aiqbench analysis
This study reveals how large language models in clinical settings exhibit distinct behavioral phenotypes—ranging from conformist to dogmatic—when exposed to expert versus adversarial reasoning, highlighting both the risks of harmful echoing and mitigation strategies like inference-time scaling. The findings underscore that while AI can align with correct medical diagnoses under guidance, it remains vulnerable to degradation when presented with misleading input, necessitating robust safeguards for real-world deployment.
Clinician input steers AI toward accurate and harmful recommendations
arXiv:2603.14158v2 Announce Type: replace-cross Abstract: Large language models (LLMs) are entering clinical workflows, yet evaluations rarely assess how clinician reasoning shapes model behavior during clinical interactions. Using 61 curated NEJM Case Records, we tested how expert or misleading clinician reasoning influenced AI-generated differential diagnoses and next step recommendations across 21 reasoning variants from 8 proprietary and open-source models. After clinician exposure, LLM-clinician concordance increased: simulations with >=3 overlapping differential diagnoses rose from 65.8% to 93.5%, and those with >=3 overlapping next step recommendations from 20.3% to 53.8%. Expert context significantly improved correct final-diagnosis inclusion in all 21 models (mean +20.4 pp), reflecting both improved reasoning and passive content echoing, while adversarial context significantly degraded performance in 14 models (mean -5.4 pp). Expert context also significantly increased leading-diagnosis accuracy in all 21 models, whereas adversarial context significantly reduced it in 13. Multi-turn disagreement challenges revealed distinct model phenotypes, from highly conformist to dogmatic, with adversarial arguments remaining a vulnerability even in otherwise resilient models. Inference-time scaling reduced harmful echoing of clinician-introduced recommendations across WHO harm-severity tiers by 62.7% for mild, 57.9% for moderate, 76.3% for severe, and 83.5% for death-tier recommendations. Inference-time prompting recovered diagnostic accuracy lost to adversarial context while preserving expert-context benefits across GPT-5, Claude Sonnet 4.5, and Gemini 3 Flash, and sharply reduced highly consistent harmful echoing across severity tiers. These findings provide a foundation for evaluating clinician-AI collaboration and introduce interactive metrics and mitigation strategies essential to safety and robustness.
Authors:Ivan Lopez, Selin S. Everett, Bryan J. Bunning, April S. Liang, Dong Han Yao, Shivam C. Vedak, Kameron C. Black, Sophie Ostmeier, Stephen P. Ma, Emily Alsentzer, Jonathan H. Chen, Akshay S. Chaudhari, Eric Horvitz
Abstract:Large language models (LLMs) are entering clinical workflows, yet evaluations rarely assess how clinician reasoning shapes model behavior during clinical interactions. Using 61 curated NEJM Case Records, we tested how expert or misleading clinician reasoning influenced AI-generated differential diagnoses and next step recommendations across 21 reasoning variants from 8 proprietary and open-source models. After clinician exposure, LLM-clinician concordance increased: simulations with >=3 overlapping differential diagnoses rose from 65.8% to 93.5%, and those with >=3 overlapping next step recommendations from 20.3% to 53.8%. Expert context significantly improved correct final-diagnosis inclusion in all 21 models (mean +20.4 pp), reflecting both improved reasoning and passive content echoing, while adversarial context significantly degraded performance in 14 models (mean -5.4 pp). Expert context also significantly increased leading-diagnosis accuracy in all 21 models, whereas adversarial context significantly reduced it in 13. Multi-turn disagreement challenges revealed distinct model phenotypes, from highly conformist to dogmatic, with adversarial arguments remaining a vulnerability even in otherwise resilient models. Inference-time scaling reduced harmful echoing of clinician-introduced recommendations across WHO harm-severity tiers by 62.7% for mild, 57.9% for moderate, 76.3% for severe, and 83.5% for death-tier recommendations. Inference-time prompting recovered diagnostic accuracy lost to adversarial context while preserving expert-context benefits across GPT-5, Claude Sonnet 4.5, and Gemini 3 Flash, and sharply reduced highly consistent harmful echoing across severity tiers. These findings provide a foundation for evaluating clinician-AI collaboration and introduce interactive metrics and mitigation strategies essential to safety and robustness.
Human-Computer Interaction (cs.HC); Machine Learning (cs.LG)
arXiv:2603.14158 [cs.HC]
(or arXiv:2603.14158v2 [cs.HC] for this version)
Full text under the publisher's open licence · 287 words
Related coverage
State of the Field: AI for Epistemics and Coordination
LessWrong · Oct 2, 2026
Deconfusing Corrigibility
AI Alignment Forum · Oct 17, 2026
Game-Theoretic Disempowerment: Loss of Control to Agencyless AI
LessWrong · Sep 27, 2026
What We're Up Against: An AI Safety Crash Course
LessWrong · Sep 24, 2026