Center for Social Information Sciences (CSIS) Seminar

Abstract: In this paper, we study the multiplayer Bandit problem where the graph of social ties between agents is random. We fully characterize the equilibria of the game, and provide a number of comparative statics of the equilibrium with respect to the model parameters and importantly the average degree of connections. We further inspect the social optimum and provide an (almost) necessary and sufficient condition for the complementarity between the initial belief and the number of exploring agents.

Joint work with Gustavo Manso.