Neural Representation of games and strategies in repeated stochastic games

Most of real life interactions are repeated, rather than isolated, encounters. Such repeated strategic interactions are modeled in game theory (GT) as stochastic games, where the players play a sequence of different single--‐shot game. GT assumes that players choose actions in a game according to strategies. The aim of this study is to identify neural representations of both game and strategy information during strategic playing, where either the game or the opponent player may change during the interaction. Forty--‐one participants played a stochastic game while undergoing functional magnetic resonance imaging (fMRI). They played with six virtual players (following different strategies) two different stage games: the Prisoner’s Dilemma and the Battle of the Sexes. At the beginning of each trial, information about the opponent player (or the game to be performed) was shown on the screen, followed by a delay in which the participant had to represent the provided information. Afterwards, information about the current game (or the opponent player) was shown, together with the choices of the player and the participant in the previous trials. Participants had to make their choice considering all provided information. fMRI data from the delay phase were analyzed to identify neural representations of either game or strategy information. Preliminary results show that game information is encoded by a confined network comprising parietal and prefrontal areas, while strategy information is represented in a much wider frontoparietal network. These findings suggest that different game variables of the same stochastic game are represented by different neural networks.