Proceedings of the Annual Meeting of the Cognitive Science Society

People reliably make two errors when making inferences aboutthree-variable causal structures: they violate what is known asthe Markov assumption (mediation) on causal chains andcommon cause structures, and fail to sufficiently ‘explainaway’ on common effect structures. Our goal for the presentstudy was to quantitatively compare these two errors aftersubjects have learned the statistical relations between threevariables using procedures designed to maximize the accuracyof their learning and inferences. Aligning with prior research,we found that subjects violated the Markov assumption, anddid not sufficiently explain away. We also found judgmentsabout mediation were worse than judgments about explainingaway for one inference, but better for another, suggesting thatpeople are not uniquely worse at reasoning about one structurethan another. We discuss the results in terms of a theory of cueconsistency.