Nature Medicine: AI Explanations May Amplify Automation Bias in Laypeople

Orson_Xu · x · 2026-08-06

A new paper published in Nature Medicine explores the divergent impacts of explainable AI (XAI) for dermatological diagnosis.

The study reveals that while LLM-generated explanations help when the AI is right, they can make AI errors more persuasive to non-experts, thereby amplifying automation bias.

The authors stress that for current GenAI deployments, more explanation does not necessarily mean safer or more trustworthy AI. Systems must support calibrated reliance rather than simply producing more convincing outputs.

Related event: Nature Study: AI Diagnostics Amplify Automation Bias in Non-Experts(3 posts)→

Original post →

More from Safety

Safety channel →