 |
| Amos Tversky And Daniel Kahneman - Maps Of Bounded Rationality.Psychology for Behavioral Economics 2002, różne |
| |
[ Pobierz całość w formacie PDF ]
MapsofBoundedRationality:
PsychologyforBehavioralEconomics
†
By
D
ANIEL
K
AHNEMAN
*
TheworkcitedbytheNobelcommitteewas
donejointlywithAmosTversky(1937–1996)
duringalongandunusuallyclosecollaboration.
Together,weexploredthepsychologyofintu-
itivebeliefsandchoicesandexaminedtheir
boundedrationality.HerbertA.Simon(1955,
1979)hadproposedmuchearlierthatdecision
makersshouldbeviewedasboundedlyrational,
andhadofferedamodelinwhichutilitymaxi-
mizationwasreplacedbysatis
Ž
cing.Ourre-
searchattemptedtoobtainamapofbounded
rationality,byexploringthesystematicbiases
thatseparatethebeliefsthatpeoplehaveandthe
choicestheymakefromtheoptimalbeliefsand
choicesassumedinrational-agentmodels.The
rational-agentmodelwasourstartingpointand
themainsourceofournullhypotheses,but
TverskyandIviewedourresearchprimarilyas
acontributiontopsychology,withapossible
contributiontoeconomicsasasecondaryben-
e
Ž
t.Weweredrawnintotheinterdisciplinary
conversationbyeconomistswhohopedthat
psychologycouldbeausefulsourceofassump-
tionsforeconomictheorizing,andindirectlya
sourceofhypothesesforeconomicresearch
(RichardH.Thaler,1980,1991,1992).These
hopeshavebeenrealizedtosomeextent,giving
risetoanactiveprogramofresearchbybehav-
ioraleconomists(Thaler,2000;ColinCamerer
etal.,forthcoming;forotherexamples,see
KahnemanandTversky,2000).
MyworkwithTverskycomprisedthreesep-
arateprogramsofresearch,someaspectsof
whichwerecarriedoutwithothercollaborators.
The
Ž
rstexploredtheheuristicsthatpeopleuse
andthebiasestowhichtheyareproneinvari-
oustasksofjudgmentunderuncertainty,includ-
ingpredictionsandevaluationsofevidence
(KahnemanandTversky,1973;Tverskyand
Kahneman,1974;Kahnemanetal.,1982).The
secondwasconcernedwithprospecttheory,a
modelofchoiceunderrisk(Kahnemanand
Tversky,1979;TverskyandKahneman,1992)
andwithlossaversioninrisklesschoice(Kah-
nemanetal.,1990,1991;TverskyandKahne-
man,1991).Thethirdlineofresearchdealtwith
framingeffectsandwiththeirimplicationsfor
rational-agentmodels(TverskyandKahneman,
1981,1986).Thepresentessayrevisitsthese
threelinesofresearchinlightofrecentad-
vancesinthepsychologyofintuitivejudgment
andchoice.Manyoftheideaspresentedhere
wereanticipatedinformallydecadesago,but
theattempttointegratethemintoacoherent
approachtojudgmentandchoiceisrecent.
Economistsoftencriticizepsychologicalre-
searchforitspropensitytogeneratelistsof
errorsandbiases,andforitsfailuretooffera
coherentalternativetotherational-agentmodel.
Thiscomplaintisonlypartlyjusti
Ž
ed:psycho-
logicaltheoriesofintuitivethinkingcannot
matchtheeleganceandprecisionofformalnor-
mativemodelsofbeliefandchoice,butthisis
justanotherwayofsayingthatrationalmodels
arepsychologicallyunrealistic.Furthermore,
thealternativetosimpleandprecisemodelsis
notchaos.Psychologyoffersintegrativecon-
ceptsandmid-levelgeneralizations,whichgain
credibilityfromtheirabilitytoexplainostensi-
blydifferentphenomenaindiversedomains.In
thisspirit,thepresentessayoffersauni
Ž
ed
ThisarticleisarevisedversionofthelectureDaniel
KahnemandeliveredinStockholm,Sweden,onDecember
8,2002,whenhereceivedtheBankofSwedenPrizein
EconomicSciencesinMemoryofAlfredNobel.Thearticle
iscopyright©TheNobelFoundation2002andispublished
herewiththepermissionoftheNobelFoundation.
