Noise in Reasoning as a Cause of the Conjunction Fallacy

The conjunction fallacy occurs when people judge a conjunction A&B as more likely than a constituent A, contrary to the rules of probability theory. We describe a model where this fallacy arises purely as a consequence of noise and random error in the probability estimation process. We describe an experiment testing this proposal by assessing the relationship between fallacy rates and the average difference between conjunction and constituent estimates (in the model, the smaller this difference, the more likely it is that the conjunction fallacy can occur due to random error), and by assessing the degree of inconsistency in people’s conjunction fallacy responses for repeated presentations of the same probability questions (in the model these responses should tend to be inconsistent due to random error, especially in cases where the average difference in estimates is low). Experimental results support both these predictions.