determinantstko

1Threekindsof'as-if'claimsAkiLehtinenPaperforthcomingintheJournalofEconomicMethodologyAbstractAs-iflocutionsareused(a)inordertoindicatethataninaccurateorunrealisticassumptionisbeingmadebecausesomeinaccuracyorunrealisticnessisnegligible.
Thiskindofclaimhastwosub-cases.
(a1)Theas-iflocutionisusedtoindicatethattheas-ifclaiminitselfisinaccurateandthatitsinaccuracydoesnotmatterforthepurposesoftheinvestigation.
(a2)Itisusedtoindicatethatclaimsaremadewithoutregardtothecausalfactorsthatareassumedtoexistbutaredeemedtobeunimportant.
As-iflocutionsmayalso(b)formulateanaccuratebehaviouralassumptionbyascribingintentionsorcognitionstoanentityinanunrealisticmanner,or(c)indicatethatthemodellerisnotcommittedtoanyparticularmentalassumptions.
Thevariouskindsofclaimsmayberecognisedbyidentifyingtheirunderlying'attributions'.
(a2),(b)and(c)maybeusedinformulatinganaccurateclaim.
1IntroductionManyeconomistsandeconomicmethodologistsqualifyassumptionswithan'as-if'locution.
Consider,forexample,thefollowingcases.
(a1)'AccordingtoFriedman,businessfirmsbehaveasiftheywerefullyinformed,rationalcalculatorsaimingattheachievementofprofitmaximisation,eventhoughclearlytheseassumptionsdonotapplytoactualfirms'(BoylanandO'Gorman1995,pp.
108-109).
'Individualsarenotoptimizers,theyareadaptiveandonlybehave"asif"theyoptimize'(Kirman2006,p.
271).
(a2)Peoplebehaveasiftheywereonlymotivatedbyself-interest.
1(b)Votersactasiftheyhaveobtainedperturbedsignalsconcerningthepreferenceprofileandasiftheyhavecomputedprobabilitiesfromthesesignalsusingstatisticalreasoning(Lehtinen2007,2008).
'Peoplemaybeinducedbysocialcodesofbehaviourtoactasiftheyhavedifferentpreferencesfromwhattheyreallyhave'(Sen1973,p.
258).
(c)Evechoosesasthoughsheweremaximizingtheexpectedvalueofautilityfunction(Binmore1998,pp.
360-1).
Whydotheyusesuchclaims,whatdotheymean,andhowdoweknowwhetherornottheyarejustifiedUsinganas-iflocutiontendstoimplythatoneismakinganunrealisticassumption.
SuchanimpressionstemsinpartfromthefactthatMiltonFriedman's(1953)methodologyas1ThemethodologicalargumentthatcouldjustifythisbehaviouralclaimisattributabletoJohnStuartMill(2000[1844],p.
98)butheneverusedthe'as-if'locutiontoexpressit.
2awholehasbecomeknownashis'as-ifmethodology',andhehasbecomeknownasachampionofunrealisticassumptions(seee.
g.
,RgebergandNordberg2005).
Theaiminthepaperistoclarifytheliteratureonunrealisticandfictionalassumptions.
Forsuchapurpose,letusdistinguishbetweenrealisticandaccuratebehaviouralclaimsorassumptions.
Anassumptionisrealisticifitisbasedonthecorrectcausesofthebehaviour,anditisaccurateifitcorrectlydescribesthebehaviour.
Iarguethatas-ifclaimsarefrequentlymadeinordertoformulateanaccurateassumption(cf.
LehtinenandKuorikoski2007).
Morespecifically,theas-iflocutionisusedinmakingthreedifferentkindsofclaim,andonlyonesub-caseofaclaim(a1)mayunequivocallybetakentoexpresstheideathataninaccurateassumptionisbeingmade.
Theothercasesarebetterunderstoodaswaysofexpressingassumptionsormakingbehaviouralclaimsinapragmaticallyorsemanticallycorrectway.
As-ifclaimsinthemselvesareneverrealisticintheaforementionedsensebecausetheyneverexplainwhytheentityofinterestbehavesinthewayitdoes.
Nevertheless,Iwillarguethatsomeas-ifclaimsmayhelpinformulatingaperfectlyrealisticbehaviouralassumption.
Letususetheterm'as-iflocution'todenotethelinguisticexpression'asif'.
As-iflocutionsareusedinorderto(a)indicatethataninaccurateorunrealisticassumptionisbeingmadebecausesomeinaccuracyorunrealisticnessisnegligible.
Thiskindofclaimhastwosub-cases.
(a1)First,theas-iflocutionisusedtoindicatethattheas-ifclaiminitselfisinaccurateandthatitsinaccuracydoesnotmatterforthepurposesoftheinvestigation.
(a2)Second,itisusedtoindicatethatclaimsaremadewithoutregardtothecausalfactorsthatareassumedtoexistbutaredeemedtobeunimportant.
Heretheas-ifclaimitselfisformulatedwiththelocution,anditisclaimedtobetrueenough.
As-iflocutionsalsoareusedto(b)formulateanaccurateorarealisticclaimconcerningthebehaviourofsomethingbyascribingintentionsorcognitionstoitinanunrealisticmanner,and(c)toindicatethatthemodellerisnotcommittedtoanyparticularmentalassumptions.
Iwillusetheshorthand'negligibility-indicating','falsementalascription'and'non-committal-indicating'forthefirst,secondandthirdkindsofas-ifclaim,respectively.
Sub-casesundera1willbecalled'negligiblyinaccurate'or'inaccuracy-indicating'as-ifclaims,andsub-casesundera2'negligibility-indicatingaccurate'as-ifclaims.
Eachoftheseusesoftheas-iflocutionisbasedondifferentfalsity-relatedreasonwhyitisused.
Musgrave(1981,p.
385)complainedsomethirtyyearsagothatthe'logicofasifstatements(whatfollowsfromthemandwhatdoesnotfollowfromthem)isterriblyunclear'.
Someeconomicmethodologists(Lagueux1994,Mki1992,1998,2000,2003,2004,2009)discusstheissue,buttheirfocusislimitedeithertofindingoutwhethercertainkindsofas-ifclaimsarecompatiblewithparticularphilosophicaldoctrines,ortointerpretingthetwoprotagonistsintheas-ifdiscussion,MiltonFriedmanandHans3Vaihinger.
2Asaresult,thelogicofas-ifclaimsineconomicpracticehasnotbeensetoutinacomprehensivemanner.
Iendeavourtodothatherebyprovidingawayofidentifyingthedifferentas-ifclaimsandtheassociatedtruthclaimstheyaremeanttoqualifyorspecify.
Mymaingoalinthepresentaccountisthustosortoutthedifferenttruthclaimsthatarebeingmade,andtoshowhowtheyaretobeevaluated.
As-ifclaimsandlocutionsarenevernecessaryfordescribingtheassumptionsofamodelmathematicallyorforderivingtheresults.
Modellersemploythemwhentheywishtospecifywhatkindofevidenceisrelevanttothetruthstatusofvariousassumptionsortospecifywhichclaimsoughtoroughtnottobeevaluatedfortruth.
Suchprecisionisonlycalledforwhenthetruthstatusischallenged,orislikelytobechallenged.
ThisisalsowhyIdiscussclaimsratherthanassumptions:inmathematicalmodels,assumptionscanbetakentobepropositionsexpressedintheformaldescriptions.
As-ifclaimsarenotassumptions,theyareratherclaimsaboutassumptions.
Strictlyspeaking,as-iflocutionsarenotnecessaryeveninsuchmethodologicaldiscussionsbecausethereasonsforusinganas-ifclaimcanalwaysbeexpresseddifferently.
3Recognisingthedifferentkindsofas-ifclaimshelpstoavoidmisunderstandingswhenthelocutionisused.
Iproposeasimplerecognitionformula:identifytherelevantattribution(seeSect.
2below).
Thestructureofthepaperisthefollowing.
Section2laysoutthebasicconceptsthroughwhichthedifferentas-ifclaimsareanalysed.
Sections3,4and5aredevotedtocases(a),(b)and(c),respectively.
Section3hastwosubsections:3.
1discussesnegligiblyinaccurateas-ifclaims,and3.
2negligibility-indicatingaccurateas-ifclaims.
Section4isalsodividedintwosubsections:Section4.
1presentsacasestudyof(b)concerningsignalextractioninanexpectedutilitymodelofvoting,andSection4.
2comparesthepresentaccounttoUskaliMki'sdistinctionconcerningas-ifclaims.
Section5discussesanothercase:KenBinmore'saccountofrevealedpreferencesandrepresentationtheorems.
Section6concludesthepaper.
2As-ifclaimsandattributionsThecanonicalformofas-ifclaimsisgivenbythefollowingbasicschema(Mki1998):AbehavesasifitwereB.
(AIC)2ForthelatestdiscussiononFriedman,seeThemethodologyofpositiveeconomics:ReflectionsontheMiltonFriedmanlegacy(ed.
UskaliMki).
Vaihinger'sas-ifaccountisdiscussedinFine(1993),Mki(2004),andBokulich(2009),forexample.
3Iamnotarguingforthisclaimhere.
Thereaderisencouragedtofindacaseinwhichtheas-iflocutionisnecessaryforformulatingaclaim.
Ihavenotfoundanysuchcases.
4Anas-ifclaimisanassertionthathastheform(AIC),andwhichconcernshowsomethingbehaves;itisabehaviouralclaim.
4Thereisasenseinwhichthemeaningofallthedifferentkindsofas-ifclaimisthesame5:theyspecifythatAbehavesasifitwereB.
Notethat(AIC)doesnotmakeanyclaimsaboutwhetherornotAisB.
ItmerelystatesthatAbehavesasBwouldbehavewereA,infact,B.
Letuscallthepropositionthat'AisB'theunderlyingattributionofanas-ifclaim,andthepropositionthat'AbehavesasifitwereB'theas-ifclaim(AIC).
6Theattributionisnotaclaimatallbecausethemodellerneverintendstoassertthatitistrue.
Yet,amathematicalmodelincludinganassumptionthatisjustifiedwithanas-iflocutionrepresentsAasbehavinglikeB.
ThepointofusingthelocutionistofocusattentiononthefactthatAisnotassertedtobeBeventhoughthemodeldescribesAasbehavinglikeB.
Thereasonwhyanassumptionisqualifiedwithanas-iflocutionisthatitprovidesunderstandingabouttherelevanttruthclaimsofthemodel.
Somethingisalsoclaimedtobecorrectinthemodel.
Inthecaseofinaccuracy-indicatingas-ifclaims,thatsomethingisthemodelresult,andinallothercasesitisthebehaviouralassumptionitself(i.
e.
theas-ifclaim).
AsFrankHindriks(2006)notes,presentingassumptionsinameaningfulwaypresupposescertainthingsabouttherelationshipbetweenthemodelandtheworld.
In(AIC)thepresumptionisthatAisnotinfactB.
IfAwereB,therewouldnotbeanyreasontousetheas-iflocutionbecauseonewouldsimplybeabletosay'AbehaveslikeBbecauseitisB'.
Theideaofapresuppositionrequiresthattheremustbeareasonforusingtheas-iflocution.
Anagentwhousesitmustpresupposethatthecorrespondingattributionisfalseorinvolvesacategorymistake.