*WoodrowWilsonSchool,PrincetonUniversity,
Princeton,NJ08544(e-mail:Kahneman@princeton.edu).
ThisessayrevisitsproblemsthatAmosTverskyandI
studiedtogethermanyyearsago,andcontinuedtodiscussin
aconversationthatspannedseveraldecades.Itbuildsonan
analysisofjudgmentheuristicsthatwasdevelopedincol-
laborationwithShaneFrederick(KahnemanandFrederick,
2002).Adifferentversionwaspublishedin
AmericanPsy-
chologist
inSeptember2003.Fordetailedcommentsonthis
versionIamgratefultoAngusDeaton,DavidLaibson,
MichaelRothschild,andRichardThaler.Theusualcaveats
apply.GeoffreyGoodwin,AmirGoren,andKurtSchoppe
providedhelpfulresearchassistance.
†
1449
1450 THEAMERICANECONOMICREVIEW DECEMBER2003
treatmentofintuitivejudgmentandchoice,
whichbuildsonanearlierstudyoftherelation-
shipbetweenpreferencesandattitudes(Kahne-
manetal.,1999)andextendsamodelof
judgmentheuristicsrecentlyproposedbyKah-
nemanandShaneFrederick(2002).Theguid-
ingideasare(i)thatmostjudgmentsandmost
choicesaremadeintuitively;(ii)thattherules
thatgovernintuitionaregenerallysimilartothe
rulesofperception.Accordingly,thediscussion
oftherulesofintuitivejudgmentsandchoices
willrelyextensivelyonvisualanalogies.
SectionIintroducesadistinctionbetween
twogenericmodesofcognitivefunction,corre-
spondingroughlytointuitionandreasoning.
SectionIIdescribesthefactorsthatdetermine
therelativeaccessibilityofdifferentjudgments
andresponses.SectionIIIrelatesprospectthe-
orytothegeneralpropositionthatchangesand
differencesaremoreaccessiblethanabsolute
values.SectionIVexplainsframingeffectsin
termsofdifferentialsalienceandaccessibility.
SectionVreviewsanattributesubstitution
modelofheuristicjudgment.SectionVIde-
scribesaparticularfamilyofheuristics,called
prototypeheuristics.SectionVIIdiscussesthe
interactionsbetweenintuitiveanddeliberate
thought.SectionVIIIconcludes.
ationsandovertbehavioralsogoeson.Wedo
notexpresseverypassingthoughtoracton
everyimpulse.Butthemonitoringisnormally
lax,andallowsmanyintuitivejudgmentstobe
expressed,includingsomethatareerroneous
(KahnemanandFrederick,2002).EllenJ.
Langeretal.(1978)providedawell-known
exampleofwhatshecalled“mindlessbehav-
ior.”Inherexperiment,aconfederatetriedto
cutinlineatacopyingmachine,usingvarious
preset“excuses.”Theconclusionwasthatstate-
mentsthathadtheformofanunquali
Ž
edre-
questwererejected(e.g.,“Excuseme,mayIuse
theXeroxmachine?”),butalmostanystatement
thathadthegeneralformofanexplanationwas
accepted,including“Excuseme,mayIusethe
XeroxmachinebecauseIwanttomakecop-
ies?”Thesuper
Ž
cialityisstriking.
