Conference on Cognitive Computational Neuroscience

To navigate our complex social world, it is crucial for people to deploy multiple learning strategies, such as learning from directly experiencing the outcomes of one’s actions – experiential learning (EL) – and learning from observing the behavior of others – observational learning (OL). Despite the prevalence of OL and EL in humans and other social animals, it remains unclear (i) how control over behavior is assigned to one strategy over the other depending on the environment, and (ii) how individuals vary in their strategy use. Here, we describe an arbitration mechanism in which the prediction errors associated with each learning strategy influence their weight over behavior. We designed an online behavioral task to test our computational model, manipulating the uncertainty of each strategy. In two independent datasets, model comparisons revealed similar and meaningful heterogeneity in how people solve this task. While a substantial proportion of participants relied on our proposed arbitration mechanism, four other groups were identified: those who use a fixed mixture between the two strategies, those who rely on a single strategy (OL or EL) and non-learners who perform an irrelevant strategy.