Analyzing Human Search Behavior When Subjective Returns are Unobservable

Abstract The exploration versus exploitation dilemma is a critical issue in human information acquisition and sequential belief formation, the multi-armed bandit problem has been widely used to address it. Because of its high descriptive accuracy, SGU model, which combines SoftMax type probabilistic selection, Gaussian process regression updating, upper confidence interval evaluation, attracted much attention. However, this model assumes that analyst access returns from people’s choices, but many realistic tasks, assumption cannot be made because only choices are observable. Moreover, subjective. authors introduce new model-fitting method overcomes barrier evaluates performance using data sets derived agent-based simulations real consumer data. This approach potential significantly broaden range issues can applied.