Judgment and Decision Making: An Interdisciplinary Reader (Cambridge Series on Judgment and Decision Making)

772 Pages·2000·106.31 MB·English

by  Terry Connolly

PUBLISHED BYTHE PRESSSYNDICATE OFTHE UNIVERSITY OFCAMBRIDGE
ThePittBuilding,TrumpingtonStreet,Cambridge,UnitedKingdom
CAMBRIDGE UNIVERSITY PRESS
TheEdinburghBuilding,CambridgeCB22RU,UK http://www.cup.cam.ac.uk
40West20thStreet,NewYork,NY10011-4211,USA http://www.cup.org
10StamfordRoad,Oakleigh,Melbourne3166,Australia
RuizdeAlarcon13,28014Madrid,Spain
©CambridgeUniversityPress2000
Thisbookisincopyright. Subjecttostatutoryexception
andtotheprovisionsofrelevantcollectivelicensingagreements,
noreproductionofanypartmaytakeplacewithout
thewrittenpermissionofCambridgeUniversityPress.
Firstpublished2000
PrintedintheUnitedStatesofAmerica
TypefacePalatino9.75/12pt. System fb.TIjX2e[TBl
Acatalogrecordforthisbookisavailablefrom theBritishLibrary.
LibraryofCongressCataloging-in-PublicationData
Judgmentanddecisionmaking:aninterdisciplinaryreader1editedby
TerryConnolly,HalR.Arkes,KennethR.Hammond. - Rev. ed.
p. em.- (Cambridgeseriesonjudgmentanddecisionmaking)
Includesbibliographicalreferencesandindex.
ISBN0-521-62355-3(hardcover).- ISBN0-521-62602-1(pbk.)
1. Decisionmaking. 2.Judgment. I.Connolly,Terry.
II.Arkes,HalR.,1945- . III.Hammond,KennethR. IV.Series.
BF441.J79 1999
302.3- dc21 98-51484
CIP
ISBN 0521623553 hardback
ISBN 0521626021 paperback
Preface to the Second Edition
The first edition of this book was, by academic standards, something of a
best-seller. Itappearedata timewheninterestinJDMtopicswasexploding
in a variety of disciplines, and the collection sampled papers from many
ofthose disciplines. Thepapersitincluded, though realprofessionalwork,
were generally accessible to advanced undergraduate and early graduate
studentswithoutextensivebackgroundinpsychology,economics,ormath
en1atics. And the editors made abalanced selectionoftheory, method, and
application papers, withbriefintroductions to each cluster. All this served
the "entry-level" student well, and the bookbecame a central resource for
n1anyundergraduateandgraduateJDMcourses.Itwasuniversallyreferred
toas "TheJDMReader."
In undertaking a revision of this popular collection, the first need was
simply to bring it up to date, while retaining the flavor and breadth ofthe
original. Agreatdealhashappened inthefield inthedecadeormoresince
liTheReader"waspublished.Newtheoreticalconcernshaveemerged,while
othersseemtohaveworkedthemselvesout.Newapplicationshavebeenre
portedinmanyareas,andcomputerapplicationshavegrownbothasatool
for descriptive research and as a decision aid in normative work. Our re
visionhas tried to reflectsome ofthis new work. More than three quarters
of the chapters included here did not appear in the first edition- indeed,
thevastmajorityhaveappearedsincetheoriginalcollectionwaspublished.
Two chapters were commissioned especially for this volume and have not
appearedpreviously.Severalothershavebeenextensivelyrevisedandhave
not previously appeared in the form they take here. All the section intro
ductionshavebeenrewrittentoaccommodatethesechanges.Thecollection,
then,iscertainly"new";wehopeitisalso "improved."
