Rigorous Systems Research Group (RSRG) Seminar

Selfish behavior often results in a suboptimal system performance from a system designer's perspective. To drive selfish users towards the designer s objective, we introduce a new incentive design framework, called "intervention", where the designer can perfectly or imperfectly monitor the actions of self-interested users and intervene in their interaction. Moreover, in many systems, due to strong negative externalities among users, the sets of feasible payoffs in one-shot games are nonconvex. Thus, it is possible to expand the set of feasible payoffs by having users choose convex combinations of different payoffs. To this end, we study intervention in repeated games and the associated design issues. We investigate systematically what the designer can achieve in terms of improving the system performance by designing suitable protocols, and how it can achieve these improvements, given its ability to monitor the users and to impact the payoffs of users. Our focus is on the scenarios with impatient users (the discount factor strictly smaller than 1), which are not well studied in the literature. <br><br> The results on the characterization of equilibrium payoffs and on the informativeness of monitoring technologies may be of independent interests in repeated games theory.