Alternatively,ifthelocutionisusedwhentheattributioncouldbetrue,theagentwishestoindicatethatheorsheisnotcommittedtoitstruth.
Otherwiseitsuseisnotwarranted.
Whatiscommontoallattributionsisthusthattheagentisnotcommittedtotheirtruth.
Findingtheattribution4InmostcasesitisreasonabletointerpretAasanobjectandBasapropertythattheobjectcouldhave.
Ifonlyobjectscanbehaveincertainways,Amustbeanobject.
However,Bneednotalwaysbeapropertyofthatobject,asIshowinthesectiononnegligibility-indicatingas-ifclaims.
Consider,forexample,theclaimthat'objectsfallasiftherewerenoairresistance'.
Thepropertyofairresistanceisrelationalbecauseitrequiresreferencetothesurroundingcircumstances(atmosphericpressure)inadditiontotheobjectitself.
TheformulationofAICisdeliberatelysilentontheinterpretationofthetermsinordertoaccommodatevariousdifferentcases.
5Iaddacaveattothisstatementinthesectiononinaccuracy-indicating(a1)as-ifclaims.
6ThereasonforusingattributionterminologyisthatpropertyBisattributedtoA.
Onemightalsosaythat'BispredicatedaboutA',orthat'AisidentifiedasbeingB'.
However,usingtheterm'attribution'seemsmorenaturalhere.
5foranygivenas-ifclaimallowsonetodeterminewhetherusingtheas-iflocutionisappropriate.
Ifitis,onecouldsaythatthepragmaticconstraintsforusingitaresatisfied.
Whentheas-ifclaimisaccurateitisnotusuallysoinabsoluteterms,butmaybeapproximatelycorrect:itmaybe'trueenough'.
Theintendedmeaningisthatalthoughactualbehaviourisnotquiteinstantiatedbythefeaturesofbehaviourinthemodel,itisnotveryfarfromit.
Wherethedivergenceisnegligibletheas-ifclaims,althoughnotstrictlytrueofthephenomenatheydenote,aretrueenough(Elgin2004,2009).
Furthermore,giventhatsatisfyingtheconstraintsimpliesthatAisnotB,AcannotbehaveasBinallrespects,andmayonlydosoinsomecontextuallysalientrespect.
As-ifclaimsarethusalwaysmaderelativetothepurposeoftheinvestigation.
Thedivergencefromthetruthofabehaviouralclaim(AIC)maybeappropriateforonepurposebutunacceptableforanother.
Itisthispragmaticnatureofas-ifclaimsthatjustifiestheneedforapractice-basedratherthanapurelysemanticaccount.
Iamnowabletostatemypointmoreprecisely:theas-ifclaimisinaccurateorfalseundersub-case(a1),andmaybeaccurateorevenrealisticundertheothercases.
WhetherornotitisindeedtruedependsnotonwhetherAisBbutonwhetherA'sbehaviourcorrespondstothatwhichitwouldexhibitifitbehavedlikeB.
Thisprovidesacriterionforevaluatingthetruthofas-ifclaims.
MyaiminthelatersectionsistodemonstratethattherearecasesinwhichAinfactbehavesasifitwereBintherelevantrespectseventhoughitclearlyisnotB.
Identifyingdifferentkindsofattributionsisalsoawayofdistinguishingbetweenthedifferentkindsofas-ifclaimsonthebasisoftheprimaryreasonforusingthelocution.
Identifyingtherightkindofattributionthusfacilitatesunderstandingofwhatkindoftruthclaimisbeingmade.
Thisisthemainbenefittobederivedfromaclassificationofas-ifclaims.
Therearetwodimensionsalongwhichthefunctionandtruthstatusofassumptionsmaybeevaluatedthataresystematicallyrelatedtothedifferentkindsofas-ifclaims.
However,thethree-partclassificationisnotbasedonthem.
Firstly,amajorreasonfordistinguishingbetweenattributionsandas-ifclaimsisthatthelattercanbeclassifiedonthebasisofwhetherornotevaluatingthetruthoftheformerprovidesrelevantinformationconcerningtheirtruth.
Thisdoesnotprovideatruthcriterion,butratherindicateswhatkindofevidenceisrelevantforevaluatingthetruth.
Secondly,addinganas-iflocutiontoasentencedescribingabehaviouralclaimmayormaynotchangetheextenttowhichthemeaningoftheattributionandtheas-ifclaimisthesame.
Iftheydonothavethesamemeaning,Iwouldsaythatintroducingthelocutionbringsaboutasemanticchange.
3Negligibility-indicatingas-ifclaims3.
1Negligiblyinaccurateas-ifclaimsThefirstuseoftheas-iflocutioninvolvesmakinginaccurateassumptions.
Theprimaryreasonforusingithereisthatthemodellerknowsthatheorsheismakinganinaccurate6assumption.
ThesimplestwayofconveyingtheideaistosaythatAisdescribedasbeingorbehavinglikeBinthemodeleventhoughitisnotsointherealworld.
OnethensaysthatAbehavesasifitwereBinthemodel.
Thiscategoryofas-ifclaimsiscloselyrelatedtoVaihinger'snotionofafiction;thereisadeviationfromreality(whichmayultimatelybeeliminated)andanawarenessthatthefictionisjustafiction.
LetusnowhaveacloserlookatKirman'sexampleofsuchas-ifclaims:theas-ifclaimcanbewrittenbrieflyasfollows.
(a1)Peoplebehaveasiftheyoptimize.
Kirmanmakesitclearthatalthoughpeopledonotbehaveoptimally,theassumptionismadeanyway.
7Thereasonwhytheas-iflocutionisusedisthatthemodellerknowsthattheinherentassumptionisfalse.
Theattributionis'actualpeopleoptimize'.
Optimizingreferstotheendeavourofpursuingoptimalresultshere.
Theendeavourmaynotbesufficientforoptimalbehaviourbecausetheindividualsmaynothavetherequisiteinformation,knowledge,foresightetc.
inordertosucceed.
Yet,eveniftheyhadtherequisiteinformationbutdidnottrytooptimize,onewouldhavetousetheas-iflocutionbecausetheattributionisfalseirrespectiveofwhethertheyhavetherequisiteinformation.
Considernowwhethertheuseoftheas-iflocutionbringsaboutasemanticchangein(a1).
Thisrequiresspecificationofwhatexactlywouldbeclaimedifitwerenotused.
Anobviousresponseisthatwithoutthelocutiontheclaimisaboutthetruthoftheattribution.
Theredoesnotseemtobeasemanticchangeunder(a1)because'actualpeopleoptimize'makesthesameclaimaboutbehaviourastheas-ifclaim'peoplebehaveasiftheyoptimize'.
Althoughthelocutionisintroducedinordertoindicatethattheattributionandthebehaviouralclaimarefalse,themeaningoftheattributionandtheas-ifclaimisthesame.
87AccordingtoFriedman(1953,pp.
36-8),onecantreatthesamefirmasifitwereaperfectcompetitorononequestion,andasifitwereamonopolyonanother.
Thisexamplequalifiesasaninaccuracy-indicatingclaim.
NotethatKirmandoesnotendeavourtojustifytheoptimizationassumption,butratherascribestheas-ifjustificationtothosewhodo.
Infact,exceptforFriedman,Ihavenotfoundexamplesofeconomictextsinwhichtheinaccuracy-indicatingas-ifclaimismadebytheauthorwhousesthecorrespondingassumptioninamodel.
Incontrast,Macdonald(2003)aswellasBoylanandO'Gorman(ibid.
)provideanexampleinwhichtheinaccuracy-indicatingas-ifclaimisattributedtoothers.
8Anotherwayofinterpreting(a1)istosaythatitmeans'peopledonotoptimize,butthatdoesnotmatterforthemodelresults'.
Ifthisweretobetheinterpretation,therewouldbeasemanticchangeafterall.
Theas-ifclaim'peoplebehaveasiftheyoptimize'wouldthenmeanthattheattribution'Peopleoptimize'is7Onemightobjectthatitdoesnotmakeanysensetoclaim,forexample,thatrealpeoplearerationalifonedoesnotbelievethattheyare.
Nevertheless,certainscholarsthinkthatthereisnoproblem.
Consider,forexample,Boland(1979,pp.
512-3):…[Friedman's,]asiftheoryofexplanation…[I]fwearetryingtoexplaintheeffectoftheassumedbehaviorofsomeindividuals(e.
g.
,thedemandcurvederivedwiththeassumptionofmaximizingbehavior),solongastheeffectisinfactobservedanditwouldbetheeffectiftheywereinfacttobehaveasweassume,wecanuseourbehavioralassumptionevenwhentheassumptionisfalse.
Thatis,wecancontinuetoclaimtheobservedeffectoftheindividuals'(unknownbutassumed)behaviorisasiftheybehavedasweassume.
Notecarefully,theindividuals'behaviorisnotclaimedtobeasiftheybehavedasweassume,butratheritistheeffectoftheirbehaviorthatisclaimedtobeasiftheybehavedaccordingtoourassumption.
(allemphasesintheoriginal)Onewayofmakingsenseofthisisbynotingthatmodellersarenotcommittedtothetruthofallassumptionstothesamedegree.
Inparticular,amodellerwhoasserts(a1)isnotcommittedtothetruthoftheas-ifclaimitself.
Ifthemodelresultisdemonstrablyrobustwithrespecttoindividualbehaviour(LehtinenandKuorikoski2007,Kuorikoski,LehtinenandMarchionni2010),suchbehaviourneednotprovidethecentralcausalmechanisminarational-choicemodel(seealsoSatzandFerejohn1994).
Onemightthuswishtoassert(a1)justbecauseone'smodelneedssomeconnectionbetweentheindividualsandthesocial-leveloutcomes.
Asserting(a1)couldthenberephrasedas'theoptimizingmodelgivesmeaconnectionbetweenindividualsandoutcomes.
Eventhoughrealpeoplearenotliketheindividualsinmymodel,itdoesnotmatterbecausethefinedetailsoftheirbehaviourareirrelevantanyway'.
Onecouldthusassert(a)eventhoughoneknowsthatrealindividualsdonotoptimize.
Anotherpossibilityisthatalthoughindividualbehaviourisnotirrelevanttotheresults,itissufficientifpartofthepopulationoptimizes.
Notethatifsuchjustificationsareappropriate,themodellermaywellnotbecommittedtoanyparticulardeterminantsofbehaviourorbeindifferenttowardsthemandtheirtruth.
Thismaymakeitdifficulttodistinguishbetweeninaccuracy-indicating(a1)andnon-committal-indicatingclaims(c).
Herethemodellerbelievesthattheas-ifclaimitselfisfalse,butforthepurposesofhisorhermodel,itisnegligiblyso.
Thetruthoftheclaimisfalse(butnegligiblyso).
Myintuitionsareslightlyinfavouroftheinterpretationprovidedinthemaintext,butIamwillingtobepersuadedtothinkotherwise.
Adoptingthisinterpretationwouldmakeanalysingas-ifclaimsmoreuniformwithrespecttosemanticchangebutlessuniformwithrespecttomeaning.
Onewouldalwayshavetolookforasemanticchange,butthemeaningoftheas-ifclaimswouldbethesameonlyincasesa2,bandc.
8notimportanttothemodellerbecauseheorsheismerelyusingthebehaviourspecifiedinordertostudysomethingelse,andtheas-iflocutionisusedmainlytoindicatethatoneshouldnotevaluatethemodelbyexaminingtheparticularassumptionthatithasqualified.
