The Conjunction fallacy: Causality vs. event probability

Abstract

AbstractUnder certain conditions people give a conjunction of events a higher probability than one of its constituents. This finding has been called the ‘conjunction fallacy’ (Tversky and Kahneman, 1983). One such condition is when the conjunction includes a possible cause and an outcome (called ‘causal conjunctions’) because the strength of the causal link biases the probability judgment. In the present experiment we investigated whether this finding could be replicated and whether it could be reasonably attributed to causality or might be attributed alternatively to the single event probabilities, as has been suggested by Yates and Carlson (1986).120 subjects read short texts each describing two diseases. These texts varied with respect to the probability of disease A and the probability of disease B (high or low), and the causal relation between the two diseases (yes or no). Ss judged the conjunctive probabilities that a person suffered from both diseases. Causality showed neither an effect upon the judged probability of conjunctions nor upon the frequency of conjunction fallacies. The single event probabilities, on the other hand, had significant effects: conjunction fallacies occured most often when the text included a combination of high and low event probability. Our results support the assumption that (a) causality is not sufficient to induce the conjunction fallacy, and (b) single event probabilities have a major influence on conjunctive probability judgments. The mental model approach is suggested as a theoretical framework that might allow the integrative and systematic examination of causal reasoning processes.