Feature-based reward learning shapes human social learning strategies

Human adaptation depends on individuals strategically choosing whom to learn from. A mosaic of social learning strategies—such as copying majorities or successful others—has been identified. Influential theories conceive of these strategies as fixed heuristics, independent of experience. However, such accounts cannot explain the flexibility and individual variability prevalent in social learning. Here we advance a domain-general reward learning framework that provides a unifying mechanistic account of pivotal social learning strategies. We first formalize how individuals learn to associate social features (for example, others’ behaviour or success) with reward. Across six experiments (n = 1,941), we show that people flexibly adjust their social learning in response to experienced rewards. Agent-based simulations further demonstrate how this learning process gives rise to key social learning strategies across a range of environments. Our findings suggest that people learn how to learn from others, enabling adaptive knowledge to spread dynamically throughout societies.