Asecondchallengeinrevisingthebookwassimplytheenormousgrowth
of the field since the early 1980s. Many of the topics that excited JDM re
searchersthenhavenowgrownintosubstantialresearchliteratures,andim
portantnewtopicsareaddedeveryyear.Itisincreasinglydifficulttopresent
evenasamplingofboththeoreticalissuesandinterdisciplinaryapplications
xvii
xviii Prefaceto theSecondEdition
in a single volume. Fortunately, the Publications Committee ofthe Society
forJudgmentandDecisionMakinghasrecentlyaddedtothisseriesofbooks
a collection edited by William Goldstein and Robin Hogarth (Goldstein &
Hogarth, 1997) that does a superb job of reflecting many of the important
theoreticalideasinthearea.Thishasallowedusmorefreedominthepresent
collectiontoemphasizechapterswhoseprimarycontributionissubstantive
or methodological. These are, of course, imprecise lines, but the difference
in flavor between the two collections will be immediately obvious. Taken
together, thecompanionvolumes providecomplementarybutnonoverlap
pingintroductions tothecoreideas ofthefield and thewiderange oftheir
application.
Evenwithtwocollectionsintheplaceofone,itshouldbeclearthatthisis
averytinysamplingofthefield.Whenweundertookthisrevision,weasked
for advice from the membership ofthe Societyfor Judgment and Decision
Making.Wereceivedover500suggestionsofchaptersweshouldadd,only
two(!)forchaptersweshoulddrop- agenerousspiritindeed,butnotmuch
help in guiding a manageable collection. We thank all who gave their ad
vice, and exonerateallfrom blame. Thefinal selection,inevitably,haslarge
measuresofarbitrarinessand, nodoubt,otherswould havemadedifferent
choices. All we can claim of the chapters included here are that they each
seemtousexcitingandinterestinginonewayoranother,andthattheyare
drawnfromasufficientlywiderangeoftopicsandapproachesthattheystart
tosuggesttheenormousrangeandinterestofJOMresearchasweapproach
theyear2000.Thefieldiswideopen,highlydiverse,fullofenergy,andfasci
nating. Its coreideasaredeep,andtheirpracticalapplicationsofenormous
importance. Here is a sampling ofpapers from the field, intended to whet
theappetite andinvite theparticipationofa newgeneration ofresearchers
andtorefreshtheinterestofthosealreadyinvolved.Welookforwardtotheir
doing together the workthat willforce anotherrevisionof liTheReader" a
decadeorsofromnow!
TerryConnolly
HaIR.Arkes
KennethR. Hammond
References
Arkes,H.R.,& Hammond,K.R. (Eds.). (1986).Judgmentanddecision making:Aninter
disciplinaryreader.NewYork:CambridgeUniversityPress.
Goldstein,W.M.,&Hogarth,R.M.(Eds.).(1997).Researchonjudgmentanddecisionmak
ing:Currents,connections,andcontroversies.NewYork:CambridgeUniversityPress.
General Introduction
All students like to believe that their particular subject is the center of the
universe. Doubtless, students ofjudgmentand decisionmaking are110 dif
ferent,buttheymayhaveagoodargumentfortheirview.Afterall,theycan
claimt11atthegreatmomentsofhistoryallturnedonsomeo11e'sjudgmentas
towhatshouldbedoneandsomeone'sdecisiontodoit.Moreover,theywill
claimthatalthoughtheir subjectisas oldas civilizationithasbeenstudied
inascientific,empiricalwayonlywithintheverylastfew decades. Indeed,
mostofthepioneersinthisfieldarestillaliveandcontributingtoit.Thefact
thatwe arenow ableto studyjudgmentand decisionmakinginascientific
manner is, these students can claim, an exciting new discovery in and of
itself.
Of course, the editors of this book al1d the authors of the chapters in it
firmly believe in this view; judgment and decisio11 making are of critical
importance, and the fact that it is possible to study them in a scientific,
empiricalmannerisanewandexcitingeventintherecenthistoryofscience.
Despiteitscentralimportanceandlonghistory,however,thefield isstill
sonew thatitwillbe usefulto turnto the dictionaryto discoverhow these
termshavebeendefined for commonuse. Webster's Third New International
Dictionarysaysthatjudgmentis"theme11talorintellectualprocessofforming
anopinionorevaluationbydiscerningandcomparing,"andthecapacityfor
judgingis"thepowerorabilitytodecideontl1ebasisofevidence."Although
thedictionaryqllotesE.L.Godkinassayingthat"judgmentisthehighestof
thehumanfaculties," italsonotes thatOliverWendellHolmes said, "some
of the sharpest men in argun1ent are notoriously unsound in judgment."
