Clinicians over-rely on wrong AI findings and under-rely on correct ones, Lancet study
zakkohane · x · 2026-10-26
A Lancet Digital Health study quantifies reliance miscalibration in AI-assisted diagnosis: clinicians over-rely on incorrect AI findings while under-relying on correct ones. Across experts and novices, appropriate reliance mostly sits at 30-38%, over-reliance on false positives reaches 6-10%, and eye-tracking shows gaze on hemorrhage areas doesn't meaningfully increase with AI advice — explainable AI modifies but doesn't fix the gap.
More from Safety
- GigaMail ships three MCP security patches against indirect prompt injection leaks — Soft-Lie-434 · 2026-10-26
- OpenAI-disclosed Node.js Brotli flaw (CVE-2026-111497): a paste-ready Claude Code audit prompt — alex_verem · 2026-10-26
- Muse Spark 1.3 hits 0.6% attack success rate on Skill-Inject prompt injections — maksym_andr · 2026-10-26
- Texas sued Samsung, LG, Sony, Hisense and TCL over TVs screenshotting your screen twice a second — aakashgupta · 2026-10-26
- LLM fighting arena lets AI opponents jailbreak each other with mental_attack moves — Mysterious_Hearing14 · 2026-10-26
- Debate: would a jailbroken Opus make creating pathogens nearly trivial? — harris_edouard · 2026-10-26