbetween a heuristicthat ordersreactionsinto "good" and "bad" and deliberatebehaviormeantto manipulatethiseffect);Robyn Dawes, et al, Clinical versusActuarialJudgment(p 716) (distinguishing between error-proneclinical evaluationand data-based actuarialjudgment). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 764 TheUniversity ofChicagoLaw Review [70:751 The responsemightwell be heuristic). eitherclass (the availability alsoleadtoerrors. sensible, butitmight B. Amending theTheory offersome KahnemanandFrederick Withtheseunderstandings, amendments to the originalpresentation by Kahneman significant thatanchoring shouldnotbe seenas a heuandTversky. Theysuggest an attribute, butby operatesnotbysubstituting risticat all;anchoring valueseemmoreplausible(p 56).Theyalso argue makinga particular is theaftoreplaceanchoring, thatthethirdgeneral-purpose heuristic, urgethat fectheuristic (id),discussedbelow.KahnemanandFrederick (p 63), punitive damageawardsare mediatedbyan outrageheuristic in action.Juheuristic see as an exampleoftheaffect whichwe might damageawards rorsdo nothavea goodsenseofhowto setpunitive andtheybegintheprocessbyaskingabouttheout(a hardquestion), Someof thedefendant's conduct(an easierquestion).54 rageousness playsa roleinpunishment undoubtedly thinglikean outrageheuristic of manydifferent kinds;thereis a largeresearchagenda judgments here. heuristic, Now turnto theauthors'focus,therepresentativeness Themostfamousofthese whichhasleadtosomelargecontroversies.55 woman involvesquestionsaboutthelikelycareerof a hypothetical "Lindais 31 yearsold,sinnamedLinda(p 62),describedas follows: As a stuand verybright. She majoredin philosophy. gle,outspoken, withissuesofdiscrimination and sodent,shewas deeplyconcerned in antinuclear Subcialjusticeand also participated demonstrations." eightpossiblefutures jectswereaskedto rank,inorderofprobability, forLinda.Six of thesewerefillers(likepsychiatric socialworkeror schoolteacher);thetwocrucialones were"bankteller" elementary and "banktellerand activein thefeminist movement." Mostpeople said thatLindawas less likelyto be a banktellerthanto be a bank tellerand activein thefeminist Thisis an obviouslogical movement. A and B in whichcharacteristics calleda conjunction error, mistake, A alone.The error are thought to be morelikelythancharacteristic 54 Here, Kahneman and Frederickdraw on workin whichI have been involved.See, for 54 Stan L Rev 1153,1157-59, Judgments, example,Cass R. Sunstein,et al, PredictablyIncoherent 1167-70 (2002) (discussinghow juries move froma determination of punitiveintentto a metric fordamagesor punishment). 55 For a treatment oftherepresentativeness heuristicand investment behavior,see Schiller, IrrationalExuberanceat 144 (cited in note 20) (reviewingmodels of expectationalfeedbackin the stockmarket).For some of thecontroversy, see Barbara Mellers,Ralph Hertwig,and Daniel EliminateConjunctionEffects?,12 Psych Sci 269 Kahneman,Do FrequencyRepresentations (2001) (presentingdisagreementover whethertestinganomalies explainKahnemanand Tversky's findingof conjunctioneffectsand whetherfrequencyformatswould eliminatesuch effects). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 765 heuristic: Linda'sdescription seems stemsfromtherepresentativeness movement" farbetter to match"banktellerand activein thefeminist than"bankteller." As KahnemanandFrederick note,people'sanswersto theLinda problemhavebeenexplainedon numerous grounds(p 67),withcritics of theproblemincreasedor perhapseven arguingthatthestructure generatedlogicalmistakes.Kahnemanand Frederickurgethatthis pointshouldbe takennotas a challengeto theclaimthatpeopleuse the representativeness heuristic, but as evidencethatundercertain circumstances, peoplewillovercometheerrorsproducedbythatheuristic(includingthe conjunction fallacyand neglectof base-rates). KahnemanandFrederick suggestthatwhentheseproblemsare overcome,it is oftenbecauseoftheoperationsofSystemII, whichworks as a kindofsupervisor. Henceintelligent people,andthosewithstatisticalsophistication, arelesslikelyto err(p 68);forsuchpeople,System II is especiallyactive. For law and policy, thegenerallessonis simple:Wheneverpossible,institutionalize SystemII, at leastwhenquestionsof factare involved.Frequently thelegalsystemdisregards thisadvice,relying on intuitions aboutprobability juriesand henceon ordinary and causation.57 The twentieth-century movementtowardgreaterrelianceon technicalexpertiseand actualdata mightwellbe seen as an implicit oftheunreliability ofordinary intuitions. recognition Indeed,thereis reasonto thinkthatexpertsthemselves are vulnerableto heuristics, and thatrelianceon actuarialdata could improveaccuracy.58 In the domainof regulation, riskanalysisis themoststraightquantitative forwardway of overcoming the errorsthatsometimesaccompany heuristics. Considerthe controversy over regulationof arsenicin The andrepresentativeness drinking water.9 availability heuristics ensurethatmanypeoplewillbe quitefrightened of arsenic,evenin exlow doses.Quantitative riskanalysiscan workas a kindof tremely II on errors. check System potential 56 Withrespectto intelligence, thereis a nice qualification:When the problemis veryhard for everyone,intelligentrespondentsare most likelyto err,because they"are more likelyto agree on a plausible errorthanto respondrandomly"(p 68). 57 See Gigerenzer,Calculated Risks at 85-86, 139, 159,229-46 (cited in note 19) (offering several suggestionsforeducatingdoctors,administrators, judges,law students,and othersabout avoidingerrorsin riskassessment). 58 See Robyn Dawes, et al, Clinical versusActuarialJudgment be(p 716) (distinguishing tweenerror-prone clinicalevaluationand data-based actuarialjudgment). 59 See Cass R. Sunstein,The Arithmetic of Arsenic,90 Georgetown L J 2255, 2255-60 (2002) (arguingthatscientificdata onlyproduce a wide "benefitrange"ofpossible arseniclevels withinwhichthe EPA mustset itsstandard). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 766 ofChicagoLaw Review TheUniversity [70:751 C. Generalizing Representativeness unsuggestthatjudgmentheuristics, Kahnemanand Frederick derstoodto involveattribute substitution, operatenotonlyto answer butalso in a diverseclassofjudgevents, questionsaboutuncertain go so faras to urgethata modest ments.Kahnemanand Frederick heuristichelps to explain of the representativeness generalization biases in valuations of publicgoodsand economic similar strikingly Kahnemanand evaluations ofpastevents.In particular, retrospective or representative Tverskyemphasizethe crucialrole ofprototypes, inmaking complexjudgments. exemplars, topaytosaveanimals?It turnsout How mucharepeoplewilling involvedandhighly sensitive to theprototypes thatpeopleare highly thatinvolves tothenumber ofanimalsat stake.A program insensitive willingness ofa popularspecieswillproducea muchhigher members ofa lesspopularspecies thatinvolves members topaythana program to pay does not differ people'swillingness (p 71). More strikingly, theirwillingness involved; with variations in the numbers large greatly to payis aboutthesameto save2,000,20,000,or 200,000birds(p 75). inusing Thereis a clearparallelheretopeople'sneglectofbase-rates Kahto makeprobability judgments. therepresentativeness heuristic pastexperiences, also showthatin evaluating nemanand Frederick or noises,painfulmedicalprocedures, suchas exposureto unpleasant (p 77). In one experifilmclips,peopleshowduration neglect horrific movieswere ment,for example,people's evaluationsof horrific in theirlength(id). In anvariations bysubstantial largelyunaffected weregreatly ofcolonoscopies otherexperiment, people'sevaluations the level of involved and also influenced bythelevel pain by highest in theduraofpainat theend,butnotmuchbysignificant variations tionof the procedure(fromfourto sixty-nine minutes).Here too, capturedin the Kahnemanand Frederickurgethatthe prototype, dominates evaluation. PeakAffect andtheEndAffect, hereis an intriguing Withrespectto law and policy, implication willcrowdoutvariablesthat,onreflecthatpeople'suseofprototypes Thereis a seriousproblemwithcontintion,haveclearimportance. gentvaluationstudiesif people'sjudgmentsdo not attendto the in numberof animalsat stake.And indeed,someof thepathologies withthisproblem. Evidencesugregulatory policydo seemconnected of moreabouttheproportion gests,forexample,thatpeople"worry A striking riskreducedthanaboutthenumberof peoplehelped."O neglectabsolute studyof thiseffectfindsthatpeople pervasively 60 JonathanBaron, Thinkingand Deciding 500 (Cambridge3d ed 2000) (explainingthe confusionbetweenrelativeand absoluterisks). generally-observable This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 767 numbers, andthatthisneglectmapsontoregulatory policy.In a similar vein,it has been shownthatwhenemotionsare involved, people neglecttwonumbers thatshouldplainlybe relevant: theprobability of 62 harmandtheextentofharm. III. EMOTIONS, CONTAGION,AND AFFECT How arejudgments, aboutthelikelihoodof especially judgments riskor benefit, influenced byemotionsand affect? Severalpapersexplorethatquestion. A. FalseContagions andPhonyCures Paul Rozin and Carol Nemeroff explore"sympathetic magical thinking," includingthe beliefsthatsome objectshave contagious properties, and thatcauses resembletheireffects.63 Many educated willnoteat foodtouchedbya sterilized Americans cockroach(p 202). Theyrefusechocolatesthathave been shapedintorealistic-looking to use sugarfroma bottlelabeled dog feces(id). Theyare reluctant "SodiumCyanide,Poison,"eveniftheyare assured,and believe,that thebottlereallycontainssugarandnevercontainedcyanide(id)-and indeedeven if theythemselves placed the label,arbitrarily, on that bottle(p 205)! In factpeople are reluctant to eat sugarlaparticular because the verywords beled,"Not SodiumCyanide,"apparently "SodiumCyanide"automatically Peobringup negativeassociations. to weara sweaterthathas been wornforfiveminple are reluctant utesbya personwithAIDS (p 207).In thiscase,as in othercases ininsensitive to dose.A sweater volvingcontagion, peopleare relatively wornforfiveminutesbysomeonewithAIDS, andthenwashed,is not muchmoreundesirable thana sweaterused bysomeonewithAIDS fora fullyear.According tomostrespondents, a singleliveAIDS virus thatentersthehumanbodyis as likelyto infectsomeonewiththevi61 See TimothyL. McDaniels, ComparingExpressedand Revealed Preferences forRisk Reduction:DifferentHazards and QuestionFrames,8 Risk Anal 593, 602-03 (1988) (findingthat bothordinarypeople and policymakersframevaluationsas percentagechangesfromsome fixed rate of deaths). 62 See Sunstein,112 Yale L J at 70-83 (cited in note 25) (discussingneglectof probability when strongemotionsare involved);ChristopherK. Hsee and Yuval Rottenstreich, Music,Pandas, and Muggers:On theAffective Psychologyof Value,workingpaper (2002) (on filewithauthor) (suggestingthatthe dualism between valuationby feelingand by calculationproduces a and ChristopherK. Hsee, Money, non-linearprobabilityweightingofvalue); Yuval Rottenstreich Kisses,and ElectricShocks:On theAffective Psychologyof Risk,12 PsychSci 185,186-88 (2001) (findingthatin threeseparate studiesundergraduatesfail adequately to account forthe extent and frequencyof harm). 63 Paul Rozin and Carol Nemeroff, SympatheticMagical Thinking:The Contagion and and oppositesin theway Similarity"Heuristics"(p 201) (reviewinglaws of contagion,similarity, individualsthink). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 768 The University of ChicagoLaw Review [70:751 rusas 10,000oreven1,000,000 viruses(p 207).Noteinthisregardthat andreal-world disgustandfeartendto"travel";inbothexperimental settings, peopleareespecially likelytospread"urbanlegends"thatinvolverisksofcontamination."4 In someofthesecases,theintuitive fearorrevulsion canbe easily I as that is no hazard. overridden, reflection reveals there real System senseofalarmor revulsion, butSystemII givesriseto an immediate will usuallyprovidea corrective(even if SystemI continuesto squawk).But notalways.Paul Slovichas foundthatmostpeopleacwiththesugcepta kindof"intuitive toxicology," showing agreement gestionthat"thereis no safelevel of exposureto a cancer-causing thenyou are agent"and that"ifyou are exposedto a carcinogen, likelyto getcancer."65 someintuitions aboutfeararepart Apparently ofeveryday aboutsocialrisks. thinking ThomasGilovichand KennethSavitsky use theidea that"like ofa widerangeoffalsebeliefs, goeswithlike"tounpackthestructure bothancientandmodern." beliefsaboutmedicine reManyprimitive flectthebeliefthatthesymptoms ofa diseaseare likelyto resemble bothitscause and itscure.Accordingto ancientChinesemedicine, thosewithvisionproblems shouldeat groundbats,on thetheory that whichmight batshaveespecially be transferred topeople goodvision, whichremains (p 619).Homeopathy, quitepopular, dependsinparton theidea thatifa substance createsdiseasesymptoms ina healthy persuffers son,it willhave a healthyeffecton someonewhocurrently fromthosesymptoms (p 620).The idea has somevalidapplications, of a diseasedo notresembleitscauseor its butoftenthesymptoms sanitation andantibiotics cure;consider (p 620). I speculatethattheimmense popularity oforganicfoodsowesa aboveall totheviewthatthere greatdealto heuristic-driven thinking, is an associationbetweenthenaturaland thehealthy, and between To theextentthatpeopletrustscientifically chemicalanddanger.67 du64 See Heath, Bell, and Stemnberg, 81 J Personality& Soc Psychat 1032-39 (cited in note 32) (presentingthreestudieswherestorieswithhighdisgustfactorstendedto spreadfasterthan others). 