Apparently, we are to understand that the capacity to make sound judg
ments requires not only intelligencebut wisdom and that the former does
notguaranteethelatter.Webster's definitionofdecision, "theactofsettlingor
terminating ...bygivingjudgment," suggests thatthere is little difference
betweenjudgmentanddecision makinginordinarydiscourse,soweshallnot
makeadistinctionhere,althoughmoreadvancedtreatmentsofthetopicdo
(see,e.g.,Goldstein& Hogarth,1997).
1
2 GeneralIntroduction
Not only the sourcesbut the nature ofsoundjudgmenthave fascinated
scholars since the beginning of self-reflection; the Greek intellectuals ap
parently mused about those topics every day. And the discussion contin
ues today among philosophers, psychologists, political scientists, lawyers,
managementscientists,andothersinsideandoutsideofacademia,because
soundjudgmentis,ofcourse,ofgreatpracticalaswellasacademicconcern.
No question will be of greater importance to the board of directors of the
industrialfirn1 that evaluates candidates for the positionofchiefexecutive
officer than the soundness ofeach candidate'sjudgment. And while mem
bersofalltheaboveacademicdepartmentsdebatetheoriesofrationalchoice
(eachgroup inhappyignorance ofthe activities ofthe others), members of
theboardofregentswillbeexercisingtheirjudgmentastheyselectthenew
presidentoftheuniversity. Indeed,thecapacityfor soundjudgmentofev
ery personwho desires a high (or even not so high) place in almost every
segment of society will be judged by those responsible for selecting them.
Thatisbecausewithinbothgovernmentandindustrythereis astrongcor
relationbetween the prominence and power ofone's position, the amount
oftimethatonespendsonproblemsrequiringjudgment,andthesalaryone
receives. At the other end of the scale, inability to make the simple judg
ments required inthe ordinarycircumstances ofday-to-daylivingleads to
thediagnosisofmentalillness.
In short, judgment and decision making are pervasive, important intel
lectualactivitiesengagedinbyallofusinacademic,professional,andsocial
pursuitsthroughouteveryday.Theabilitytoformgoodjudgmentsandmake
wiseandeffectivedecisionsgenerallyisconsideredthemarkofasuccessful
personinthesmalleraswellas thelargermattersofliving.Apparently,the
samehasbeentrueofeveryhumansociety.
What do we know about this salientfeature ofourlives? Thisbookwill
not try to answer that question completely, but it will provide a general
introductionto our knowledge ofjudgmentand decisionmaking and pro
videguidepostsforthosewhomaywishtopursuetheirinquiryfurther. Al
thoughthestudyofjudgmentanddecisionmakingisafieldinitsownright,
it finds application in virtually every known hun1an endeavor. (A recent
surveyconductedbyoneofusshowedthatarticlesrelatedtojudgmentand
decisionmakingappearedinmorethan500differentprofessionaljournals.)
Therefore,wel'lavechosentogroupstudiesofjudgmentanddecisionmaking
withinthosemajorfieldsinwhichstudiesofjudgmentanddecisionmaking
are currently being conducted. These include judgment and social policy,
economics,law,medicineandotherfieldsindicatedintheTableofContents.
Itiseasytofindexamplesoftheimportanceofthe"intellectualprocessof
forminganopinio11orevaluationbydiscerningorcomparing"or"decid[ing]
onthebasisofevidence."ThedecisiontodroptheatomicbombonHiroshima
withoutwarningisperhapsthemostdramaticexampleofanactofjudgn'lent
inthe20thcentury.Otherexamplesincludechangesinhealthpolicy(e.g.,the
GeneralIntroduction 3
decision to institute a National Health Service in Britain), economic policy
(e.g., the deregulationofairlinesinthe UnitedStates),legalpolicy(e.g., the
use of plea bargaining), and environmental policy (e.g., the protection of
wildlife and pristine areas, the control of toxic waste), and the reduction
of risk (e.g., the nationwide 55-mph speed limit in the United States); all
provide examples ofthe attempt to exercise soundjudgment. And inwhat
followsweprovideexamplesofeffortstostudysuchjudgments,bothinthe
controlledconditionsofthelaboratoryandintheworldoutside.