Thusthemodellermaybelievethatoptimizationentailsbehaviourthatisdifferentfromthewayinwhichpeopleactuallybehave,butthewayinwhichtheybehaveisultimatelyirrelevanttotheresultsofthemodel.
Incontrast,incase(c)themodelleraimstoformulateanaccurateratherthananegligiblyfalsebehaviouralassumption.
3.
2Negligibility-indicatingaccurateas-ifclaimsHereAissomesystemofinterest,andBreferstosomesetofidealconditions.
Mki(1992,2003,2009)providesthefollowingdescriptionofthefunctionofsuchusage:9(a2)Phenomenabehaveasifcertainidealconditionsweremet:conditionsunderwhichonlythoserealforcesthataretheoreticallyisolatedareactive.
Heretheas-ifclaimreferstotheideathatvariousdisturbingfactorsfoundintherealworldareassumedtobeabsent.
Thecentralforcesidentifiedinthemodelaredepictedasiftheyweretheonlyforces.
Theattributionisthatrealitysatisfiestheidealconditions,i.
e.
thattherearenodisturbingfactorsotherthanthoseidentifiedinthemodel.
Introducingtheas-iflocutioninducesasemanticchangebecausethesentencenowmakesatrueenoughclaimaboutthebehaviourofsomethinginactualconditionsratherthanafalseclaimconcerningwhetherornottherealityexhibitsidealconditions.
Theconstraintsaresatisfiedbecauseweknowthattheidealconditionsdonotholdintherealworld.
Furthermore,theas-ifclaimdirectsattentiontowhatisachievedinassumingtheabsenceofsuchfactors:wegettoknowwhatkindofbehaviouroroutcomeswouldresultifthecentralforceswereactingalone.
'Peoplebehaveasiftheywereonlymotivatedbyself-interest'providesawell-knownexamplefromeconomics.
Inthisexamplethenegligiblecausalfactorsarementalstates,butaswewillnowsee,theyalsocanbephysicalforces.
Theas-ifclaimexpressedby(a2)maybeaccurateortrueenough.
Consider,forexample,theoftenusedexampleofGalileo'slawoffallingbodies,whichstatesthat'undervacuum,thedistancecoveredbyafallingobjectisgivenbys=gt2'.
Onecouldthensay,withFriedman,thatgoodpredictionsmaybeobtainedbyassumingthat9Friedman(1953,p.
40)alsoprovidesanexampleofthiskindofas-ifclaim.
Becker(1962,p.
5)writes'Impulsivehouseholdsareassumedtoact"asif"theyonlyconsultedaprobabilitymechanism.
'Itisnotclear,however,whathemeansbecausehehadalreadystated'Therefore,householdscanbesaidtobehavenotonly"asif"theywererationalbutalso"asif"theywereirrational(p.
4).
'Becker'slackofprecisionmakesmewonderwhetherheintendedtopresentanargumentforrationalchoicetheory(cf.
MoscatiandTubaro2011).
9(a2')actualobjectsfallasiftheywerefollowingGalileo'slaw.
Theas-ifclaimistrueenoughiftheairpressureissufficientlyclosetozero,andtheobjectsatisfiescertaincharacteristicssuchashavingsufficientdensity.
Theattributionisthattheairpressureiszero.
Giventhatitmustbefalse,onedoesnotneedtousetheas-iflocutionatallifonehasartificiallycreatedavacuum,andmakesstatementsconcerningobjectsfallinginit.
Thereis,however,arangeofconditionsunderwhichitmakessensetosaythat'anobjectbehavesasifthoseconditionscorrespondedtoavacuum'.
Here,evaluatingthetruthoftheattributionisrelevantforevaluatingthetruthoftheas-ifclaim,whichisafeaturethat(a1)and(a2)share.
Thisdoesnotseemtobeacaseinwhichonewouldsay'AbehavesasifitwereBeventhoughitclearlyisnotB'becausewhetherornotA(anobjectinactualatmosphericconditions)behavesinsuchawaydependsonhowcloseAistobeingB(anobjectinaperfectvacuum).
AsIwilldemonstrateinlatersections,however,otherkindsofas-ifclaimsaredifferentinthisrespect.
Friedman(1953,p.
18)emphasisesthatitisnotassumedthatweliveinavacuum.
10Wemerelyassume(orclaim)thatobjectsfallasiftheywereinavacuum.
Introducingtheas-iflocutionbringsaboutasemanticchangethattransformsthefalseclaimabouttherebeingavacuumintoatrueenoughclaimaboutbehaviour.
Musgrave(1981)suggeststhattheas-iflocutioncouldbeusedtoexpressanegligibilityassumption:theclaim'airresistanceisnegligible'couldbeexpressedinthewords'bodiesfallasiftheywereinavacuum'(seealsoMki2000).
Thesetwoclaimsarenotidentical,however,becausetheformerexpressesreasonswhythebehaviouralclaimmightbeaccuratewhereasthelatterprovidesadescriptiveclaimaboutthebehaviourofsomething.
Considernowanotherwell-knownexample,Boyle'slawofidealgases,whichstates,roughly,thatataconstanttemperaturethepressureofafixedamountofgasvariesinverselywithitsvolume.
AccordingtoMichaelStrevens(2008),onemightsaythat(a2'')gasesbehaveasiftherearenocollisionsofmolecules.
Thisisassertedonthegroundsthatwhetherornottherearesuchcollisions,thedemonstrationofBoyle'slawgoesthrough.
ItisclearthatcollisionsareonlynegligibleforthepurposesofexplainingBoyle'slaw.
Theyarenotnegligibleforotherpurposes(e.
g.
,modelingdiffusion).
Theattributionisthattherearenocollisions.
Notethatthefalsityofthisclaimisnotrelevantinevaluatingthetruthoftheas-ifclaim.
Weknowthatitisfalse,butwhatreallymattersistheargumentgivenforthenegligibilityclaim.
StrevenscitesMcQuarrieandSimon(1997,p.
1015),whoarguethat'anycollisionthat10NotethatFriedmanusestheterm'assumption'tomeanwhatiscalleda'claim'inthispaper.
Ifassumptionsexpresspropositionsinmodels,onewouldsaythat'themodelassumesavacuumbutthemodellerdoesnotclaimthatweliveinavacuum'.
10deflectsthepathofamoleculefrom[thepathassumedinthederivation]willbebalancedbyacollisionthatreplacesthemolecule.
'Therearethusthreekindsofnegligibilityclaim.
Thefirstisthatcausalforcesthatareknowntoaffectthebehaviourofinterestareweakenoughtobenegligibleinstrength(e.
g.
,airpressure).
Secondly,forcesthataretakentobenon-negligibleinstrengthandpotentiallyrelevanttothebehaviourinquestionhavenegligibleneteffects(duetosomecounterbalancingofforces,aswithmolecularcollisions).
Thirdly,anymodelignoresaninfinitenumberofcausalfactorsbecausethosefactorsarethoughttobeirrelevant.
Inthislastcase,theas-iflocutionisnotused.
Onedoesnotsay,forexample,that'stonesfallasifswallowsdonotmigratetowarmerterritoriesduringtheautumn'becauseonepresupposesthatthemigrationbehaviourofthesebirdscouldnotaffectthestonesinanyway.
Thereisalsoadifferencebetween(a1)and(a2).
Afeatherdoesnotsatisfytherelevantcharacteristics,andoneshouldnotsaythat'thefeatherfallsasifitwerefollowingGalileo'slaw'becausetheas-ifclaimwouldbefalse.
Contrastthiswith(a1).
Ifthereasonforusingtheas-iflocutionisindeedonlytoindicatethatthetruthof(a1)doesnotmatter,thefalsityoftheclaimwouldnotpreventthemodellerfromusingthelocution.
Heorshewoulduseitevenknowingthatpeopledonotoptimize.
Thedifferencebetween(a1)and(a2)isthattheformerdoesnotexpressaclaimthatistakentobecausallyimportantforthemodel,butthelatterdoes.
Incasea2,theas-iflocutionisneededforspecifyingtheexactcontentofthebehaviouralas-ifclaim,whereasincasea1itmerelyindicatesthatthebehaviouralclaimisnottobeevaluatedfortruth.
Mki(1998,2003)suggeststhat(AIC)couldalsobetakentoexpressepistemicuncertaintyconcerningwhetherAisB.
Itseemsthatifthisisapplicableatall,itismostsuitableforinaccurateas-ifclaims(casea1).
Onewouldsaythat'peoplebehaveasiftheywererational'becauseoneisuncertainwhetherornottheyarerational.
Note,however,thatsuchuseviolatesthepragmaticconstraintforusingtheas-iflocution.
Onecannotpresupposethattheattributionisfalseifoneissimultaneouslyuncertainaboutitstruth.
11Onemightarguethatprovisionallyexpressingepistemicuncertaintywithrespecttotheattributioncountsasnotbeingcommittedtoitstruthandthusthattheconstraintissatisfiedafterall.
Butthiswillnotdobecause,whoeverassertstheaforementionedas-ifclaiminordertoexpressepistemicuncertaintyisultimatelycommittedtothetruthoftheattribution:iflaterinvestigationsshowthattheattributionisfalse,heorsheisnolongerwillingtoassertthebehaviouralclaim.
Thatiswhatbeingcommittedtothetruthoftheattributionmeans.
Distinguishingbetweendifferentkindsofas-ifclaimsonthebasisofwhethertheirtruthdependsonthetruthoftheattributionisimportantforspecifyingtheobjectofepistemicuncertainty.
Itmayconcerntheattribution,theas-ifclaimandtheexplanationofwhether11NotethatVaihingerdistinguishesbetweenfictionsandhypothesesonthegroundsthattheformerareknowntobefalsewhereasthelatterareamatterofinvestigation.
11theclaimmaybetrue,andifso,why.
Incases(a1)and(a2)epistemicuncertaintyconcerningtheattributionimpliesepistemicuncertaintyconcerningtheas-ifclaim.
BecauseMkidoesnotdistinguishbetweenas-ifclaimsandattributions,hedoesnotdistinguishbetweendifferentobjectsofuncertainty.
Heisthusledtoconflateepistemicuncertaintyconcerning'whyitisthatAbehavesinthewayitdoes'withuncertaintyconcerning'whetherAisorisnotB'(Mki1998,p.
26).
Isuggestthatas-iflocutionsarenotusedforexpressingepistemicuncertaintypreciselybecauseepistemicuncertaintyconcerningtheattributiondoesnotimplyepistemicuncertaintyconcerningtheas-ifclaiminthemostgenuinecases(bandc).
Usingitforsuchpurposeswouldperpetuatethefalseimpressionthatthetruthofanattributionisalwaysrelevanttothetruthoftheas-ifclaim.
12As-ifclaimsneverexplain'whyitisthatAbehavesthewayitdoes'inthemselves.
Thisfeatureisdifficulttoreconcilewithrealism,evenconcerningcase(a2).
Indeed,theuseofas-iflocutionscanbejustifiedbecausetheygiveanindicationthattheclaimitselfdoesnotprovideanexplanation.
4Falseascriptionas-ifclaimsThethirduseoftheas-iflocutioninvolvesascribingfalseintentionsorcognitionstosomethinginordertodescribeitsbehaviour.
