Validating Kasparov’s Law Through Human–AI Collaboration in Clinical Diagnosis

Artificial intelligence (AI) is increasingly integrated into clinical practice through Clinical Decision Support Systems (CDSSs). While such systems can approach or even surpass human expert performance, their true value lies in how they enhance human–AI teams. This study investigates whether structured interaction protocols can improve diagnostic accuracy in a simulated radiology double-reading task. Sixteen radiologists collaborated with an AI system under eight different coordination strategies. Results demonstrate that protocols such as Accuracy-Oriented, Confidence-Oriented, and Presumptuous produced the highest overall accuracy (up to 97% among strong clinicians and 92% among weak ones), outperforming majority voting and single-metric-optimized approaches. Critically, weaker clinicians with superior protocols outperformed stronger clinicians with inferior ones, validating Kasparov’s Law: Weak Human + Machine + Better Process > Strong Human + Machine + Inferior Process. These findings highlight process design as central to effective CDSS deployment, advocating for a paradigm shift toward process-centric evaluation and design.