Integrating individual and social learning: accuracy and evolutionary viability

Much of what we know, we know thanks to our interactions with others. There is a variety of ways in which we learn from others. We sometimes simply adopt the viewpoints of those we regard as experts, but we also sometimes change our viewpoints in more subtle ways based on the viewpoints of people we regard as our peers. Both forms of social learning have been receiving increasing attention. However, studies investigating how best to combine them, and how to combine the two with individual forms of learning, are still few and far between. This paper looks at ways to integrate various forms of social learning with learning at an individual level within a broadly Bayesian framework. Using agent-based models, we compare the different ways in terms of accuracy of belief states as well as in terms of evolutionary viability. The outcomes of our simulations suggest that agents are best off spending most of their time engaging in social learning, reserving only a limited amount of time for individual learning.