Menu
10/13/2026
Reading Guides

Reading Guide: Teaching Clinical Reasoning in the Age of AI

Verity Schaye, MD, MHPE
Tags
Reading Guides
Share
Back To Top

Suggested

Reading Guide: Introduction to Clinical Reasoning
10/12/26
Reading Guide: Patient Use of AI in Medical Diagnosis
10/12/26
Reading Guide: Teaching Clinical Reasoning in the Age of AI
10/12/26
Reading Guide: Diagnostic Excellence Starter Set
10/12/26

Summary

This is an overview of the state of the science of teaching clinical reasoning in the age of generative AI. If you are a faculty or a trainee interested in learning more about how advances in AI impact how to teach or learn clinical reasoning, this is for you. 

For a guided summary of these resources from Dr. Schaye, watch her presentation for Sketchy, “Teaching clinical reasoning in the age of generative AI."
For more from Dr. Schaye, watch her CODEX webinar, "Artificial Intelligence-Based Assessment to Enhance Internal Medicine Resident Clinical Reasoning Practices." 
 

AI and Diagnostic Reasoning 

Study after study shows that under controlled circumstances (meaning curated information is fed into an LLM), AI outperforms physicians on diagnosis. However, real-life clinical medicine is completely different from “controlled circumstances,” and this doesn’t fully translate to performance in the real clinical environment.


AI and Human Reasoning 

Integrating AI into clinical reasoning without an explicit structure does not necessarily improve performance. Studies show that practicing physicians do not achieve greater diagnostic accuracy simply by using LLMs. Similarly, encouraging medical learners to use AI without guidance is unlikely to enhance diagnostic performance unless they receive training in effective prompting strategies and how to appropriately incorporate AI into the diagnostic process. Additionally, without proper instruction, LLM use may reinforce existing cognitive biases.  


The New Clinical Reasoning Process – AI as Diagnostic Team Member  

Generative AI can play an infinite number of roles in diagnostic teams (e.g., asking for other viewpoints, reviewing clinical notes, triggering questions to prevent potential errors). Preliminary research shows structured integration of AI + Clinician can improve diagnostic performance in simulated cases and, in real-life cases, provide a diagnostic time-out. However, LLMs are sycophantic and have the tendency to agree with users, which can be detrimental when asking LLMs for a second opinion. Without training in how physicians prompt AI, adding AI to a physician does not necessarily improve performance compared to AI alone. 


Considerations for Integration of AI into Curriculum & Clinical Care  

Given that AI is part of almost every medical learner's life today, it is imperative to train learners to use AI effectively. However, we don’t fully understand the impact of integrating AI into clinical skill development, including the potential for de-skilling, mis-skilling, and never-skilling. In particular, we don’t know when to introduce AI to medical learners and how to integrate it into their diagnostic process. Early learners need to commit to their thinking before receiving AI output to develop their clinical reasoning. Potential strategies include demonstrating a certain level of competence in a clinical skill before integrating AI, allowing AI integration only at a certain stage of training, and assessing human-only and human-plus-AI performance on that clinical skill.  

Risks of De-skilling or Mis-killing  


Potential for Upskilling  


When and How to Integrate AI Across the Learner Journey


Find more Reading Guides here.

Verity Schaye, MD, MHPE

Assistant Dean for Education in the Clinical Sciences, Office of Medical Education
Assistant Director of Curricular Innovation, Institute for Innovations in Medical Education
Associate Professor of Medicine, NYU Grossman School of Medicine

Dr. Schaye’s work focuses on the integration of technology and artificial intelligence in the teaching and assessment of clinical reasoning, and she has been grant-funded by the NBME Stemmler fund to support this work focusing on AI assessment of clinical reasoning documentation. Recently, she was selected as a National Academy of Medicine Scholar in Diagnostic Excellence focusing on developing an AI-based diagnostic performance feedback system.

Back To Top