How does the reliability of a model affect children's choice to learn socially or individually?

Abstract The effect of model reliability on children's choices to learn socially versus individually is pertinent to theories addressing cultural evolution and theories of selective trust. Here the effect of a reliable versus unreliable model on children's preferences to learn socially or individually was examined, as well as their subsequent imitation on a puzzle box task. Experiment One (N=156) found children were more likely to ask to learn socially when presented with a novel task, after witnessing an unreliable rather than a reliable model. Experiment Two (N=40) found children select a new unknown model, over the previously unreliable model, suggesting a preference to learn socially was created, although not specifically from the unreliable model. Experiment Three (N=48) replicated children's learning preference in Experiment One with a new task, and showed children's attention is drawn towards other sources of social information (another adult model) when viewing an unreliable model, and also found a reliable model caused more fidelity of imitation. Together these results suggest that model unreliability causes greater social learning requests and attention to other, even novel, models when they are available. These findings evidence human children's strong propensity to learn socially compared with non-human animals; and suggest there is a more complicated relationship between learning preference, model reliability and selective trust than has been captured in previous research.