Cognitive hierarchies and emotions in behavioral game theory Colin F. Camerer 1, 2 Alec Smith 1 1 Division of Humanities and Social Sciences 2 Computation and Neural Systems California Institute of Technology 4/25/11 11:15 am. Comments welcome. Prepared for Oxford Handbook of Thinking and Reasoning (K. J. Holyoak & R. G. Morrison, Editors). This research was supported by The Betty and Gordon Moore Foundation and by National Science Foundation grant NSF‐SES 0850840. Correspondence to: [email protected].
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Cognitivehierarchiesandemotions
inbehavioralgametheory
ColinF.Camerer1,2
AlecSmith1
1DivisionofHumanitiesandSocialSciences
2ComputationandNeuralSystems
CaliforniaInstituteofTechnology
4/25/1111:15am.Commentswelcome.PreparedforOxfordHandbookofThinkingand
Reasoning(K.J.Holyoak&R.G.Morrison,Editors).ThisresearchwassupportedbyThe
BettyandGordonMooreFoundationandbyNationalScienceFoundationgrantNSF‐SES
0850840.Correspondenceto:[email protected].
1
ABSTRACT
Untilrecently,gametheorywasnotfocussedoncognitively‐plausiblemodelsofchoicesin
humanstrategicinteractions.Thischapterdescribestwonewapproachesthatdoso.The
firstapproach,cognitivehierarchymodeling,assumesthatplayershavedifferentlevelsof
partiallyaccuraterepresentationsofwhatothersarelikelytodo,whichvaryfromheuristic
andnaïvetohighlysophisticatedandaccurate.Thereisreasonableevidencethatthis
approachexplainschoices(betterthantraditionalequilibriumanalysis)indozensof
experimentalgamesandsomenaturally‐occurringgames(e.g.,aSwedishlottery,auctions,
andconsumerreactionstoundisclosedqualityinformationaboutmovies).Measurementof
eyetrackingandfMRIactivityduringgamesisalsosuggestiveofacognitive.Thesecond
approach,psychologicalgames,allowsvaluetodependuponchoiceconsequencesandon
beliefsaboutwhatwillhappen.Thismodelingframeworkcanlinkcognitionandemotion,
andexpresssocialemotionssuchas“guilt”.Inapsychologicalgame,guiltismodeledasthe
negativeemotionofknowingthatanotherpersonisunpleasantlysurprisedthatyour
choicedidnotbenefitthem(astheyhadexpected).Ourhopeisthatthesenew
developmentsinatraditionallycognitivefield(gametheory)willengageinterestof
psychologistsandothersinterestedinthinkingandsocialcognition.
KEYWORDS
Boundedrationality,cognitivehierarchy,emotions,gametheory,psychologicalgames,
strategicneuroscience
2
I.Introduction
Thischapterisaboutcognitiveprocessesinstrategicthinking.Thetheoryofgames
providesthemostcomprehensiveframeworkforthinkingaboutthevaluedoutcomesthat
resultfromstrategicinteractions.Thetheoryspecifieshow“players”(that’sgametheory
jargon)mightchoosehigh‐valuestrategiestoguesslikelychoicesofotherplayers.
Traditionally,gametheoryhasbeenfocusedonfinding“solutions”togamesbasedon
highlymathematicalconceptionsofrationalforecastingandchoice.Morerecently(starting
withCamerer,1990),behavioralgametheorymodelshaveextendedtherationaltheories
toincludestochasticresponse,limitsoninferringcorrectlywhatotherplayerswilldo,
socialemotionsandconsiderationssuchasguilt,anger,reciprocity,orsocialimage,and
modulatingfactorsincludinginferencesaboutothers’intentions.Twogeneralbehavioral
modelsthatmightinterestcognitivepsychologistsarethefocusofthischapteri:Cognitive
hierarchymodeling,andpsychologicalgametheory.
Conventionalgametheoryistypicallyabstract,mathematicallyintimidating,
computationallyimplausible,andalgorithmicallyincomplete.Itisthereforenotsurprising
thatconventionaltoolshavenotgainedtractionincognitivepsychology.Ourhopeisthat
themorepsychologicallyplausiblebehavioralvariantscouldinterestcognitive
psychologists.Oncelimitedstrategicthinkingisthefocus,questionsofcognitive
representation,categorizationofdifferentstrategicstructures,andthenatureofsocial
cognition,andhowcooperationisachievedallbecomemoreinterestingresearchable
questions.Thequestionofwhetherornotpeopleareusingthedecision‐makingalgorithms
3
proposedbythesebehavioralmodelscanalsobeaddressedwithobservables(suchas
responsetimesandeyetrackingofvisualattention)familiarincognitivepsychology.
Numericalmeasuresofvalueandbeliefderivedinthesetheoriescanalsobeusedas
parametricregressorstoidentifycandidatebraincircuitsthatappeartoencodethose
measures.Thisgeneralapproachhasbeenquitesuccessfulinstudyingsimpler
nonstrategicchoicedecisions(Glimcher,Camerer,Fehr,&Poldrack,2008)buthasbeen
appliedinfrequentlytogames(seeBhatt&Camerer,inpress).
Whatisagame?
Gametheoryisthemathematicalanalysisofstrategicinteraction.Ithasbecomea
standardtoolineconomicsandtheoreticalbiology,andisincreasinglyusedinpolitical
science,sociology,andcomputerscience.Agameismathematicallydefinedasasetof
players,descriptionsoftheirinformation,afixedorderofthesequenceofchoicesby
differentplayers,andafunctionmappingplayers’choicesandinformationtooutcomes.
Outcomesmayincludetangibleslikecorporateprofitsorpokerwinnings,aswellas
intangibleslikepoliticalgain,status,orreproductiveopportunities(inbiologicaland
evolutionarypsychologymodels).Thespecificationofagameiscompletedbyapayoff
functionthatattachesanumericalvalueor“utility”toeachoutcome.
Thestandardapproachtotheanalysisofgamesistocomputeanequilibriumpoint,
asetofstrategiesforeachplayerwhicharesimultaneouslybestresponsestooneanother..
ThisapproachisdueoriginallytoJohnNash(1950),buildingonearlierworkbyVon
NeumannandMorgenstern(1947).Solvingforequilibriummathematicallyrequires
solvingsimultaneousequationsinwhicheachplayer'sstrategyisaninputtotheother
4
player'scalculationofexpectedpayoff.Thesolutionisacollectionofstrategies,onefor
eachplayer,whereeachplayer’sstrategymaximizeshisexpectedpayoffgiventhe
strategiesoftheotherplayers.
Fromthebeginningofgametheory,howequilibriummightarisehasbeenthe
subjectofongoingdiscussion.Nashhimselfsuggestedthatequilibriumbeliefsmight
resolvefromchangesin“massaction”aspopulationslearnaboutwhatothersdoand
adjusttheirstrategiestowardoptimization.ii
MorerecentlygametheoristshaveconsideredtheepistemicrequirementsforNash
equilibriumbytreatinggamesasinteractivedecisionproblems(cf.Brandenburger1992).
ItturnsoutthatNashequilibriumforn‐playergamesrequiresverystrongassumptions
abouttheplayers’mutualknowledge:thatallplayersshareacommonpriorbeliefabout
chanceevents,knowthatallplayersarerational,andknowthattheirbeliefsarecommon
knowledge(Aumann&Brandenburger1995).iiiThelatterrequirementimpliesthat
rationalplayersbeabletocomputebeliefsaboutthestrategiesofcoplayersandallstates
oftheworld,beliefsaboutbeliefs,andsoon,adinfinitum.
TwoBehavioralApproaches:CognitiveHierarchyandPsychologicalGames
Cognitivehierarchy(CH)andpsychologicalgames(PG)modelsbothmodify
assumptionsfromgametheorytocapturebehaviormorerealistically.
TheCHapproachassumesthatboundedlyrationalplayersarelimitedinthe
numberofinterpersonaliterationsofstrategicreasoningtheycan(orchoose)todo.There
arefiveelementstoanyCHpredictivemodel:
5
1. Adistributionofthefrequencyofleveltypesf(k)
2. Actionsoflevel0players;
3. Beliefsoflevel‐kplayers(fork=1,2,…)aboutotherplayers;
4. Assessingexpectedpayoffsbasedonbeliefsin(3).
5. Astochasticchoiceresponsefunctionbasedontheexpectedpayoffsin(4)
Thetypicalapproachistomakepreciseassumptionsaboutelements(1‐5)andsee
howwellthatspecificmodelfitsexperimentaldatafromdifferentgames.Justasintesting
acookingrecipe,ifthemodelfailsbadlythenitcanbeextendedandimproved.
InCamerer,HoandChong(2004),thedistributionoflevelktypesisassumedto
followaPoissondistributionwithameanvalueτ.Oncethevalueofτischosen,the
completedistributionisknown.ThePoissondistributionhasthesensiblepropertythatthe
frequenciesofveryhighleveltypeskdropsoffquicklyforhighervaluesofk.(Forexample,
iftheaveragenumberofthinkingstepsτ=1.5,thenlessthan2%ofplayersareexpectedto
dofiveormorestepsofthinking.)
Tofurtherspecifythemodel,level0typesareusuallyassumedtochooseeach
strategyequallyoften.ivIntheCHapproach,levelkplayersknowthecorrectproportions
oflower‐levelplayers,butdonotrealizethereareotherevenhigher‐levelplayers(perhaps
reflectingoverconfidenceinrelativeability).Analternativeassumption(called“levelk”
modeling)isthatalevelkplayerthinksallotherplayersareatlevelk‐1.
Undertheseassumptions,eachlevelofplayerinahierarchycanthencomputethe
expectedpayoffstodifferentstrategies:Level1’scomputetheirexpectedpayoff(knowing
6
whatlevel0’swilldo);level2’scomputetheexpectedpayoffgiventheirguessaboutwhat
level1’sand0’sdo,andhowfrequentthoseleveltypesare;andsoforth.Inthesimplest
formofthemodel,playerschoosethestrategywiththehighestexpectedpayoff(the“best
response”);butitisalsoeasytousealogisticorpowerstochastic“betterresponse”
function(e.g.,Luce,1959).Becausethetheoryishierarchical,itiseasytoprogramand
solvenumericallyusinga“loop”.
Psychologicalgamesmodelsassumethatplayersarerationalinthesensethatthey
maximizetheirexpectedutilitygivenbeliefsandtheutilityfunctionsoftheotherplayers.
However,inpsychologicalgamesmodels,payoffsareallowedtodependdirectlyupon
player’sbeliefs,theirbeliefsabouttheircoplayers’beliefs,andsoon,adependencethatis
ruledoutinstandardgametheory.Theincorporationofbelief‐dependentmotivations
makesitpossibletocaptureconcernsaboutintentions,socialimage,orevenemotionsina
game‐theoreticframework.Forexample,inpsychologicalgamesoneperson,Conor(C)
mightbedelightedtobesurprisedbytheactionofanotherplayer,Lexie(L).Thisis
modeledmathematicallyasClikingwhenL’sstrategyisdifferentthanwhathe(C)
expectedLtodo.Someofthesemotivationsarenaturallyconstruedassocialemotions,
suchasguilt(e.g.,apersonfeelsbadchoosingastrategywhichharmedanotherpersonP
whodidnotexpectit,andfeelslessbadifPdidexpectit).
