Inferring non-observed correlations from causal scenarios: The role of causal knowledge

Abstract

This work aimed at demonstrating, first, that naı̈ve reasoners are able to infer the existence of a relationship between two events that have never been presented together and, second, the sensitivity of such inference to the causal structure of the task. In all experiments, naive participants judged the strength of the causal link between a cue A and an outcome O in a first phase and between a second cue B and the same outcome O in a second phase. In the final test, participants estimated the degree of correlation between the two cues, A and B. Participants perceived the two cues as significantly more highly correlated when they were effects of a common potential cause (Experiment 1a and 2) than when they were potential causes of a common effect (Experiment 1b and 2). This effect of causal directionality on inferred correlation points out the influence of mental models on human causal detection and learning, as proposed by recent theoretical models.