GATSBY COMPUTATIONAL NEUROSCIENCE UNIT

Causal Models: Right about Reasoning, Wrong about Learning (and More)

How much cognition can Causal Bayes’ net theory explain? The theory provides a strong foundation for explaining qualitatively how people (and rats) make inferences about causal systems. In particular, the theory makes a fundamental distinction between inferences from observation and from intervention, a distinction that cognition respects. The theory does less well accounting for how people learn causal systems (not from covariational data) how people make token causal ascriptions, and how we reason when our conclusions have the potential to tell us about ourselves.