Ofthetwoapproaches,CHandlevel‐kmodelingareeasytouseandapplyto
empiricalsettings.Psychologicalgamesaremoregeneral,applyingtoabroaderclassof
games,butaremoredifficulttoadapttoempiricalwork.
II.TheCHmodel
7
Thenextsectionwillgivessomemotivatingempiricalexamplesofthewidescopeof
gamestowhichthetheoryhasbeenappliedwithsomesuccess(includingtwokindsoffield
data),andconsistencywithdataonvisualfixationandfMRI.TheCHapproachisappealing
asapotentialcognitivealgorithmforfourreasons:
1. Itappearstofitalotofexperimentaldatafrommanydifferentgamesbetter
thanequilibriumpredictionsdo(e.g.,Camereretal.,2004;Crawford,Costa‐
Gomes,&Iriberri,2010).
2. Thespecificationofhowthinkingworksandcreateschoicesinvites
measurementofthethinkingprocesswithresponsetimes,visualfixationson
certainpayoffs,andtransitionsbetweenparticularpayoffs.
3. TheCHapproachintroducesaconceptofskillintobehavioralgametheory.
IntheCHmodel,theplayerswiththehighestthinkinglevels(higherk)and
mostresponsivechoices(higherλ)areimplicitlymoreskilled.(In
equilibriummodels,allplayersareperfectlyandequallyskilled.)
NextwewilldescribeseveralempiricalgamesthatillustratehowCHreasoning
works.
Example1:p‐beautycontest
AsimplegamethatillustratesapparentCHthinkinghascometobecalledthe“p‐
beautycontestgame”(orPBC).ThenamecomesfromafamouspassageinJohnMaynard
Keynes’sbookTheGeneralTheoryofEmployment,InterestandMoney.Keyneswrote:
8
“Professionalinvestmentmaybelikenedtothosenewspapercompetitionsinwhich
thecompetitorshavetopickoutthesixprettiestfacesfromahundredphotographs,
theprizebeingawardedtothecompetitorwhosechoicemostnearlycorrespondsto
theaveragepreferencesofthecompetitorsasawhole;sothateachcompetitorhas
topick,notthosefaceswhichhehimselffindsprettiest,butthosewhichhethinks
likeliesttocatchthefancyoftheothercompetitors,allofwhomarelookingatthe
problemfromthesamepointofview.Itisnotacaseofchoosingthosewhich,tothe
bestofone'sjudgment,arereallytheprettiest,noreventhosewhichaverage
opiniongenuinelythinkstheprettiest.Wehavereachedthethirddegreewherewe
devoteourintelligencestoanticipatingwhataverageopinionexpectstheaverage
opiniontobe.Andtherearesome,Ibelieve,whopractisethefourth,fifthandhigher
degrees.”
IntheexperimentalPBCgamepeoplechoosenumbersfrom0to100simultaneously
withouttalking.vThepersonwhosenumberisclosesttoptimestheaveragewinsafixed
price.
Atypicalinterestingvalueofpis2/3.Thenthewinnerwantstobetwo‐thirdsofthe
waybetweentheaverageandzero.Butofcourse,theplayersallknowtheotherplayers
wanttopick2/3oftheaverage.InaNashequilibrium,everyoneaccuratelyforecaststhat
theaveragewillbeX,andalsochoosesanumberwhichis(2/3)X.ThisimpliesX=(2/3)Xor
X*=0.
9
Intuitively,supposeyouhadnoideawhatotherpeoplewoulddo,soyouchose2/3
of50=33.Thisisareasonablechoicebutisnotanequilibrium,sincechoosing33while
anticipating50leavesagapbetweenexpectedbehaviorofothersandlikelybehaviorby
oneself.Soapersonwhothinks“Hey!I’llpick33”shouldthenthink(toadheretothe
equilibriummath)“Hey!They’llpick33”andthenpick22.Thisprocessofimagining,
choosingandrevisingdoesnotstopuntileveryoneexpects0tobechosen,andalsopicks0.
Figure1showssomedatafromthisgameplayedwithexperimentalsubjectsandin
newspaperandmagazinecontests(wherelargegroupsplayforasinglelargeprize).There
issomeevidenceof“spikes”innumberscorrespondingto50p,50p2andsoon.
Example2:Bettingonselfishrationalityofothers
AnothersimpleillustrationoftheCHtheoryisshowninTable1.Inthisgamearow
andcolumnplayerchoosefromoneoftwostrategies,TorB(forrow)orLorR(for
column).Thecolumnplayeralwaysgets20forchoosingLand18forchoosingR.Therow
playergetseither30or10fromT,andasure20fromB.
Ifthecolumnplayeristryingtogetthelargestpayoff,sheshouldalwayschooseL(it
guarantees20insteadof18).ThestrategyLiscalleda“strictlydominantstrategy”because
ithasthehighestpayoffforeverypossiblechoicebytherowplayer.
Therowplayer’schoiceisalittletrickier.Shecanget20forsurebychoosingB.
ChoosingTistakingasocialgamble.Ifsheisconfidentthecolumnplayerwilltrytoget20
andchooseL,sheshouldinferthatP(L)ishigh.ThentheexpectedvalueofTishighandshe
shouldchooseT.However,thisinferenceisessentiallyabetontheselfishrationalityofthe
10
otherplayer.Therowplayermightthinkthecolumnplayerwillmakeamistake,oris
spiteful(andprefersthe(10,18)cellbecauseshegetslessabsolutepayoffbutahigher
relativepayoffcomparedtotherowplayer).Thereisacrucialcognitivedifferencein
playingL—whichistherightstrategyifyouwantthemostmoney—andplayingT—which
istherightstrategyifyouarewillingtobetthatotherplayersareverylikelytochooseL
becausetheywanttoearnthemostmoney.
WhatdoestheCHapproachpredicthere?Supposelevel0playersrandomize
betweenthetwostrategies.Ifτ=1.5,thenf(0|τ=1.5)=.22.Thenhalfofthelevel0players
willchoosecolumnRandrowB,whichis.11%ofthewholegroup.
Level1playersalwayschooseweaklydominantstrategies,sotheypickcolumnL
(infact,allhigherlevelcolumnplayersdotoo).Sincelevel1rowplayersthinkLandR
choicesareequallylikely,theirexpectedpayofffromTis30(.5)+10(.5)=20,whichisthe
sameastheBpayoff;soweassumetheyrandomizeequallybetweenTandB.Since
f(1|τ=1.5)=.33,thismeanstheunconditionaltotalfrequencyofBplayforthefirsttwo
levelsis.11+.33/2=.27.
Level2rowplayersthinktherelativeproportionsoflowertypesare
g2(0)=.22/(.22+.33)=.40andg2(1)=.33/(.22+.33)=.60.Theyalsothinkthelevel0’splay
eitherLorR,butthelevel1’schooseLforsure.Together,thisimpliesthattheybelieve
thereisa.20chancetheotherpersonwillchooseR(=.5(.40)+0(.60))andan.80chance
theywillchooseL.Withtheseodds,theyprefertochooseT.Thatis,theyaresufficiently
confidenttheotherplayerwill“figureitout”andchoosetheself‐servingLthatTbecomesa
goodbettoyieldthehigherpayoffof30.
11
Puttingtogetherallthefrequenciesf(k)andchoicepercentages,theoverall
expectedproportionofcolumnRplayis.11androwBplayis.27.Notethatthese
proportionsgointhedirectionoftheNashprediction(whichiszeroforboth),butaccount
morepreciselyforthechanceofmistakesandmisperceptions.Importantly,choicesofR
shouldbelesscommonthanchoicesofB.Rchoicesarejustcareless,whileBchoicesmight
becarelessormightbesensibleresponsestothinkingtherearealotofcarelessplayers.
Table1showsthatsome(unpublished)datafromCaltechundergraduateclassroom
games(formoney)overthreeyearsaregenerallyclosetotheCHprediction.TheRandB
choicefrequenciesaresmall(asbothNashandCHpredict)butBismorecommonthanR.
[InsertTable1abouthere]
OnepotentialadvantageofCHmodelingisthatthesamegeneralprocesscould
applytogameswithdifferenteconomicstructures.Inbothofthetwoexamplesabove,a
Nashequilibriumchoicecanbederivedbyrepeatedapplicationoftheprincipleof
eliminating“weaklydominated”strategies(i.e.,strategieswhichareneverbetterthan
anotherdominatingstrategy,forallchoicesbyotherpeople,andisactuallyworseforsome
choicesbyothers).Hence,thesearecalled“dominancesolvable”games.Indeed,the
beauty‐contestexampleisamongthosethatmotivatedCHmodelinginthefirstplace,since
eachstepofreasoningcorrespondstoonemorestepindeletionofdominatedstrategies.
Hereisanentirelydifferenttypeofgame,called“asymmetricmatchingpennies”.In
thisgametherowplayerearnspointsifthechoicesmatch(H,H)or(T,T).Thecolumn
12
playerwinsiftheymismatch.Thereisnopairofstrategiesthatarebestresponsestoeach
other,sotheequilibriumrequireschoosingaprobabilistic“mixture”ofstrategies.Here,
equilibriumanalysismakesabizarreprediction:TherowplayershouldchooseHandT
equallyoften,whilethecolumnplayershouldshyawayfromH(asifpreventingRowfrom
gettingthebiggerpayoffof2)andchooseT2/3ofthetime.(Evenmorestrangely:Ifthe2
payoffisx>1ingeneral,thenthemixtureisalways50‐50fortherowplayer,andisx/(x+1)
onTforthecolumnplayer!Thatis,intheorychangingthepayoffof2onlyaffectsthe
columnplayer,anddoesnotaffecttherowplayerwhomightearnthatpayoff.
TheCHapproachworksdifferentlyvi.Thelowerlevelrowplayers(1‐2)areattracted
tothepossiblepayoffof2,andchooseH.However,thelowlevelcolumnplayersswitchto
T,andhigherlevelrowplayers(levels3‐4)figurethisoutandswitchtoT.Thepredicted
mixture(fortau=1.5)isactuallyratherclosetotheNashpredictionforthecolumnplayer
(P(T)=.74comparedtoNash.67),sincethehigher‐leveltypeschooseTmoreandnotH.
Andindeed,datafromcolumnplayerchoicesinexperimentsareclosetobothpredictions.
TheCHmixtureofrowplay,averagedacrosstypefrequencies,isP(H)=.68,closetothedata
averageof.72.Thus,thereasonablepartoftheNashprediction,whichislopsidedplayofT
andHbycolumnplayers,isreproducedbyCHandisconsistentwiththedata.The
unreasonablepartoftheNashprediction,thatrowplayerschooseHandTequallyoften,is
notreproducedandthedifferingCHpredictionismoreempiricallyaccurate.
[InsertTable2abouthere]
Entrygames
13
Insimple“entry”games,Nplayerssimultaneouslychoosewhethertoentera
marketwithdemandC,ornot.Iftheystayout,theyearnafixedpayoff($.50).Iftheyenter,
thenalltheentrantsearn$1ifthereareCorfewerentrants,andearn0iftherearemore
thanCentrants.ItiseasytoseethattheequilibriumpatternofplayisforexactlyCpeople
toenter;thentheyeachearn$1andthosewhostayoutearn$.50.Ifoneofthestayer‐
outersswitchedandentered,shewouldtipthemarketandcausetheC+1entrantstoearn
0.Sincethiswouldlowerherownpayoff,shewillstayput.Sothepatternisanequilibrium.
