Neural correlates of Bayesian social belief updating in the medial prefrontal cortex.

Social learning, a hallmark of human behavior, entails integrating other's actions or ideas with one's own. While it can accelerate the learning process by circumventing slow and costly individual trial-and-error learning, its effectiveness depends on knowing when and whose information to use. In this study, we explored how individuals use social information based on their own and others' levels of uncertainty. Participants performed a social information use task in which they could revise their initial estimate after viewing a peer's estimate. Uncertainty was manipulated by varying the amount of information provided before their decision and by manipulating the peer's reported confidence. As expected, adjustments were larger when individuals themselves were less certain and the peer was more confident. Through a combination of Bayesian computational modeling and neuroimaging analyses, we were able to identify regions in the anterior and ventral medial prefrontal cortex where neural activity overlapped in response to lower personal certainty, higher peer confidence, and larger belief updates after viewing the peer's estimate. We discuss how these regions in the medial prefrontal cortex likely serve as a convergence zone for the preparation and execution of integrating certainty estimates.