The Foundations of Cooperation

University Of Pennsylvania, Philadelphia PA

Investigators

Abstract

The project continues and extends the principal investigators' previous work on the organization of society, and draws heavily on our previous National Science Foundation projects in this area. The results of the project will provide a better understanding of cooperation in large groups. There is a large body of work on dynamic games that demonstrates the circumstances under which cooperation can be enforced in groups of agents behaving strategically. Much of this work has been motivated by oligopolistic problems with relatively few players; the insights from these models provide a less than completely satisfactory understanding of cooperation in large groups. The project consists of two parts. The first part addresses authority - what it is, how a person might acquire it, maintain it, and use it. The aim is to provide an understanding of authority that centers on an individual's ability to coordinate other players' actions. A good model of authority should satisfy several desiderata, including: (i) An individual has authority when a group of agents coordinates their behavior in response to announcements by that individual; (ii) There is a tension between the incentives of the individual with authority to maintain that authority and to take advantage of his authority; (iii) Too much flexibility or unpredictability can destroy authority; (iv) Authority need not reflect some special ability; (v) Authority does not arise from players simply coordinating on an equilibrium identified by the "authority" figure. Punishment is a central ingredient of most models of cooperation in dynamic games, with punishment typically taking the form of coordinated actions following perceived deviations from prescribed play. The coordination in punishing deviations often rests on common knowledge, or near common knowledge, among the players about past signals of deviations. Such consensus is unlikely to arise from players' direct observations in large groups, and any consensus is likely to involve communication among the members of the group. The second part of the project involves adding communication to dynamic games. The mere possibility of communication will not ensure cooperation; the communication protocol must give agents an incentive to truthfully reveal their information for cooperation to be enforced. The notion of informational size of agents has been recently introduced to capture the importance of information held privately by a single agent, taking the information of other agents as given. The project investigates the connection between informational size and the degree to which incentives can be constructed to induce agents to reveal private information. Broader Impacts The research will provide a better understanding of how cooperation obtains in large social groups. A central question in economics and other social sciences as well is why some countries or groups do less well than others. The variation in performance seems not due primarily to differences in available resources, suggesting that differences in social organization are likely to be important. The project focuses on particular aspects of social organization - authority and communication - as important determinants of the level of cooperation possible within a society. The research will provide a foundation for more applied work aimed at understanding differences among societies by providing models linking social structure to social performance.