Thebasicformofsuchaclaimisthusthefollowing:(b)AbehavesasifithadintentionsorcognitionsB.
Thepointistodescribebehaviouraccuratelywhenthereisnoneedtoconcernoneselfaboutwhethertheintentionalorthecognitiveattributionsarecorrect.
Notethatincontrasttocase(a1),hereoneaimstoassertthatthebehaviourformulatedbymeansoftheas-ifclaimisaccurate.
ManyofMiltonFriedman'sexamplesexemplifysuchunrealisticattributions:onecandescribethetrajectoriesofbilliardballsaccuratelyundertheassumptionthatexpertplayersactasiftheyknowcomplicatedmathematicalformulasderivedfromphysicaltheory(1953,p.
21).
Thistheoryattributestheabilitytomakesuchcomplicatedcalculationstotheplayers.
Theattributionisfalsebecausetheplayersdonotconceptualisetheissueintermsofthephysicaltheory.
Nevertheless,onecouldderivetrajectoriesforthebilliardballsbyusingthatphysicaltheory.
Similarly,Friedman(ibid.
,p.
19)arguesthatthepositionoftheleavesonatreemaybeaccuratelydescribedontheassumptionthateachindividualleafactsasifitistryingtomaximisetheamountof12Naturally,IagreewithMkithatepistemicuncertaintyisconsistentwithrealism.
Iammerelyconcernedtoshowthatexpressingsuchuncertaintybymeansofas-iflocutionsdoesnotsatisfymypragmaticconstraints.
Inthissense,itisnotsurprisingthatMkihasnotprovidedanyconcreteexamplesofsuchuses.
12sunlightitreceives.
Lehtinen&Kuorikoski(2007)positthatinsuchcasestheassumptionisbehaviourallyrealisticbutintentionallyunrealisticbecauseitdescribesthebehaviourtruthfullybutattributesintentionsuntruthfully.
Ascribingintentionstothetreefailstoaccuratelyidentifythecausesoftheleafpattern.
13InFriedman'sexamplesofbilliardsplayersandleavestheattributionisclearlyfalse,andthetruthoftheas-ifclaimdoesnotdependonthetruthoftheattribution.
Examiningthetruthoftheattributiondoesnotproviderelevantinformationaboutthetruthoftheas-ifclaim.
Friedmanalsousesthiskindofas-ifclaiminarguingthatfirmsbehaveasiftheyweremaximisingprofits(ibid.
,p.
21).
Hewasperfectlyawareofthefactthatmanagersdonotintentionallytrytoequalisemarginalcostandbenefit,andthattheyoftendonotevenknowthecostfunction.
Thepointofusingtheas-ifclaim,however,istoarguethateveniftheydonotintentionallytrytomaximiseprofits,theirbehaviourissoclosetotheprofit-maximisingassumptionthatonecouldobtaingoodpredictionsonthatbasis.
Thesepredictionsdonot,ofcourse,concernprofitmaximisinginitself,butratherfocusonotherthingssuchastheeffectsofapricechangeorataxonsupply.
Friedmanwasarguingthattheprofit-maximisationassumptionwas,infact,accurateforthepurposesforwhichitwasmeantinstandardpricetheory.
Itakeitthatsuchaninterpretationisnolongerconsideredradical,giventhatHoover(2009)alsoarguesthatFriedmancaresabouttheaccuracyofassumptions.
Iftheattributiondoesnotproviderelevantinformationforevaluatingthetruthofthiskindofas-ifclaim,whatdoesFriedmanisknowntoresthiscaseonanevolutionary-selectionargument:onlyfirmswhosebehaviourmimicsprofitmaximisingwillsurviveinthelongrun.
TheselectionargumentiscommonlytakentofacilitateexplanationofwhythebehaviourofAcouldmimicthatofBevenwhenAisnotBinvariousdifferentcontexts.
MauriceLagueux(1994)argues,however,thatas-ifclaimsarealwaysjustifiedbysomemoregeneraltheory,buthedoesnotrequirethistheorytobethatofevolutionaryselection.
Thisseemsright,giventhattheredoesnotseemtobeanyneedforaselectionargumentinthecaseofGalileo'slawoffallingbodies.
Thereisageneraltheory,butunlikeinFriedman'sotherexamples(firms,leavesonatree,billiardsplayers),itisnotseparatefromtheas-ifclaimbecauseitistheverytheoryofgravitythatexplainswhyonecandescribeobjectsasiftheyfollowedGalileo'sformulations=gt2.
Theselectionargumenttendstobeconsideredlessthanfullycompelling(seeVromen2009foranaccountandreferences),andthereforetheprofit-maximisationassumptionis13Ananonymousreviewersuggesteddescribingthisexampleintermsofincorrectcauses.
Itmightbepossibletogeneralisecase(b)suchthatitconcernsallfalseascriptionsofcauses:'AbehavesasifitweregovernedbycausalforcesB'.
Thisisclearlypossibleatleastinprinciple,butgiventhatIhaveonlyseenexamplesof(b)inwhichthefalseattributionconcernsmentalstates,Ihesitatetoargueforthismoregeneralformulation.
Similarconsiderationsapplytocase(c).
13oftenconsideredinaccurateafterall.
EvenBecker(1962,p.
10)arguesthattheassumptionisnotaccurate.
Inmyjudgmentthegreatachievementofthe"survival"argumentadvancedbyAlchianandothers'isnotademonstrationthatsurvivingfirmsmustactasiftheyweretryingtomaximizeprofits,forcounterexamplescaneasilybedeveloped,butratherademonstrationthatthedecisionsofirrationalfirmsarelimitedbyabudgetaryconstraint.
(myitalics)ThefactthatFriedmanemploysthesurvivalargumentgivessupporttotheinterpretationthatFriedmantriedtoarguefortheaccuracyoftheprofitmaximisationassumptionandthusthathisas-ifclaimconcerningfirmsistobeinterpretedintermsof(b)ratherthan(a1).
Yet,IwouldhesitatetoarguethatBoland'sinterpretationofFriedmanintermsof(a1)iswrongbecauseFriedmanthoughtthatdiscussingthe'realism'ofassumptionsisonlymeaningfulrelativetothepurposeoftheorizing.
Thus,althoughheseemstohavethoughtthatprofitmaximisingisaccurateenoughforthepurposesofthetheoryofthefirmasitwaspractisedinthe1950's14,heprobablythoughtthatdiscussingwhetherornottheassumptionisaccurateinanabsolutesenseismeaningless.
IdonotwishtorelyonacorrectinterpretationofFriedman.
Iwillinsteadexaminetheuseoftheclaiminanothereconomicmodelinordertoshowthatitcouldbeusedtodescribebehaviouraccuratelyandevenrealistically.
Thus,thefailureofFriedman'sexamplefromeconomicsshouldnotleadonetobelievethatthiskindofas-ifclaimisalwaysinaccurate.
Giventhatthereisnoneedtoprovideanykindofselectionargumentorgeneraltheoryinthiscase,itwillalsodemonstratethatsuchargumentsandtheoriesarenotnecessaryfortherealisticdescriptionofbehaviourthroughas-ifclaims.
4.
1Acasestudy:signalextractioninanexpectedutilitymodelofstrategicvotingConsiderLehtinen's(2006,2007,2008)useofasignalextractionmodelaspartofanexpectedutilitymodelofstrategicvoting.
Strategicvotingisusuallydefinedasgivingone'svotetoacandidateoralternativethatisnotconsideredtobethebest.
Onlythecasewithpair-wisecomparisonsisanalysedhereinordertosimplifythepresentation.
Lehtinenassumesthatvotersdonotknowothervoters'preferences,butthattheyactasiftheyhadobtainedinformativesignalsconcerningthem.
LetNdenotethetotalnumberofvotersandnjkthenumberofvoterswhopreferalternativejtok.
Ifnjk>N/2,jwillbeatkinapair-wisecomparisonunderthemajorityruleifeveryonevotessincerely.
Themodelissetinacomputer-simulationframeworkinwhichvoters'preferencesaregeneratedrandomly.
Votersareassumedtoactasiftheyhaveobtainedperturbedinformationontherealisedvalueofnjkasfollows:anindividualvoter'ssignalsi(jk)isgivenbythesumofthestandardisednjkandarandomterm:14ThisaspectisevenmoreevidentinMachlup(1946).
14s!
jk=!
!
!
"!
!
!
+εR!
,(1)whereRidenotesastandardnormalrandomvariableandεdenotesaparameterthatreflectsthequalityoftheinformation.
Heshows(Lehtinen2006)thatanindividual'sprobabilitypi(jk)forthepropositionthatalternativejbeatsalternativekisgivenbyp!
jk=1Φ!
!
!
"!
!
!
!
!
,(2)whereΦdenotesthestandardnormaldistributionfunction.
Theexistenceofsuchsignalsisnotafictionbecauseinmasselectionspre-electionpollsmaybeinterpretedassomesortofsignalconcerningthewinningprobabilitiesofvariouscandidates,andeveninparliamentaryvotingtherepresentativeshavevarioussourcesofinformationonothervoters'preferences,suchaspartyaffiliationandthecontentofargumentsmadebydifferentrepresentatives.
Yet,variousfeaturesofthesignalsarefictions:thederivationintheappendixconsistsoffourpagesofmanipulationsofrandomvariablesandtheirdensities.
Itwouldbefoolhardytoclaimthatvotersareabletomakesuchcalculations,letalonethattheywouldactuallyperformthem.
Realvotersdonotengageincomputingsuchexpressionsatall.
Similarly,theydonotliterallycomputeexpectedutilitiesbymultiplyingprobabilitiesbyutilities.
Furthermore,intherealworldtheydonotobtainperturbedsignalsinthekindofforminwhichtheyarerepresentedinthemodel,inwhichasignalisgivenbyanexactnumeral.
Forexample,asignalcouldbegivenbysi(jk)=-0.
298234,andaprobabilitybypi(jk)=0.
267543.
Thesemodelsofstrategicvotingthusdepictvotersasiftheyhadobtainedperturbedsignalsinanexactnumericalform,asiftheywereabletoperformcomplicatedcalculations,andasiftheyactuallycarriedoutthecomputationsandusedtheminanexpected-utilitycalculusfordeterminingtheirdecisions.
Nowcomparethefollowingtwostatements:(b')Votersactasiftheyhaveobtainedperturbedsignalsconcerningthepreferenceprofileandasiftheyhavecomputedprobabilitiesfromthesesignalsusingstatisticalreasoning.
(b*)Votersobtainperturbedsignalsconcerningthepreferenceprofileandcomputeprobabilitiesfromthesesignalsusingstatisticalreasoning.
(b*)isclearlyfalsebut(b')istrueenoughinthesensethatactualvoters'behaviourreallycorrespondstothatwhichwouldensueiftheyactuallyobtainedsignals,computedprobabilitiesandsoon.
Introducingtheas-iflocutionconvertsafalsestatementintoatrueoneandthesemanticchangeisobvious.
Thesestrategicvotingmodelsareintentionallyandcognitivelyunrealisticinthesensethatvotershavecognitionsandintentions(signalsandcomputing)inthemodelthattheydonothaveinreality.
However,themodelmayneverthelessbeperfectlyaccurateinaccountingfortheirbehaviour,andevenrealisticinthesenseofcorrectlyidentifyingthefactorsthatinfluencethebehaviourandthewayinwhichtheydoso.