However,thereisaproblemremaining(it’sacommononeingametheory):How
doesthegroupcollectivelydecide,withouttalking,whichoftheCpeopleenterandearn
$1?EverybodywouldliketobeintheselectgroupofCentrantsiftheycan;butiftoomany
entertheyallsuffer.viiThisisafamiliarproblemof“coordinating”toreachoneofmany
differentequilibria.
Thefirstexperimentsonthistypeofentrygameweredonebyateamofeconomists
(JamesBranderandRichardThaler)andapsychologist,DanielKahneman.Theywere
neverfullypublishedbutweredescribedinachapterbyKahneman(1988).Kahneman
saystheywereamazedhowclosethenumberoftotalentrantswastotheannounced
demandC(whichvariedovertrials).“Toapsychologist”,hewrote,“itlookedlikemagic”.
Sincethen,acoupleofdozenstudieshaveexploredvariantsofthesegamesandreported
similardegreesofcoordination(e.g.,Duffy&Hopkins,2005).
Let’sseeifcognitivehierarchycanproducethemagic.Supposelevel0playersenter
andstayoutequallyoften,andignoreC.Iflevel1playersanticipatethis,theywillthink
therearetoomanyentrantsforC<(N/2)andtoofewifC>(N/2)‐1.Level1playerswill
14
thereforeenterathighvaluesofC.Noticethatlevel1playersarehelpingthegroupmove
towardtheequilibrium.Level1’sundothedamagedonebythelevel0’s,whoover‐enterat
lowC,bystayingoutwhichreducestheoverallentryrateforlowC.Theyalsoexploitthe
opportunitythatremainsforhighC,byentering,whichincreasestheoverallentryrate.
Combiningthetwolevels,therewillbelessentryatlowCandmoreentryathighC(itwill
looklikeastepfunction;seeCamereretal.,2004).
Furthermore,itturnsoutthataddinghigher‐levelthinkerscontinuestopushthe
populationprofiletowardanoverallentrylevelthatisclosetoC.Thetheorymakesthree
sharppredictions:(1)Plottingentryrates(asa%ofN)againstC/Nshouldyielda
regressivelinewhichcrossesat(.5,.5).(2)EntryratesshouldbetoohighforC/N<.5and
toolowforC/N>.5.(3)EntryshouldbeincreasinginC,andrelativelyclose,evenwithout
anylearningatall!(e.g.,inthefirstperiodofthegame).
Figure3illustratesaCHmodelpredictionwithτ=1.25,single‐perioddatawithno
feedbackfromCamereretal.(2004),andtheequilibrium(a45‐degreeline).Exceptfor
somenonmonotonicdipsintheexperimentaldata(easilyaccountedforbysampling
error),thepredictionsareroughlyaccurate.
Thepointofthisexampleisthatapproximateequilibrationcanbeproduced,asifby
“magic”,purelyfromcognitivehierarchythinkingwithoutanylearningorcommunication
needed.Thesedataarenotsolidproofthatcognitivehierarchyreasoningisoccurringin
thisgame,butdoesshowhow,inprinciple,thecognitivehierarchyapproachcanexplain
bothdeviationsfromNashequilibrium(inthebeautycontest,betting,andmatching
15
penniesgamesthatweredescribedabove),andalsosurprisingconformitytoNash
equilibrium(inthisentrygame).
Privateinformation
Thetrickiestclassofgameswewilldiscuss,briefly,involve“privateinformation”.
Thestandardmodelingapproachistoassumethereisahiddenvariable,X,whichhasa
possibledistributionp(X)thatiscommonlyknowntobothplayersviii.TheinformedplayerI
knowstheexactvaluexfromthedistributionandbothplayersknowthatonlyIknowsthe
value.Forexample,incardgameslikepoker,playersknowthepossiblesetofcardstheir
opponentmighthave,andknowthattheopponentknowsexactlywhatthecardsare.
Thecognitivechallengethatisspecialtoprivateinformationgamesistoinferwhat
aplayer’sactions,whethertheyareactuallytakenorhypothetical,mightrevealabouttheir
information.Variousexperimentalandfielddataindicatethatsomeplayersarenotvery
goodatinferringhiddeninformationfromobservedaction(oranticipatingtheinferable
information).
Asimpleandpowerfulexampleisthe“acquire‐a‐company”problemintroducedin
economicsbyAkerlof(1970)andstudiedempiricallybyBazermanandSamuelson(1983).
Inthisgame,aprivately‐heldcompanyhasavaluewhichisperceivedbyoutsiderstobe
uniformlydistributedfrom0to100(i.e.,allvaluesinthatrangeareequallylikely).The
companyknowsitsexactvalue,andoutsidersknowthatthecompanyknows(duetothe
commonpriorassumption).
16
Abiddercanoperatethecompanymuchbetter,sothatwhateverthehiddenvalueV
is,itisworth1.5Vtothem.Theymakeatake‐it‐or‐leave‐it(Boulwarean)offerofapriceP.
Thebargainingcouldhardlybesimpler:ThecompanysellsifthepricePisabove“hidden”
valueV—whichthebidderknowsthatthecompanyknows—andkeepsthecompany
otherwise.Thebidderwantstomaximizetheexpected“surplus”gainbetweentheaverage
ofthevalues1.5Vtheyarelikelytoreceiveandtheprice.
Whatwouldyoubid?Theoptimalbidissurprising,thoughthealgebrabehindthe
answerisnottoohard.ThechanceofgettingthecompanyisthechancethatVislessthan
P,whichisP/100(e.g.,ifP=60then60%ofthetimethevalueisbelowPandthecompany
changeshands).Ifthecompanyissold,thenthevaluemustbebelowP,sotheexpected,
valuetotheselleristheaverageofthevaluesintheinterval[0,P],whichisP/2.Thenet
expectedvalueistherefore(P/100)timesexpectedprofitifsold,whichis1.5*(P/2)‐P=
‐1/4P.Thereisnowaytomakeaprofitonaverage.Theoptimalbidiszero!
However,typicaldistributionsofbidsarebetween50and75.Thisresultsina
“winner’scurse”inwhichbidders“win”thecompany,butfailtoaccountforthefactthat
theyonlywonbecausethecompanyhadalowvalue.Thisphenomenonwasfirstobserved
infieldstudiesofoil‐leasebidding(Capenetal1971)andhasbeenshowninmanylaband
fielddatasetssincethen.Thegeneralprinciplethatpeoplehaveahardtimeguessingthe
implicationsofprivateinformationforactionsotherswilltakeshowsupinmanyeconomic
settings(akindofstrategicnaivete;e.g.Brocasetal.,2009).
17
TheCHapproachcaneasilyexplainstrategicnaiveteasaconsequenceoflevel1
behavior.Iflevel1playersthinkthatlevel0players’choicesdonotdependonprivate
information,thentheywillignorethelinkbetweenchoicesandinformation.
Eyetrackingevidence
Apotentialadvantageofcognitivehierarchyapproachesisthatcognitivemeasures
associatedwiththealgorithmicstepsplayersareassumedtouse,intheory,couldbe
collectedalongwithchoices.Forpsychologiststhisisobviousbut,amazingly,itisarather
radicalpositionineconomicsandmostareasofgametheory!
Theeasiestandcheapestmethodistorecordwhatinformationpeoplearelooking
atastheyplaygames.Eyetrackingmeasuresvisualfixationsusingvideo‐basedeyetracking,
typicallyevery5‐50msec.Cameraslookintotheeyeandadjustforheadmotiontoguess
wheretheeyesarelooking(usuallywithexcellentprecision).Mosteyetrackersrecord
pupildilationaswell,whichisusefulasameasureofcognitivedifficultyorarousal.
Sincegametheoryisaboutinteractionsamongtwoormorepeople,itisespecially
usefultohavearecordingtechnologythatscalesuptoenablerecordingofseveralpeople
atthesametime.Onewidely‐usedmethodiscalled“Mouselab”.InMouselab,information
thatisusedinstrategiccomputations,intheory,ishiddeninlabeledboxes,which“open
up”whenamouseismovedintothem.ix
18
Severalstudieshaveshownthatlookuppatternsoftencorrespondroughly,and
sometimesquiteclosely,todifferentnumbersofstepsofthinking.We’llpresentone
example(seealsoCrawfordetal.,2010).
Example1:Alternating‐offerbargaining
Apopularapproachtomodelingbargainingistoassumethatplayersbargainovera
knownsumofjointgain(sometimescalled“surplus”,likethevaluablegapbetweenthe
highestpriceabuyerwillpayandthelowestpriceasellerwillaccept).However,astime
passestheamountofjointgain“shrinks”duetoimpatienceorothercosts.Players
alternatemakingoffersbackandforth(Rubinstein,1982).
Athree‐periodversionofthisgamehasbeenstudiedinmanyexperiments.The
amountdividedinthefirstroundis$5,whichthenshrinksto$2.50,$1.25,and0inlater
rounds(thelastroundisan“ultimatumgame”).Ifplayersareselfishandmaximizetheir
ownpayoffs,andbelievethatothersaretoo,the“subgameperfect”equilibrium(SPE)offer
bythefirstpersonwhooffers(player1),toplayer2,shouldbe$1.25.However,deriving
thisoffereitherrequiressomeprocessoflearningorcommunication,orananalysisusing
“backwardinduction”todeducewhatofferswouldbemadeandacceptedinallfuture
rounds,thenworkingbacktothefirstround.Earlyexperimentsshowedconflictingresults
inthisgame.Neelinetal.(1988)foundthataverageofferswerearound$2,andmanywere
equalsplitsof$2.50each.Earlier,Binmoreetal.(1985)foundsimilarresultsinthefirst
roundofchoices,butalsofoundthatasmallamountofexperiencewith“rolereversal”
(player2’sswitchingtotheplayer1first‐offerposition)movedofferssharplytowardthe
SPEofferof$1.25.Otherevidencefromsimplerultimatumgamesshowedthatpeopleseem
19
tocareaboutfairness,andarewillingtorejecta$2offeroutof$10abouthalfthetime,to
punishabargainingpartnertheythinkhasbeenunfairandgreedy(Camerer,2003).
Soaretheoffersaround$2duetocorrectanticipationoffairness‐influenced
behavior,ortolimitedunderstandingofhowthefutureroundsofbargainingmightshape
reactionsinthefirstround?Tofindout,Camereretal.(1993)andJohnsonetal.(2002)did
thesametypeofexperiment,buthidtheamountsbeingbargainedoverineachroundin
boxesthatcouldbeopened,or“lookedup”,bymovingamouseintothoseboxes(an
impoverishedexperimenter’sversionofvideo‐basedeyetracking).Theyfoundthatmost
peoplewhoofferedamountsbetweentheequalsplitof$2.50andtheSPEof$1.25werenot
lookingaheadatpossiblefuturepayoffsasbackwardinductionrequires.Infact,in10‐20%
ofthetrialstheround2andround3boxeswerenotopenedatall!
Figure4illustratesthebasicresults.Thetoprectangular“icongraphs”visually
representtherelativeamountsoftimebargainersspentlookingateachpayoffbox(the
shadedarea)andnumbersofdifferentlookups(rectanglewidth).Theboldarrowsindicate
therelativenumberoftransitionsfromoneboxtothenext(withaveragesoflessthanone
transitionomitted).
Eachcolumnrepresentsagroupoftrialsthatarepre‐classifiedbylookuppatterns.
