Interpreting Covariation in Causal Structure Learning


Recent studies have shown that people use covariation information to infer causal structure. However, there is little information about how people derive causal directionality from covariation. The present study is designed to provide further evidence about the role of covariation in causal structure learning. In Experiment 1, where covariation between two variables was systematically manipulated, participants were asked to observe the states of bacteria (present or absent) and to infer their causal relationship. We found that judgments of causal structure varied as a function of covariation, and that participants interpreted covariation according to necessity of causation. In Experiment 2, participants who received information about high causal strength interpreted covariation according to sufficiency of causation. These results demonstrate that prior knowledge modulates interpretation of covariation and suggest that domain-general covariation information and domain-specific prior knowledge of causal relations interact in causal structure learning.

Back to Table of Contents