Feature-based reward learning shapes human social learning strategies

Human adaptation relies on individuals strategically selecting whom to learn from. A rich mosaic of social learning strategies has been identified, such as copying majorities or successful others. Influential theories conceive of these strategies as a disjointed set of fixed heuristics that are independent of experience. By ignoring underlying mechanisms, these theories cannot explain the widely observed variation and flexibility in social learning. Here, we advance a domain-general reward learning framework which provides a unifying and mechanistic explanation of pivotal social learning strategies. We first formalise how individuals learn which social features, such as others’ behaviour and success, are associated with rewards. Through four experiments (n=1,226) we then confirm that people adjust their social learning upon experiencing rewards of associated social features. Finally, we present agent-based simulations showing that such social feature learning can parsimoniously explain the emergence of key social learning strategies across a range of environments. By elucidating how people learn to learn from others, our findings help explain how adaptive knowledge can be transmitted and spread throughout societies.