Thefirstcolumn(N=129trials)averagespeoplewholookedmoreoftenattheperiod1box
thanatthefutureperiodboxes(indicating“level‐0”planning).Thesecondcolumn(N=84)
indicatespeoplewholookedlongeratthesecondboxthanthefirstandthird(indicating
“levelOne”planningwithsubstantialfocusonestepahead).Thethirdcolumn(N=27)
indicatesthesmallernumberof“equilibrium”trialsinwhichthethirdboxislookedatthe
20
most.NotethatinthelevelOneandEquilibriumtrials,therearealsomanytransitions
betweenboxesoneandtwo,andboxestwoandthree,respectively.Finally,thefourthand
lastcolumnshowspeoplewhowerebrieflytrainedinbackwardinduction,thenplayeda
computerizedopponentthat(theyweretold)plannedahead,actedselfishlyandexpected
thesamefromitsopponents.
Themainpatterntonoticeisthatofferdistributions(shownatthebottomofeach
column)shiftfromright(fairer,indicatedbytherightdottedline)toleft(closertoselfish
SPE,theleftdottedline)asplayersliterallylookaheadmore.Thelinkbetweenlookupsand
higher‐than‐predictedoffersclearlyshowsthatoffersabovetheSPE,inthedirectionof
equalsplitsofthefirstroundamount,arepartlyduetolimitsonattentionandcomputation
aboutfuturevalues.Eveninthefewequilibriumtrials,offersarebimodal,clusteredaround
$1.25and$2.20.However,offersarerathertightlyclusteredaroundtheSPEpredictionof
$1.25inthe“trained”condition.Thisresultindicates,importantly,thatbackwardinduction
isnotactuallythatcognitivelychallengingtoexecute(afterinstruction,theycaneasilydo
it),butinsteadisanunnaturalheuristicthatdoesnotreadilyspringtothemindsofeven
analyticalcollegestudents.
fMRIevidence
Severalneuralstudieshaveexploredwhichbrainregionsaremostactivein
differenttypesofstrategicthinking.Theearlieststudiesshoweddifferentialactivation
whenplayingagameagainstacomputercomparedtoarandomizedopponent(e.g.,
(Gallagheretal.,2002;McCabeetal.,2001;Coricelli&Nagel,2009).
21
Oneofthecleanestresults,andanexemplarofthecompositepictureemergingfrom
otherstudies,isfromCoricelli&Nagel’s(2009)studyofthebeautycontestgame.Their
subjectsplayed13differentgameswithdifferenttargetmultipliersp(e.g.,p=2/3,1,3/2
etc.).Oneachtrial,subjectschosenumbersintheinterval[0,100]playingagainsteither
humansubjectsoragainstarandomcomputeropponent.Usingbehavioralchoices,most
subjectscanbeclassifiedintoeitherlevel1(n=10;choosingptimes50)orlevel2(n=7;
choosingptimesptimes50,asifanticipatingtheplayoflevel1opponents).
Figure7showsbrainareasthatweredifferentiallyactivewhenplayinghuman
opponentscomparedtocomputeropponents,andinwhichthathuman‐computer
differentialislargerinlevel2playerscomparedtolevel1players.Thecrucialareasare
bilateraltemporo‐parietaljunction(TPJ),MPFC/paracingulateandVMPFC.xTheseregions
arethoughttobepartofageneralmentalizingcircuit,alongwithposteriorcingulate
regions(Amodio&Frith,2006).
Inrecentstudies,atleastfourareasarereliablyactivatedinhigher‐levelstrategic
thinking:dorsomedialprefrontalcortex(DMPFC),precuneus/posteriorcingulate,insula,
anddorsolateralprefrontalcortex(DLPFC).Nextwesummarizesomeofthesimplest
results.
DMPFCactivityisevidentinFigure7.Itisalsoactiveinresponsetononequilibrium
choices(wheresubjects’guessesaboutwhatotherswilldoarewrong;Bhatt&Camerer,
2005),anduncertaintyaboutstrategicsophisticationofanopponent(Yoshidaetal.,2009).
Inaddition,DMPFCactivityisrelatedtothe“influencevalue”ofcurrentchoicesonfuture
rewards,filteredthroughtheeffectofaperson’sfuturechoicesonanopponent’sfuture
22
choices(Hampton,Bossaerts,&O’Doherty,2009;a/k/a“strategicteaching”Camerer,Ho
andChong2002).Amodio&Frith(2006)suggestanintriguinghypothesis:that
mentalizing‐valueactivationforsimplertomorecomplexactionvaluecomputationsare
differentiallylocatedalongaposterior‐to‐anterior(back‐to‐front)gradientinDMPFC.
Indeed,thelatterthreestudiesshowactivationroughlyinaposterior‐to‐anteriorgradient
(Tailarachy=36,48,63;andy=48inCoricelli&Nagel,2009)thatcorrespondstoincreasing
complexity.
Activityintheprecuneus(adjacenttoposteriorcingulate)isassociatedwith
economicperformanceingames(“strategicIQ”;Bhatt&Camerer2005)anddifficultyof
strategiccalculations(Kuoetal.2009).Precuneusisabusyregion,withreciprocal
connectionstoMPFC,cingulate,andDLPFC.Itisalsoactivatedbyawidevarietyofhigher‐
ordercognitions,includingperspective‐takingandattentionalcontrol(aswellasthe
“defaultnetwork”activeatrest;seeBhatt&Camerer,inpress).Itislikelythatprecuneusis
notactivatedinstrategicthinking,perse,butonlyinspecialtypesofthinkingwhich
requiretakingunusualperspectives(e.g.,thinkingaboutwhatotherpeoplewilldo)and
shiftingmentalattentionbackandforth.
Theinsulaisknowntobeinvolvedininteroceptiveintegrationofbodilysignalsand
cognition.Disgust,physicalpain,empathyforothersinpain,andpainfromsocialrejection
activateinsula(Eisenbergeretal.2003,Krossetal.2011).Financialuncertainty
(Preuschoff,Quartz,Bossaerts,2008),interpersonalunfairness(Sanfeyetal.,2003;Hsuet
al.,2008),avoidanceofguiltintrustgames(Changetal.,2011),and“coaxing”orsecond‐
trysignalsintrustgamesalsoactivateinsula.Instrategicstudies,BhattandCamerer
23
(2005)foundthathigherinsulaactivityisassociatedwithlowerstrategicIQ
(performance).
TheDLPFCisinvolvedinworkingmemory,goalmaintenance,andinhibitionof
automaticprepotentresponses.Differentialactivitythereisalsoassociatedwiththelevel
ofstrategicthinking(Yoshidaetal.,2009)withstrongerresponsetohumanopponentsin
higher‐levelstrategicthinkers(Coricelli&Nagel,2009),andwithmaintaininglevel‐2
deceptioninbargaininggames(Bhattetal.,2009).xi
Dothinkingstepsvarywithpeopleorgames?
Towhatextentdostepsofthinkingvarysystematicallyacrosspeopleorgame
structures?Fromacognitivepointofview,itislikelythatthereissomeintrapersonal
stabilitybecauseofdifferencesinworkingmemory,strategicsavvy,exposuretogame
theory,experienceinsportsbettingorpoker,taskmotivation,etc.However,itisalsolikely
thattherearedifferencesinthedegreeofsophistication(measuredbyτ)acrossgames
becauseofaninteractionbetweengamecomplexityandworkingmemory,orhowwellthe
surfacegamestructuremapsontoevolutionarilyfamiliargamesxii.
Todate,thesesourcesofleveldifferenceshavenotbeenexploredverymuch.Chong,
HoandCamerer(2005)notesomeeducationaldifferences(Caltechstudentsareestimated
todo.5stepsofthinkingmorethansubjectsfromanearbycommunitycollege)andan
absenceofagendereffect.Otherstudieshaveshowedmodestassociations(r=.3)between
strategiclevelsandworkingmemory(digitspan;Devetag&Warglien,2003)andthe“eyes
ofthemind”testofemotiondetection(Georganas,Healy,&Weber2010).
24
Manypapershavereportedsomedegreeofcross‐gametypestabilityinlevel
classification.Studiesthatcompareachoiceinonegamewithonedifferentgamereport
lowstability(Georganasetal.,2010;Burchardi&Penczynski2010).However,asiswell‐
knowninpersonalitypsychologyandpsychometrics,intrapersonalreliabilitytypically
increaseswiththenumberofitemsusedtoconstructascale.Otherstudiesusingmore
gamechoicestoclassifyreportmuchhighercorrelations(comparabletoBig5personality
measures)(Bhui&Camerer,2011).
Asoneillustrationofpotentialtype‐stability,Figure5belowshowsestimatedtypes
forindividualsusingthefirst11gamesina22‐gameseries(x‐axis)andtypesforthesame
individualsusingthelast11games.Thecorrelationisquitehigh(r=.61).Thereisalsoa
slightupwarddriftacrossthegames(theaveragelevelishigherinthelast11games
comparedtothefirst),consistentwithatransferorpracticeeffect,eventhoughthereisno
feedbackduringthe22games(seealsoWeber,2003).
Fielddata
Sincecontrolledexperimentationcamelatetoeconomics(c.1960)comparedto
psychology,thereisalong‐standingskepticismaboutwhethertheoriesthatworkinsimple
labsettingsgeneralizetonaturally‐ocurringeconomicactivity.Fivestudieshaveapplied
CHorlevel‐kmodelingtoauctions(Gillen,2009),strategicthinkinginmanagerialchoices
(GoldfarbandYang,2009;GoldfarbandXiao,inpress),andboxofficereactionwhen
moviesarenotshowntocriticsbeforerelease(Brown,Camerer&Lovallo,2011).
Onestudyisdescribedhereasanexample(Ostlingetal.,2011).In2007the
SwedishLotterycreatedagameinwhichpeoplepay1eurotoenteralottery.Eachpaying
25
entrantchoosesaninteger1‐99,999.Thelowestuniquepositiveinteger(hence,the
acronymLUPI)winsalargeprize.
Thesymmetricequilibriumisaprobabilisticprofileofhowoftendifferentnumbers
arechosen(a“mixed”equilibrium).Thelowestnumbersarealwayschosenmoreoften
(e.g.,1ischosenmostoften);therateofdeclineinthefrequencyofchoiceisaccelerating
uptoasharpinflectionpoint(number5513);andtherateofdeclineslowsdownafter
5513.
Figure6showsthedatafromonlythelowest10%ofthenumberrange,from1‐
10,000(highernumberchoicesarerare,asthetheorypredicts).ThepredictedNash
equilibriumisshownbyadottedline—aflat“shelf”ofchoiceprobabilityfrom1to5513,
thenasharpdrop.AfittedversionoftheCHmodelisindicatedbythesolidline.CHcan
explainthelargefrequencyoflownumberchoices(below1500),sincethesecorrespondto
lowlevelsofstrategicthinking(i.e.,peopledon’trealizeeveryoneelseischoosinglow
numberstoo).Sincelevel‐0typesrandomize,theirbehaviorproducestoomanyhigh
numbers(above5000).Sincethelowestandhighestnumbersarechosentoooften
accordingtoCH,comparedtotheequilibriummixture,CHalsoimpliesagapbetween
predictedandactualchoicesintherange2500‐5000.Thisbasicpatternwasreplicatedina
labexperimentwithasimilarstructure.WhiletherearecleardeviationsfromNash
equilibrium,consistentwithevidenceoflimitedstrategicthinking,inourviewtheNash
theorypredictionisnotbadconsideringthatusesnofreeparameters,andcomesfroman
equationwhichiselegantinstructurebutdifficulttoderiveandsolve.
