Dr Christoph Korn | Modelling multistep reward-based decisions and social learning about other persons’ character traits

Social interactions pose many challenges for human decision-making and learning. Here, I focus on two pertinent aspects: First, many social encounters require decisions between uncertain alternatives for the near and far future. Second, getting to know another person entails learning the character traits of that person. However, formal models that adequately capture the neuro-cognitive mechanisms of such complex decision-making and learning processes are lacking.

In the first part of the talk, I will present a series of partly published studies that show how humans combine optimal and heuristic policies to maximize rewards in non-social multistep decision scenarios. Results obtained from behavioral modelling and functional neuroimaging suggest a role of the medial prefrontal cortex in the computation of the employed policies and of the uncertainty associated with relying on these policies. To conclude the first part of the talk, I will discuss the relevance of these findings for multistep decisions in social contexts.

In the second part of the talk, I will describe unpublished experiments that outline how humans update their estimation of other persons’ character traits (e.g., how polite, helpful, and reliable is another person?). The best-fitting models combine principles derived from reinforcement learning algorithms with participants’ world knowledge about the distributions and interrelations of different character traits. Re-analyses of a published functional neuroimaging dataset show that these interrelations between character traits are represented in the medial prefrontal cortex.

Taken together, the to-be-presented projects aim at providing neuro-computational accounts of decision-making and learning processes that are integral to successful social cognition.