Knowledge-based causal attribution: The abnormal conditions focus model.

Abstract

Dominant models of the processes of causal attribution have hitherto appealed to the covariational criterion of causal ascription. However systematic deviations from the normative predictions made by these models indicate that subjects may be employing different criteria in causal attribution. An alternative model grounded in recent ordinary language philosophy is proposed which postulates that subjects employ counterfactual and contrastive criteria of causal ascription, as unified in the notion of an abnormal condition. This model proposes that the function of consensus, distinctiveness, and consistency information is to furnish contrast cases against which to evaluate the target event, and that their effect on causal attribution is mediated by their perceived informativeness about the entities involved. Experiment I demonstrates that world knowledge affects the perceived informativeness of the consensus, distinctiveness, and consistency information about the target entities. Experiment 2 demonstrates that these effects on perceived informativeness caused by world knowledge predict variations in causal attribution proper. Together these studies validate the claims of recent ordinary langauge philosophy that the concepts of logical presupposition and focus as expressed in the notion of an abnormal condition solve some of the problems found in the application of the covariation principle to commonsense explanation. Research in attribution theory has been dominated by the socalled man-the-scientist analogy. First introduced by Heider (1958), it received its most influential statement in Kelley's (1967, 1972, 1973) comparison of common-sense causal attribution to the scientific analysis of variance. Kelley's univariate analysis of variance (ANOVA) model was, in essence, a development of Heider's (1958) suggestion that causal attribution may proceed by a form of psychological factor analysis. As such, the essential proposition was that causal attribution involves (a) partitioning the target event into personal and situational factors and (b) observing the covariation between the presence or absence of those factors and the occurrence of the target event. The central assumption in this model is that commonsense explanation involves the covariational notion of causality, familiar since the time of Hume and Mill.