Iwillnowattempttodemonstratethis.
15WhatwouldarealisticdescriptionofstrategicvotingbehaviourlooklikeOneshouldgetthedeterminantsofthebehaviourright.
ConsiderAnthonyDowns'verbaldescriptionoftheincentivesforstrategicvoting:Assumethattherearethreeparties:Right,Center,andLeft.
VoterXprefersRighttoCenterandCentertoLeft,buthebelievesthatRighthastheleastchanceofwinning.
IfhegreatlyprefersRighttoCenterandisalmostindifferentbetweenCenterandLeft,heislesslikelytoswitchhisvotefromRighttoCenterthanifheslightlyprefersRighttoCenterbutabhorsLeft(Downs1957,p.
49).
Theimplicationhereisthatstrategicvotingdependsonthewinningchancesofvariouscandidatesandonpreferenceintensities.
LetussaythatvotersengageinDownsiandeliberationiftheyconceptualisevotinginsuch(qualitative)terms.
Itinvolvesthementaloperationofweighingthedesirabilitiesoftheoptionswiththeprobabilities.
Thisoperationcouldmoregenerallybecalledexpected-utilitydeliberation.
Lehtinen'smodelsarebasedonEnelow's(1981)andMcKelvey&Ordeshook's(1972)expectedutilitymodelsforexplainingvoterbehaviourundervariousvotingrules.
Thesemodelsshowhowbeliefsandwinningprobabilitiesoughttocombinesoastotakeintensitiesandprobabilitiesintoaccountintherightway.
Ifallonewishestodoistoprovideanabstractaccountofpeople'sstrategicvotingbehaviour,suchmodelsaresufficient.
However,inordertostudytheaggregate-levelconsequencesofsuchbehaviouritisnecessarytogivespecificvaluestoindividualutilityfunctionsandbeliefs,andthequestionthatthenarisesishowtheyshouldbedetermined.
Lehtinengoesontogenerateutilitiesrandomlyandusesthesignalextractionmodeltoderivethebeliefs.
Giventhatbehaviourdependsonvoters'beliefs,hismodelshouldgeneratebeliefsinarealisticway.
Hereitseemsplausibletostatethatthereareinterpersonaldifferencesintheaccuracyofknowledgeconcerningthewinningchancesofvariouscandidates,andintheconfidencedifferentvotershaveintheirknowledge.
IfonebelievesthattheRightdoesnothaveachanceofwinningbutiswrongaboutthisandvotesstrategically,andiftheoutcomethenturnsouttobeone'ssecond-bestoptionratherthanthebestoneduetothatstrategicvote,onewouldhavehurtoneselfbyvotingstrategically(seeLehtinen2007foranexample).
Hence,itseemsreasonabletorequirethatpeoplevotestrategicallyonlyiftheycanbereasonablyconfidentthattheirinformationiscorrect.
Furthermore,andevenmoreobviously,thebeliefsderivedfromthesignalextractionmodelshouldberesponsivetovariationsintherealisedpreferenceprofile.
Thisispreciselywhatthesignalextractionmodelshows(seeesp.
Lehtinen2006fordetails).
Eventhoughvotersmustbesaidtoactonlyasiftheyhadreceivedsignalsandasiftheyhadcomputedbeliefsonthebasisofthosesignals,thesignalextractionmodelprovidesaperfectlyrealisticaccountoftheirdeliberationandthedeterminantsofstrategicvoting.
Thedegreeofconfidenceinandtherealqualityoftheinformationaretheactual16determinantsoftheirbeliefswhentheyvotestrategically,andthemodelcapturestheminarealisticmanner.
15Itisthusrealisticwithrespecttosomeintentionsbutunrealistic(andfictional)withrespecttosomeothers.
Notethatwhereastheattributionstatesthatvotersintentionallycalculatetheseprobabilities,theas-ifclaimmerelystatesthattheactions(orbeliefs)derivedfromthemodelmaybetakentodescribeactualvoters'actions(orbeliefs).
Furthermore,itisdefinitelynotassumedthatvotershaveparticularlygoodcognitivecapacities.
Ifεislarge,somevotersmayhavehighlyinaccuratebeliefs.
Itwouldbesillytocriticisethismodelforattributingunrealisticallysophisticatedbehaviourtovotersonthegroundsthatitdescribestheirbehaviourasiftheyhadgonethroughcomplicatedcalculations.
Usingtheas-iflocutionhighlightsthisfact.
Formulatingthebehaviouralassumptioninstatementssuchasvotersactasiftheycomputedprobabilitiesonthebasisofperturbedsignalsdoesnotindicatewhetherornotthebehaviourwillberealisticallymodelled.
Theas-ifclaiminitselfdoesnotjustifysuchthinking,itmerelyformulatesthebehaviouralclaim.
Nevertheless,thefactthattheattribution(b*)isunrealisticisentirelyirrelevantintermsofevaluatingtherealisticnessoftheas-ifclaim:itisknownthatvotersdonotgothroughcomplexcomputationswhentheyvote,butthatisirrelevant.
However,merelystatingthat(b')istrueenoughisnotsufficientbecauseitdoesnotprovideanyexplanationofwhyoneshouldthinkthatitis.
Itstruthisrathertobedemonstrated.
Hence,as-ifclaimsaretypicallyincompleteinthatshowingthatthemodelisrealisticalwaysrequiressomethingbeyondtheclaimitself.
Otherkindsofas-ifclaimsalsoseemtobeincompleteinthisway.
Asdemonstratedabove,whethernegligibility-indicating(a2)as-ifclaimsinvolvemakingaccurateassumptionsdependsonthedegreetowhichtheattributionistrue.
Inthiscase(b)thereisnosimilarcontinuumofclaimsfromtheperfectlyaccuratetotheutterlyinaccurate.
Theclaimthatvotersobtainsignalsexpressedintermsofaseven-digit15Lehtinen'smodelsarenotintendedtoberealisticwithrespecttoallaspectsofbehaviour.
Onemajorfeatureisthatinhisanalysesofstrategicbehaviourhecomparesthecaseinwhicheveryvoterengagesinsuchbehaviourtooneinwhichnobodydoesso.
Whatispresentedhereappliestovotersinsofarastheyengageinstrategicvoting.
Inreallifesomevotersengageinstrategicbehaviourandsomedonot.
Herelaxesthisassumption,however,inLehtinen(2010),buthereisanothersenseinwhichthesignalextractionmodelisnotentirelyrealistic.
Ithasbeenarguedthatvoterssystematicallyoverestimatethewinningchancesofcandidatestheysupport,butthemodelimplicitlyassumesnon-biasedestimates.
Note,however,thatsuchbiasescouldbetakenintoaccount,andiftheywereitwouldbeviaincorporatingsomeparametersorfunctionsintotheformalmodel,andonewouldstillhavetousetheas-iflocutioninpresentingthecomputingpart.
17number,computeprobabilitiesonthisbasis,andchooseanactionafterconsideringtheexpectedutilityofthevariousoptionsisafictioninotherwordsknowntobefalse.
However,Iarguethatthesignalextractionmodeldoesaccuratelycapturethedeterminantsofbehaviour.
Thereisnogeneraltheory(suchasevolutionaryselection)thatexplainswhyindividualbeliefsshouldcorrespondtothosegeneratedinthemodel.
Ihavearguedthatthefictionofcomputingbehaviouranditssignalsallowsforarealisticrepresentationofvotingbehaviour.
Ultimately,Iassertthattheas-ifclaimisrealisticonthegroundsthatitcorrectlydescribesDownsiandeliberation,which,Ialsoassert,providesatrueaccountofthereasonsforstrategicvoting.
Morrison(2009)claimsthatthewayinwhichfictionalrepresentationsprovidereliableinformationdiffersfromonecasetoanother.
Itdoesnotseempossibletogeneraliseonwhy,whetherandwhentheyfacilitatetherealisticdescriptionofbehaviour,however,whichcouldbeattributabletothefactthatfictionsplayacrucialroleinrepresentingtheworld(seealsoTeller2009).
Inthiscaseitisnecessarytorepresentthesignals,theutilitiesandtheprobabilitieswithprecisenumeralsinordertorepresentDownsiandeliberationinacomputermodel,andthefunctionoftheas-iflocutionistoshowthatthisrepresentationisnotintendedtoprovidealiterallytruedescriptionofallthedetailsofthatdeliberation(cf.
Coddington1979).
4.
2Adigression:Mki'sdistinctionUskaliMki(1992,1998,2000,2003,2004,2009)distinguishes(a2)(a2)Phenomenabehaveasifcertainidealconditionsweremet:conditionsunderwhichonlytherealforcesthataretheoreticallyisolatedareactive.
from(b~)thus:(b~)Phenomenabehaveasifthoseforceswerereal.
IwillnowattempttoclarifyhowtheclassificationproposedherediffersfromMki'sdistinction.
Mki'sdistinctionconcernsthereferenceoftheas-iflocution:hedistinguishesbetweencasesinwhichsomethingismodelledasifcertainminorforceswereabsent,andthoseinwhichthecentralforcesinthemodelarebasedonfictions.
Intheformercasetheisolationofforcesisfalse(i.
e.
,theyarenottheonlyforcesthatoperateinreality),whereasinthelattercasethecentralforcespickedoutarefictional(i.
e.
,theyarefoundintherepresentationbutnotintherealworld)(Mki2009).
Thisdistinctioncannotbeincorporatedintotheclassificationschemeproposedherebecausethebasisonwhichitrestsisdifferent.
Mki'sdistinctionisbasedonwhatthefictiontheas-iflocutionexpressesisabout,whereashereitisbasedonthepragmaticreasonfor18usingthelocution.
Mkidoesnotgiveanyreasonwhyonewouldwanttomodelthecentralcausalforceswithafiction.
16AccordingtoMki(e.
g.
,1998),(b~)invitesaninstrumentalistreadingbecauseitpostulatesfictionalforces.
Inmyexampleofsignalextractionindividualbehaviourisdescribedintermsoffictionseventhoughitisacentralcausalfactorinthemodel.
Mki'sdistinctionthusrequiresthat(b')mustbeunderstoodintermsof(b~),and(b~)shouldthusbeunderstoodininstrumentalterms.
Ibelievethisconclusionshouldberesisted.
Ihavearguedinthissectionthatanas-ifclaimmaywellberealisticeveninacaseinwhichcentralforcesarerepresentedbyfictions.
IfIhavesucceededinshowingthatvotersreallybehaveasthoughtheyweregovernedbysomefictionalforces(computingprobabilitiesandexpectedutilities)identifiedinthemodel,thereseemstobelittlereasonforarealisttobeworried.
Therelevantquestionishowtherealisticallydescribedbehaviourspecifiedintheas-ifclaimisjustifiedorexplainedgiventhattheclaimitselfdoesnotprovideanexplanation.
Itisobviousthatsuchanexplanationmustbefoundelsewhere,andwhatIhavewritteninthissectionprovidesit.
Thedifferencebetweenrealistsandinstrumentalistsispresumablythattheformerrequirethatanexplanationorjustificationisprovidedsomewherewhereasthelatterdonotcareaslongasthepredictionsprovidedbythetheoryarecorrect.
Onlyafoolishbrandofrealismwouldrequirethateachandeveryassumptionmustbeformulatedinsuchawaythatanexplanationisalsoprovidedintheveryformulation.
