Figure 3. Inference on agent's strategy in the stag-hunt game.

We assumed agents used unmatched strategies, in which the first agent used a fourth order strategy and the second agent used a first order strategy (A), and matched strategies - both agents used the fourth order strategy (B). The left panels show four state trajectories of 32 moves simulated using (or generated from) value-functions in Figure 2B. The right panels show the conditional probabilities of the second agent's strategy over a model-space of as a function of time.