Frederick(2003,personalcommunication)
hasusedsimplepuzzlestostudycognitiveself-
monitoring,asinthefollowingexample:“Abat
andaballcost$1.10intotal.Thebatcosts$1
morethantheball.Howmuchdoestheball
cost?”Almosteveryonereportsaninitialten-
dencytoanswer“10cents”becausethesum
$1.10separatesnaturallyinto$1and10cents,
and10centsisabouttherightmagnitude.Fred-
erickfoundthatmanyintelligentpeopleyieldto
thisimmediateimpulse:50percent(47/93)ofa
groupofPrincetonstudentsand56percent
(164/293)ofstudentsattheUniversityofMich-
igangavethewronganswer.Clearly,thesere-
spondentsofferedtheirresponsewithout
Ž
rst
checkingit.Thesurprisinglyhighrateoferrors
inthiseasyproblemillustrateshowlightlythe
outputofeffortlessassociativethinkingismon-
itored:peoplearenotaccustomedtothinking
hard,andareoftencontenttotrustaplausible
judgmentthatquicklycomestomind.Re-
markably,Frederickhasfoundthaterrorsin
thispuzzleandinothersofthesametype
weresigni
Ž
cantpredictorsofhighdiscount
rates.
Intheexamplesdiscussedsofar,intuition
wasassociatedwithpoorperformance,butin-
tuitivethinkingcanalsobepowerfulandaccu-
rate.Highskillisacquiredbyprolonged
practice,andtheperformanceofskillsisrapid
andeffortless.Theproverbialmasterchess
playerwhowalkspastagameanddeclares
“whitematesinthree”withoutslowingisper-
formingintuitively(SimonandWilliamG.
Chase,1973),asistheexperiencednursewho
I.TheArchitectureofCognition:TwoSystems
Thepresenttreatmentdistinguishestwo
modesofthinkinganddeciding,whichcorre-
spondroughlytotheeverydayconceptsofrea-
soningandintuition.Reasoningiswhatwedo
whenwecomputetheproductof17by258,
Ž
ll
anincometaxform,orconsultamap.Intuition
isatworkwhenwereadthesentence“Bill
Clintonisashyman”asmildlyamusing,or
whenwe
Ž
ndourselvesreluctanttoeatapiece
ofwhatweknowtobechocolatethathasbeen
formedintheshapeofacockroach(PaulRozin
andCarolNemeroff,2002).Reasoningisdone
deliberatelyandeffortfully,butintuitivethoughts
seemtocomespontaneouslytomind,without
conscioussearchorcomputation,andwithout
effort.Casualobservationandsystematicre-
searchindicatethatmostthoughtsandactions
arenormallyintuitiveinthissense(DanielT.
Gilbert,1989,2002;TimothyD.Wilson,2002;
SeymourEpstein,2003).
Althougheffortlessthoughtisthenorm,
somemonitoringofthequalityofmentaloper-
VOL.93NO.5 KAHNEMAN:MAPSOFBOUNDEDRATIONALITY
1451
F
IGURE
1.T
HREE
C
OGNITIVE
S
YSTEMS
detectssubtlesignsofimpendingheartfailure
(GaryKlein,1998;AtulGawande,2002).
Thedistinctionbetweenintuitionandreason-
inghasrecentlybeenatopicofconsiderable
interesttopsychologists(see,e.g.,Shelly
ChaikenandYaacovTrope,1999;Gilbert,
2002;StevenA.Sloman,2002;KeithE.
StanovichandRichardF.West,2002).Thereis
substantialagreementonthecharacteristicsthat
distinguishthetwotypesofcognitiveprocesses,
forwhichStanovichandWest(2000)proposed
theneutrallabelsofSystem1andSystem2.
TheschemeshowninFigure1summarizes
thesecharacteristics.TheoperationsofSystem
1arefast,automatic,effortless,associative,and
oftenemotionallycharged;theyarealsogov-
ernedbyhabit,andarethereforedif
Ž
cultto
controlormodify.TheoperationsofSystem2
areslower,serial,effortful,anddeliberately
controlled;theyarealsorelatively
�
exibleand
potentiallyrule-governed.
