Game Dynamics on the Internet: Incentives, Asynchrony, and Beyond

Abstract: Motivated by real-life Internet environments we study dynamic multi-agent systems in which the prescribed behaviour for each agent (whether human or machine) is simple and natural (e.g., repeated best-response, minimizing "regret"). We use ideas from both Game Theory and Computer Science to tackle the following questions: (1) When is the prescribed behavior for each agent *incentive compatible*, i.e., when is in the best interest of each agent to follow the prescribed behaviour? and (2) When are the resulting dynamics guaranteed to converge to an equilibrium even in the presence of *asynchrony*? We consider the surprising implications of our results across a wide variety of interesting and timely applications: today's protocols for routing and congestion control on the Internet, social networks, cost-sharing mechanisms, and more. 

Michael Schapira
The Hebrew University of Jerusalem
Elath Hall, 2nd floor, Feldman Building, Edmond J. Safra Campus
Sunday, December 23, 2012 - 16:00 to 18:00