US Health AI Courses Skip Key Training on Data Origins, Study Finds
The AI that could diagnose you is trained on data—but are its creators taught to check it?
Deep Dive
A new cross-sectional analysis of publicly advertised curricula finds that U.S. health AI education programs widely taught ethics but rarely taught how clinical data are produced.
Only 8 of 36 programs covered data provenance, and just 6 covered structural provenance; no program addressed sustainability, and none addressed the environmental cost of computation. The article notes that failures of deployed clinical AI systems have often originated in how their training data were generated.
Key Points
- Only 22% of U.S. health AI programs teach data provenance—how to check where training data comes from.
- None of the master's programs met the study's standard for teaching data provenance, and none covered AI's environmental impact.
- Most instructors are from computational fields, so students may miss social and ethical perspectives.
Why It Matters
Flawed data in medical AI can lead to misdiagnoses; better training means safer AI for your health.