Thedifferenceineffortprovidesthemost
usefulindicationsofwhetheragivenmental
processshouldbeassignedtoSystem1orSys-
tem2.Becausetheoverallcapacityformental
effortislimited,effortfulprocessestendtodis-
rupteachother,whereaseffortlessprocesses
neithercausenorsuffermuchinterferencewhen
combinedwithothertasks.Forexample,adriv-
er’sabilitytoconductaconversationisasen-
sitiveindicatoroftheamountofattention
currentlydemandedbythedrivingtask.Dual
taskshavebeenusedinhundredsofpsycholog-
icalexperimentstomeasuretheattentionalde-
mandsofdifferentmentalactivities(fora
review,seeHaroldE.Pashler,1998).Studies
usingthedual-taskmethodsuggestthattheself-
monitoringfunctionbelongswiththeeffortful
operationsofSystem2.Peoplewhoareoccu-
piedbyademandingmentalactivity(e.g.,at-
temptingtoholdinmindseveraldigits)are
muchmorelikelytorespondtoanothertaskby
blurtingoutwhatevercomestomind(Gilbert,
1989).Thephrasethat“System2monitorsthe
activitiesofSystem1”willbeusedhereas
shorthandforahypothesisaboutwhatwould
happeniftheoperationsofSystem2weredis-
rupted.Forexample,itissafetopredictthatthe
percentageoferrorsinthebat-and-ballquestion
willincrease,iftherespondentsareaskedthis
questionwhileattemptingtokeepalistof
wordsintheiractivememory.
Inthelanguagethatwillbeusedhere,the
perceptualsystemandtheintuitiveoperations
1452 THEAMERICANECONOMICREVIEW DECEMBER2003
ofSystem1generate
impressions
oftheat-
tributesofobjectsofperceptionandthought.
Theseimpressionsarenotvoluntaryandneed
notbeverballyexplicit.Incontrast,
judgments
arealwaysexplicitandintentional,whetheror
nottheyareovertlyexpressed.Thus,System2
isinvolvedinalljudgments,whethertheyorig-
inateinimpressionsorindeliberatereasoning.
Thelabel“intuitive”isappliedtojudgments
thatdirectlyre
�
ectimpressions.
Figure1illustratesanideathatguidedthe
researchthatTverskyandIconductedfromits
earlydays:thatintuitivejudgmentsoccupya
position—perhapscorrespondingtoevolution-
aryhistory—betweentheautomaticoperations
ofperceptionandthedeliberateoperationsof
reasoning.Allthecharacteristicsthatstudents
ofintuitionhaveattributedtoSystem1arealso
propertiesofperceptualoperations.Unlikeper-
ception,however,theoperationsofSystem1
arenotrestrictedtotheprocessingofcurrent
stimulation.LikeSystem2,theoperationsof
System1dealwithstoredconceptsaswellas
withpercepts,andcanbeevokedbylanguage.
Thisviewofintuitionsuggeststhatthevast
storeofscienti
Ž
cknowledgeavailableabout
perceptualphenomenacanbeasourceofuseful
hypothesesabouttheworkingsofintuition.The
strategyofdrawingonanalogiesfrompercep-
tionisappliedinthefollowingsection.
F
IGURE
2.E
XAMPLESOF
D
IFFERENTIAL
A
CCESSIBILITY
dismantledisnotperceptuallyaccessible,
thoughitcanbeestimatedbyadeliberatepro-
cedure,suchasmultiplyingtheareaofablock
bythenumberofblocks.Ofcourse,thesitua-
tionisreversedwithFigure2b.Nowtheblocks
arelaidoutandanimpressionoftotalareais
immediatelyaccessible,buttheheightofthe
towerthatcouldbeconstructedwiththese
blocksisnot.
Somerelationalpropertiesareaccessible.
