人在面对机器与人时的决策差异

SocialnormsandhumancooperationErnstFehrandUrsFischbacherUniversityofZurich,InstituteforEmpiricalResearchinEconomics,Blu¨mlisalpstrasse10,CH-8006Zurich,SwitzerlandTheexistenceofsocialnormsisoneofthebigunsolvedproblemsinsocialcognitivescience.Althoughnootherconceptisinvokedmorefrequentlyinthesocialsciences,westillknowlittleabouthowsocialnormsareformed,theforcesdeterminingtheircontent,andthecognitiveandemotionalrequirementsthatenableaspeciestoestablishandenforcesocialnorms.Inrecentyears,therehasbeensubstantialprogress,however,onhowcooperationnormsareenforced.Herewereviewevidenceshowingthatsanctionsaredecisivefornormenforcement,andthattheyarelargelydrivenbynon-selfishmotives.Moreover,theexplicitstudyofsanc-tioningbehaviorprovidesinstrumentsformeasuringsocialnormsandhasalsoledtodeeperinsightsintotheproximateandultimateforcesbehindhumancooperation.Humansocietiesrepresentaspectacularoutlierwithrespecttoallotheranimalspeciesbecausetheyarebasedonlarge-scalecooperationamonggeneticallyunrelatedindividuals[1].Inmostanimalsocieties,cooperationiseitherordersofmagnitudelessdevelopedcomparedwithhumans,oritisbasedonsubstantialgeneticrelatedness.Cooperationinhumansocietiesismainlybasedonsocialnorms,includinginmodernsocieties,whereaconsider-ableamountofcooperationisduetothelegalenforcementofrules.Legalenforcementmechanismscannotfunctionunlesstheyarebasedonabroadconsensusaboutthenormativelegitimacyoftherules–inotherwords,unlesstherulesarebackedbysocialnorms.Moreover,theveryexistenceoflegalenforcementinstitutionsisitselfaproductofpriornormsaboutwhatconstitutesappropriatebehaviour.Thus,itisnecessarytoexplainsocialnormstoexplainhumancooperation.Socialnormsarestandardsofbehaviourthatarebasedonwidelysharedbeliefshowindividualgroupmembersoughttobehaveinagivensituation[2–5].Thegroupinwhichsocialnormsprevailcanbeafamily,apeergroup,anorganizationorevenawholesociety.Thegroupmembersmightobeythenormvoluntarilyiftheirindividualgoalsareinlinewiththenormativelyrequiredbehaviour,ortheymightbeforcedtoobeythenormbecausetheirindividualgoalsdifferfromthenormativelyrequiredbehavior,inwhichcasetheenforcementofthenormpresupposesthatnormviolationsarepunished.Accordingtoawidelyshared[6]butnotuncontested[7]view,thedemandforasocialnormariseswhenactionscausepositiveornegativeside-effectsforotherpeople.Environ-mentalpollution,oranindividualemployee’seffortwhentheteamispaidaccordingtotheteam’stotaloutput,areexamplesthatleadtosuchside-effects.Intheteamcase,amemberwhocontributestotheteam’soutputalsoraisesthepayoftheothermembers.Itisthereforeintheinterestofeachindividualteammemberthattheothermembersworkhard.Thisinterestintheothers’actionsisallegedtocreatethedemandforasocialnorm.Positiveornegativeside-effectsofindividualactionstypicallygiverisetoacooperationor‘public-good’problem.Thedefiningcharacteristicofapublicgood,suchascleanairorteamoutput,isthatnogroupmembercanbeexcludedfromtheconsumptionofthegood.Therefore,allpartiesarebetteroffifthepublicgoodisprovidedandthegroupmemberssharethecost,buteachindividualalsohasaneconomicincentivetofreeride;thatis,tocontributenothingtowardsprovidingthegood(seeBox1).Thismeansthatifthegroupmembersbehaveaccordingtotheireconomicincentives,theydonotcooperateand,therefore,thepublicgoodwillnotbeprovided.Itisclearthatasocialnormsuchas‘Youshouldnottakeadvantageofyourteammembersbyshirking’contributestotheprovisionofthepublicgood.Box1.PublicgoodsandPrisoners’DilemmaexperimentsBydefinition,allgroupmemberscanconsumeapublicgood,eventhosewhodonotbearthecostofprovidingthegood.Therefore,eachmemberhasanincentivetofreerideonthecontributionsofothers.Inapublicgoodexperiment,subjectsformgroupsofn$2individualsandeachindividualisgivenamonetaryendowmentE.SubjectssimultaneouslydecidehowmuchofEtheykeepforthemselvesandhowmuchtheyspendonagroupproject.Theexperimenterthenmultipliesthetotalamountspentonthegroupprojectbyanumber,denotedbyb,thatisgreaterthan1butsmallerthann.Themultipliedsumofthemember’scontributionconstitutestheproceedsfromthegroupproject.Theseproceedsarethendistributedequallyamongthenmembers.Thismeansthateachgroupmember,includingthecontributingsubject,earnsb/nMUsforeverymoneyunit(MU)spentontheproject;thisislessthan1becauseofb,n.Yet,thecon-tributingsubjecthasacostof1,meaningthataselfishsubjectnevercontributesanythingtotheprojectinaone-shotexperiment.Thispredictionholds,althoughitwouldbecollectivelyrationaltocon-tributeeverythingbecauseifallsubjectskeeptheirendowments,theyearnEMUseach,whereasifallcontributetheirendowmentsthesumofcontributionsisnE,yieldinganincomeof(b/n)nE¼bE,whichisgreaterthanEforeachgroupmemberbecauseb.1.Forexample,ifE¼20,b¼2,subjectsearn20ifnobodycontributesand40ifeverybodycontributesthewholeendowmenttothegroupproject.Formally,thePrisoners’Dilemma(PD)isaspecialcaseofthepublic-goodgamewithn¼2andtwoavailableactions:contributingnothing(i.e.defect)orcontributingeverything(i.e.cooperate).Therefore,eachplayerinthePDisbetteroffifhedefects(becauseb/n¼b/2,1)regardlessofwhattheopponentdoes.Correspondingauthor:ErnstFehr([e-mail])