The reader will notice that all of these examples of studies of judgment
and decision making are recent. The systematic empirical study of judg
mentand decisionmakingbeganto emerge as a disciplineinits ownright
only in the 1960s. This occurred together with a strong surge ofinterestin
the larger, more general field of cognitive psychology, which includes the
studyofmemory,thinking,problemsolving,mentalimagery,andlanguage.
The explosion ofresearch in cognitive psychology marked a sharp shift in
interest from the concentration on motivation in psychological research to
a concentration on "mental activity." There are two main reasons for this.
First,somethingdroppedout;by1960strictstimulus- responsebehaviorism
lostcredibilityamongmanylaboratoryscientists,andFreudianpsychology
basedlargelyonunconsciousmotivationlostcredibilitywithalmostevery
one.Second,somethingdroppedin,namely,theelectroniccomputer,which
immediatelyprovidedacrediblemetaphorformentalactivity.Thus,within
a decade of the introduction of the computer, psychologists were talking
aboutandstudying "humaninformationprocessing." Asone psychologist
(GeorgeMiller) putit, "themindcameinonthebackofthe machine." The
arrivalofthecomputermadeitpossibletocarryoutresearchonhumanin
formationprocessing(includingjudgmentanddecisionmaking)innewand
powerfulways.Forexample,thoseinterestedinproblemsolvingwereableto
buildcomputermodelsthatsimulatedhumaninformation-processingactiv
ity,andthisledrapidlytothecreationofthenewfieldofartificialintelligence.
Thoseinterestedinconstructingmathematicalmodelsofthejudgmentand
decision-makingprocesscouldrapidlytestavarietyofsuchmodelsfortheir
abilitytorepresentand/orevaluatetherationalityofhumanjudgmentand
decisionmaking.Bythe1980s,workthatwouldhavebeenutterlyimpossible
priortothe computerbecamecommonplace.
Because two types of mathematical representations of judgment and
decision-making behavior are frequently used, we present the basic ideas
that underlie them in this General Introduction. The mathematical opera
tionsofbothapproachesareeasytograsp;aknowledgeofsimplealgebrais
all thatis required. We first describe the approach known as decision analy
sis- whichinvolvesanaprioridecompositionofthedecisionprocess- and,
second, we describe the approach known as judgment analysis - which in
volvesanaposterioridecompositionofthejudgmentprocess.Althoughthe
distinctionbetweendecisionandjudgmentissomewhatarbitraryandneed
4 GeneralIntroduction
notconcernushere,thedistinctionbetweenaprioriandaposterioridecom
position is importantand should be kept in mind, for itwill be illustrated
ofteninthechaptersthatfollow.
DecisionAnalysis: APrioriDecomposition
A priori decomposition refers to separating the decision process into its
components before the decision is made. Such components include (a) the
probabilities or likelihood of occurrence of each alternative considered and
(b) the utility attached to each alternative. The decision process is greatly
aidedwhentheseconceptsareusedinthecontextofadecision tree.
Construction of a decision tree prior to making the decision is an easy
wayofguidingandsimplifyingthedecisionprocessbecauseitdiagramsthe
decomposition of the decision process into probabilities and utilities and
thusprovidesaclearpictureoftheprocessanditscomponents.
The decision maker needs only four types ofinformation to construct a
decisiontree:
1. Whataremypossiblecoursesofaction? (Alternatives)
2. What are the events that might follow from those actions? (Out
comes)
3. Whatisthelikelihoodofeachevent?
4. Whatisthevalueofeacheventto me?
Hereisadecision-makingsituationsimilartoonethatactuallyconfronted
anelderlymanknowntous. Themanhadaveryseriousmedicalproblem.
Hisphysiciansaidthatadifficultoperationwasnecessarytoremedythesit
uation.Thephysicianadded,however,that,giventheman'sveryadvanced
age and thenatureoftheoperation, therewasa40% probabilitythat thepa
tient would not recover from the operation. If the operation were not per
formed, the seriousmedicalproblemwouldlinger,causingthe patient dis
comfortand impairinghis mobility. There was no chance that the problem
would"go away," andthere was a 20% probabilitythatwithoutthe opera
tionthemanwoulddiewithinthenext6months.Whatshouldthemando?
