9310276 Blume Traditional models of strategic interaction assume that a population of highly rational players interacts with one another in a non-cooperative fashion to determine a collective outcome. While these models improve our understanding of centralized socialized interaction such as that seen in well-organized financial asset markets, they do not address the kind of decentralized non- cooperative competition that arises in labor markets, markets for retail services, and many non-market social settings. Such phenomena are best modeled as games with a spatial structure. Players live at given locations. Each player competes (or, more generally, interacts) only with his near neighbors. Nonetheless the choice of distant players has an indirect effect through a string of players who share neighbors who share neighbors, etc. Social interaction through this decentralized spatial or geographic structure stands in contrast to centralized social interaction wherein everyone interacts directly with everyone else. This research will investigate how strategic choice evolves over time when interaction is decentralized. Preliminary research has shown that the sharp rationality hypotheses of traditional game theory are not necessary to understand how competition evolves and what outcomes are ultimately achieved. Consequently the investigation will study the consequences of abandoning the traditional models of hyper-rationality in favor of looser strictures on individual behavior, and study how the spatial structure of the player population enhances or inhibits the organization of play (e.g., the evolution of cooperation in so- called "coordination games"). Another theme of the project is the contrast between social outcomes that can be reached through centralized social interaction and those that are achieved with decentralized interaction. Finally, the project will study the ramifications of different "interaction technologies," such as random matching and fixed partners, for the evolution of collective choice. ***