26
TheLUPIgamewasplayedinSwedenfor49daysinarow,andresultswere
broadcastonanightlyTVshow.Analysisindicatesanimitate‐the‐winnerfictivelearning
process,sincechoicesononedaymoveinthedirectionof600‐numberrangearoundthe
previousday’swinner.Theresultofthisimitationisthateverystatisticalfeatureofthe
numberschosenmovestowardtheequilibriumacrossthesevenweeks.Forexample,in
thelastweektheaveragenumberis2484,within4%ofthepredictedvalueof2595.
IIIPsychologicalgames
Inmanystrategicinteractions,ourownbeliefsorbeliefsofotherpeopleseemto
influencehowwevalueconsequences.Forexample,surprisingapersonwithawonderful
giftthatisperfectforthemismorefunforeveryonethanifthepersonhadaskedforit.
Someofthatpleasurecomesfromthesurpriseitself.
Thistypeofpatterncanbemodeledasa“psychologicalgame”(Geanakoplos,Pearce
&Stacchetti,1989andBattigalli&Dufwenberg,2009).PGsareanextensionofstandard
gamesinwhichtheutilityevaluationsofoutcomescandependonbeliefsaboutwhatwas
thoughttobelikelytohappen(aswellastypicalmaterialconsequences).Thisapproach
requiresthinkingandreasoningsincethebeliefisderivedfromanalysisoftheother
person’smotives.Togetherthesepapersprovidetoolsforincorporatingmotivationssuch
asintentions,socialnorms,andemotionsintogame‐theoreticmodels.
Emotionsareanimportantbeliefdependentmotivation.Anxiety,
disappointment,elation,frustration,guilt,joy,regret,andshame,amongother
emotions,canallbeconceivedofasbelief‐dependentincentivesormotivationsand
27
incorporatedintomodelsofbehaviorusingtoolsfrompsychologicalgametheory.
Oneexampleisguilt:Baumeister,Stillwell,andHeatherton(1994)write:“If
peoplefeelguiltyforhurtingtheirpartners…andforfailingtoliveuptotheir
expectations,theywillaltertheirbehavior(toavoidguilt).”BattigalliandDufwenberg
(2007)operationalizethenotionthatpeoplewillanticipateandavoidguiltintheir
modelofguiltaversion.Intheirmodel,playersderivepositiveutilityfromboth
materialpayoffsandnegativeutilityfromguilt.Playersfeelguiltyiftheirbehavior
disappointsaco‐playerrelativetohisexpectations.xiii
ConsiderFigure8,whichillustratesasimpletrustgame.Player1maychoose
either“Trust”or“Don’t.”Inthefirstcaseplayer1getsthemove,whileafterachoiceof
“Don’t”thegameendsandeachplayergetspayoff1.Ifplayer2getsthemove,she
choosesbetween“Grab”and“Share.”ThepayoffstoGrabare0forplayer1and4for
player2.
ThesubgameperfectequilibriumofthisgameforselfishplayersisforPlayer2to
chooseGrabifshegetsthemovesinceitresultsinahigherpayoffforherthanchoosing
Share.Player1anticipatesthisbehaviorandchoosesDon’ttoavoidreceiving0.Both
playersreceiveapayoffof1,whichisinefficient.
NowsupposethatPlayer2isguiltaverse.Thenherutilitydependsnotonlyonher
materialpayoff,butalsoonhowmuchshe“letsdown”player1relativetohisexpectations.
Letpbetheprobabilitythatplayer1assignsto“Share.”Letp’representPlayer2’s(point)
beliefregardingp,andsupposethat2’spayofffromGrabisthen4‐θp,wheretheta
28
representsplayer2’ssensitivitytoguilt.Ifplayer1choosesTrustitmustbethatpis
greaterthan½‐otherwiseplayer1wouldchooseDon’t.Thenifθ≥2,player2willchoose
SharetoavoidtheguiltfromlettingdownPlayer1.Knowingthis,player1willchoose
Trust.Inthisoutcomebothplayersreceive2(insteadof1intheselfishsubgameperfect
equilibrium),illustratinghowguiltaversioncanfostertrustandcooperationwhereselfish
behaviorleadstoinefficiency.
Anumberofexperimentshavestudiedguiltaversioninthecontextoftrustgames,
includingDufwenberg&Gneezy(2000),Charness&Dufwenberg(2006,2011),and
Reubenetal.(2009).Allofthesepapersfindevidencethatadesiretoavoidguiltmotivates
playerstobehaveunselfishlybyreciprocatingtrust(foracontraryopinionseeEllingsenet
al.,2010).RecentfMRIevidence(Changetal,inpress)suggeststhatavoidingguiltintrust
gamesisassociatedwithincreasedactivityintheanteriorinsula.
Psychologicalgametheoryalsomaybeemployedtomodelothersocialemotions
suchasshame(Tadelis,2008)oranger(Smith,2009)ortoimportexistingmodelsof
emotionssuchasdisappointment,elation,regret,andrejoicing(Bell,1982,1986;
Loomes&Sugden,1982,1985)intogames.xivBattigalli&Dufwenberg(2009)provide
someexamplesoftheseapplications.Thesemodelsarejustaglimpseofthepotential
applicationsofpsychologicalgametheorytotheinteractionofemotionandcognitionin
socialinteractions.
Anotherimportantapplicationofpsychologicalgametheoryissociological
concerns,suchasreciprocity(whichmaybedrivenbyemotions).Inanimportantwork,
Rabin(1993)modelsreciprocityviafunctionsthatcaptureaplayer’s“kindness”tohis
29
coplayerandtheotherplayer’skindnesstohim.Thesekindnessfunctionsdependonthe
players’beliefsregardingeachother’sactions,andtheirbeliefsabouteachother’sbeliefs.
Dufwenberg&Kirchsteiger(2004)andFalk&Fischbacher(2006)extendRabin’smodelto
sequentialgames.
Psychologicalgametheoryprovidesausefultoolkitforincorporatingpsychological,
social,andculturalfactorsintoformalmodelsofdecision‐makingandsocialinteractions.
Manyapplicationsremaintobediscoveredandtestedviaexperiment.
Conclusions
Comparedtoitsimpactonotherdisciplines,gametheoryhashadlessimpactin
cognitivepsychologysofar.Thisislikelybecausemanyoftheanalyticalconceptsusedto
derivepredictionsabouthumanbehaviordonotseemtocorrespondcloselytocognitive
mechanisms.Somegametheoristshavealsocomplainedaboutthisunrealism.EricVan
Damme(1999)wrote:
Withouthavingabroadsetoffactsonwhichtotheorize,thereisacertaindangerof
spendingtoomuchtimeonmodelsthataremathematicallyelegant,yethavelittle
connectiontoactualbehavior.Atpresentourempiricalknowledgeisinadequate
anditisaninterestingquestionwhygametheoristshavenotturnedmore
frequentlytopsychologistsforinformationaboutthelearningandinformation
processesusedbyhumans.
30
Butrecently,anapproachcalledbehavioralgametheoryhasbeendevelopedwhich
usespsychologicalideastoexplainbothchoicesinmanydifferentgames,andassociated
cognitiveandbiological(Camerer2003;Bhatt&Camerer,2011)
Thischapterdiscussedtwoelementsofbehavioralgametheorythatmightbeof
mostinteresttocognitivepsychologists:Thecognitivehierarchyapproach;and
psychologicalgamesinwhichoutcomevaluescandependonbeliefs,oftenaccompaniedby
emotions(e.g.,alowbargainingoffercouldcreateangerifyouexpectedmore,orjoyifyou
expectedless).
Thecognitivehierarchyapproachassumesthatsomeplayerschooserapidlyand
heuristically(“level0”)andhigher‐levelplayerscorrectlyanticipatewhatlower‐level
playersdo.Thetheoryhasbeenusedtoexplainbehaviorinlabgameswhichisbothfar
fromandclosetoequilibjriumindifferentgames,issupportedbyevidencefromvisual
eyetrackingandMouselab,isevidentin“theoryofmind”circuitryduringfMRI,andalsocan
explainsomepatternsinfielddata(suchastheSwedishLUPIlottery).
Researchonpsychologicalgamesislesswelldevelopedempirically,buthasmuch
promiseforunderstandingphenomenalike“socialimage”,normenforcement,how
emotionsarecreatedbysurprises,andtherelationshipbetweenemotion,cognition,and
strategicbehavior.
FutureDirections
Therearealotofopenresearchquestionsinwhichcombiningcognitivescienceand
gametheorywouldbeuseful.Hereareafew:
31
1. Canthedistributionofleveltypesbederivedendogeneouslyfrommorebasic
principlesofcognitivedifficultyandperceivedbenefit,orperhapsfrom
evolutionaryconstraintonworkingmemoryandtheoryofmind(e.g.Stahl,1993).
2. CHmodelshavethepotentialtodescribedifferencesinskillorexperience.Skill
arisesineverydaydiscussionsaboutevensimplegameslikerock,paperand
scissors,ingameswithprivateinformationsuchaspoker,andgamesthattax
workingmemorysuchaschess.Areskilldifferencesgeneralordomain‐specific?
Canskillbetaught?Howdoesskilldevelopmentchangecognitionandneural
activity?
3. Thecomputationalapproachtostrategicthinkinginbehavioralgametheorycould
beusefulforunderstandingthesymptoms,etiologyandtreatmentofsome
psychiatricdisorders.Disorderscouldbeconceptualizedasfailurestocorrectly
anticipatewhatotherpeopledoandfeelinsocialinteractions,ortomakegood
choicesgivensensiblebeliefs.Forexample,inrepeatedtrustgamesKing‐Casaset
al.,(2008)foundthatborderlinepersonalitydisorder(BPD)didnothavetypical
activityininsulacortexinresponsetobeingmistrusted,andearnedlessmoney
becauseoftheinabilitytomaintainsteadyreciprocaltrustbehaviorally.Chiu
(2008)foundthatautismpatientshadlessactivityinaregionofanteriorcingulate
thattypicallyencodessignalsofvaluationduringone’sownstrategicchoices
(comparedtochoicesofothers).
4. Asmallemergingapproachinthestudyofliteraturefocusesonthenumberof
mentalstatesthatreaderscantrackandtheireffect(e.g.,Zunshine,2006).One
theoryisthatthreementalstatesareasociallyimportantreasonablenumber(e.g.,
32
lovetriangles)andarethereforenarrativelyengaging.Workoncognitiveschemas,
socialcategorization,computationallinguistics,andgametheorycouldthereforebe
ofinterestinthestudyofliterature.
5. Formalmodelsconnectingemotionswithbeliefs,actions,andpayoffscanilluminate
therelationshipsbetweenaffectivestatesandbehavior.Theutilityfunction
approachtomodelingemotionsmakesclearthatemotionsinfluencebehavioronly
whenthehedonicbenefitsofemotionalbehavioroutweighthecosts.Thisapproach,
whichconsidersevenemotion‐drivenbehaviorastheoutcomeofanoptimization
problem(perhapssculptedbyhumanevolutionratherthanconsciouscost‐benefit,
ofcourse).,promisestoopenupnewavenuesofresearchstudyingtherelationship
betweenemotionandstrategicchoices.
