Context and the inverse base rate effect

Abstract

Recent categorization research set in abstract medical diagnosis contexts has demonstrated an ‘inverse base rate effect’, whereby subjects make diagnoses which are consistent with base rates when presented with some sets of symptoms, but inconsistent with base rates when presented with other sets of symptoms. This paper reports three experiments which demonstrate that whether or not the inverse base rate effect is observed depends on the context in which categorization takes place. Inverse base rate effects are replicated in the abstract medical diagnosis context used in prior research, but not in a relatively realistic financial auditing setting, a less realistic financial auditing setting, or a very abstract generic setting. These results indicate that the inverse base rate effect may not generalize to applied settings, suggesting that interventions designed to mitigate the inverse base rate effect should not be instituted without first determining the existence of the effect in the particular setting in question.