65 NancyKraus,TorbjornMalmfors, and Paul Slovic,IntuitiveToxicology:Expertand Lay Judgments of ChemicalRisks,in Slovic,ed, The Perceptionof Risk285,290-91 (citedin note 22). 66 Thomas Gilovichand KennethSavitsky, Like Goes withLike: The Role of Representativenessin Erroneousand Pseudo-Scientific heuristicBeliefs(p 617) (reviewingrepresentative related errorsin medical beliefs,astrology,graphology,and psychoanalysis).Some of these themesare illuminatingly addressedinThomasGilovich,How WeKnow WhatIsn'tSo: TheFallibilityofHuman Reason in EverydayLife(Free Press 1995) (positingthattherepresentative heuristic,the clustering illusion,and erroneousperceptionof randomdispersionsaccountforeveryday intuitivefallacies). 67 For criticism of thatassociation,see JamesP. Collman,NaturallyDangerous:Surprising Factsabout Food, Health,and theEnvironment (UniversityScience 2001) (debunkingthebelief thatnaturaland organicqualitiescorrespondto safetyand wholesomeness). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 769 it is oftenbecause theyare neglecting bious curesand treatments, heuristic, and misbase-rates, makingselectiveuse oftheavailability of inevitably produces apthe effects randomness, which perceiving butit can be corrected parentpatterns."' Here SystemI is theculprit, bySystemII. B. TheAffect Heuristic In emphasizing theaffect refer heuristic, KahnemanandTversky This to thechapterofthattitlebyPaul Slovicandseveralcoauthors.69 andsuggestive in thevolume.It chapteris one ofthemostinteresting law and polalso createsnumerouspuzzles,manyof theminvolving icy. 70 responseto objects People oftenhave a rapid,largelyaffective and situations, consumerproducts, animals, including job applicants, A an action. have immediate negative jurymight cars,and causes of in a personalinjurycase; a judgemighthave a reactionto a plaintiff reactionto an equal protection claim;an employer positiveintuitive mightinstantly like,or dislike,someonewho has appliedfora job.71 is a "heuristic"? But whatdoes itmeanto saythataffect Slovic,et al, and responsesoccurrapidlyand automatically, urgethatouraffective fora moresysthatpeople use theirfeelingsas a kindof substitute It is in thissensethatattribjudgment.72 tematic, all-things-considered maybe at ute substitution, as meantby Kahnemanand Frederick, fora morereflective assesswork;affecttowardan objectsubstitutes 68 See Nassim Nicholas Taleb, Fooled by Randomness:The Hidden Role of Chance in the examplesof how randomMarketsand in Life (Texere2001) (givingmanyentertaining areperceived as inevitable HowWeKnowat 133-45 effects consequences); Gilovich, generated newage,andalternative beliefsinholistic, medicine). (citedinnote66) (debunking 69 PaulSlovic, et al,TheAffect Heuristic (p 397). 70 I haveelsewhere discussedan earlierandlesselaborateversionofSlovic'sworkon afTheLawsofFear,115HarvL Rev here.See CassR. Sunstein, andI drawonthatdiscussion fect, 1119(2002)(reviewing ofRisk(citedinnote22)). Slovic,ed,ThePerception 71 See Timothy David B. Centerbar, andNancyBrekke, MentalContamination D. Wilson, in employment decimentalcontamination and theDebiasingProblem(pp 198-99)(discussing as itrelatestoracialdiscrimination andTitleVII recovery). sions,especially 72 Fortheviewthatemotions see MarthaNussbaum, area formofcognition, Upheavals of Thought:The Intelligenceof Emotions 19 (Cambridge 2001) (arguing that emotions"involve in which, an external aboutimportant objectas salient judgments appraising judgments things, we acknowledge ourownneedinessandincompleteness beforepartsof forourownwell-being, Fortheviewthatcognition theworldthatwe do notfully control"). playsa largeroleinproducing emotions,see JonElster,Alchemiesof theMind: Rationalityand the Emotions(Cambridge withtheseviews.Slovicdoesnotdenythat 1999).It is notclearthatSlovic'sworkis inconsistent It is clear,however, or eventhataffect is a formofcognition. has cognitive affect antecedents, as Nussbaum canlead toerrors, as indeedcanmanyemotional thattheaffect heuristic reactions, in whatsenseemotional at 46-48,51 (discussing agrees.See Nussbaum, UpheavalsofThought andimpressions judgments maybe falseormistaken). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 770 TheUniversity ofChicagoLaw Review [70:751 mentof theobject.Affectis an exampleof SystemI in operation quickbuterror-prone. to treat"afButthereis an obvioussenseinwhichitis unhelpful forsomeone'sattitude towardobjects.In some fect"as an explanation oris,thatveryattitude, andtherefore cannot affect settings, represents, toexplainTom'sromanexplainoraccountforit.(Woulditbe helpful toAnnebysayingthatAnneproducesa favorable affect ticattraction thatsometimes inTom?)Slovic,et al,mustbe urging affect worksin andrepresentativeness: In manycontexts, thesamewayas availability fora morecarefulinpeople'semotionalreactionsare substituting quiryintothe(factual?)issuesat stake. The simplest wayto establishthiswouldbe to proceedas Kahforexample, thatpeonemanandTversky originally did,byshowing, ple assessquestionsofprobability byreference to affect, andthatthis errors. Whatis theprobability of death methodleads to predictable If people'saffect or eatingpesticides? fromsmoking, driving, flying, matchedtheirprobability towardtheseactivities judgments, producitwouldcertainly be plausibleto speakofan afingsystematic error, fectheuristic. Slovic,et al, do nothave data of exactlythissort.But of an affect theydo have some closelyrelatedevidence,suggestive in thedomainofrisk(pp 410-13).Whenaskedto assessthe heuristic risksand benefitsassociatedwithcertainitems,people tendto say andthatbeneficial activities thatriskyactivities containlowbenefits, as bothhighly containlowrisks.It is rarethattheywillsee an activity beneficialand quitedangerous, or as bothbenefit-free and dangerfree.Becauseriskandbenefit aredistinct this seems concepts, finding to suggestthat"affect" comesfirst, andhelpsto"direct" of judgments bothriskandbenefit. thishypothesis ofthese Twostudiesfortify (pp 411-12).Thefirst newinformation abouttherisksassociatedwithsome testswhether itemalterspeople'sjudgments aboutthebenefits associatedwiththe item-and whethernew information about benefitsalterspeople's The motivation aboutrisks. forthisstudyis simple.Ifpeojudgments andcalculative, wereanalytical information aboutthe ple'sjudgments shouldnotproducea judggreatbenefits of,say,foodpreservatives mentthattherisksare low-just as information aboutthegreatrisks of,say,naturalgas shouldnotmakepeoplethinkthatthebenefits are low.Strikingly, information aboutbenefits altersjudgments however, aboutrisks, andinformation aboutrisksaltersjudgments aboutbenefits. Whenpeoplelearnaboutthelowrisksofan item,theyaremoved to thinkthatthebenefits are high-andwhentheylearnaboutthe