Asdemonstratedabove,theattributionisplainlyfalse,andtheas-iflocutionisneededinordertoeffectasemanticchangethatallowsthecorrectunderstandingofthebehaviouralassumption.
Ifrealistscanacceptunrealisticassumptionsiftheyareproperlyjustified,theyshouldnotbanishrealisticassumptionsmerelybecausetheyarerepresentedbyfictions.
Mkicouldperhapstakeadifferenttack,andarguethatmyexampledoesnotcountasacaseof(b~)because,althoughobtainingsignalsandmultiplyingutilitiesarecentralforcesinthemodel,theyarereal.
Itistruethattheas-ifclaimdoesnotqualifytheexistenceoftheseforces.
However,neitherdoesitindicateisolationornegligibility.
Itdoesnotmeanthattheremightbeotherrelevantforcesthatareignoredforthepurposesofinvestigation.
Theas-iflocutionisratherneededtoindicatethatthemodellerdoesnot16Theinstrumentalusefulnessoffictionscouldbetakenasareasonwhyas-iflocutionsareusedincase(b*).
Given,however,thatstatingthatafictionisinstrumentallyusefuldoesnotreallyexplainwhyitis,itisratherbeggingthequestion,anditmightbebettertostatethatMkidoesnotprovideanyreasonforusingtheas-iflocutioninthiscase.
Itisbecauseinstrumentalusefulnessisquestion-beggingthatitisnotincludedasaseparatetypeofas-ifclaim,eventhoughFriedmansometimesusestheas-iflocutiontoindicatethatahypothesisisusefulifitgeneratescorrectpredictions(FriedmanandSavage1948,p.
298,1952,p.
473).
19claimthatvoterscomputeprobabilitiesormultiplythembyutilities,albeitthefictionsofliterallymanipulatingdensityfunctionsandofliterallymultiplyingutilitiesbybeliefsareneededforformulatingthebehaviouralassumption.
Itwouldbemisleadingtoassertthatvoterscomputeprobabilitiesormultiplythembyutilities.
ThesefictionsarebestcomparedtousingphysicaltheorytodeterminethetrajectoriesofbilliardballsinFriedman'sexample:itisknownthatthesefictionsarenottheunderlyingcausesofthephenomenaofinterest.
IfMkiweretoadoptthissecondlineofreasoning,hewouldendupinadilemma:hewouldeitherhavetoexplainwhyusingphysicaltheorydoesnotprovideafictionalaccountofbilliardsplayers'intentionsorshowhowFriedman'sexampleisrelevantlydifferentfrommysignal-extractionexample.
5Non-committal-indicatingas-ifclaimsDennett(1996,p.
27)suggeststhattheintentionalstanceisthestrategyofinterpretingthebehaviourofanentity(person,animal,artefact)bytreatingitasifitwerearationalagentwhogovernedits'choice'of'action'through'consideration'ofits'beliefs'and'desires'.
Heretheas-ifclaimexpressesanon-committalattitudetowardswhethertheentityinquestionhasbeliefsanddesires.
Notethateventhoughtheas-iflocutionisusedtobrackettheexistenceofbeliefsanddesires,theassertionaboutbehaviourisintendedtobeaccurate.
ThisiswhatmakesDennett'suseoftheas-iflocutiondifferentfromthatofinaccurateas-ifclaims(casea1).
Giventhatathermostatreallybehavesasifitwantedtokeepthetemperaturefixed,suchas-ifclaimsmaybeusedtodescribebehaviouraccuratelyeveniftheattribution(thermostatshavedesires)weretobeblatantlyfalse.
Whatmakesthebehaviouralclaimaboutthermostatstrueisthattheyhavebeendesignedtodojustthat:keepthetemperaturefixed.
Whentheintentionalstanceisappliedtorealpeople,therelevantattributionisthat'peoplehavedesiresandbeliefs'.
Forsomeonepersuadedthatpeoplehavedesiresandbeliefs,thiswouldseemtoprovidecounterexampletothethesisthatattributionsarealwaysfalse.
However,Dennettwouldusetheas-iflocutionevenifhisaudiencethoughtthatpeoplehadbeliefsanddesires.
Hewouldsaythatthisattributioncannotbeproventobetrue,andthattheas-iflocutionisusedpreciselytoindicatethattheexistenceofdesiresandbeliefsisbracketed.
Thusitisultimatelynotthefalsityoftheattributionthatpragmaticallyjustifiesusingtheas-iflocutionbutratherwhethertheuttererofthelocutioniscommittedtoitstruth.
Letusnowconsideranexamplefromeconomicswhichismoredifficulttointerpret.
(c)OrthodoxrevealedpreferencetheorythenprovidesconsistencyconditionsforEve'sbehaviorinAtobedescribedbysayingthatshechoosesasthoughmaximizingtheexpectedvalueofautilityfunctiondefinedonC,relativetoasubjectiveprobabilitydistributiondefinedonB(Binmore1998,pp.
360-1).
Inordertoseetherelevantattributionforthisclaim,itisnecessarytolookatthenotionofutilitymoreclosely.
Classicalutilitariansusedtheterm'utility'todenoteeithera20mentalstate(degreeofsatisfaction)orafeeling(e.
g.
,pleasure).
'Maximisingutility'couldthenbeunderstoodasanintentionalactivitywherebyanagentendeavourstomaximisethedegreetowhichhisorherdesiresaresatisfied.
If'utility'weretobeinterpretedinsuchamanner,onewouldsaythattheattributionisthat'Evemaximisesdesiresatisfaction',whichwouldbetrueifsheactuallymaximisedherdesiresatisfaction.
NotethatifEveactuallymaximisedutilityinthissense,suchaninterpretationwouldconstituteacounterexampletotheclaimthattheattributionisalwaysfalse.
Here,however,Binmorewouldusetheas-iflocutionindescribingEve'sactionsevenifEvedeliberatelyandconsciouslyendeavouredtomaximisesatisfaction.
Bywayofillustration,considerthewayinwhich'utility'and'maximisation'areinterpretedinmodernutilitytheory,whichessentiallyconsistsofrepresentationtheorems.
17Suchtheoremsspecifythatifpreferencessatisfyasetofconsistencyconditions,thentheymayberepresentedwithutilityfunctions.
LetPidenoteindividuali'sstrictpreferenceordering.
Thestandardinterpretationisthatforsomealternativesxandy,xPiymeansthaticonsidersxbetterthany.
TherepresentationtheoremthenstatesthatUi(x)>Ui(y)ifandonlyifxPiy.
Inotherwords,functionUassignsahighernumbertoxthantoyiftheindividualstrictlyprefersxtoy.
Itcouldthenbesaidthattheindividual'sbehaviourisdescribedasifheorsheweremaximisingautilityfunction.
Thetheoryfeaturesutilityfunctionsthatmaybeusedtodescribethebehaviourofanindividualwhosepreferencessatisfytheconditions.
Sometimestheterm'utility'or'util'isusedforthevaluesofthesefunctions.
Giventhat'utility'referstoanumberhere,itconsistsneitherinmaterialwelfarenorinanykindofmentalstatesuchaspleasure.
Binmorewouldbemakinganonsensicalstatementthatinvolvedacategorymistakeifheweretodroptheas-iflocution.
Itisnonsensicaltosaythat'apersontriestomaximiseutility'whenutilityisinterpretedinitsmodernsenseasadescriptionoftheperson'spreferencesorchoices.
Maximisationisnotamentaloperationcarriedoutbythepersonwhosebehaviourisdescribed,itisratherapropertyofamathematicalmodelthatrepresentsthatperson'sbehaviourintermsofmaximisation.
Thecategorymistakewouldthusconsistinstatingthat'apersonaimsto(purposefully)maximisehisorherutilityfunction'.
However,theperson'sactionscouldbedescribedasifheorsheweremaximisingautilityfunctionbecausethemorepreferred(orchosen)outcomesareassignedhigherutilitiesthanthelesspreferred.
AsBinmorestates,'ourstorymakesitnonsensetosaythatAdamchoosestheformerbecauseitsutilityisgreater',(Binmore2009,pp.
19-21).
Sayingsowouldamounttocommittingthe'causalutilityfallacy'.
Inotherwords,utilitydoesnotprovideanyreasonsforchoosingonewayortheother,andmerelyrepresentsanindividual'spreferences.
17Theoriginalcontributionofthissectionislimitedtotheroleofas-iflocutionsinexpectedutilitytheory.
TheideaspresentedherewerefirstclearlyformulatedbyLuce&Raiffa(1957),andsimilarargumentsarepresentedinKavka(1991)andLehtinen&Kuorikoski(2007).
21BinmorecouldnotusethelocutioninordertoclaimthatactualbehaviourcorrespondedtothatwhichwouldensueifEvemaximisedherdesiresatisfactionbecauseheclaimsthatmoderndecisiontheory'makesavirtueofassumingnothingwhateveraboutthepsychologicalcausesofourchoicebehavior'(Binmore2009,pp.
8-9).
Representationtheoremsdonotsayanythingaboutwhypeoplearesupposedtomaximiseutility.
Unlikeincase(b),BinmoredoesnotattributefalseintentionstoEveinordertodescribeherbehaviouraccurately.
NeitherdoesBinmorequalifythebehaviouralclaimwithanas-iflocutionbecausehedeniesthatEvemaximisesutility(inthemodernsense).
HehasalreadyassumedthatEveisrational:Althoughtherevealedpreferenceinterpretationofutilitywillbemaintainedthroughoutthisbook,halfofthelaborofconstructingautilityfunctionfromanagent'schoicebehaviorwillbeskipped.
Itwillbeassumedthatapreferencerelationhasalreadybeenconstructedandthatitremainsonlytoshowthatitcanberepresentedusinganappropriateutilityfunction(Binmore1994,p.
268).
Eventhoughhereadilyacceptselsewherethat'humanbehaviorisoftendownrightirrational'(Binmore2005,p.
75),hecannotdoubtEve'srationalityherebecausepreferencerelationscannotbeconstructedatallunlessindividualsarerational.
Thelocutionisthususedinordertoformulateanaccuratebehaviouralassumption.
Onemightproposethatmaximisingexpectedutilityistobeinterpretedasexpectedutilitydeliberation.
'Evemaximisesexpectedutility'wouldthenbetherelevantattribution,andmaximisationwouldbeinterpretedintermsofthismentaloperation.
Thiswouldnotbecorrecteither,however,becausethecriterionoftruthfortheas-ifclaim'Eve'sbehaviourcanbedescribedasifsheweremaximisingtheexpectedvalueofautilityfunction'isnotwhetherherbehaviourcorrespondstothatwhichwouldensueifsheengagedinexpected-utilitydeliberation.
Itisratherwhetherherpreferencessatisfytheconditionsforrepresentingutility.
Whetherornotaperson'sactionscouldbedescribedasifheorsheweremaximisingexpectedutilitydoesnotdependonwhetherheorsheengagesinsomekindofmentaloperation.
Similarly,whetherornotamodelincorporatingtheexpectedutilityassumptionisaccuratedoesnotdependonwhetherthepersonistryingtomaximisesomethingorwhetherheorsheisengaginginexpected-utilitydeliberation,butonlyonwhetherhisorherpreferencessatisfytheconditions.
