Agent-Based Models and Experiments in a Coordination Game: the Dying Seminar

SEMESHENKO VIKTORIYA

Montevideo

Workshop; DINAMICA ECONOMICA: TEORIA Y APLICACIONES; 2016

Universidad de la Republica

In this presentation, I will present an experimental study of a coordination game with N players carried out under different information treatments. Usually, the coordination games in the game theory are limited to homogeneous players. However, players are likely to have heterogeneous preferences. In this case the equilibria of the game cannot be merely extrapolated from the individual players' intentions by self-analysis [Schelling 1978]. In a decentralized environment, information provided to the players may help to coordinate their choices. We choose the "dying seminar" introduced by T. Schelling as a coordination game with N heterogeneous individuals, and explore the effects of information available to them on the outcomes of the game. This game represents situations in which collective activities are driven by personal motivations. The dying seminar represents many everyday lifestyle situations, where social or economic activities are considered interesting or worthwhile only if enough others participate in them. We designed the computer interface as to play this game in the laboratory in controlled environment. This game has been tested under various information treatments. In particular, we explore the effect of information on the emergence of Pareto-efficient outcomes, by means of a gradual decrease of the information content provided to the subjects. We observe that efficient coordination is possible with private information alone, although not on a Pareto-optimal equilibrium. This outcome is in contrast to the commonresults known in the literature on coordination games. Then, we build the learning, agent-based models, that fit the experimental results. In particular, reinforcement-based learning models reproduce the qualitative trends of theobtained results.