CAREER: New Directions in Computing Game-Theoretic Solutions: Commitment and Related Topics

Duke University, Durham NC

Investigators

Abstract

Game Theory occupies an important place in the foundations of multi-agent systems in artificial intelligence. Research under this award focuses on settings where one agent can commit to her (possibly randomized) strategy before the other agent moves. We consider how to compute optimal strategies to commit to in games with a combinatorial structure, extensive-form games, and repeated/stochastic games. Among other topics are connections to learning in games and mechanism/environment design, and the implications of commitment for the efficiency of computing game-theoretic solutions more generally. The main objective of the research is to make scientific contributions to artificial intelligence, multi-agent systems, and computational game theory, but to also advance real-world applications. For example, other researchers have expanded on the PI's prior theoretical research (with his PhD advisor) to apply the commitment framework to security applications, such as the placement of checkpoints and canine units at Los Angeles International Airport and the scheduling of Federal Air Marshals. The research performed under this award aims to provide a solid scientific foundation for improving and expanding this and related applications. The award also helps to build connections between computer science and economics, the traditional home of Game Theory. This interdisciplinary link can help diversify the computer science community, intellectually and demographically. Research is tightly integrated with the PI's educational efforts, which include the development of courses in Computational Microeconomics and Game Theory and an approved Computational Economics minor.