The mathematics of strategic decisions, where the best choice for each player depends on what the others choose. It covers competition and cooperation, from auctions and negotiations to arms races. In AI it is used to build systems that play games, to design training where models compete, and to think about how many AI agents might behave together.
Technical view
Established as a field by John von Neumann and Oskar Morgenstern’s 1944 book Theory of Games and Economic Behavior, with John Nash’s equilibrium concept (1950) at its centre: a set of strategies from which no player gains by changing alone. AI uses include self-play and equilibrium finding in poker and other games of hidden information, the two-player training of generative adversarial networks, Shapley values for attributing predictions to features, auction design, and analysis of multi-agent systems. Its limits are that real agents are not perfectly rational and that equilibria can be many, inefficient or hard to compute.
Related
Part of AI Terminology Index, section: Adjacent Fields.