维基百科 囚徒困境 英文版解释

Prisoner'sdilemmaFromWikipedia,thefreeencyclopediaJumpto:navigation,searchThisarticleisaboutgametheory.Forthe1988novel,seePrisoner'sDilemma(novel).FortheDoctorWhoaudiobook,seeThePrisoner'sDilemma.Forthe2001play,seeThePrisoner'sDilemma(play).Thisarticlehasanunclearcitationstyle.Thereferencesusedmaybemadeclearerwithadifferentorconsistentstyleofcitation,footnoting,orexternallinking.(October2012)Theprisoners'dilemmaisacanonicalexampleofagameanalyzedingametheorythatshowswhytwopurelyrationalindividualsmightnotcooperate,evenifitappearsthatitisintheirbestinterests[citationneeded]todoso.ItwasoriginallyframedbyMerrillFloodandMelvinDresherworkingatRANDin1950.AlbertW.Tuckerformalizedthegamewithprisonsentencerewardsandgaveitthenameprisoner'sdilemma(Poundstone,1992),presentingitasfollows:Twomembersofacriminalgangarearrestedandimprisoned.Eachprisonerisinsolitaryconfinementwithnomeansofspeakingtoorexchangingmessageswiththeother.Thepoliceadmittheydon'thaveenoughevidencetoconvictthepairontheprincipalcharge.Theyplantosentencebothtoayearinprisononalessercharge.Simultaneously,thepoliceoffereachprisoneraFaustianbargain.Eachprisonerisgiventheopportunityeithertobetraytheother,bytestifyingthattheothercommittedthecrime,ortocooperatewiththeotherbyremainingsilent.Here'showitgoes:IfAandBbothbetraytheother,eachofthemserves2yearsinprisonIfAbetraysBbutBremainssilent,AwillbesetfreeandBwillserve3yearsinprison(andviceversa)IfAandBbothremainsilent,bothofthemwillonlyserve1yearinprison(onthelessercharge)Itisimpliedthattheprisonerswillhavenoopportunitytorewardorpunishtheirpartnerotherthantheprisonsentencestheyget,andthattheirdecisionwillnotaffecttheirreputationinthefuture.Becausebetrayingapartneroffersagreaterrewardthancooperatingwiththem,allpurelyrationalself-interestedprisonerswouldbetraytheother,andsotheonlypossibleoutcomefortwopurelyrationalprisonersisforthemtobetrayeachother.[1]Theinterestingpartofthisresultisthatpursuingindividualrewardlogicallyleadsbothoftheprisonerstobetray,whentheywouldgetabetterrewardiftheybothcooperated.Inreality,humansdisplayasystematicbiastowardscooperativebehaviorinthisandsimilargames,muchmoresothanpredictedbysimplemodelsofrationalself-interestedaction.[2][3][4][5]Amodelbasedonadifferentkindofrationality,wherepeopleforecasthowthegamewouldbeplayediftheyformedcoalitionsandthentheymaximizetheirforecasts,hasbeenshowntomakebetterpredictionsoftherateofcooperationinthisandsimilargamesgivenonlythepayoffsofthegame.[6]Thereisalsoanextendediteratedversionofthegame,wheretheclassicgameisplayedoverandoverbetweenthesameprisoners,andconsequently,bothprisonerscontinuouslyhaveanopportunitytopenalizetheotherforpreviousdecisions.Ifthenumberoftimesthegamewillbeplayedisknowntotheplayers,then(bybackwardinduction)twoclassicallyrationalplayerswillbetrayeachotherrepeatedly,forthesamereasonsasthesingleshotvariant.Inaninfiniteorunknownlengthgamethereisnofixedoptimumstrategy,andPrisoner'sDilemmatournamentshavebeenheldtocompeteandtestalgorithms.Theprisoner'sdilemmagamecanbeusedasamodelformanyrealworldsituationsinvolvingcooperativebehaviour.Incasualusage,thelabelprisoner'sdilemmamaybeappliedtosituationsnotstrictlymatchingtheformalcriteriaoftheclassicoriterativegames:forinstance,thoseinwhichtwoentitiescouldgainimportantbenefitsfromcooperatingorsufferfromthefailuretodoso,butfinditmerelydifficultorexpensive,notnecessarilyimpossible,tocoordinatetheiractivitiestoachievecooperation.Contents[hide]1Strategyfortheclassicprisoners'dilemma2Generalizedformo2.1Specialcase:Donationgame3Theiteratedprisoners'dilemmao3.1Strategyfortheiteratedprisoners'dilemmao3.2Stochasticiteratedprisoner'sdilemma3.2.1Zero-determinantstrategieso3.3Continuousiteratedprisoners'dilemmao3.4EmergenceofStableStrategies4Real-lifeexampleso4.1Inenvironmentalstudieso4.2Inanimalso4.3Inpsychologyo4.4Ineconomicso4.5Insporto4.6Multiplayerdilemmaso4.7Armsraces5Relatedgameso5.1Closed-bagexchangeo5.2FriendorFoe?o5.3Iteratedsnowdrift6Seealso7References8Furtherreading9ExternallinksStrategyfortheclassicprisoners'dilemma[edit]Thenormalgameisshownbelow:PrisonerBstayssilent(cooperates)PrisonerBbetrays(defects)PrisonerAstayssilent(cooperates)Eachserves1yearPrisonerA:3yearsPrisonerB:goesfreePrisonerAbetrays(defects)PrisonerA:goesfreePrisonerB:3yearsEachserves2yearsHere,regardlessofwhattheotherdecides,eachprisonergetsahigherpay-offbybetrayingtheother(defecting).Thereasoninginvolvesanargumentbydilemma:Bwilleithercooperateordefect.IfBcooperates,Ashoulddefect,sincegoingfreeisbetterthanserving1year.IfBdefects,Ashouldalsodefect,sinceserving2yearsisbetterthanserving3.Soeitherway,Ashoulddefect.ParallelreasoningwillshowthatBshoulddefect.Intraditionalgametheory,someveryrestrictiveassumptionsonprisonerbehaviouraremade.Itisassumedthatbothunderstandthenatureofthegame,andthatdespitebeingmembersofthesamegang,theyhavenoloyaltytoeachotherandwillhavenoopportunityforretributionorrewardoutsidethegame.Mostimportantly,averynarrowinterpretationofrationalityisappliedindefiningthedecision-makingstrategiesofthepriso