How AI Models Learn Bias From Medical Imaging Data

AI Ethics with Fexingo: Bias, Safety, and Responsible Artificial Intelligence · Jul 11, 2026 · 9m

Episode 105 of AI Ethics with Fexingo: Bias, Safety, and Responsible Artificial Intelligence. Hosts Lucas and Luna examine how AI diagnostic tools absorb racial and demographic bias from medical imaging datasets. They focus on a 2025 Nature Medicine study showing that chest X-ray models trained on predominantly white populations misdiagnose pneumonia at 40 percent higher rates for Black patients. The conversation covers the root cause — historical disparities in healthcare access reflected in training data — and explores solutions like federated learning and demographic rebalancing that…