Risk, Unexpected Uncertainty, and Estimation Uncertainty Bayesian Learning in Unstable Settings 英文参考文献.docVIP

Risk,UnexpectedUncertainty,andEstimation Uncertainty:BayesianLearninginUnstableSettings ElisePayzan-LeNestour1*,PeterBossaerts2,3 1UniversityofNewSouthWales,Sydney,Australia,2CaliforniaInstituteofTechnology,Pasadena,California,UnitedStatesofAmerica,3SwissFinanceInstituteatEPFL, Lausanne,Switzerland Abstract Recently, evidence has emerged that humans approach learning using Bayesian updating rather than (model-free) reinforcement algorithms in a six-arm restless bandit problem. Here, we investigate what this implies for human appreciationofuncertainty.Inourtask,aBayesianlearnerdistinguishesthreeequallysalientlevelsofuncertainty.First,the Bayesian perceives irreducible uncertainty or risk: even knowing the payoff probabilities of a given arm, the outcome remainsuncertain.Second,thereis(parameter)estimationuncertaintyorambiguity:payoffprobabilitiesareunknownand needtobeestimated.Third,theoutcomeprobabilitiesofthearmschange:thesuddenjumpsarereferredtoasunexpected uncertainty.We documenthowthethreelevelsofuncertainty evolvedduring thecourse ofourexperiment andhowit affectedthelearningrate.Wethenzoominonestimationuncertainty,whichhasbeensuggestedtobeadrivingforcein exploration,inspiteofevidenceofwidespreadaversiontoambiguity.Ourdatacorroboratethelatter.Wediscussneural evidence that foreshadowed the ability of humans to distinguish between the three levels of uncertainty. Finally, we investigate theboundaries ofhuman capacity toimplement Bayesianlearning. We repeat theexperiment withdifferent instructions, reflecting varying levels of structural uncertainty. Under this fourth notion of uncertainty, choices were no better explained by Bayesian updating than by (model-free) reinforcement learning. Exit questionnaires revealed that participantsremainedunawareofthepresenceofunexpecteduncertaintyandfailedtoacquiretherightmodelwithwhich toimplementBayesianupdating. Citation:Payzan-LeNestourE,BossaertsP(2011)Risk,UnexpectedUncertainty,andEstimationUncertainty:BayesianLearni