In search of rationality in human causal learningevidence of non-normative strategies when combining causes

Resumen

Causal learning models make different assumptions about how people should combine the influence of different potential causes presented in combination. Based on the linear integration rule, some models propose that the causal impact of a compound should equal the linear sum of each of the causes presented in isolation. Other models such as the Power PC theory are based on a different integration rule, the noisy-OR, suggesting that the rational way of computing the causal impact of a compound involves correcting the sum of the causes by subtracting the overlap between them. The present experiments tested which integration rule people use. Four different cover stories were used to ensure that the participants understood the independence of the causes. The experiments used different sets of probabilities and several formats for presenting information. The results of most experiments do not confirm the predictions of the noisy-OR integration rule. Only one experiment (of ten) supports the predictions of the noisy-OR rule. In spite of having mixed evidence, people do not appear to spontaneously use this rule. We discuss the implications of our results and alternative explanations for our pattern of data, including inhibitory mechanisms and an averaging heuristic.