Proceedings of the Annual Meeting of the Cognitive Science Society

In groups and organizations, agents use both individual and so-cial learning to solve problems. The balance between these twoactivities can lead collectives to very different levels of perfor-mance. We model collective search as a combination of simplelearning strategies to conduct the first large-scale comparativestudy, across fifteen challenging environments and two differ-ent network structures. In line with previous findings in thesocial learning literature, collectives using a hybrid of individ-ual and social learning perform much better than specialistsusing only one or the other. Importantly, we find that collec-tive performance varies considerably across different task en-vironments, and that different types of network structures canbe superior, depending on the environment. These results sug-gest that recent contradictions in the social learning literaturemay be due to methodological differences between two sepa-rate research traditions, studying disjoint sets of environmentsthat lead to divergent findings.