December 3, 2009: Alison Gopnik

The statistical social learner: Using causal inference to learn about action sequences and personality traits.

Recent work shows that children can infer causal structure from statistical data. We will describe two different ways that this kind of statistical causal learning can help children learn about other people. In the first series of studies, we explore how children learn about how to segment and interpret goal-directed action sequences, and how this shapes their imitation and observational learning.

In the second series of studies we showed 4 and 6 year old children different patterns of covariation between actions and either individual people or situations. 4-year-olds inferred that the actions were caused by an enduring trait of the person or by the external situation depending on covariance. 6-year-olds were also sensitive to covariation evidence but they had developed a consistent prior bias towards person explanations.

These findings suggest that statistically based causal inference may be responsible for the development of social knowledge. In combination, these studies show that children use statistical causal learning to make important inferences about other people – inferences that shape their social experience.