Correlated symptoms and simulated medical classification

Area of Science:

• Cognitive Psychology
• Artificial Intelligence
• Medical Decision Making

Background:

• Category learning theories diverge on whether judgments rely on configural information or additive summation.
• Understanding these models is crucial for explaining human categorization and decision-making processes.

Purpose of the Study:

• To investigate whether category learning is sensitive to configural information versus additive summation.
• To test predictions of different category learning models in a simulated medical diagnosis task.

Main Methods:

• Subjects learned about fictitious diseases from case studies with correlated and independent symptoms.
• Four experiments involved judging case likelihood or disease presence based on learned patterns.

Main Results:

• Subjects demonstrated sensitivity to configural information, favoring correlated symptoms.
• Judgments between case pairs prioritized preserving symptom correlations.
• Diagnoses for single cases were primarily driven by correlated symptoms.

Conclusions:

• Findings support category learning models that incorporate configural information.
• The results have implications for understanding human categorization and diagnostic reasoning.