Learning of equilibria and misperceptions in hypergames with perfect observations

This paper studies the learning of equilibria in adversarial situations when players may have misperceptions about the game they are involved in with their opponents. We use the concept of high-level hypergames to model these scenarios. By drawing connections with the theory of ordinal potential games, we establish that players in a hypergame can individually learn their perceived equilibria using any improving adjustment scheme. We investigate how players can incorporate the information gained from observing the opponents´ actions by updating different levels of her perception. We introduce high-level perception updating algorithms for resolving inconsistencies in perception using self-blaming or opponent-blaming strategies. Finally, we establish that when all players are rational and have perfect observation about past outcomes, repeated play converges to an equilibrium.