This presentation reviews empirical findings on causal learning and diagnostic inference, primarily in humans but also in rats. Some findings are consistent with Bayes nets; others are not. The findings will be discussed with respect to an associationist and a causal approach. The two approaches differ on whether reasoners make tacit generic assumptions about unobservable causal events. The assumptions enable a differentiation between causation and mere association. They also have implications for statistics and lead to novel predictions that go beyond intuition. Deviations from Bayes nets may be explained by a process that is adaptive in view of attention and memory limitations.
Back to Graduate Summer School: Probabilistic Models of Cognition: The Mathematics of Mind