Domain Generality and Specificity in Children's Causal Inference About Ambiguous Data

In 5 experiments the authors examined children's understanding of causal mechanisms and their reasoning about base rates across domains of knowledge. Experiment I showed that 3-year-olds interpret objects activating a machine differently from it novel agent liking each object; children are more likely to treat the latter as indicating the objects with the causal property possessed an internal property. Experiment 2 suggested that 3-year-olds potentially use this mechanistic knowledge to reason about ambiguous data in terms of base rate information. Experiments 3, 4a, and 4b showed that these inferences are not the result of children being more interested in an agent's desires. Instead, children integrate domain-specific knowledge (i.e., reasoning about an agent vs. I machine) with the nature of that inference within that domain (i.e., reasoning about desires vs. other mental states). The authors suggest that a particular computational approach, based on Bayesian inference, best describes these inferences. This approach offers a description of how children might integrate domain-specific mechanism knowledge into a more general model of causal inference based on observing covariation data among events.