33
Tablecaptions
Table1:Payoffsinbettinggame,predictions(NashandCH),andresultsfromclassroom
demonstrationsin2006‐08.UpperleftistheuniqueNashequilibrium.
Table2:PayoffsfromHandTchoiceina“matchingpennies”game,predictions,anddata.
Table1:
predictions Data
L R Nash CH 2006+07+08 Average
T 30, 20 10, 18 1.00 .73 .81+.86+.78 .82
B 20, 20 20, 18 .00 .27 .19+.14+.22 .18
Nash 1.00 0
CH .89 .11
2006+07+08 .95+.95+.75 .05+.05+.25
average .88 .12
Table2:
predictions
H T Nash CH Levels
1-2
Levels
3-4
data
H 2,0 0,1 .50 .68 1 0 .72
T 0,1 1,0 .50 .32 0 1 .28
34
Nash .33 .67
CH .26 .74
data .33 .67
Figurecaptions
Figure1:Choicesin“2/3oftheaverage”game(Nagel,2005?)
Figure2:Predictedandobservedbehaviorinentrygames
Figure3:ThegameboardfromHeddenandZhang(2002)
Figure4:Anicongraphofvisualattentioninthreeroundsofbargaining(1,2and3)and
correspondingdistributionsofoffers.Eachcolumnrepresentsadifferent“type”ofperson‐
trialclassifiedbyvisualattention.
Figure5:Estimatedstrategicleveltypesforeachindividualintwosetsof11different
games(Chong,Camerer,Ho&Chong,2005).Estimatedtypesarecorrelatedintwosets
(r=.61)
Figure6:Numberschoseninweek1ofSwedishLUPIlottery(Napproximately350,000).
DottedlineindicatesmixedNashequilibrium.Solidlineindicatestochasticcognitive
hierarchy(CH)modelwithtwofreeparameters.Best‐fittingaveragestepsofthinkingisτ
=1.80andλ=.0043(logitresponse).
35
Figure7:Brainregionsmoreactiveinlevel2reasonerscomparedtolevel1reasoners
(classifiedbychoices),differentiallyinplayinghumancomparedtocomputeropponents
(fromCoricelliandNagel,2009,FigureS2a).
Figure8:Asimpletrustgame(DufwenbergandGneezy,2008)
36
Figure1
Beauty contest results (Expansion, Financial Times, Spektrum)
0.000.050.100.150.20
numbers
rela
tiv
e
fre
qu
en
cie
s
22 50 10033
average 23.07
0
37
Figure2
How entry varies with demand (D), experimental data and thinking model
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Demand (as % of # of players)
% e
ntry
entry=demandexperimental data!=1.25
Figure3
38
Figure4
39
Figure5
40
Figure6
41
Figure7
Figure8
42
References
Addis,D.R.,McIntosh,A.R.,Moscovitch,M.,Crawley,A.P.&McAndrews,M.P.(2004).
Characterizingspatialandtemporalfeaturesofautobiographicalmemoryretrieval
networks:apartialleastsquaresapproach.NeuroImage,23,1460‐1471.
Addis,D.R.,Wong,A.T.&Schacter,D.L.(2007),Rememberingthepastandimaginingthe
future:commonanddistinctneuralsubstratesduringeventconstructionand
elaboration,Neuropsychologia,45,1363‐1377.
Akerlof,G.A.(1970).“TheMarketfor"Lemons":QualityUncertaintyandtheMarket
Mechanism.TheQuarterlyJournalofEconomics,84,488‐500.
Amodio,D.M.&Frith,C.D.(2006),Meetingofminds:themedialfrontalcortexandsocial
cognition,NatureReviewsNeuroscience,7,268‐277.
Aumann,R.andBrandenburger,A.(1995),“EpistemicconditionsforNashEquilibrium,”
Econometrica,63,1161‐1180.
Baker,S.C.,Rogers,R.D.,Owen,A.M.,Frith,C.D.,Dolan,R.J.,Frackowiak,R.S.&Robbins,T.W.
(1996),Neuralsystemsengagedbyplanning:aPETstudyoftheTowerofLondontask,
Neuropsychologia,34,515‐526.
Baumeister,R.,Stillwell,A.&Heatherton,T.(1994),“Guilt:AnInterpersonalApproach”,
PsychologicalBulletin,115,243‐67.
Battigalli,P.,andDufwenberg,M.(2007),“GuiltinGames”,AmericanEconomicReview,
43
97,170‐176.
Battigalli,P.,andDufwenberg,M.(2009),“DynamicPsychologicalGames”,Journalof
EconomicTheory,144,1‐35.
Bazerman,M.H.,&Samuelson,W.F.(1983).IWontheAuctionButDon'tWantthePrize.
JournalofConflictResolution,27,618‐634.
Benabou,R.andTirole,J.(2006),“IncentivesandProsocialBehavior”,AmericanEconomic
Review,96,1652‐1678.
Bell,D.E.(1982),RegretinDecisionMakingunderUncertainty,OperationsResearch
30,961‐981.
BellD.E.(1985),DisappointmentinDecisionMakingunderUncertainty,Operations
Research,33,1‐27.
Bernheim,B.D.(1994),“ATheoryofConformity”,TheJournalofPoliticalEconomy,102,
841‐877.
Bernheim,B.D.&Andreoni,J.(2009),“SocialImageandthe50‐50Norm:Theoryand
ExperimentalEvidence.”Econometrica,77,1607–1636.
Bhatt,M.&Camerer,C.F.(2005),Self‐referentialthinkingandequilibriumasstatesofmind
ingames:fMRIevidence,GamesandEconomicBehavior,52,424.
Bhatt,M.&Camerer,C.F.Inpress.“Thecognitiveneuroscienceofstrategicthinking”,inJ.
CacciopoandJ.Decety(Eds.),HandbookofSocialNeuroscience.Oxford,England:
OxfordUniversityPress.
44
Bhui,R&Camerer,C.F.,(2011),“Thepsychometricsofmeasuringstabilityofcognitive
hierarchytypesinexperimentalgames”,workingpaper.
Binmore,K.,Shaked,A.,&Sutton,J.(1985).TestingNoncooperativeBargainingTheory:A
PreliminaryStudy.TheAmericanEconomicReview,75,1178‐1180.
Burchardi,K.B.,andPenczynski,S.P.,(2010),“OutOfYourMind:ElicitingIndividual
ReasoninginOneShotGames”,workingpaper.
Brandenburger,A.(1992),“KnowledgeandEquilibriuminGames”,JournalofEconomic
Perspectives,6(4),83–101.
Brocas,I.,Carrillo,J.D.,Wang,S.&Camerer,C.F.(2009),Measuringattentionandstrategic
behavioringameswithprivateinformation,Mimeoedn,Pasadena.
Brown,A.L.,Camerer,C.F.,&Lovallo,D.(2011).“ToRevieworNotReview?Limited
StrategicThinkingattheMovieBoxOffice”,workingpaper.
Camerer,C.,(1990).Behavioralgametheory.InR.Hogarth(Ed.),InsightsinDecision
Making:ATributetoHillelJ.Einhorn.Chicago:UniversityofChicagoPress.
Camerer,C.F.(2003),Behavioralgametheory,Princeton:PrincetonUniversityPress.
Camerer,C.&Ho,T.H.(1999),Experience‐weightedattractionlearninginnormalform
games,Econometrica,67,827‐874.
Camerer,C.F.,Ho,T.‐H.,&Chong,J.‐K.(2002).SophisticatedExperience‐Weighted
AttractionLearningandStrategicTeachinginRepeatedGames.JournalofEconomic
Theory,104,137‐188.
45
Camerer,C.F.,Ho,T.H.&Chong,J.K.(2004),Acognitivehierarchymodelofgames,
QuarterlyJournalofEconomics,119,861‐898.
Camerer,C.F.,Johnson,E.,Rymon,T.&Sen,S.(1993),“CognitionandFraminginSequential
BargainingforGainsandLosses.”InFrontiersofGameTheory,eds.K.G.Binmore,A.P.
Kirman&P.Tani,MITPress,Cambridge,27‐47.
Capen,E.C.,R.V.Clapp,andW.M.Campbell,(1971).CompetitiveBiddinginHigh‐Risk
Situations.JournalofPetroleumTechnology,23,641‐653.
Caplin,A.andLeahy,J.(2004).TheSupplyofInformationbyaConcernedExpert.The
EconomicJournal,114,487‐505
Carter,C.S.(1998).AnteriorCingulateCortex,ErrorDetection,andtheOnlineMonitoringof
Performance,Science,280,747‐749.
Cavanna,A.E.&Trimble,M.R.(2006),Theprecuneus:areviewofitsfunctionalanatomy
andbehaviouralcorrelates,Brain:Ajournalofneurology,129(3),564‐583.
Chang,L.,Smith,A.,Dufwenberg,M.,andSanfey,A.(2011).Triangulatingtheneural,
psychological,andeconomicbasesofguiltaversion.Neuron,inpress.
Chapman,H.A.,Kim,D.A.,Susskind,J.M.,Anderson,A.K.(2009).Inbadtaste:Evidencefor
theoraloriginsofmoraldisgust.Science,323,1222‐1226.
Charness,G.&Rabin,M.(2002).Understandingsocialpreferenceswithsimpletests.
QuarterlyJournalofEconomics,117,817‐869.
46
Charness,G.,andDufwenberg,M.(2006).PromisesandPartnership.Econometrica
74,1579‐1601.
Charness,G.,andDufwenberg,M.(2011).Participation.AmericanEconomicReview,
forthcoming.
Cheng,P.W.,&Holyoak,K.J.(1985).Pragmaticreasoningschemas.CognitivePsychology,
17,391‐416.
Chiu,P.H.,Kayali,M.A.,Kishida,K.T.,Tomlin,D.,Klinger,L.G.,Klinger,M.R.&Montague,P.R.
(2008).Selfresponsesalongcingulatecortexrevealquantitativeneuralphenotypefor
high‐functioningautism.Neuron,57,463‐473.
Chong,J.,Camerer,C.F.&Ho,T.H.(2006).Alearning‐basedmodelofrepeatedgameswith
incompleteinformation.GamesandEconomicBehavior,55,340‐371.
Coricelli,G.&Nagel,R.(2009).Neuralcorrelatesofdepthofstrategicreasoninginmedial
prefrontalcortex.PNAS,106,9163‐8.
Costa‐Gomes,M.,Crawford,V.P.&Broseta,B.(2001).Cognitionandbehaviorinnormal‐
formgames:Anexperimentalstudy.Econometrica,69,1193‐1235.
Costa‐Gomes,M.A.&Crawford,V.P.(2006).CognitionandBehaviorinTwo‐Person
GuessingGames:AnExperimentalStudy.AmericanEconomicReview,96,1737‐1768.
Chiu,P.H.,Kayali,M.A.,Kishida,K.T.,Tomlin,D.,Klinger,L.G.,Klinger,M.R.,etal.(2008).
SelfResponsesalongCingulateCortexRevealQuantitativeNeuralPhenotypeforHigh‐
FunctioningAutism.Neuron,57,463‐473.
47
Craig,A.D.(2002).Howdoyoufeel?Interoception:thesenseofthephysiologicalcondition
ofthebody.NatureReviewsNeuroscience,3,655‐666.
