InmemoryofAmosTversky Contents Introduction PartlTwoSystems 1.ThecharactersoftheStory 2.Attentionandeffort 3.TheLazycontroller 4.TheAssociativemachine 5.CognitiveEase 6.Norms,Surprises,andcauses 7.AMachineforJumpingtoConclusions 8.HowJudgmentsHappen 9.AnsweringanEasierQuestion Partl.heuristicsandbiases 10.ThelawofsmallNumbers ≤5> 11.Anchors 12.TheScienceofAvailability 13.Availability,Emotion,andRisk 14.TomWsSpecialty 15.Linda:Lessismore 16.CausesTrumpstatistics 17.RegressiontotheMean 18.TamingIntuitivePredictions Partll.Overconfidence 19.TheIllusionofunderstanding 20.TheIllusionofvalidity 21Intuitionsvs.Formulas 22.ExpertIntuition:WhenCanWeTrustIt? 23.Theoutsideview 24.TheEngineofcapitalism PartⅣ.Choices 25.Bernoulli'serrors 26.ProspectTheory 27.Theendowmenteffect 28,BadEvents 29.TheFourfoldPattern 30.RareEvents 31RiskPolicies 32.Keepingscore 33.Reversals 34FramesandReality artvtwoselves 35.TwoSelves 36.LifeasaStory 37.ExperiencedWell-Being 38.ThinkingAboutLife Conclusions AppendixA:JudgmentUnder Uncertainty AppendixB:Choices,Values,andFrames Acknowledgments Notes Index Introduction Everyauthor,Isuppose,hasinmindasettinginwhichreadersofhisorher workcouldbenefitfromhavingreadit.Mineistheproverbialoffice watercooler,whereopinionsaresharedandgossipisexchanged.Ihope toenrichthevocabularythatpeopleusewhentheytalkaboutthe judgmentsandchoicesofothers,thecompany'snewpolicies,ora colleaguesinvestmentdecisions.Whybeconcernedwithgossip? Becauseitismucheasier,aswellasfarmoreenjoyable,toidentifyand labelthemistakesofothersthantorecognizeourown.Questioningwhat ebelieveandwantisdifficultatthebestoftimes,andespeciallydifficult whenwemostneedtodoit,butwecanbenefitfromtheinformedopinions ofothers.Manyofusspontaneouslyanticipatehowfriendsandcolleagues willevaluateourchoices;thequalityandcontentoftheseanticipated judgmentsthereforematters.Theexpectationofintelligentgossipisa powerfulmotiveforseriousself-criticism,morepowerfulthanNewYear resolutionstoimproveonesdecisionmakingatworkandathome Tobeagooddiagnostician,aphysicianneedstoacquirealargesetof labelsfordiseases.eachofwhichbindsanideaoftheilnessandits symptoms,possibleantecedentsandcauses,possibledevelopmentsand consequences,andpossibleinterventionstocureormitigatetheillness Learningmedicineconsistsinpartoflearningthelanguageofmedicine.A deeperunderstandingofjudgmentsandchoicesalsorequiresaricher vocabularythanisavailableineverydaylanguage.Thehopeforinformed gossipisthattherearedistinctivepatternsintheerrorspeoplemake Systematicerrorsareknownasbiases,andtheyrecurpredictablyin particularcircumstances.Whenthehandsomeandconfidentspeaker boundsontothestage,forexample,youcananticipatethattheaudience willjudgehiscommentsmorefavorablythanhedeserves.Theavailability ofadiagnosticlabelforthisbias-thehaloeffect--makes dierto anticipate,recognize,andunderstand Whenyouareaskedwhatyouarethinkingabout,youcannormally swer.Youbelieveyouknowwhatgoesoninyourmind,whichoften consistsofoneconsciousthoughtleadinginanorderlywaytoanother.But thatisnottheonlywaythemindworks,norindeedisthatthetypicalway Mostimpressionsandthoughtsariseinyourconsciousexperiencewithout yourknowinghowtheygotthere.Youcannottracryd>ehowyoucameto thebeliefthatthereisalamponthedeskinfrontofyou,orhowyou detectedahintofirritationinyourspousesvoiceonthetelephone,orhow youmanagedtoavoidathreatontheroadbeforeyoubecameconsciously awareofit.Thementalworkthatproducesimpressions,intuitions,and