Heterogeneity in frequency-dependent social learning under cognitive load
Social learning can be radically heterogeneous, both within, as well as across individuals. However, we do not yet fully understand how, when, and why individuals vary in their use of social learning strategies and how this heterogeneity precisely affects evolutionary processes. In our behavioural experiment we examine different factors influencing frequency-dependent social learning. We focus on cognitive load, a factor that, to our knowledge, has not been previously studied in frequency-dependent social learning. Participants performed a binary choice task. They were assigned to one of two roles. One role, which we call social learner, observed the choice distribution of participants in the second role, which we call demonstrator, before they made their choice. In addition to this distinction between social learners and demonstrators, participants were assigned to one of two groups. Both of these two groups consisted of demonstrators and social learners. Our study manipulated whether social learners were similar or dissimilar to the demonstrators of their ingroup in terms of winning points for a given choice and whether they observed ingroup or outgroup demonstrators. These two treatment variations created four conditions that were informationally equivalent. In other words, if social learning of participants was completely flexible, they could perform equally well in all conditions. In addition, we implemented a cognitive load treatment that required half of the social learners to memorise a multi-digit number during the social learning task. Memorising the number was incentivised. We find that social learning was not flexible. Social learners performed worse when being dissimilar and they performed worse when observing outgroup demonstrators. Surprisingly, cognitive load did not affect social learners' performance. Our study indicates that other factors outweigh cognitive load as determinants of social learning. Crucially, certain social learning biases, such as similarity bias or in-group bias, might be so robust that not even cognitive distractions affect them.