The dilution effect: judgmental bias, conversational convention, or a bit of both?

This study explored competing normative interpretations of the dilution effect: the tendency for people to underutilize diagnostic evidence in prediction tasks when that evidence is accompanied by irrelevant information. From the normative vantage point of the intuitive statistician, the dilution effect is a judgmental bias that arises from the representativeness heuristic (similarity-matching of causes and effects). From the normative prospective of the intuitive politician, however, the dilution effect is a rational response to evidence presented in a setting in which Gricean norms of conversation are assumed to hold. The current experiment factorially manipulated conversational norms, the degree to which diagnostic evidence was diluted by irrelevant evidence, and the accountability of subjects for their judgments. Accountable subjects demonstrated a robust dilution effect when conversational norms were explicitly primed as well as in the no-priming control condition, but no dilution when conversational norms were explicitly deactivated. Non-accountable subjects demonstrated the dilution effect across norm activation conditions, with the strongest effect under the activation of conversational norms. Although the results generally support the conversational-norm interpretation of dilution, the significant dilution effect among non-accountable subjects in the norm-deactivated condition is more consistent with the judgmental-bias interpretation.