Itakethistobethestandardpositionamongdecisiontheorists,butIwillprovideasimpleargumentinfavourofit:Evemayviolatetheconditionsevenifsheisengagedinexpected-utilitydeliberationif,forexample,shemakeserrorsincalculatingexpectedutilities.
Intheprevioussectiontheattributionwas:(b*)Votersobtainperturbedsignalsconcerningthepreferenceprofileandcomputeprobabilitiesfromthesesignalsusingstatisticalreasoning.
22Amorefullyspecifiedattributionwouldhavebeen:(b**)Votersobtainperturbedsignals,computeprobabilities,andliterallycalculateexpectedutilitiesbyweighingutilityvalueswithprobabilities.
Literallycalculatingexpectedutilitiesbymultiplyingprobabilitiesandutilitiesexpressedinseven-digitnumbersisthusamentaloperationthatisalsofalselyattributedtovoters.
Giventhat(b)and(c)bothconcerncaseswithexpectedutilitymaximisation,onemightarguethattheyarenotdifferentenoughtojustifyputtingthemintoseparatecategories.
Lehtinen(2011)posits,however,thatthecredibilityofhismodelcruciallydependsontheplausibilityoftheassumptionthatvotersreallyengageinDownsiandeliberation,andthatwhetherornottheirpreferencessatisfytheconsistencyconditionsmattersverylittleinthesensethattheresultsofthemodeldonotdependonit.
Infact,itseemsthatsatisfyingtheconsistencyconditionsisnotsufficienttoguaranteethatvotersengageinDownsiandeliberation.
ThisisatleastwhatoneiscompelledtothinkifoneacceptsBinmore'sargumentthatmodernutilitytheorymakesnopsychologicalassumptions.
Completelydifferentthingsarerequiredinthetwocasesfortheas-ifclaimsconcerningexpectedutilitymaximisationtobetrue.
Incase(c),theclaimistrueiftheindividualsatisfiestheconsistencyconditions.
Incase(b),however,whetherornotitistrueenoughdependsonwhetherornotvotersengageinDownsiandeliberation.
Thesimilaritybetween(b)and(c)liesinthefactthat,inbothcases,themodellerusesanas-iflocutioninordertoclarifywhatkindofclaimsaboutbehaviourandmentalstatesaremade.
Binmoreusestheas-iflocutioninordertoalertthereadertothefactthat'maximisation'doesnotrefertoanykindofmentaloperation.
Heavoidsmakingthefalseclaimthatmaximisationissuchanoperationbyusingthelocution.
Letusnowreturntoc).
Oncetherightattributionisspecified,itisclearthatitisindeedalwayseitherfalseorinvolvesacategorymistake.
Isuggesttwopossibilities.
ThefirstisthatitconcernsEve'sintentions:'Eveendeavourstomaximiseexpectedutility'whenutilityisunderstoodinitsmodernsenseasanumberdescribingpreferences.
Thisattributionincorporatesacategorymistake.
ItisnotaconsequenceofanempiricalfailuretoactinacertainmannerbutratherfromthefactthatutilityisnotsomethingthatEvecantrytomaximiseasamatterofthelogicalformoftheterm'utility'.
ThesecondpossibilityisthatitconcernsBinmore'sbeliefsconcerninghisaudience.
Hecouldbetakentoassumethatnoteveryoneinhisaudienceknowsthattheterm'utility'willbeusedinitsmodernsense.
Hewouldthusbeusingtheas-iflocutionbecausehebelievesthatsomepeopleinhisaudiencemightthinkthat'utilitymeanspleasureinthetheorythatBinmoreispresenting'.
Thisisagainfalse.
Itisalsoevidentthatevaluatingthetruthofeitheroftheseattributionsisentirelyirrelevanttowhetherornottheas-ifclaimisaccuratebecauseneithersaysanythingabouttheconsistencyconditions.
Onemightarguethatifthefirstattributionisappropriate,claimslike(c)shouldbetreatedasaseparatefourthkindofas-ifclaimonthegroundsthatavoidingcategorymistakesisnotintrinsicallyrelatedtoexpressingnon-committaltoanyparticular23determinantsofbehaviour.
Infurthersupportofsuchanargumentonecouldpointoutthattheabovequotationfrom1994commitsBinmoretoEve'srationality,andthatrationalityisadeterminantofbehaviour.
Iresistsuchaninterpretationmainlybecauserationalityisnottherightkindofdeterminant.
Insofarasthe'preferencerelationhasalreadybeenconstructed',thebehaviourhasalsoalreadytakenplace,andifitisdescribedintermsofutilityfunctions,itdoesnotmatterwhatitsdeterminantswere.
Furthermore,onceoneunderstandswhyBinmoreusestheas-iflocution,onealsounderstandshowavoidingcategorymistakesamountstobeingnon-committalaboutthedeterminantsofbehaviour.
As-ifclaimsusedinexpectedutilitytheorymaywellprovideaccurateaccountsofbehaviour.
However,heretheuseoftheas-iflocutiontellsusabsolutelynothingaboutwhetherornotthisisthecase.
Ithasnothingtodowiththequestionofwhetherornotitisaccuratetodescribepeople'sactualbehaviourasutility-maximising.
Thisquestioncanbeevaluatedinmanydifferentways,ofwhichoneistoevaluatewhetherpeopleinsomeactualsituationstrytomaximisepleasure.
Thisissobecausethereasonabilityoftheconsistencyconditionsdependsonhowrealpeopleact,andoneoftherelevantconsiderationsinassessingwhethertheirbehavioursatisfiestheconditionsiswhetherornottheytrytomaximisepleasure.
Insomecircumstancestheydosatisfytheconditions,inotherstheydonot,butinanycaseoneneedstousetheas-iflocutionevenwhenexpectedutilitymaximisationisanaccurateassumption.
Thisiswhatmakes(c)differentfromcase(a1).
Ifonewishedtosaythatthedifferencebetweenactualbehaviourandthatspecifiedbyexpectedutilitytheoryisnegligible,onewouldhavetosaythat'peopleactasiftheywereactingasiftheymaximisedexpectedutility'.
Naturally,suchclaimsarenevermadebecausetherearemoreelegantwaysofconveyingtheidea.
Onecouldsay,forexample,that'Eveactsasifshewereautilitymaximiser'.
18Thisisbecausetherelevantattribution'Eveisautilitymaximiser'isfalseifshedoesnotactaccordingtotheconditions.
Somedecisiontheorists,includingBinmore,apparentlythinkthatmerechoicesratherthanpreferencesultimatelyneedtobeconsistentforthetheoryofexpectedutilitytoapply.
Onecouldthenstatethat'anagentactsasifheorshehadconsistentpreferences',andthathisorherchoicescouldbedescribedintermsofutilityfunctions.
Heretheveryexistenceofpreferencesisbracketed.
Notethatifexpectedutilitytheoryisinterpretedinsuchrevealed-preferenceterms,statingthat'anagentactsasifheorshehadconsistentchoices'wouldmeanthattheas-ifclaimistobeinterpretedasindicatinginaccuracy(a1).
Incases(b)and(c)thereasonforusingtheas-iflocutionisthatitallowsonetospecifyexactlywhatonewantstosayaboutbehaviouralassumptions.
DennettandBinmoreuseittoemphasisethatnointentionalascriptionsneedtobemade,whereasLehtinenusesittoformulatearealisticbehaviouralassumptionandtoshowthatthecredibilityofthemodeldoesnothingeontheassumptionthatvotersactuallyengageincomputingexpectedutilitiesorreceivenumericallyprecisesignals.
Thesecasesaresimilarinthat18'Eveisamaximiser'hereispresupposedtomeanthatshesatisfiesthevNMconditions.
24usingtheas-iflocutionallowsthemodellertoavoidmakingsomeparticularmentalattributions.
6ConclusionsAs-iflocutionscanbeusedinavarietyofways.
Ihavearguedthattheycanbeusedto(a1)indicatethatthebehaviouralclaimtheyqualifyisinaccuratebutnegligiblyso,and(a2)toindicatethatthemodelleracknowledgestheexistenceofsomecausalfactorsthatarenotexplicitlytakenintoaccountinthemodelbecausetheyaredeemedtobenegligible.
Inthemostgenuinecases,theyareusedfor(b)expressingthecontentofaccuratebehaviouralclaimsbyascribingintentionsonentitiesinanunrealisticmanner,and(c)forindicatingthatdescriptiveclaimsmadeaboutbehaviourdonotcommitonetoanyparticularmentalassumptions.
Theclaimscouldthusbeputbrieflyasfollows.
(a1)Inaccuracy-indicatingclaims.
AIC-orrather,AdoesnotactuallybehaveasB,butthatisunimportantinthismodel.
(a2)Negligibility-indicatingaccurateclaims.
AICbecauseAislikeBinsomerespectsbutnotinall.
Theignoredrespectsarenotimportantinthismodel.
(b)Claimsindicatingfalsementalascription.
AICbecauseAdoesnothave(mental)propertiesBanddoeshavepropertiesC.
YetitsbehaviourcouldbedescribedundertheassumptionthatithaspropertiesB.
(c)Claimsindicatingnon-committal.
AICbecause,althoughAhasnothingtodowithB,AbehavesasBwhatevertheactualdeterminantsofA'sbehaviour.
However,as-ifclaimsalsohavesomecommonalities.
Firstly,theirmostimportantfunctionistohelpinspecifyingwhichpartofamodeloughttobetrueandwhatcanbeallowedtobefalse.
Secondly,theywouldappeartobeinvariablyrelatedtomakingclaimsaboutordescribingbehaviour.
Thirdly,theyallexpresslackofcommitmenttothetruthoftheirunderlyingattributions.
Fourthly,inthemselvestheyarealwaysmethodologicallyincompleteinthesensethattheydonotprovideanyexplanationofwhybehaviourshould,inreality,correspondtothatspecifiedinthemodel.
As-ifclaimsarelikelytocreateagreatdealofconfusion.
Inviewoftheirincompletenessitisalwaysnecessarytofindoutwhetherornotagivenoneisrealistic.
Furthermore,itmaybedifficulttoseethedifferencebetweenthedifferentkindsofclaim.
Onecouldeasilybemisledintobelievingthatanas-ifclaimisrealisticevenwhenbehaviourisnotrealisticallymodelled,orthatallas-ifclaimsarefalsebecausetheirattributionsarealwaysfalse.
Friedmanusesthelocutionforallofthereasonsdescribedinthispaper.
Giventhatthedifferentusesarebasedondifferentmethodologicalcommitmentsandleadtodifferentwaysofevaluatingthetruthoftheclaims,itisnotclearthatitmakesanysensetotalkabout'Friedman'sas-ifmethodology'19.
Somepeoplemayprefertominimisetheuseofas-ifclaimsjustbecausetheyaresoconfusingandincomplete.
Inallthecasesinwhichtheyinvolveasemanticchange,theyclearlyhaveausefulfunction.
Usuallythechangeinvolvestransformingafalse19I,too,havebeenguiltyofusingthisexpression(LehtinenandKuorikoski2007,p.
123).
25attributionintoatrueenoughas-ifclaim.
Yet,theyarenevernecessary.
Ifonemerelywishestoexpresstheideathattherealisticnessofanassumptiondoesnotmatter,orthatoneisawareofmakinganunrealisticassumption,Isuggestthatitwouldbebettertoavoidusingtheas-iflocutionbecausetherearemorenaturalandlessconfusingwaysofmakingsuchclaims.
