A generalized approach to belief learning in repeated games

We propose a methodology that is generalizable to a broad class of repeated games in order to facilitate operability of belief-learning models with repeated-game strategies. The methodology consists of (1) a generalized repeated-game strategy space, (2) a mapping between histories and repeated-game beliefs, and (3) asynchronous updating of repeated-game strategies. We implement the proposed methodology by building on three proven action-learning models. Their predictions with repeated-game strategies are then validated with data from experiments with human subjects in four, symmetric 2×2 games: Prisoner's Dilemma, Battle of the Sexes, Stag-Hunt, and Chicken. The models with repeated-game strategies approximate subjects' behavior substantially better than their respective models with action learning. Additionally, inferred rules of behavior in the experimental data overlap with the predicted rules of behavior.