The University of British Columbia

We consider the problem of computing a fair communication equilibrium in a repeated game while making efficient use of communicative and computational resources. At each stage of a game, several controllers (game players) observe a random process and evaluate the utilities that they associate with all possible joint actions given their observations. These utilities are reported to a central communication device, which computes and recommends a fair joint action, subject to game-theoretic equilibrium constraints. It is shown how, under a class of generalized coordination games, such a communication device fulfill its role, while trading off performance and efficiency.