manydecisionsgoesoninsilenceinourmind Muchofthediscussioninthisbookisaboutbiasesofintuitionhowever thefocusonerrordoesnotdenigratehumanintelligence,anymorethan theattentiontodiseasesinmedicaltextsdeniesgoodhealthmostofus arehealthymostofthetime,andmostofourjudgmentsandactionsare appropriatemostofthetime.Aswenavigateourlives,wenormallyallow ourselvestobeguidedbyimpressionsandfeelings,andtheconfidence wehaveinourintuitivebeliefsandpreferencesisusuallyjustified.Butnot always.Weareoftenconfidentevenwhenwearewrong,andanobjective observerismorelikelytodetectourerrorsthanweare Sothisismyaimforwatercoolerconversations:improvetheabilityto identifyandunderstanderrorsofjudgmentandchoice,inothersand eventuallyinourselves,byprovidingaricherandmorepreciselanguageto discussthem.Inatleastsomecases,anaccuratediagnosismaysuggest aninterventiontolimitthedamagethatbadjudgmentsandchoicesoften cause Origins Thisbookpresentsmycurrentunderstandingofjudgmentanddecision making,whichhasbeenshapedbypsychologicaldiscoveriesofrecent decades.However,Itracethecentralideastotheluckydayin1969when askedacolleaguetospeakasaguesttoaseminarIwasteachinginthe DepartmentofPsychologyattheHebrewUniversityofJerusalem.Amos Tverskywasconsideredarisingstarinthefieldofdecisionresearch- indeed,inanythinghedid-soIknewwewouldhaveaninterestingtime ManypeoplewhoknewAmosthoughthewasthemostintelligentperson theyhadevermet.Hewasbrilliant,voluble,andcharismatic.Hewasalso blessedwithaperfectmemoryforjokesandanexceptionalabilitytouse themtomakeapoint.TherewasneveradullmomentwhenAmoswas around.Hewasthenthirty-two;Iwasthirty-five Amostoldtheclassaboutanongoingprogramofresearchatthe UniversityofMichiganthatsoughttoanswerthisquestion:Arepeople goodintuitivestatisticians?Wealreadyknewthatpeoplearegood intuitivegramm atagefourachildeffortlesslyconformstotherules ofgrammarasshespeaks,althoughshehasnoideathatsuchrulesexist Dopeoplehaveasimilarintuitivefeelforthebasicprinciplesofstatistics? Amosreportedthattheanswerwasaqualifiedyes.Wehadalivelydebate intheseminarandultimatelyconcludedthataqualifiednowasabetter answer AmosandIenjoyedtheexchangeandconcludedthatintuitivestatistics wasaninterestingtopicandthatitwouldbefuntoexploreittogether.That FridaywemetforlunchatCafeRimon,thefavoritehangoutofbohemians andprofessorsinJerusalem,andplannedastudyofthestatistical tuitionsofsophisticatedresearchersWehadconcludedintheseminar thatourownintuitionsweredeficient.Inspiteofyearsofteachingand usingstatistics,wehadnotdevelopedanintuitivesenseofthereliabilityof statisticalresultsobservedinsmallsamples.Oursubjectivejudgments werebiased:wewerefartoowillingtobelieveresearchfindingsbasedon inadequateevidenceandpronetocollecttoofewobservationsinourown research.Thegoalofourstudywastoexaminewhetherotherresearchers suferedfromthesameaffliction Wepreparedasurveythatincludedrealisticscenariosofstatistical ssuesthatariseinresearch.Amoscollectedtheresponsesofagroupof expertparticipantsinameetingoftheSocietyofMathematical Psychology,includingtheauthorsoftwostatisticaltextbooks.Asexpected wefoundthatourexpertcolleagues,likeus,greatlyexaggeratedthe likelihoodthattheoriginalresultofanexperimentwouldbesuccessfully replicatedevenwithasmallsample.Theyalsogaveverypooradvicetoa fictitiousgraduatestudentaboutthenumberofobservationssheneeded tocollect.Evenstatisticianswerenotgoodintuitivestatisticians Whilewritingthearticlethatreportedthesefindings,Amosand