ofan item,theyaremovedtothinkthattherisksarelow. highbenefits The conclusion is thatpeople assessproductsand activities through affect -and thatinformation thatimprovespeople's affective re- This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 771 ofthoseprodof all dimensions sponsewillimprovetheirjudgments uctsandactivities. The secondstudyasked people to make decisionsundertime is claimis thatthe affectheuristic pressure(p 412). The motivating esin thesensethatitpermits thananalyticprocessing moreefficient Slovic,et al,hypotheUndertimepressure, peciallyrapidassessments. besize thattherewouldbe an unusuallystronginversecorrelation becauseaffectwillbe thedetweenjudgedriskand judgedbenefit, and peoplewillhaveless timeto undertake of assessment, terminant thekindof analysisthatcouldbeginto pullthetwoapart(p 412). In whentimeis scarce,and in otherwords,SystemI is mostimportant The hymonitor. suchcircumstances, SystemII willbe a less effective theinversecorrelation Undertimepressure, is pothesisis confirmed: timepressure. thanwithout evenstronger The affectheuristic castsa numberoffactsin a newlight.Backinmany decisionsandreactions groundmood,forexample,influences factthatstockpricesincreasesigConsidertheremarkable domains.73 on sunnydays,a factthatis hardto explainin termsthatdo nificantly 74 notrelyon affect.Anotherstudysuggeststhatwhenpeopleare anxprocessing, theyarelesslikelytoengageinsystematic iousandfearful, Note herethatthereis and henceSystemII is especiallyunreliable.75 If andtheemotions: betweensocialinfluences an evidentrelationship increase emotionsweakensystematic theysimultaneously processing, Fear itselfis likelyto to theapparentviewsof others.76 susceptibility to theacceptanceoffaulty logicand to presmakepeoplesusceptible sureto conform.77 implications Theauthorsemphasizeanotherpointwithimportant Whenan outcomeis accompaniedbystrongemoforriskregulation: littleweighton peoin probability have surprisingly tions,variations 78 ple's decisions.Whatmattersare theimagesassociatedwiththere73 and DecisionMaking,in WilliamM. Goldsteinand See Alice M. Isen,Positive Affect and Decision Making:Currents, Connections,and Robin M. Hogarth,eds,Researchon Judgment onthinkaffect "theimpactofmildpositive Controversies 1997)(exploring 509,512(Cambridge ingandmotivation"). 74 See David Hirshleifer and the StockReturns andTylerShumway, Good Day Sunshine: Weather 19-21, 30, Dice Center WorkingPaper No 2001-3 (2001), online at a strongcorrelation between (visitedFeb 18,2003) (finding http://papers.ssm.com/id=265674 incitiesacrosstheworld). sunshine foreachofseveralexchanges andmorning stockreturns 75 See Chaikenand Trope,Dual-ProcessTheories 19-20 (citedin note52) (notingthat stress orotherwise toa healththreat, experiencing "peoplewhoareanxiousaboutorvulnerable ofhealth-relevant orlessextensive information"). processing mayengageinlesscareful 76 4 Personality& Soc See Robert Baron, Arousal, Capacity,and IntenseIndoctrination, arousalleadstohigher emotional thatintense susceptibilPsychRev238,244-46(2000)(finding itytoindoctrination). 77 See id at 244. 78 See Cass R. Sunstein, a neglectofprob112Yale L Jat 70-82(citedinnote25) (finding andreal-world basedon a reviewofexperimental evidence). ability This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 772 TheUniversity ofChicagoLaw Review [70:751 in a studyofpeoconfirmation sult.The pointhas receivedempirical or to payto avoidelectricshocks, to be able to kiss ple's willingness In onestudy, topaytoavoid moviestars.7 favorite people'swillingness an electricshockvariedlittle,dependingon whether itsprobability or99 percent!80 Withrespecttohope,thosewhooperate was1 percent casinosandstatelotteries are well-aware oftheunderlying gambling mechanisms. sense Theyplayon people'semotionsin theparticular thattheyconjureup palpablepictures ofvictory andeasyliving. With respectto risks,insurancecompaniesand environmental groupsdo thesame. exactly It followsthatifgovernment is seekingto encouragepeopleto avoidlargerisks,and to worryless oversmallrisks, it mightwellattemptto appealto theiremotions, perhapsbyemphasizing theworstIt shouldbe no surprise case scenario. thatsomeofthemosteffective efforts to controlcigarette smokingappeal to people'semotions, by themfeelthatiftheysmoke,theywillbe dupesofthetobacco making companiesor imposeharmson innocent thirdparties.8 Thereis also hereto tryto activateSystemII, bypromoting an opportunity critical of reactionsthatare based on "affective scrutiny ties"in cases in whichpeopleareneglecting seriousrisksorexaggerating them. IV. ARE PEOPLEUNREALISTICALLY OPTIMISTIC? Withrespectto mostoftherisksoflife,peopleappearto be unThisclaimis closelyrelatedto thesuggestion, realistically optimistic.8 withprominent advocatesin economics, thatpeoplemayattempt to reducecognitive dissonancebythinking thattheriskstheyfaceare lowerthantheyare in fact.&3 Ifpeoplesystematically understate risks, thereis a seriousproblemforlaw and policy, and a seriousproblem 79 Rottenstreich and Hsee, 12 PsychSci at 186-88 (cited in note 62) (findingthatwhen variationsin probability). strongemotionsare present,individualsignoreimportant 80 See id at 188 (also reporting thattherewas a largespread,on thebasis ofprobability, for loss of $20, where the median willingnessto pay was $1 fora 1 percent the less "affect-rich" chanceof loss and $18 fora 99 percentchanceof loss). 81 See Lisa K. Goldmanand Stanton A. Glantz,Evaluation ofAntismoking Advertising Campaigns,279 JAMA 772 (1998) (findingthatmore aggressiveanti-smoking advertisements thatportraycigarettemakersas manipulativeand emphasizethedangersof secondhand smoke are farmore effectivethan ads emphasizingyouthaccess,romanticrejection,and the known shortand longtermeffectsof smoking). 82 See ShelleyE. Taylor,PositiveIllusions:CreativeSelf-Deception and theHealthyMind611 (Basic 1989) (proposingthata healthyself-conception is biased towardoptimism). 83 See GeorgeA. Akerlof andWilliam T. Dickens,TheEconomicConsequences ofCogni- tiveDissonance,in George A. Akerlof,ed, An Economic Theorist'sBook of Tales: Essays That EntertaintheConsequencesof NewAssumptionsin Economic Theory123-44 (Cambridge1984) (advancingan economicmodel of cognitivedissonancewithimplicationsforstandardeconomic puzzlessuchas thesalienceofnoninformational and thepopularityofSocial Security advertising legislation). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 773 too forthosewho acceptthe rationalactormodelin the social sciences. A. Evidence involverelative ofoptimism findings The mostwell-documented (as opposedto absolute)risk.About90 percentof driversthinkthat theyare saferthantheaveragedriverandlesslikelyto be involvedin a seriousaccident.8People generallythinkthattheyare less likely thanotherpeople to be divorced,to have heartdisease,to be fired glance,a beliefin relativeimmufroma job,and muchmore.8At first does notesbutbyitselfthisfinding nityfromriskseemsdisturbing, the risksthattheyactuallyface. tablishthatpeople underestimate oftheirownstatistiPerhapspeoplehave an accurateunderstanding thatotherpeople are cal riskseveniftheysay and believe,wrongly, Withrespectto absoluterisk,theevithantheyare.86 morevulnerable is lessclear,as DanielArmorandSheldenceforunrealistic optimism Forsignificant to thiscollection. leyTaylorshowin theircontribution people unwanted pregnancy, relevantevents, including andpersonally of theirsusceptibility (p 335).With showan accurateunderstanding risks life-threatening events, including respectto somelow-probability theirownsuscepsuchas AIDS, peopleactuallytendto overestimate 87 bias (id). One surandin thatsenseseemto showpessimistic tibility, of personalrisklevelsforsuchhazveyfindsgeneraloverestimates ardsas breastcancer(wherewomenratetheiractualriskas 40 percent,withthe actualriskbeingroughly10 percent);prostatecancer withtheactualrisk (wheremenranktheiractualriskas 40 percent, at 35 percent, 10 percent);lungcancer(estimated againbeingroughly comparedto an actualriskof under20 percent);and stroke(estimatedat 45 percent,comparedto an actualriskof roughly20 percent).88 See Taylor,PositiveIllusionsat 10-11 (cited in note 82). to HealthProblems,10 J See Neil D. Weinstein,UnrealisticOptimismabout Susceptibility Behav Med 481, 486 (1987) (listingresultsfromsurveyobtainingcomparativeriskjudgments from297 individualsabout a broad rangeof hazards). 86 See W. Kip Viscusi,Smoke-FilledRooms: A Postmortem on the Tobacco Deal 162-66 (Chicago 2002) (using surveydata to show that smokersdo not ignore risksto themselvesas muchas underestimatethemin relationto othersmokers). 87 Armorand Taylordoubtthisconclusion,suggesting that"these estimatesmaysimplyreand reportingextremepossibilities"(p 335). flectdifficulties interpreting 88 See HumphreyTaylor,Perceptionsof Risks:The Public Overestimates theRisks of Most Major Diseases and TypesofAccidents-Breast and ProstateCancer in Particular(Jan27, 1999), (visitedFeb 18, 2003) online at http://www.harrisinteractive.com/harris_poll/index.asp?PID=44 to hazards salient (reportingsurveydata showingthatthe public overestimatesits susceptibility in the media). 84 85 This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions ofChicagoLaw Review TheUniversity 774 [70:751 theirstatistical people do underestimate But in some domains, overestifinancial expertsconsistently risk.Forexample, professional and businessschoolstudents their matelikelyearnings, overestimate thattheywillreceive salariesandthenumberofoffers likelystarting theirlikelihoodof beingin(pp 334-35).People also underestimate andtheirownfailuretobuy accident,89 volvedin a seriousautomobile is at leastconsistent withthe insuranceforfloodsand earthquakes Theevidenceofoptimisoptimistic. viewthatpeopleareexcessively to raise questions tic bias,bothrelativeand absolute,is sufficient interventions. aboutinformational andregulatory B. Debiasing? ofapparently andWilliamKleinexplorea variety NeilWeinstein biaswithrespectto relative promising strategies to reduceoptimistic The punchline? None of thesestrategies worked.One study risk.9' asked people to generatetheirown listof personal"factors"that mighteitherincreaseor decreasetheirriskof developinga weight thatan problem.The authorshypothesized problemor a drinking biasinmanycases; offactors woulddecreaseoptimistic identification was observed. Norwas optimistic biasreducedbyaskno sucheffect toreadaboutmajorriskfactors forcertainhazards, to ingparticipants and to offeran reporttheirstandingwithrespectto thesefactors, tothelistoffactors. Theauthors afterresponding overallriskestimate targets(such concludethat"healthcampaigns emphasizing high-risk ofsmokers) interventions thatshowunattractive as smoking pictures aboutundesirable actions(as information and campaigns conveying withpamphlets factors thatraisetheriskfora particular health listing worsentheverybiasestheyaredesignedto problem)mayunwittingly reduce"(p 323). has beenfoundto reduce As theauthorsnote,one intervention bias: information about theirownstanding on optimistic giving people riskfactors or abouttheirpeers'standing on riskfactors. Buttheyobto serve,sensibly enough,thatit is noteasyto adaptthisinformation mediacampaigns designedto improvehumanhealth.Thisis a valuof literature on thepossibility ablepaper,addingto thestill-emerging 89 ChristineJolls,BehavioralEconomicsAnalysisof Redistributive Legal Rules,51 Vand L Rev 1653,1660-61(1998). 90 See id at 1658-62 (discussingindividuals'failureto insureas a functionof theiroveroptimism).Note thatthe availabilityheuristiccan counteractthisproblemby leadingpeople to insureagainstsalientrisks. 91 Neil D. Weinsteinand WilliamM. Klein,Resistanceof PersonalRisk Perceptions to DeincludbiasingIntervention, (pp 313-22) (discussingfourstrategiesforavoidingover-optimism, individualsof theirtruerisk,requiringcomparisonswithlow ratherthanhighrisk inginforming waysto altertheirriskfactors). groups,and havingindividualsbrainstorm This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 775 ofSystemII). Butbecausethefocusis on debiasing(or theactivation thefindings do not offerclearguidance the"above average"effect, aboutcampaignsdesignedto givepeoplea bettersenseofthestatistiIt wouldbe valuableto learnmoreaboutthattopic. cal reality. Fools? C. Optimistic withsomeobviDavid ArmourandShelleyTaylorareconcerned ous puzzles:If people are excessively optimistic, whydon'ttheypurandblunder?Whydon'talertpeople sue ambitious goalsrecklessly -take systematic psychologists? entrepreneurs? advantageof human optimism? Thisis whatArmorandTaylorsee as the"dilemmaofun-the likelihoodthatif real,thisbias wouldprorealisticoptimism" If optimism werewidespread, duce extremely seriousharmful effects. we shouldprobablysee farmorerecklessness and failurethanwe generally observe. The authorsresolvethedilemmabygivinga morerefinedsense of the natureof optimistic bias.In theirview,people are not indisor Theirpredictions are usuallywithin criminately blindlyoptimistic. reasonablebounds(p 346). People are less likelyto be optimistic whentheconsequencesof errorare severe(p 339). In addition, optimismdecreasesiftheoutcomewillbe knownin thenearfuture(id). stateof Optimismalso decreaseswhenpeople are in a predecisional Whenpeoplearechoosingamonggoals,or amongpossideliberation. ble coursesofaction,thebiasis attenuated, anditincreasesagainonly afterpeople have selectedgoals and beginto implement theirplans (p 340).Thereis also evidencethatoptimistic bias,whenit exists,can be adaptive,72 the leadingto (almost)self-fulfilling policies,increasing likelihoodofsuccess(p 341). Theseclaimsraiserealdoubtsabouttheviewthatoptimistic bias To be sure,we providesa good reasonforpaternalistic interventions. knowenoughaboutoptimistic biasto giveseriousconsideration to informational campaignsto ensurethatpeoplewillnothavean inflated In the contextof smoking, beliefin theirown immunity. statistical ofrisks"3 knowledge mightbe inadequateifpeoplebelievethemselves immune.94 But in viewofthearguments relatively byArmorandTayis generally lor,theidea thatpaternalism justifiedby optimistic bias Ifpeoplearenotexcesmustbe regardedas an unproven speculation. whentheconsequencesoferroraresevere,ifthebias sivelyoptimistic A generaltreatmentis Taylor,PositiveIllusions (cited in note 82). See Viscusi,Smoke-FilledRooms at 221 (cited in note 86) (proposingas a way to reduce effortsat providingcomparativeriskinformation"). smoking"thatwe targetour informational 94 See JohnZ. Ayanian and Paul D. Cleary,PerceivedRisks of Heart Disease and Cancer among CigaretteSmokers,281 JAMA 1019,1020-21 (1999) (findingthatmostsmokersthinkthat theirrisksare average or below average). 92 93 This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 776 The University of ChicagoLaw Review [70:751 is smallor nonexistent whendecisionsareactually beingmade,andif peopleoverstate low-probability risks, thereis no problemforthelaw tocorrect. V. MORALHEURISTICS? The heuristics-and-biases literature was originally focusedon issues of probability, and whilemanyof thechaptersgo beyondthat topic,theydo notmuchdeal withnormative questions-with therole in ofheuristics informing judgments aboutmorality andpolitics." It is therulesofmorality naturalto wonderwhether also haveheuristics (isn't thatinevitable?), and whetherthe normative judgmentsinvolvedinlawandpoliticsare alsoproneto heuristics, or to rapidSysII override. andtopossibleSystem temI assessments Wecanimaginesomeambitious claimshere.Armedwithpsychologicalfindings, utilitarians mightbe temptedto claimthatordinary moralcommitments are a setofmentalshortcuts thatgenerally work well,butthatalsoproducesevereandsystematic errors(Is retribution a cognitive error?Is Kantianism?). On one view,muchof everyday withfairness, concerned shouldbe seenas a setof morality, nominally heuristics fortherealissue,whichis howto promoteutility. Fortheir couldeasilyturnthetables.Deontologists part,deontologists might wellclaimthattherulesrecommended are consistent, byutilitarians muchofthetime,withwhatmorality -but also thatutilitarirequires anism,takenseriously, producesmistakesin somecases.Theselarge to be tractable, debatesare unlikely and simplybecauseutilitarians tobe convinced aremostunlikely that deontologists bythesuggestion theirdefining commitments aremereheuristics. Butinsomecases,we thehypothesis that mightbe able to makeprogressby entertaining certainwidelyacceptedrulesofmorality are heuristics. for Consider, example,theidea thatone should"neverlie" or"neversteal"-good butinjunctions rulesofthumb, thatbadlymisfire certainly, (sanepeois neededtoprotect thedeathsof ple think)whenthelie,or thetheft, innocent ofrelepeople."'I turnto severalpossible"moralheuristics" vancetolaw.97 95 An exceptionis the suggestivediscussionby PhilipTetlock.See PhilipE. Tetlock,IntuitivePoliticians,Theologians,and Prosecutors(pp 596-98) (urgingthatmanypeople believe in "taboo tradeoffs," and thatwe mightsee suchpeople notas defectiveintuitiveeconomists, butas defendersofsacredvalues). 96 Note the rule-utilitarian defenseof these ideas: They mightmisfirein particularcases, butit mightbe best forpeople to treatthemas firmrules,because a case-by-caseinquirywould warrantan exceptionto the prove even more errors.If people ask whetherthe circumstances prohibitionon lyingor stealing,theremightwell be excessiveor self-serving lyingand stealing. The strongvoice ofconscience-callingforadherenceto whatI am callingmoralheuristicseven - probablyservessome valuable social functions. in cases in whichtheybadlymisfire For fallible humanbeings,a decision to go rightto the issue of consequences,withoutfirmmoral rules of This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 777 A. Pointless Punishment aboutpunishment withthe seemdisconnected People'sintuitions in a waythatsuggestsa moralheuristic consequencesofpunishment, studyof people's is at work.Consider,forexample,an intriguing harmsfromvaccinesand judgments aboutpenaltiesincasesinvolving birthcontrolpills.98In one case,subjectsweretoldthattheresultofa higherpenaltywouldbe to makecompaniestryharderto makesafer In an adjacentcase,subjectsweretoldthattheconsequence products. ofa higherpenaltywouldbe to makethecompanymorelikelytostop withtheresultthatless safeproductswouldbe makingtheproduct, a groupofjudges,gave the on the market.Most subjects,including samepenaltiesin bothcases.Can thisoutcomebe defendedin principle? Perhapsitcan,butitis moresensibleto thinkthatpeopleare opto theeffect thatpenaltiesshouldbe a proeratingundera heuristic, oftheact,andshouldnotbe portionalresponseto theoutrageousness considerations. basedon consequential If thisclaimseems too adventurous, considera similartestof whichaskedsubjects, including judgesandlegpunishment judgments, In one islators,to choose penaltiesfordumpinghazardouswaste.99 thumb,would likelyproduce seriousproblems. 97 For a discussionofmoralissues thatappeals to intuitions, see F.M. Kamm,Responsibility and Collaboration,28 Phil & Pub Aff169, 173 (1999) (discussingconsequentialismand collaborationwithevil). Kamm's treatmentis extremelyimpressive, but it seems to place undue emphasis on moralintuitionsabout exoticcases of thekindneveror rarelyencounteredin ordinarylife. I believe thatthe relevantintuitionsordinarilyworkwell,but thatwhen theyare wrenchedout of familiarcontexts,in whichtheymake a greatdeal of sense,theirreliability, forpurposesof legal and moral analysis,is unclear.Consider the followingintuition:Do not kill an innocentperson, even if thisis necessaryto save others.In all likelihood,a societydoes much betterifmost if onlybecause judgmentsabout necessityare likelyto be unreliable people have thisintuition, But in a hypotheticalcase, in whichit reallyis necessaryto killan innocentperand self-serving. son to save fiveothers,our intuitionsmightwell turnout to be unclearand contested.And ifour intuitionsabout the hypotheticalcase turnout to be very firm(do not kill innocentpeople, ever!), theymightnot deserve to be so firm,simplybecause theyhave been wrenchedout of the real worldcontext,whichis wheretheyneed to be to make sense. I wonderwhethersome legal and philosophicalanalysis,based on exotic moral dilemmas, mightnot be replicatingthe early work of Kahneman and Tversky:uncoveringsituationsin whichintuitions, normallyquite sensible,turnout to misfire.The ironyis that Kahneman and some philosophers,includTverskymeantto devise cases thatwould demonstratethe misfiring; ing Kamm, devise cases withthe thoughtthat the intuitionsare reliable and should formthe buildingblocks forsound moral judgments.An understandingof how heuristicswork suggests even when theyare veryfirm.Much workrereason to doubt the reliabilityof those intuitions, mains to be done on thiscomplicatedtopic;mydiscussionin thisPart is intendedas a tentative start.For a more detailed discussion,see Cass R. Sunstein,Moral Heuristics(forthcoming). 