Learning and Generalization under Ambiguity: An fMRI Study
Figure 2
Subject's predicted value (expected reward) for guessing ‘purple’, according to model-based (M1, red-dashed) and model-free (M2, black) schemes.
M1 tracks current information when necessary (AC), and otherwise exploits generalization to limit the impact of spurious outcomes on action (GC). M2 is ignorant about each new individual and myopically chases reward. Red circles indicate the actual guesses of a typical subject.