Assessing biases in medical decisions via clinician and AI chatbot responses to patient vignettes.
Artificial intelligence (AI) is increasingly used in healthcare, but concerns have been raised that it can exacerbate existing disparities because of underlying biases in AI tools. In this observational study, researchers evaluated biases in clinician versus AI chatbox responses to 19 clinical vignettes involving cardiology, emergency medicine, rheumatology, and dermatology. Findings indicate that both AI chatboxes and clinicians provide different clinical recommendations based on a patient’s gender, race/ethnicity, and socioeconomic status under certain clinical scenarios.