Crawford,V.P,Costa‐Gomes,M.A.,andIriberri,N.,(2010),“StrategicThinking,”working
paper.
Critchley,H.D.(2005).Neuralmechanismsofautonomic,affective,andcognitive
integration.TheJournalofcomparativeneurology,493,154‐166.
Crockett,M.J.,Clark,L.,Tabibnia,G.,Lieberman,M.D.,Robbins,T.W.(2008).Serotonin
modulatesbehavioralreactionstounfairness.Science,320,1739.
Culham,J.C.,Brandt,S.A.,Cavanagh,P.,Kanwisher,N.G.,Dale,A.M.&Tootell,R.B.(1998).
CorticalfMRIactivationproducedbyattentivetrackingofmovingtargets.Journalof
neurophysiology,80,2657‐2670.
D'Argembeau,A.,Ruby,P.,Collette,F.,Degueldre,C.,Balteau,E.,Luxen,A.,Maquet,P.&
Salmon,E.(2007)."Distinctregionsofthemedialprefrontalcortexareassociatedwith
self‐referentialprocessingandperspectivetaking.Journalofcognitiveneuroscience,19,
935‐944.
Devetag,G.,&Warglien,M.(2003).Gamesandphonenumbers:Doshort‐termmemory
boundsaffectstrategicbehavior?JournalofEconomicPsychology,24,189‐202.
Duffy,J.,&Hopkins,E.,(2005).Learning,information,andsortinginmarketentrygames:
theoryandevidence.GamesandEconomicBehavior,51,31‐62.
48
Dufwenberg,M.andGneezy,U.(2000).MeasuringBeliefsinanExperimentalLostWallet
Game.GamesandEconomicBehavior,30,163‐182
Dufwenberg,M.&Kirchsteiger,G.(2004).ATheoryofSequentialReciprocity.Gamesand
EconomicBehavior.47,268‐298.
Duffy,J.,andHopkins,E.(2005).Learning,information,andsortinginmarketentrygames:
theoryandevidence.GamesandEconomicBehavior,51,31‐62.
Eisenberger,N.I.,Lieberman,M.D.,&Williams,K.D.(2003).DoesRejectionHurt?AnfMRI
StudyofSocialExclusion.Science,302,290‐292.
Ellingsen,T.,Johannesson,M.,Tjotta,S.GauteTorsvik,G.(2010).Testingguiltaversion.
GamesandEconomicBehavior,68,95‐107.
Farrer,C.&Frith,C.D.(2002).Experiencingoneselfvs.anotherpersonasbeingthecauseof
anaction:theneuralcorrelatesoftheexperienceofagency.NeuroImage,15,596‐603.
Fehr,E.&Camerer,C.F.(2007).Socialneuroeconomics:theneuralcircuitryofsocial
preferences.TrendsinCognitiveSciences,11,419‐27.
Fehr,E.&Schmidt,K.M.(1999).ATheoryofFairness,Competition,andCooperation.
QuarterlyJournalofEconomics,114,817‐868.
Fiddick,L.,Cosmides,L,&Tooby,J.(2000).Nointerpretationwithoutrepresentation:the
roleofdomain‐specificrepresentationsandinferencesintheWasonselectiontask.
Cognition,77,1‐79.
49
Fletcher,P.C.,Frith,C.D.,Baker,S.C.,Shallice,T.,Frackowiak,R.S.&Dolan,R.J.(1995).The
mind'seye‐‐precuneusactivationinmemory‐relatedimagery.NeuroImage,2,,195‐
200.
Fudenberg,D.&Levine,D.(1998).TheoryofLearninginGames.MITPress,Cambridge,MA.
Gallagher,H.L.,Jack,A.I.,Poepstorff,A.&Frith,C.D.(2002).ImagingtheIntentionalStance
inaCompetitiveGame.NeuroImage,16,814‐821.
Geanakoplos,D.,Pearce,D.,andStacchetti,E.(1989).Psychologicalgamesandsequential
rationality.GamesandEconomicBehavior,1,60‐79.
Gill,D.andStone,R.(2010).Fairnessanddesertintournaments.GamesandEconomic
Behavior,69,346‐364.
Gillen,B.,(2009).IdentificationandEstimationofLevel‐kAuctions.Workingpaper.
Glimcher,P.W.,ColinCamerer,ErnstFehr,RussellPoldrack(Eds.)(2008).Neuroeconomics:
decisionmakingandthebrain.Academic,London.
Georganas,S.,Healy,P.J.,andWeber,R.,(2010).Onthepersistenceofstrategic
sophistication.Workingpaper.
Goldfarb,A.,andXiao,M.,(2011).WhoThinksabouttheCompetition:ManagerialAbility
andStrategicEntryinUSLocalTelephoneMarkets.Forthcoming,AmericanEconomic
Review.
Goldfarb,A.,andYang,B.(2009).AreAllManagersCreatedEqual?JournalofMarketing
Research,46,612‐622.
50
Hampton,A.,Bossaerts,P.&O’Doherty,J.(2008).Neuralcorrelatesofmentalizing‐related
computationsduringstrategicinteractionsinhumans.PNAS,105,6741‐6746.
Harsanyi,J.C.(1967).GameswithIncompleteInformationPlayedby‘Bayesian’Players,I‐
III.PartI.TheBasicModel.ManagementScience,14,159‐182
Hayden,B.,Pearson,M.&Platt,M.L.(2009).Fictivelearningsignalsinanteriorcingulate
cortex.Science,324,948‐950.
Hsu,M.,Anen,C.,&Quartz,S.R.(2008).TheRightandtheGood:DistributiveJusticeand
NeuralEncodingofEquityandEfficiency.Science,320,1092‐1095.
Izuma,K.,Saito,D.N.&Sadato,N.(2010).Processingoftheincentiveforsocialapprovalin
theventralstriatumduringcharitabledonation.Journalofcognitiveneuroscience,22,
621‐631.
Izuma,K.,Saito,D.N.&Sadato,N.(2008).Processingofsocialandmonetaryrewardsinthe
humanstriatum.Neuron,58,284‐294.
Johnson,E.J.,Camerer,C.,Sen,S.,&Rymon,T.(2002).DetectingFailuresofBackward
Induction:MonitoringInformationSearchinSequentialBargaining.Journalof
EconomicTheory,104,16‐47.
Kahneman,Daniel,(1988)."ExperimentalEconomics:APsychologicalPerspective,"inR.
Tietz,W.Albers,andR.Selten,eds.,Boundedrationalbehaviorinexperimentalgames
andmarkets(pp.11‐18).NewYork:Springer‐Verlag.
51
Keynes,J.M.,(1936).TheGeneralTheoryofEmployment,Interest,andMoney.London:
Macmillan.
Keysers,C.&Gazzola,V.(2007).Integratingsimulationandtheoryofmind:fromselfto
socialcognition.Trendsincognitivesciences,11,194‐196.
King‐Casas,B.,Sharp,C.,Lomax‐Bream,L.,Lohrenz,T.,Fonagy,P.&Montague,P.R.(2008).
Theruptureandrepairofcooperationinborderlinepersonalitydisorder.Science,321,
806‐810.
Knoch,D.,Pascual‐Leone,A.,Meyer,K.,Treyer,V.&Fehr,E.(2006).DiminishingReciprocal
FairnessbyDisruptingtheRightPrefrontalCortex.Science,314,829–832.
Kross,E.,Berman,M.G.,Mischel,W.,Smith,E.E.,&Wager,T.D.(2011).Socialrejection
sharessomatosensoryrepresentationswithphysicalpain.ProceedingsoftheNational
AcademyofSciences,108(15),6270‐6275.
Kuo,W.J.,Sjostrom,T.,Chen,Y.P.,Wang,Y.H.&Huang,C.Y.(2009).Intuitionand
deliberation:twosystemsforstrategizinginthebrain.Science324,519‐522.
Le,T.H.,Pardo,J.V.&Hu,X.(1998).4T‐fMRIstudyofnonspatialshiftingofselective
attention:cerebellarandparietalcontributions.Journalofneurophysiology,79,1535‐
1548.
Li,C.S.,Huang,C.,Constable,R.T.&Sinha,R.(2006).Imagingresponseinhibitioninastop‐
signaltask:neuralcorrelatesindependentofsignalmonitoringandpost‐response
processing.TheJournalofneuroscience:theofficialjournaloftheSocietyfor
Neuroscience,26,186‐192.
52
Lohrenz,T.,McCabe,K.,Camerer,C.F.&Montague,P.R.(2007).Neuralsignatureoffictive
learningsignalsinasequentialinvestmenttask.PNAS,104,9493‐9498.
Loomes,G.andSugden,R.(1982).RegretTheory:AnAlternativeTheoryofRationalChoice
UnderUncertainty.TheEconomicJournal,92,805‐824
Loomes,G.andSugden,R.(1986).DisappointmentandDynamicConsistencyinChoice
underUncertainty.TheReviewofEconomicStudies,53,271‐282.
LuceR.D.(1959).Individualchoicebehavior.Oxford,England:JohnWiley.
Luhmann,C.C.,Chun,M.M.,Yi,D.J.,Lee,D.&Wang,X.J.(2008).NeuralDissociationofDelay
andUncertaintyinIntertemporalChoice.TheJournalofneuroscience,28,14459‐
14466.
Lundstrom,B.,Petersson,K.M.,Andersson,J.,Johansson,M.,Fransson,P.&Ingvar,M.
(2003).Isolatingtheretrievalofimaginedpicturesduringepisodicmemory:activation
oftheleftprecuneusandleftprefrontalcortex.NeuroImage,20,1934.
McCabe,K.,Houser,D.,Ryan,L.,Smith,V.&Trouard,T.(2001).AFunctionalImagingStudy
ofCooperationinTwo‐PersonReciprocalExchange.ProceedingsoftheNational
AcademyofSciencesoftheUnitedStatesofAmerica,98,11832‐11835.
Mobbs,D.,Yu,R.,Meyer,M.,Passamonti,L.,Seymour,B.,Calder,A.J.,Schweizer,S.,Frith,
C.D.,Dalgleish,T.(2009).Akeyroleforsimilarityinvicariousreward.Science,324,p,900,
Nagahama,Y.,Okada,T.,Katsumi,Y.,Hayashi,T.,Yamauchi,H.,Sawamoto,N.,Toma,K.,
Nakamura,K.,Hanakawa,T.,Konishi,J.,Fukuyama,H.&Shibasaki,H.(1999).Transient
53
neuralactivityinthemedialsuperiorfrontalgyrusandprecuneustimelockedwith
attentionshiftbetweenobjectfeatures.NeuroImage,10,193‐199.
Nagel,R.(1995).UnravelinginGuessingGames:AnExperimentalStudy.TheAmerican
EconomicReview,85,1313‐1326.
Nash,J.F.(1950).EquilibriumPointsinn‐PersonGames.ProceedingsoftheNational
AcademyofSciencesoftheUnitedStatesofAmerica,36,48‐49.
Neelin,J.,Sonnenschein,H.,&Spiegel,M.(1988).AFurtherTestofNoncooperative
BargainingTheory.AmericanEconomicReview,78,824‐836.
Ochsner,K.,Hughes,B.,Robertson,E.Gabrieli,J.,Cooper,J.&Gabrieli,J.(2009).Neural
SystemsSupportingtheControlofAffectiveandCognitiveConflicts.Journalof
CognitiveNeuroscience,21,1841‐1854.