discoveredthatweenjoyedworkingtogether.Am funny,andinhispresenceIbecamefunnyaswell,sowespenthoursof solidworkincontinuousamusement.Thepleasurewefoundinworking togethermadeusexceptionallypatient;itismucheasiertostrivefor perfectionwhenyouareneverbored.Perhapsmostimportant,we checkedourcriticalweaponsatthedoor.BothAmosandIwerecritical andargumentative,heevenmorethanlbutduringtheyearsofour collaborationneitherofuseverrejectedoutofhandanythingtheother said.Indeed,oneofthegreatjoysIfoundinthecollaborationwasthat Amosfrequentlysawthepointofmyvagueideasmuchmoreclearlythan Amoswasthemorelogicalthinker,withanorientationtotheoryand unfailingsenseofdirection.Iwasmoreintuitiveandrootedinthe psychologyofperception,fromwhichweborrowedmanyideas.Wewere sufficientlysimilartounderstandeachothereasily,andsufficientlydifferent tosurpriseeachother.Wedevelopedaroutineinwhichwespentmuchof ourworkingdaystogether,oftenonlongwalks.Forthenextfourteenyears ourcollaborationwasthefocusofourlives,andtheworkwedidtogether duringthoseyearswasthebesteitherofuseverdid Wequicklyadopteda cethatwemaintainedformanyyears.our researchwasaconversation,inwhichweinventedquestionsandjointly examinedourintuitiveanswers.Eachquestionwasasmallexperiment andwecarriedoutmanyexperimentsin gleday.Wewerenot seriouslylookingforthecorrectanswertothestatisticalquestionswe posed.Ouraimwastoidentifyandanalyzetheintuitiveanswer,thefirst onethatcametomind,theoneweweretemptedtomakeevenwhenwe knewittobewrong.Webelieved-correctly,asithappened-thatany intuitionthatthetwoofussharedwouldbesharedbymanyotherpeople aswell,andthatitwouldbeeasytodemonstrateitseffectsonjudgments Weoncediscoveredwithgreatdelightthatwehadidenticalsillyideas boutthefutureprofessionsofseveraltoddlerswebothknewWecould identifytheargumentativethree-year-oldlawyer,thenerdyprofessor,the empatheticandmildlyintrusivepsychotherapist.Ofcoursethese predictionswereabsurd,butwestillfoundthemappealing.Itwasalso clearthatourintuitionsweregovernedbytheresemblanceofeachchildto theculturalstereotypeofaprofession.Theamusingexercisehelpedus developatheorythatwasemerginginourmindsatthetime,abouttherole ofresemblanceinpredictions.Wewentontotestandelaboratethat theoryindozensofexperiments,asinthefollowingexample Asyouconsiderthenextquestion,pleaseassumethatStevewas selectedatrandomfromarepresentativesample Anindividualhasbeendescribedbyaneighborasfollows Steveisveryshyandwithdrawn,invariablyhelpfulbutwithlittle interestinpeopleorintheworldofreality.Ameekandtidysoul hehasaneedfororderandstructurutandstre,andapassionfor detail.IsStevemorelikelytobealibrarianorafarmer? TheresemblanceofSteve'spersonalitytothatofastereotypicallibrarian strikeseveryoneimmediately,butequallyrelevantstatistica considerationsarealmostalwaysignored.Diditoccurtoyouthatthere aremorethan20malefarmersforeachmalelibrarianintheUnited States?Becausetherearesomanymorefarmers,itisalmostcertainthat more"meekandtidysoulswillbefoundontractorsthanatlibrary nformationdesks.However,wefoundthatparticipantsinourexperiments ignoredtherelevantstatisticalfactsandreliedexclusivelyonresemblance leproposedthattheyusedresemblanceasasimplifyingheuristic (roughly,aruleofthumb)tomakeadifficultjudgment.Therelianceonthe heuristiccausedpredictablebiases(systematicerrors)intheir predictions Onanotheroccasion,amosandiwonderedabouttherateofdivorce amongprofessorsinouruniversity.Wenoticedthatthequestiontriggered