Collective search on rugged landscapes: A cross-environmental analysis

In groups and organizations, agents use both individual and social learning to solve problems. The balance between these two activities can lead collectives to very different levels of performance. We model collective search as a combination of simple learning strategies to conduct the first large-scale comparative study, across fifteen challenging environments and two different network structures. In line with previous findings in the social learning literature, collectives using a hybrid of individual and social learning perform much better than specialists using only one or the other. Importantly, we find that collective performance varies considerably across different task environments, and that different types of network structures can be superior, depending on the environment. These results suggest that recent contradictions in the social learning literature may be due to methodological differences between two separate research traditions, studying disjoint sets of environments that lead to divergent findings.