Thus,itisobviousataglancethatFigures2a
and2caredifferent,butalsothattheyaremore
similartoeachotherthaneitheristoFigure
2b.Andsomestatisticalpropertiesofensembles
areaccessible,whileothersarenot.Foran
example,considerthequestion“Whatisthe
averagelengthofthelinesinFigure3?”This
questioniseasy.Whenasetofobjectsofthe
samegeneralkindispresentedtoanobserver—
whethersimultaneouslyorsuccessively—arep-
resentationofthesetiscomputedautomatically,
whichincludesquitepreciseinformationabout
theaverage(DanAriely,2001;Sang-Chul
ChongandAnneTreisman,2003).Therepre-
sentationoftheprototypeishighlyaccessible,
andithasthecharacterofapercept:weforman
impressionofthetypicallinewithoutchoosing
todoso.TheonlyroleforSystem2inthistask
istomaptheimpressionoftypicallengthonto
theappropriatescale.Incontrast,theanswerto
thequestion“Whatisthetotallengthofthe
linesinthedisplay?”doesnotcometomind
withoutconsiderableeffort.
Astheexampleofaveragesandsumsillus-
trates,someattributesaremoreaccessiblethan
others,bothinperceptionandinjudgment.At-
tributesthatareroutinelyandautomatically
producedbytheperceptualsystemorbySystem
II.TheAccessibilityDimension
Ade
Ž
ningpropertyofintuitivethoughtsis
thattheycometomindspontaneously,likeper-
cepts.Thetechnicaltermfortheeasewith
whichmentalcontentscometomindis
acces-
sibility
(E.ToryHiggins,1996).Tounderstand
intuition,wemustunderstandwhysome
thoughtsareaccessibleandothersarenot.The
remainderofthissectionintroducestheconcept
ofaccessibilitybyexamplesdrawnfromvisual
perception.
ConsiderFigures2aand2b.Aswelookat
theobjectinFigure2a,wehaveimmediate
impressionsoftheheightofthetower,thearea
ofthetopblock,andperhapsthevolumeofthe
tower.Translatingtheseimpressionsintounits
ofheightorvolumerequiresadeliberateoper-
ation,buttheimpressionsthemselvesarehighly
accessible.Forotherattributes,noperceptual
impressionexists.Forexample,thetotalarea
thattheblockswouldcoverifthetowerwere
VOL.93NO.5 KAHNEMAN:MAPSOFBOUNDEDRATIONALITY
1453
wasinitiatedwithinafractionofasecond,well
beforethemeaningofthestimuluswascon-
sciouslyregistered,theemotionalvalenceofthe
wordhadasubstantialeffect.Participantswere
relativelyfasterinpullingalevertowardthem-
selves(approach)forpositivewords,andrela-
tivelyfasterpushingtheleverawaywhenthe
wordwasaversive.Thetendenciestoapproach
oravoidwereevokedbyanautomaticprocess
thatwasnotunderconsciousvoluntarycontrol.
Severalpsychologistshavecommentedonthe
in
�
uenceofthisprimordialevaluativesystem
(hereincludedinSystem1)ontheattitudesand
preferencesthatpeopleadoptconsciouslyand
deliberately(Zajonc,1998;Kahnemanetal.,
1999;PaulSlovicetal.,2002;Epstein,2003).
Theprecedingdiscussionestablishesadi-
mensionofaccessibility.Atoneendofthis
dimensionwe
Ž
ndoperationsthathavethe
characteristicsofperceptionandoftheintuitive
System1:theyarerapid,automatic,andeffort-
less.Attheotherendareslow,serial,and
effortfuloperationsthatpeopleneedaspecial
reasontoundertake.Accessibilityisacontin-
uum,notadichotomy,andsomeeffortfulop-
erationsdemandmoreeffortthanothers.Some
ofthedeterminantsofaccessibilityareprobably
genetic;othersdevelopthroughexperience.The
acquisitionofskillgraduallyincreasestheac-
cessibilityofusefulresponsesandofproductive
waystoorganizeinformation,untilskilledper-
formancebecomesalmosteffortless.Thiseffect
ofpracticeisnotlimitedtomotorskills.A
masterchessplayerdoesnotseethesameboard
asthenovice,andvisualizingthetowerinan
arrayofblockswouldalsobecomevirtually
effortlesswithprolongedpractice.
