Problem structure and the use of base-rate information from experience

Area of Science:

• Cognitive Psychology
• Decision Making
• Medical Diagnosis

Background:

• Base-rate information, derived from the frequency of occurrence of categories, is crucial for accurate classification.
• Simulated medical decision-making tasks explore how individuals use base-rate information when diagnosing diseases based on symptoms.

Purpose of the Study:

• To investigate the extent to which participants utilize base-rate information in ambiguous medical diagnosis scenarios.
• To compare the predictive capabilities of existing classification models (prototype vs. exemplar-based) regarding base-rate information usage.

Main Methods:

• Five experiments involving simulated medical diagnosis where participants learned disease-symptom relationships.
• Transfer tests presented ambiguous symptom patterns suggesting multiple diseases with varying base rates.
• Analysis of participants' predictions to determine their use of base-rate information.

Main Results:

• Participants demonstrated varied use of base-rate information: appropriate use, ignorance, and inappropriate use (favoring rare diseases).
• The pattern of base-rate information usage was complex and dependent on category structure and test ambiguity.
• Existing classification models failed to predict the observed complex pattern of results.

Conclusions:

• No current categorization model adequately explains the observed human behavior in using base-rate information.
• A novel model integrating property competition and context-sensitive retrieval is proposed to account for these findings.