Whetherornotasemanticchangeisthusentailedcouldthenbetakentofacilitaterecognitionofthewarrantedusesofas-ifclaims.
Thispapermaybeconsideredsomewhatoddbecausetheauthorisusinghisownworkasaneconomistinaphilosophicalcasestudy.
However,themotivationforwritingthispaperarosefromtheexperienceofwritingthepapersonvotingandsignalextraction.
Economicjournalsdonotallowauthorstoexplaintheas-ifjustificationsoftheirassumptionsatlength.
The'as-if'locutionactuallyneverappearsinanyofmypapersonvotingandsignalextraction.
ItdoesnotappearbecauseIwastooafraidthatIwouldbeinterpretedaspromotingmymodelintermsofinaccuracy-indicatingas-ifclaims.
Thisinterpretationseemedtobeverycommonamongtheeconomists,andIwassodeeplyconcernedaboutbeingmisunderstoodthatIdecidednottousethelocutionatall.
TheseexperiencesalsoexplainwhyIhavenotbeencontentmerelytodescribehowas-iflocutionsareusedineconomics,butratheralsotriedtodelimittheirusewiththepragmaticconstraints:Ibelievethatamoreregimenteduseofthelocutionwouldhelpinavoidingconfusions.
Giventhat(a2),(b)and(c)representcasesinwhichtheas-ifclaimmaybeaccurateorevenrealistic,restrictingusagetothesethreekindsalsoencourageevaluationofthecriteriaastotheirtruth.
Themostimportantconsiderationhereiswhetherornotthetruthoftheattributionisrelevanttothetruthoftheas-ifclaim.
Theexampleofsignalextractionwasintendedtoshowthatinonecaseinwhichitisnotrelevant,i.
e.
whenitinvolvesafictionofafalseintentionalascription,itmaywellberealisticdespitebeingbasedonafiction,andthegroundsforevaluatingitstruthneednotcomefromamoregeneraltheory.
ReferencesBecker,G.
S.
(1962),"IrrationalBehaviorandEconomicTheory,"JournalofPoliticalEconomy,70(1),1-13.
Binmore,K.
(2009),RationalDecisions,Princeton;Oxford:PrincetonUniversityPress.
Binmore,K.
(2005),NaturalJustice,Oxford;NewYork:OxfordUniversityPress.
Binmore,K.
(1998),GameTheoryandtheSocialContract:JustPlaying(Vol.
2),London:TheMITPress.
Binmore,K.
(1994),GameTheoryandtheSocialContract:PlayingFair(Vol.
1),Cambridge,Mass.
:MITPress.
Bokulich,A.
(2009),"ExplanatoryFictions,"inFictionsinScience;PhilosophicalEssaysonModelingandIdealization,ed.
M.
Suárez,NewYork,London:Routledge,pp.
91-109.
Boland,L.
A.
(1979),"ACritiqueofFriedman'sCritics,"JournalofEconomicLiterature,17(2),503-522.
26Boylan,T.
A.
andO'Gorman,P.
F.
(1995),BeyondRhetoricandRealisminEconomics:TowardsaReformulationofEconomicMethodology,London:Routledge.
Coddington,A.
(1979),"Friedman'sContributiontoMethodologicalControversy,"BritishReviewofEconomicIssues,2(4),1-13.
Dennett,D.
C.
(1996),KindsofMinds:TowardsanUnderstandingofConsciousness,London:Weidenfeld&Nicolson.
Downs,A.
(1957),AnEconomicTheoryofDemocracy,NewYork:Harper.
Elgin,C.
Z.
(2009),"Exemplification,Idealization,andScientificUnderstanding,"inFictionsinScience;PhilosophicalEssaysonModelingandIdealization,ed.
M.
Suárez,NewYork,London:Routledge,pp.
77-90.
Elgin,C.
Z.
(2004),"Trueenough,"PhilosophicalIssues,14(1),113-131.
Enelow,J.
M.
(1981),"SavingAmendments,KillerAmendments,andanExpectedUtilityTheoryofSophisticatedVoting,"JournalofPolitics,43(4),1062-1089.
Fine,A.
(1993),"Fictionalism,"MidwestStudiesinPhilosophy,XVIII,1-18.
Friedman,M.
(1953),"TheMethodologyofPositiveEconomics,"inEssaysinPositiveEconomics,Chicago:UniversityofChicagoPress,pp.
3-43.
Friedman,M.
andSavage,L.
J.
(1952),"TheExpected-UtilityHypothesisandtheMeasurabilityofUtility,"JournalofPoliticalEconomy,60(6),463-474.
Friedman,M.
andSavage,L.
J.
(1948),"TheUtilityAnalysisofChoicesInvolvingRisk,"JournalofPoliticalEconomy,56(4),279-304.
Hindriks,F.
A.
(2006),"TractabilityAssumptionsandtheMusgrave-MkiTypology,"JournalofEconomicMethodology,13(4),401-423.
Hoover,K.
D.
(2009),"MiltonFriedman'sStance:TheMethodologyofCausalRealism,"inTheMethodologyofPositiveEconomics:ReflectionsontheMiltonFriedmanLegacy,ed.
U.
Mki,Cambridge:CambridgeUniversityPress,pp.
303-320.
Kavka,G.
(1991),"RationalMaximizinginEconomicTheoriesofPolitics,"inTheEconomicApproachtoPolitics:ACriticalReassessmentoftheTheoryofRationalAction,ed.
K.
R.
Monroe,NewYork:HarperCollins,pp.
371-385.
Kirman,A.
P.
(2006),"DemandTheoryandGeneralEquilibrium:FromExplanationtoIntrospection,aJourneyDowntheWrongRoad,"HistoryofPoliticalEconomy,38,246-280.
Kuorikoski,J.
,Lehtinen,A.
andMarchionni,C.
(2010),"EconomicModellingasRobustnessAnalysis,"BritishJournalforthePhilosophyofScience,61(3),541-567.
Lagueux,M.
(1994),"Friedman's'Instrumentalism'andConstructiveEmpiricisminEconomics,"TheoryandDecision,37(2),147-174.
Lehtinen,A.
(2011),"AWelfaristCritiqueofSocialChoiceTheory,"JournalofTheoreticalPolitics,23(3),359-381.
Lehtinen,A.
(2010),"BehaviouralHeterogeneityUnderApprovalandPluralityVoting,"inHandbookonApprovalVoting,eds.
J.
LaslierandM.
R.
Sanver,Heidelberg:Springer,pp.
285-310.
Lehtinen,A.
(2008),"TheWelfareConsequencesofStrategicBehaviourUnderApprovalandPluralityVoting,"EuropeanJournalofPoliticalEconomy,24,688-704.
Lehtinen,A.
(2007),"TheWelfareConsequencesofStrategicVotinginTwoCommonlyusedParliamentaryAgendas,"TheoryandDecision,63(1),1-40.
Lehtinen,A.
(2006),"SignalExtractionforSimulatedGameswithaLargeNumberofPlayers,"ComputationalStatisticsandDataAnalysis,50,2495-2507.
27Lehtinen,A.
andKuorikoski,J.
(2007),"UnrealisticAssumptionsinRationalChoiceTheory,"PhilosophyoftheSocialSciences,37(2),115-138.
Luce,D.
R.
andRaiffa,H.
(1957),GamesandDecisions;IntroductionandCriticalSurvey,NewYork:Wiley.
MacDonald,P.
K.
(2003),"UsefulFictionOrMiracleMaker:TheCompetingEpistemologicalFoundationsofRationalChoiceTheory,"AmericanPoliticalScienceReview,97(4),551-565.
Machlup,F.
(1946),"MarginalAnalysisandEmpiricalResearch,"TheAmericanEconomicReview,36(4),519-554.
Mki,U.
(2009),"UnrealisticAssumptionsandUnnecessaryConfusions:RereadingandRewritingF53asaRealistStatement,"inTheMethodologyofPositiveEconomics.
ReflectionsontheMiltonFriedmanLegacy,ed.
U.
Mki,Cambridge:CambridgeUniversityPress,pp.
90-116.
Mki,U.
(2004),"RealismandtheNatureofTheory:ALessonfromJHVonThünenforEconomistsandGeographers,"Environment&PlanningA,36(10),1719-1736.
Mki,U.
(2003),"'TheMethodologyofPositiveEconomics'(1953)doesNotGiveUstheMethodologyofPositiveEconomics,"JournalofEconomicMethodology,10(4),495-505.
Mki,U.
(2000),"KindsofAssumptionsandtheirTruth:ShakinganUntwistedF-Twist,"Kyklos,53(3),317-335.
Mki,U.
(1998),"Asif,"inTheHandbookofEconomicMethodology,eds.
J.
B.
Davis,D.
W.
HandsandU.
Mki,Cheltenham:EdwardElgar,pp.
25-27.
Mki,U.
(1992),"FriedmanandRealism,"ResearchintheHistoryofEconomicThoughtandMethodology,10,171-195.
McKelvey,R.
D.
andOrdeshook,P.
C.
(1972),"AGeneralTheoryoftheCalculusofVoting,"inMathematicalApplicationsinPoliticalScience(Vol.
IV),eds.
J.
F.
HerndonandJ.
L.
Bernd,Charlottesville:UniversityPressofVirginia,pp.
32-78.
McQuarrie,D.
A.
andSimon,J.
D.
(1997),PhysicalChemistry:AMolecularApproach,Sausalito:UniversityScienceBooks.
Mill,J.
S.
(2000[1844]),EssaysonsomeUnsettledQuestionsofPoliticalEconomy,(2nded.
),Kitchener:Batochebooks.
Morrison,M.
(2009),"Fictions,Representations,andReality,"inFictionsinScience;PhilosophicalEssaysonModelingandIdealization,ed.
M.
Suárez,NewYork,London:Routledge,pp.
110-135.
Moscati,I.
andTubaro,P.
(2011),"BeckerRandomBehaviorandtheas-ifDefenseofRationalChoiceTheoryinDemandAnalysis"JournalofEconomicMethodology,18(2),107-128.
Musgrave,A.
(1981),"'UnrealAssumptions'inEconomicTheory:TheF-TwistUntwisted,"Kyklos,34(3),377-387.
Rgeberg,O.
andNordberg,M.
(2005),"ADefenceofAbsurdTheoriesinEconomics,"JournalofEconomicMethodology,12(4),543-562.
Satz,D.
andFerejohn,J.
A.
(1994),"RationalChoiceandSocialTheory,"JournalofPhilosophy,91(2),71-87.
Sen,A.
K.
(1973),"BehaviourandtheConceptofPreference,"Economica,40(159),241-259.
28Strevens,M.
(2008),Depth:AnAccountofScientificExplanation,Cambridge,Mass.
;London:HarvardUniversityPress.
Teller,P.
(2009),"Fictions,Fictionalization,andTruthinScience,"inFictionsinScience;PhilosophicalEssaysonModelingandIdealization,ed.
M.
Suárez,NewYork,London:Routledge,pp.
235-247.
Vromen,J.
J.
(2009),"Friedman'sSelectionArgumentRevisited,"inTheMethodologyofEconomics.
ReflectionsontheMiltonFriedmanLegacy,ed.
U.
Mki,Cambridge:CambridgeUniversityPress,pp.
257-284.