Theimpressionsthatbecomeaccessiblein
anyparticularsituationaremainlydetermined,
ofcourse,bytheactualpropertiesoftheobject
ofjudgment:itiseasiertoseeatowerinFigure
2athaninFigure2b,becausethetowerinthe
latterisonlyvirtual.Physicalsaliencealsode-
terminesaccessibility:ifalargegreenletterand
asmallblueletterareshownatthesametime,
“green”willcometomind
Ž
rst.However,sa-
liencecanbeovercomebydeliberateattention:
aninstructiontolookforthesmallobjectwill
enhancetheaccessibilityofallitsfeatures.
Analogouseffectsofsalienceandofsponta-
neousandvoluntaryattentionoccurwithmore
abstractstimuli.Forexample,thestatements
“TeamAbeatteamB”and“TeamBlostto
F
IGURE
3.D
IFFERENTIAL
A
CCESSIBILITY
OF
S
TATISTICAL
P
ROPERTIES
1,withoutintentionoreffort,havebeencalled
naturalassessments
(TverskyandKahneman,
1983).KahnemanandFrederick(2002)com-
piledapartiallistofthesenaturalassessments.
Inadditiontophysicalpropertiessuchassize,
distance,andloudness,thelistincludesmore
abstractpropertiessuchassimilarity,causal
propensity,surprisingness,affectivevalence,
andmood.
Theevaluationofstimuliasgoodorbadisa
particularlyimportantnaturalassessment.The
evidence,bothbehavioral(JohnA.Bargh,
1997;RobertB.Zajonc,1998)andneurophys-
iological(e.g.,JosephE.LeDoux,2000),is
consistentwiththeideathattheassessmentof
whetherobjectsaregood(andshouldbeap-
proached)orbad(shouldbeavoided)iscarried
outquicklyandef
Ž
cientlybyspecializedneural
circuitry.Aremarkableexperimentreportedby
Bargh(1997)illustratesthespeedoftheevalu-
ationprocess,anditsdirectlinktoapproachand
avoidance.Participantswereshownaseriesof
stimulionascreen,andinstructedtorespondto
eachstimulusassoonasitappeared,bymoving
aleverthatblankedthescreen.Thestimuliwere
affectivelychargedwords,somepositive(e.g.,
LOVE)andsomeaversive(e.g.,VOMIT),but
thisfeaturewasirrelevanttotheparticipant’s
task.Halftheparticipantsrespondedbypulling
thelevertowardthemselves,halfrespondedby
pushingtheleveraway.Althoughtheresponse
[ Pobierz całość w formacie PDF ]
zanotowane.pldoc.pisz.plpdf.pisz.plrumian.htw.pl
|
| |
|
|
|
| Linki |
| |
- Strona startowa
- American Brownies, ciasta, przepisy różne-ciasta i nie tylko(1)
- Amedeo Minghi - Cantare è D'amore, Teksty piosenek różne
- Ami's First Love, PEŁNE WERSJE - RÓŻNE
- Americas Army 2 - komendy gracza, Różne
- Angorka 2008-02-17, Różne
- Anne Bronte - Agnes Grey, Różne
- Anioły - opis i klasyfikacja, żydzi, różne
- Anne Rice - Przebudzenie Śpiącej Królewny, ebooki kindle, kindle
- Antologia - O kobietach . Czeskie opowiesci, Ebooki RTF
- Angst, Teksty piosenek, Die Toten Hosen
- zanotowane.pl
- doc.pisz.pl
- pdf.pisz.pl
- stardollblog.htw.pl
|
| |
|