There are two central reasons for studying cuasal learning in children.
Adults,especially the undergraduates in most studies already have extensive knowledge and sometimes formal tuition in causal inference. With children we are more likely to tap fundamental learning mechanisms.
Second, we know that, in fact, children learn an enormous amount of causal structure and information before they are five. Arguably, they learn substantially more in this period than they ever will again. If causal learning mechanisms are in place this may explain this remarkable acquisition of knowledge. Over the last seven years there has been a large body of work using the causal Bayes net framework to explore causal learning in children from 8 months to four years old. This work shows that children can normatively learn simple causal strength, more complex causal structure involving common effects of potential causes, causal direction, common cause, common effect and chain structures among three variables, unobserved causes, and causal schemata. They learn from deterministic and indeterministic data, involving both observations of dependence, the experimental interventions of others, and their own experimental interventions. They use these kinds of data to overturn prior domain-specific knowledge. They also integrate prior knowledge and new evidence in a Bayesian way. Finally, they can use these techniques to learn "real" theories such as theories of false belief and of personality traits.
While cognitive development has been informed by computational theory it has also inspired new elaborations of that theory - particularly accounts of the nature of experimental intervention and of higher-order acquisition of causal schemas and frameworks. I will list a number of still unanswered empirical and computational questions for further research.
Copyright. All Rights Reserved.