Commitment to Correlated Strategies

Without commitment, this game is solvable by iterated Game theory provides a mathematical framework for rational strict dominance: U strictly dominates D for player 1; after action in settings with multiple agents. As such, algorithms removing D, L strictly dominates R for player 2. So for computing game-theoretic solutions are of great the iterated strict dominance outcome (and hence the only interest to the multiagent systems community in AI. equilibrium outcome) is (U, L), resulting in a utility of 1 for It has long been well known in game theory that being player 1. However, if player 1 can commit to a pure strategy able to commit to a course of action before the before player 2 moves, then player 1 is better off committing other player(s) move(s)--often referred to as a Stackelberg to D, thereby incentivizing player 2 to play R, resulting model (von Stackelberg 1934)--can bestow significant in a utility of 2 for player 1. Even better for player 1 is to advantages. In recent years, the problem of computing commit to a mixed strategy of (.49U,.51D); this still incentivizes an optimal strategy to commit to has started to receive player 2 to play R and results in an expected utility a significant amount of attention, especially in the multiagent of.49