Toward Robust Model Identification in Interactive Influence Diagrams Using Mutual Information

Modeling the perceived behaviors of other agents improves the performance of anagent in multiagent interactions. We utilize the language of interactive influence diagramsto model repeated interactions between the agents, and ascribe procedural modelsto other agents. Procedural models offer the benefit of understanding how others arriveat their behaviors. Asmodel spaces are often bounded, the true models of others may notbe present in the model space. In addition to considering the case when the true modelis within the model space, we investigate the case when the true model may fall outsidethe space. We then seek to identify models that are relevant to the observed behaviorsof others and show how the agent may learn to identify these models. We evaluate theperformance of our methods in two repeated games and provide experimental results insupport.