资源描述:
《[博弈论书籍].the.theory.of.learning.in.games,drew fudenberg,david k. levine,the mit press》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、11.Introduction1.1.IntroductionThisbookisaboutthetheoryoflearningingames.Mostofnon-cooperativegametheoryhasfocusedonequilibriumingames,especiallyNashequilibrium,anditsrefinementssuchasperfection.Thisraisesthequestionofwhenandwhywemightexpectthatobservedplayinagamewillc
2、orrespondtooneoftheseequilibria.Onetraditionalexplanationofequilibriumisthatitresultsfromanalysisandintrospectionbytheplayersinasituationwheretherulesofthegame,therationalityoftheplayers,andtheplayers’payofffunctionsareallcommonknowledge.Bothconceptuallyandempirically,
3、these1theorieshavemanyproblems.Thisbookdevelopsthealternativeexplanationthatequilibriumarisesasthelong-runoutcomeofaprocessinwhichlessthanfullyrationalplayersgropeforoptimalityovertime.Themodelswewilldiscussservetoprovideafoundationforequilibriumtheory.Thisisnottosayth
4、atlearningmodelsprovidefoundationsforalloftheequilibriumconceptsintheliterature,nordoesitarguefortheuseofNashequilibriumineverysituation;indeed,insomecasesmostlearningmodelsdonotleadtoanyequilibriumconceptbeyondtheveryweaknotionofrationalizability.Nevertheless,learning
5、models1First,amajorconceptualproblemoccurswhentherearemultipleequilibria,forintheabsenceofanexplanationofhowplayerscometoexpectthesameequilibrium,theirplayneednotcorrespondtoanyequilibriumatall.Whileitispossiblethatplayerscoordinatetheirexpectationsusingacommonselectio
6、nproceduresuchasHarsanyiandSelten’s[1988]tracingprocedure,leftunexplainedishowsuchaprocedurecomestobecommonknowledge.Second,wedoubtthatthehypothesisofexactcommonknowledgeofpayoffsandrationalityapplytomanygames,andrelaxingthistoanassumptionofalmostcommonknowledgeyieldsm
7、uchweakerconclusions.(Seeforexample.DekelandFudenberg[1990],Borgers[1994].)Third,equilibriumtheorydoesapoorjobexplainingplayinearlyroundsofmostexperiments,althoughitdoesmuchbetterinlaterrounds..Thisshiftfromnon-equilibriumtoequilibriumplayisdifficulttoreconcilewithapur
8、elyintrospectivetheory.2cansuggestusefulwaystoevaluateandmodifythetraditionalequilibriumconcepts.Learningmodelsleadto