10.2 Representations of Games

To be able to reason about a multiagent interaction, we represent the options available to the agents and the payoffs for their actions. There are many representation schemes for games, and multiagent interactions in general, that have been proposed in economics and AI. In AI, these representation schemes typically try to represent some aspect of games that can be exploited for computational gain.

We present three representations; two of these are classic representations from economics. The first abstracts away all structure of the policies of the agents. The second models the sequential structure of games and is the foundation for much work on representing board games. The third representation moves away from the state-based representation to allow the representation of games in terms of features.