Shouldhehavetheoperation?
Figure1.1 depicts the decision tree for this situation. The box represents
a "decision node." The two possible courses of action emanate from this
node. They are "operate" and "don'toperate," and thus they comprise the
firstofthefourtypesofinformationneededtoconstructadecisiontree.The
circlesarechancenodes. (Theyarecalledchancenodesbecausenodecision
can be made to cause one of the outcomes to occur rather than the other.
Their occurrence is therefore left to chance.) The events emanating from
these circles are the possible events that might occurfollowing the courses
of action. This is the second type of information needed. Preceding each
GeneralIntroduction 5
Successful
Cured
Operate .60
Unsuccessful
Dead
Operateornot? .40
D
Problemlingers
Lingering
Don'toperate .80
Problenlgetsworse
Dead
.20
Figure1.1. Decisiontreeshowingprobabilitiesandoutcomesassociatedwithal
ternativeactions.
possibleeventisitsprobabilityofoccurrence,basedonthephysician'sbest
estimate. Thisisthe third essentialtype ofinformation. Finally,weneed to
knowwhatvaluethepatientplacesoneachoftheseoutcomes.
Becausethevalueofanycommodityisjudgeddifferentlybyeveryone,the
termutilityratherthanvalueisused.Thistermcapturesthesubjectivenature
oftheevaluation;aparticularamountofmoneymayhavedifferentutilityfor
methanforyou.Evenhealthmayhavedifferentutilitiesfordifferentpeople.
Tocalculatetheutilityofeachoutcome,weshallcalltheworstoutcome0.0
onautilityscaleandthebestoutcome1.0.
The patientassigned "death" the former value and "completecure" the
lattervalue. Usingthisrange(0.0-1.0),thepatientfeltthatlivingindiscom
fortandhavingdecreasedmobilitywasworth.6tohim.Itwasastatecloser
to completecure"thanto"death,"butnotbymuch.
1/
We now have all the information needed to make a decision tree. First,
itisnecessaryto examineeachoutcome. Theutilityofeach.outcomeneeds
tobeweightedaccordingtoitslikelihood.Anoutcomeof1.0("wonderful")
that has a high probability of occurring should definitely be preferred to
another outcome of utility 0.0 that has very little likelihood of occurring.
Toaccomplishthismathematically,theutilityofeachoutcomeismultiplied
byits probability of occurrence. This product is the expected utility ofeach
outcome.Figure1.2containsthesecalculations,whicharelocatedattheright
edgeofeachbranchofthedecisiontree.
Allthatremainsnowistheprocesscalled"foldingback,"whichconsists
ofpruningallbutthemostpreferredcourseofactionateachdecisionnode.
Therearetwopossiblecoursesofaction:operateanddon'toperate.Foreach
ofthesetwooptionswe addtogetheralloftheexpectedutilitiesassociated
withthatoption.Forexample,theoption"don'toperate"hasassociatedwith
o.
ittwoexpectedutilitiesof.48and Theirsum,.48,istheexpectedutilityof
the "don'toperate"courseofaction.Becausethe"operate"courseofaction
ishigher(.6),thepreferredcourseofactionistooperate.
Oneimmensevirtueofadecisiontreeisthatitisawonderfullygeneralde
cisionaid.AslongasthefOLlrtypesofinformationareavailable,anydecision
6 GeneralIntroduction
Successful
Operate .60 1.0)
ED : (.6x1.0) + (.4x0)
Unsuccessful
0.0 - .60
-Op-e-ra-te-o-r-no1t?0 .40
Problemlingers
Don'toperate .80 0.6,
Problemgetsworse U = (.8x .6) + (.2x0)
0.0 = .48
.20
Figure1.2. Calculationofexpectedutilityforeachalternative.
canbeanalyzedbyuseofthe tree. ShouldImove toMinneapolisto accept
thisnewjob?ShouldIswitchinsurancepoliciesfromXtoY?Shouldweget
aloantogetanewcarnow,orshouldwelimpalongwiththeonewehave?
