Inferring Repeated Game Strategies From Actions: Evidence From Trust Game Experiments

This paper is empirical study, using new experimental data, of repeated game strategies in trust games; its goal is to identify strategies that people use in repeated games. We develop a strategy inference method that maps observed actions to a set of best fitting unobserved repeated game stategies. Data analysis shows the ability of the method to infer distinct but intuitive and theoretically justified sets of strategies across finitely an indefinitely repeated games. In indefinitely repeated trust games we infer trigger strategies that are consistent with equilibria. In finitely repeated games we infer strategies with end-game effects. Almost all strategies inferred are best responses to the inferred strategies of opponents. For the first time we hypothesize repeated game strategies based on observed behavior, and characterize observed behavior using the core game theory concept of repeated-game strategies.