Fig. 2. Choice behavior and model fits.

(A, D, and G) Demonstrator choice behavior in the learning phase in studies 1 to 3. Demonstrators showed both a reward effect (slope of lines) and a group bias (distance between lines). Error bars indicate SEM. (B, E, and H) Observers in studies 1 to 3 showed a matching reward and group effect in their own test phase choices. Solid lines indicate choice behavior, and dashed lines indicate predictions from the best-fitting hybrid model. (C, F, and I) Comparison of model fits from reward learning, action learning, and hybrid families for studies 1 to 3. Dots indicate individual participants’ model fit, bold horizontal lines indicate mean AIC values, and box plots indicate 50% interquartile ranges.