One difficulty often encountered in constructing a decision tree is that
likelihoodsandutilitiesareoftennoteasytoassess.Whenanexpertopinion
isavailable,asinthecaseofourmedicalexample,reasonablelikelihoodscan
oftenbeprovided.Becausethehelpfulnessofadecisiontreeisbasedlargely
onthe accuracyofthelikelihoods and utilities used, everyeffortshouldbe
madetoobtaingoodestimates.
Occasionallythe decisionmakeris uneasyaboutthe "verdict" ofthe de
cision tree. In our medical example, the "operate" option was only .12 su
periortothe "don'toperate" option. "WhatifIlaterdecidethatlivingwith
discomfort and decreased mobility isn'tsobad? Maybe it's worth a .7and
notameasly.6,"thinkstheelderlymanasthemorningoftheoperationap
proaches. Aquickcalculationwillrevealthat"operate"isstillthepreferred
choice, even if .7 is deemed the utility of an uncomfortable and imn10bile
existence. Modifying the probabilities and utilities in this way is called a
"sensitivity analysis," because such manipulating of the probabilities and
utilities tests how sensitive the final choice is to the numbers initially as
signed.Reasonablemodificationsoftheprobabilitiesandutilitiesoftenleave
thedecisionunchanged.Thedecisionmakercanthenrestcomfortablywith
thedecisionthathasbeenreached.
Concernabouttheassignmentofaccurateprobabilities,thecalculationof
expected utilities, and the performance of various arithmetic tasks should
notobscurewhatmaybe the greatestvirtue ofa decision tree: Itforces the
decisionmakertomakeexplicitallthebasesforthedecision. Inthetreeare
contained all the courses ofaction, allthe probabilities, all the utilities, and
alltheoutcomesofwhichthedecisionmakerisaware- or,atleast,thosethat
heorsheplanstoconsider. Everyanalysisworksonasimplifiedversionof
therealsituation/butthetreeatleastmakesitexplicitwhatisbeingincluded,
whatexcluded. Italsomakesitexplicithowfar intothefuture thedecision
makeristhinkingabouttheconsequencesofthisaction.Thetime-frameissue
isimportantwhentheutilityofanoutcomeshiftsovertime.Inexercising,for
GeneralIntroduction 7
example,ashorttime-framestressesnegativeoutcomeslikefeelingtired,but
alonger time-framestressesmorepositiveoneslikegoodhealthand vigor.
Merely having to generate the informationnecessary to draw the tree may
forcethedecisionmakertoconfrontthesituationinamuchmoreorganized
and thoughtfulwaythanwouldotherwisebethecase.
JudgmentAnalysis: APosterioriDecomposition
If a priori decomposition implies decomposing the decision process prior
to its occurrence, then a posteriori decomposition obviously implies that
decomposition will take place after a series ofjudgments have beenmade.
Asweshallsee,aperson'sjudgmentpolicycanbe captured"afterjudgments
II
aremaderegardinghypotheticalcases;thepolicymaythenbeappliedtoreal
cases.
Theprincipalconceptsofajudgmentanalysisarebestillustratedbyref
erencetothemodelofthejudgmentsituationpresentedinFigure1.3,which
indicatesthatjudgmentisacognitiveprocesssimilartoinductiveinference.
That is, judgment is a cognitive or intellectual process in which a person
drawsaconclusion,oraninference(Ys),aboutsomething(Ye),whichcannot
be seen, onthe basis ofdata (Xi), whichcan be seen. Inother words, judg
ments are made from tangible data, which serve as cues to intangibleevents
andcircumstances.Thewide-rangingarcconnectingY andY (labeledr in
s e a
ACHIEVEMENT(!(,)
r",J~rs.,
~ ..
;
~~~';;~(s.
~
VALIDITY (r" i) CUE UTILIZATION (r.)
I $,1
CUES (Xi)
Figure1.3. The "LensModel," a schematicillustrationofhow asubjectfroms a
judgment (Ys) of some criterion variable (Ye) on the basis of a set of imperfect
cues, Xi. Cue validity (re,;) measures the extent to which each cue reflects the
valueofthecriterion;cueutilization(rs.;)measurestherelationbetweeneachcue
andthesubject'sjudgment;andachievement,ra,measureshowwellthesubject's
judgmentscorrespondtotheactualcriterionvalues.