Östling,R.,Wang,J.T.‐y.,Chou,E.,&Camerer,C.F.,(2011).TestingGameTheoryinthe
Field:SwedishLUPILotteryGames.Forthcoming,AmericanEconomicJournal:
Microeconomics.
Preuschoff,K.,Quartz,S.R.&Bossaerts,P.(2008).Humaninsulaactivationreflectsrisk
predictionerrorsaswellasrisk.JournalofNeuroscience,28,2745‐2752.
Rabin,M.(1993).IncorporatingFairnessintoGameTheoryandEconomics.American
EconomicReview,83,1281‐1302.
54
Raichle,M.E.,MacLeod,A.M.,Snyder,A.Z.,Powers,W.J.,Gusnard,D.A.&Shulman,G.L.
(2001).Adefaultmodeofbrainfunction.ProceedingsoftheNationalAcademyof
SciencesoftheUnitedStatesofAmerica,98,676‐682.
ReubenE.,SapienzaP.,ZingalesL.(2009).Ismistrustself‐fulfilling?EconomicsLetters,104,
89‐91.
Ridderinkhof,K.R.,Ullsperger,M.,Crone,E.A.&Nieuwenhuis,S.(2004),Theroleofthe
medialfrontalcortexincognitivecontrol.Science,306,443‐447.
Rotemberg,J.J.(2008).MinimallyAcceptableAltruismandtheUltimatumGame.Journalof
EconomicBehavior&Organization,66,nos.3‐4,457‐476
Ruby,P.&Decety,J.(2001).Effectofsubjectiveperspectivetakingduringsimulationof
action:aPETinvestigationofagency.Natureneuroscience,4,546‐550.
Sanfey,A.G.,Rilling,J.K.,Aronson,J.A.,Nystrom,L.E.&Cohen,J.D.(2003).TheNeuralBasis
ofEconomicDecision‐MakingintheUltimatumGame.Science,300,1755‐1758.
Saxe,R.&Powell,L.J.(2006).It'sthethoughtthatcounts:specificbrainregionsforone
componentoftheoryofmind.Psychologicalscience,17,692‐699.
Sbriglia,P.(2008).Revealingthedepthofreasoninginp‐beautycontestgames.
ExperimentalEconomics,11,107‐121.
Shallice,T.,Fletcher,P.C.,Frith,C.D.,Grasby,P.,Frackowiak,R.S.&Dolan,R.J.(1994).Brain
regionsassociatedwithacquisitionandretrievalofverbalepisodicmemory.Nature,
368,633‐635.
55
Simon,O.,Mangin,J.,Cohen,L.,LeBihan,D.&Dehaene,S.(2002).TopographicalLayoutof
Hand,Eye,Calculation,andLanguage‐RelatedAreasintheHumanParietalLobe.
Neuron,33,475‐487.
Smith,A.(2009).Belief‐DependentAngerinGames.Workingpaper.
Stahl,D.O.(1993).EvolutionofSmartnPlayers.GamesandEconomicBehavior,5,604‐617.
Stahl,D.,O.,&Wilson,P.,W.(1995).OnPlayers'ModelsofOtherPlayers:Theoryand
ExperimentalEvidence.GamesandEconomicBehavior,10,218‐254.
Tadelis,S.(2011).ThePowerofShameandtheRationalityofTrust.Workingpaper.
Takahashi,H.,Kato,M.,Matsuura,M.,Mobbs,D.,Suhara,T.&Okubo,Y.(2009).Whenyour
gainismypainandyourpainismygain:neuralcorrelatesofenvyandschadenfreude.
Science,323,937‐939.
Tomlin,D.,Kayali,M.A.,King‐Casas,B.,Anen,C.,Camerer,C.F.,Quartz,S.R.&Montague,P.R.
(2006).Agent‐specificresponsesinthecingulatecortexduringeconomicexchanges.
Science,312,1047‐1050.
Tricomi,E.,Rangel,A.,Camerer,C.F.&O'Doherty,J.P.(2010).Neuralevidencefor
inequality‐aversesocialpreferences.Nature,463,1089‐1091.
Unterrainer,J.M.,Rahm,B.,Kaller,C.P.,Ruff,C.C.,Spreer,J.,Krause,B.J.,Schwarzwald,R.,
Hautzel,H.&Halsband,U.(2004).Whenplanningfails:individualdifferencesand
error‐relatedbrainactivityinproblemsolving.Cerebralcortex(NewYork,N.Y.:1991),
14,1390‐1397.
56
VanDamme,E.(1999),"Gametheory:Thenextstage,"inL.A.Gérard‐Varet,A.P.Kirman
andM.Ruggiero(eds.),EconomicsbeyondtheMillennium(pp.184‐214).Oxford
UniversityPress.
Vogeley,K.,Bussfeld,P.,Newen,A.,Herrmann,S.,Happe,F.,Falkai,P.,Maier,W.,Shah,N.J.,
Fink,G.R.&Zilles,K.(2001).Mindreading:neuralmechanismsoftheoryofmindand
self‐perspective.NeuroImage,14,170‐181.
Vogeley,K.,May,M.,Ritzl,A.,Falkai,P.,Zilles,K.&Fink,G.R.(2004).Neuralcorrelatesof
first‐personperspectiveasoneconstituentofhumanself‐consciousness.Journalof
cognitiveneuroscience,16,817‐827.
vonNeumann,J.,&Morgenstern,O.(1944).Theoryofgamesandeconomicbehavior.
Princeton:PrincetonUniv.Press.
Wang,J.T.,Spezio,M.&Camerer,C.F.(2010).Pinocchio'sPupil:UsingEyetrackingand
PupilDilationtoUnderstandTruth‐tellingandDeceptioninGames.AmericanEconomic
Review,100,984‐1007.
Weber,R.A.(2003).'Learning'withnofeedbackinacompetitiveguessinggame.Games
andEconomicBehavior,44,134‐144
Yoshida,W.,Seymour,B.,Friston,K.J.&Dolan,R.(2009).Neuralmechanismofbelief
inferenceduringcooperativegames.JournalofNeuroscience,30,10744‐10751.
Yoshida,W.,Dolan,R.J.&Friston,K.J.(2008),"Gametheoryofmind",PLoScomputational
biology,4(12),e1000254.Formixedprospects,however,
57
Zunshine,Lisa.WhyWeReadFiction:TheoryofMindandtheNovel.Columbus:TheOhio
StateUniversityPress,2006.
iAnotherimportantcomponentofbehavioralgametheoryislearningfromrepeatedplay(perhapsusingreinforcementrulesaswellasmodel‐based“fictivelearning”(Camerer&Ho,1999).Learningmodelsarewidelystudiedbutliebeyondthescopeofthischapter(seee.g.Fudenberg&Levine,1998;Camerer,2003,chapter6).
iiOthermechanismsthatcouldproduceequilibrationincludelearningfromobservation,introspection,calculation(suchasfirmshiringconsultantstoadviseonhowtobidonauctions),imitationofattention‐gettingorsuccessfulstrategiesorpeople,oraprocessofpre‐playtalkingaboutfuturechoices.Thelearningliteratureiswelldeveloped(e.g.,Camerer,2003,chapter6)butthestudyofimitationandpre‐playtalkingcouldcertainlyusemorecollaborationbetweengametheoristsandpsychologists.
iiiCommonknowledgerequires,fortwoplayers,thatAknowsthatBknowsthatAknows…adinfinitum.
ivAmoregeneralviewisthatlevel0’schooseintuitivelyor“heuristically”(perhapsbasedonvisuallysalientstrategiesorpayoffs,or“luckynumbers”),butthattopichasnotbeenexploredverymuch.
vRestrictingcommunicationisnotmeanttoberealisticandcertainlyisnot.Insteadcommunicationisrestrictedbecausechoosingwhattosayisitselfa“strategy”choicewhichcomplicatesanalysisofthegame—itopensaPandora’sboxofpossibleeffectsthatlieoutsidethescopeofstandardgametheory.However,gametheoristsarewellawareofthepossiblepowerfuleffectsofcommunicationandhavebeguntostudyitinsimpleways.InherthesisNagel(1995)reportssomesubjectdebriefingwhichareillustrativeofCHthinking,andSbriglia(2008)reportssomeprotocolstoo.BurchardiandPenczynski(2010)alsousedchatmessagingandteamchoicetostudycommunicationandreportevidencelargelyconsistentwithCHreasoning.
viThisanalysisassumesτ=1.5butthegeneralpointholdsmorewidely.
viiNotethatthisisacloserelativeofa“thresholdpublicgoods”game.Inthatgame,apublicgoodiscreated,whichbenefitseveryone,ifTpeoplecontribute,butifevenoneperson
58
doesnotethepublicgoodisnotproduced.Inthatcase,everyonewouldliketobeintheN‐Tgroupofpeoplewhobenefitwithoutpaying.
viiip(X)isa“commonprior.”Anexampleisagameofcards,inwhicheveryoneknowsthecontentsofthecarddeck,donotknowwhatface‐downcardsotherplayersareholding,butalsoknowthattheotherplayersdoknowtheirownface‐downcards.
ixThereareseveralsubtlevariants.IntheoriginalMouselab,boxesopenandcloseautomaticallywhenthemouseentersandexits.Costa‐Gomesetal(2001)wantedmoredeliberateattentionsotheycreatedaversioninwhichaclickisrequiredtoopenabox.Brocasetal(2010)createdaversionthatrequiresthemousebuttontobehelddowntoviewboxcontents(ifthebuttonpressishaltedtheinformationdisappears).
xTherostralACC,labeledrACC,ismoreactiveinlevel1thaninlevel2playersinthehuman‐computercontrast.
xiDLPFCisalsoinvolvedincognitiveregulationofemotions(e.g.,Ochsneretal.,2009)
xiiWhatwehaveinmindhereissimilartoHolyoakandCheng(1985),Fiddick,CosmidesandTooby(2000)argumentsaboutthedifferencebetweenabstractlogicperformanceandcontextualizedperformance.Forexample,gamesthatresemblehidingfoodandguardinghiddenlocationsmightmaproughlyontosomethinglikepoker,whereasalotofgamesconstructedforchallengeandentertainment,suchaschess,donothaveclearcounterpartsinancestraladaptiveenvironments.
xiiiOthermodelsofbelief‐dependentutilitycanbeplacedinthegeneralframeworkofBattigalliandDufwenberg(2009).Forexample,CaplinandLeahy(2004)modeldoctor‐patientinteractionswhereuncertaintymaycausepatientanxiety.Thedoctorisconcernedaboutthepatient’swellbeingandmustdecidewhetherornottoprovide(potentially)anxiety‐causingdiagnosticinformation.Bernheim(1994)proposesamodelofconformitywhereplayerscareaboutthebeliefstheircoplayershaveregardingtheirpreferences.Themodelcanproducefadsandadherencetosocialnorms.RelatedworkbyBenabouandTirole(2006)modelsplayerswhoarealtruistic,andalsocareaboutother’sinferencesabouthowaltruistictheyare.GillandStone(2010)modelplayerswhocareaboutwhattheyfeeltheydeserveintwo‐playertournaments.Theplayers’perceivedentitlementsdependupontheirowneffortlevelandtheeffortsofothers.
xivA(singleperson)decisionprobleminvolvinganyoftheseemotionsmaybemodeledasapsychologicalgamewithoneplayerandmovesbynature.
Related Documents
