The value of rational analysis:
Steven A. Sloman, Philip M. Fernbach · 2008
Abstract This chapter provides a skeptical analysis of the technical machinery of causal Bayesian networks. It points out that this machinery provides valuable insights into how the representational power of probabilistic method is of crucial importance. But the learning algorithms for such networks do not provide a good model of human causal learning. It argues that human causal learning is rational, to some degree — but also exhibits large and systematic biases. Models of causality learning that are exclusively based on rational principles are unlikely to be successful.