Unveiling the Risks and Benefits of AI in Hospital Diagnoses
A study evaluated the impact of artificial intelligence (AI) models on clinician diagnostic accuracy in the diagnosis of hospitalized patients. The results showed that when clinicians were provided with standard AI model predictions and explanations, diagnostic accuracy increased by 4.4%. However, when clinicians were shown systematically biased AI model predictions, diagnostic accuracy decreased by 11.3%, and the explanations did not mitigate the negative effects. This suggests that while AI models can improve diagnostic accuracy, systematic bias in AI models can worsen clinician accuracy, and image-based AI model explanations may not help clinicians recognize biased models.