98 See JonathanBaron and Ilana Ritov,Intuitions about Penaltiesand Compensationin the Contextof TortLaw, 7 J Risk & Uncertainty17, 17 (1993) (reportingthatpeople use overgeneralized rulesto determinepenalties,ignoringtheirdeterrenteffectson futurebehavior). 99 See JonathanBaron, et al, AttitudesTowardManaging Hazardous Waste,13 Risk Anal judges,law183, 183 (1993) (reportingthatin a studyof CEOs, economists,environmentalists, makers,and hazardous waste policyexperts,retributivepenaltieswere preferredover welfare- This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 778 TheUniversity ofChicagoLaw Review [70:751 case,thepenaltywouldmakecompanies tryharderto avoidwaste.In thepenaltywouldlead companiesto cease makinga benefianother, in cial product.Mostpeople did notpenalizecompaniesdifferently to require the twocases.Perhapsmoststrikingly, people preferred companiesto clean up theirown waste,even if the wastedid not threaten anyone, insteadofspending thesameamounttocleanup far moredangerous wasteproducedbyanother, now-defunct company. It is reasonabletobelievethatinthinking aboutpunishment, peopleuse a simpleheuristic, the now-familiar Thisheuristic outrageheuristic. butin somecases, producesreasonableresultsin mostcircumstances, itseemstoleadtosystematic errors. B. Aversion to Cost-Benefit Analysis An automobilecompanyis decidingwhetherto take certain foritscars.In deciding todo so,itconducts whether safety precautions a cost-benefit in whichit concludesthatcertainprecautions analysis, are notjustified-because, say,theywouldcost$100millionandsave of$10million hasa "ceiling" onlyfourlives,andbecausethecompany per lifesaved.How willordinary peoplereactto thisdecision?The answeris thattheywillnotreactfavorably.'?? Theytendtopunishcomon that base their decisions cost-benefit evenifa high panies analysis, valuationis placedon humanlife.By contrast, theydo notmuchpunishcompaniesthatare willingto imposea "risk"on people.'0' What underlies thesemoraljudgments? It is possiblethatwhenpeopledisapprove oftrading moneyfor froma setofmoralprinciples risks, theyaregeneralizing thataregenerallysound,and even useful,but thatworkpoorlyin some cases. moralprinciple: Do not knowingly Considerthe following cause a humandeath.People disapproveof companiesthatfailto improve safetywhentheyare fullyaware thatdeathswillresult-whereas peopledo notdisapproveof thosewhofailto improvesafetywhile notto know,forcertain, thatdeathswillensue.Companies appearing thatfailto do cost-benefit butthatareawarethata "risk"exanalysis, or to jurors,thattheycaused ists,do notmakeclear,to themselves thatthiswas whattheyweregoingto do. deathswithfullknowledge above all, of companiesthatcause deathknowPeople disapprove, maximizingpenaltiesforknowingpolluters). 100 See Viscusi,52 Stan L Rev at 586-90 (citedin note41) (concludingthatjurorsholda varietyof biases thathinderappropriateapplicationof cost-benefit analysis,includingpossiblya bias againsttheuse ofcost-benefit analysisitself). 101 See id. See also PhilipE. Tetlock,CopingWithTradeoffs: PsychologicalConstraints and PoliticalImplications, in ArthurLupia, MathewD. McCubbins,and Samuel L. Popkin,eds, ElementsofReason: Cognition,Choice,and theBounds ofRationality 239,252-57 (Cambridge2000) different "taboo tradeoff'behaviorsand theirpoliticalimplications). (summarizing This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 779 is at work,one thatimI suggest, then,thata genuineheuristic ingly. engagein acts on thosewho knowingly poses moralcondemnation thatwillresultinhumandeaths. to cause death The problemis thatit is notalwaysunacceptable fewand an unintended at leastifthedeathsare relatively knowingly, of generallydesirableactivity. Much of whatis done,by byproduct is likelyto resultin one or more both industry and government, to deaths.Of courseit wouldmakesense,in manyof thesedomains, does notsupport takeextrastepsto reducerisks.Butthatproposition fromthemoralpoint theimplausible claimthatwe shoulddisapprove, ofview,ofanyactiontakenwhendeathsareforcseeable. in principle, thewideI believethatit is impossibleto vindicate, Butto adapta claim to cost-benefit balancing. spreadsocialantipathy heuristic byStephenJayGould(p 68),"a abouttherepresentativeness in myhead continuesto jumpup and down,shoutlittlehomunculus tradingdollarsfora analysis, ing at me" thatcorporatecost-benefit The voice of the knownnumberof deaths,is morallyunacceptable. I am suggesting, butinis nottheresultof conscience, homunculus, steadofa crudebutquitetenaciousmoralheuristic. C. ActsandOmissions ofwhether andwhythedistincTherehas beenmuchdiscussion tionbetweenacts and omissionsmightmatterforlaw and policy.In one case,forexample,a patientmightask a doctornotto providelifethusensuringthe patient'sdeath.In another equipment, sustaining case,a patientmightask a doctorto injecta substancethatwillimmemoralintuidiatelyendthepatient'slife.Peopleseemtohavea strong andeventhe equipment, tionthatthefailuretoprovidelife-sustaining -but that is acceptableandlegitimate ofsuchequipment, withdrawal And indeedconstitutional law retheinjectionis morallyabhorrent. But whatis themorallyrelevantdifto thiseffect.'02 flectsjudgments ference? thatthe action-omission It is worthconsidering the possibility asfora morecomplexanddifficult distinction operatesas a heuristic sessmentof themoralissuesat stake.Fromthemoralpointofview, in termsof worsethanharmful omissions, harmful actsare generally and thelikelyconsequences boththestateofmindofthewrongdoer of the wrong.But harmful acts are not always worsethanharmful The moralpuzzlesarisewhenlife,or a cleverinterlocutor, omissions. comesup witha case inwhichthereis no morallyrelevantdistinction drivenby a but whenmoralintuitions, betweenacts and omissions, 102 See Washington v Glucksberg,521 US 702,724-25 (1997) (upholdingthe stateofWashington'slaw thatprohibitsaidinga suicide). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 780 TheUniversity ofChicagoLaw Review [70:751 heuristic thatopposesactionsmorethanomissions, strongly suggest In suchcases,we might thattheremustbe sucha difference. hypotheof principles size thatmoralintuitions reflectan overgeneralization thatusuallymakesense-but thatfailto makesensein theparticular moralintuitions case.103In otherwords, reflect SystemI; theyneedto be corrected bySystemII. I believethatthepersistent acceptanceof of life-saving withdrawal equipment,alongsidepersistentdoubts abouteuthanasia, is a demonstration ofthepoint. Considerin thisregardthedisputeovertwowell-known problemsin moralphilosophy.04 The first, calledthetrolley problem, asks people to supposethata runawaytrolleyis headedforfivepeople, whowillbe killedifthetrolleycontinues on itscurrent course.The questionis whether youwouldthrowa switchthatwouldmovethe trolley ontoanothersetoftracks, killingone personratherthanfive. Most people wouldthrowthe switch. The second,called the footis thesameas thatjustgiven,butwithone difference: bridgeproblem, The onlywayto save thefiveis to throwa stranger, nowon a footbridgethatspansthetracks, intothepathof thetrolley, killingthat but preventing the trolleyfromreachingthe others.Most stranger peoplewillnotkillthestranger. Butwhatis thedifference betweenthetwocases?A greatdealof workhas been doneon thisquestion, oftensuggesting philosophical thatourintuitions canbe defended inprinciple. Let mesuggest a simAs a matterofprinciple, pleranswer. thereis no difference between thetwocases.People'sdifferent reactions arebasedon moralheuristicsthatcondemnthethrowing ofthestranger butsupport thethrowIn thefootbridge ingoftheswitch. case,theheuristic says,Do notlay hands on people in orderto cause theirdeaths.Such heuristicsgener- Buttheymisfire a distincin drawing allypointin therightdirection. tionbetweenthetwocases.In thissense,theaction-omission distinctionleadsto systematic errors. Is thereanything to be saidto thosewhobelievethattheirmoral thetrolley judgments, andfootbridge areendistinguishing problems, andreflect no heuristic at all? Consideran intrigutirely deliberative, ingexperiment, designedto see howthehumanbrainresponds to the twoproblems. The authorsdo notattempt to answerthemoralquestionsinprinciple, buttheyfind"thattherearesystematic in variations 103 See Jonathan Baron,Nonconsequentialist Decisions,17 Beh and BrainSci 1,1 (1994) ("I suggestthatnonconsequentialist principlesarise fromovergeneralizing rulesthatare consistent withconsequentialismin a limitedset ofcases."). 104 See JoshuaD. Greene,et al, An fMRI Investigation of EmotionalEngagementin Moral 293 Sci 2105,2105-06(2001) (relyingon a studyofthetwofamiliarmoraldilemmasto Judgment, argue thatmoraldilemmasrequireemotionalthoughtprocessesto varyingdegrees,influencing moraljudgments). This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 2003] HazardousHeuristics 781 ofemotionsin moraljudgment,"'105 theengagement and thatbrainareas associatedwithemotionare farmoreactivein contemplating the thetrolley As in footbridge problemthanin contemplating problem.'06 thecase offear,wherean identifiable regionofthebrainmakeshelpfullyimmediate butnotentirely reliablejudgments,'07 andwhereother, in the so too,perhaps, also identifiable regionscan supplycorrectives, contextofmorality andlaw. D. Betrayals A betrayaloftrustis likelytoproducea greatdeal ofoutrage.Ifa babysitter neglectsa child,or ifa security guardstealsfromhis emthaniftheidenticalactswereperformed ployer, peoplewillbe angrier by someonein whomtrusthas notbeen reposed.So far,perhaps,so thatpeople willfavorgreater good.And it shouldnot be surprising Perforbetrayalsthanforotherwiseidenticalcrimes.108 punishment can be justifiedon thegroundthatthebetrayalof haps thedisparity trustis an independent harm,one thatwarrants greaterdeterrence fromthefactthattrust, -a pointthatdrawsstrength and retribution thatis harder once lost,is noteasilyregained.But considera finding to explain:People are especiallyaverseto risksof deaththatcome fromproductsdesignedto promotesafety, so muchso thatpeople as a resultofachavebeenfoundto prefera greaterchanceofdying, lowerchanceofdyingin a crash cidentsfroma crash,to a significantly as a resultofa malfunctioning airbag.109 Indeed,"mostpeoplearewillingto double theirchanceof dyingto avoid incurring a verysmall chanceofdyingviabetrayal.""10 Whatexplainsthisseemingly bizarreand self-destructive preferis at work:Punish,and neverreward, ence? I suggestthata heuristic The heuristic workswell.Butitmisfires in betrayals oftrust. generally somecases,as whenthosewho deployit end up increasing therisks face.An airbagis nota security theythemselves guardor a babysitter, It is a prodthosewhomtheyhavebeenhiredto protect. endangering 105 Id at 2106. Id (presentingevidence thatdifferent areas of the brain are affectedforthe two classic moralproblems). 107 See LeDoux, The EmotionalBrain at 157-69 (cited in note 17) (discussingthe amygdala as a regionof the brainresponsibleforperceivingfear). 108 See JonathanJ.Koehler and Andrew D. Gershoff, BetrayalAversion:WhenAgentsof ProtectionBecome AgentsofHarm,Org Beh & Human Dec Processes (forthcoming 2003) (relying on fiveempiricalstudiesto findthatacts of betrayalelicitedstrongerdesired punishments thanotherbad acts). 109 Id at 40 (findingthat"when faced witha choice among pairs of safetydevices ... most people preferredinferioroptions (in termsof risk exposure) to options that included a slim (0.01 %) riskof betrayal"). 110 Id at 33-34. 106 This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions 782 TheUniversity ofChicagoLaw Review [70:751 uct,to be chosenifit decreasesaggregate risks.If an airbag makes peoplesaferon balance,itshouldbe used,evenifina tinypercentage ofcasesitwillcreatea riskthatwouldnototherwise exist.To rejectair ofbetrayal bagson grounds is irrational butunderstandable -the sort ofmistake towhichheuristics oftenleadhumanbeings. Thedistinctive oftheanti-betrayal feature heuristic is thatitinvolves moralandlegal offact. judgments rather thanjudgments Theseare speculative on somecomplexsubjects. remarks Butif heuristics andsometimes lead people playa roleinfactual judgments, to makesystematic errors, thereis reasonto believethatheuristics also help producenormative judgments, bothmoraland legal,and sometimes produceerrorsthereas well.If thisis harderto demonitis largely strate, becausewe areableto agreeaboutwhatconstitutes errorin thedomainoffacts,and oftenless able to agreeaboutwhat constitutes errorin thedomainofvalues.I believethatHeuristics and Biases: The Psychologyof Intuitive willilluminateproblems Judgment oflaw andpolicyformanyyearsto come.Andwe shouldnotbe surprisediftheideas of attribute substitution, and of thecorrection of assessments rapid,intuitive by morereflective processes, have anaas well. loguesinmoralandlegalintuitions This content downloaded from 86.136.89.205 on Tue, 28 May 2013 13:35:04 PM All use subject to JSTOR Terms and Conditions