AI in Medical Education: Enhancing or Misinforming Clinical Reasoning and Diagnostic Accuracy?

The impact of artificial intelligence (AI) on medical students’ learning has been examined through two studies, one that focused on real-world AI use and another on AI explanations versus misinformation risks. The findings showed that AI's effects vary by individual and setting, influencing how medical curricula should be designed and how clinical competence should be redefined. The findings were published in npj Digital Medicine, a Nature Portfolio journal.

Doctor in a white coat interacting with a handheld tablet. Ai-inspired images superimposed.
Study: Reconciling how clinical reasoning is learned in the age of artificial intelligence. Image Credit: raker/Shutterstock.com

The AI Tutor Dilemma in Medicine

AI is rapidly permeating healthcare, with narrow AI tools already supporting clinical practice across specialties and large language models (LLMs) sparking a surge of innovative applications since late 2022. These technologies are increasingly adopted as unsupervised "pocket tutors" by medical students, raising urgent pedagogical concerns.

Recent literature warns of dual risks. One is "deskilling" experienced clinicians as a result of "never-skilling" or "mis-skilling" trainees who rely on AI without developing foundational reasoning. Superficial analogies to past technological panics, such as calculators in 1960s classrooms, prove imperfect as modern AI encroaches on judgment, interpretation, and communication.

Despite these concerns, empirical evidence on AI's actual educational impact remains sparse. Prior work has largely focused on technical performance or expert perspectives, lacking a systematic investigation into learner-centered outcomes in authentic educational contexts. Crucially, the variability of AI's effects across individuals and learning environments remains underexplored, leaving educators without evidence-based guidance for curriculum design.

Download this article as a PDF for later.

Supervision, Engagement, and Cognitive Surrender

The first study discussed by the authors followed 372 senior students through a longitudinal survey using supervised AI tools during clinical rotations, finding that higher AI engagement predicted increased AI literacy, which in turn correlated with improved critical thinking over time.

Conversely, the second study reviewed in this paper performed a randomized trial with 111 pre-clinical students, revealing that even verified AI explanations failed to boost diagnostic accuracy, while plausible misinformation significantly degraded performance and disrupted confidence calibration. 

These findings are not contradictory when examining engagement conditions. The authors suggest that the first study's structured clinical environment fostered active interrogation of AI outputs, whereas the second's test-taking scenario promoted passive acceptance. The concept of "cognitive surrender" explains why learners may uncritically adopt AI responses, losing internal uncertainty signals that prompt help-seeking.

The studies highlight risks of "never skilling" (dependency before competency development) and "mis-skilling" (adopting AI errors). The diagnostic-accuracy trial’s error analysis showed 70.3% of misinformation-group errors aligned with misleading AI endorsements, versus 24.7% in controls, while confidence remained inappropriately high regardless of accuracy. 

Together, these papers suggest that AI's educational impact depends critically on the quality of supervision and the mode of learner engagement, advocating for curricula that teach active AI appraisal rather than passive reliance.

Literacy Gaps and Unsupervised AI Risks

The authors examined AI literacy's role in medical education and the equity implications of AI-assisted learning. AI literacy - understanding AI principles, and critically evaluating its outputs - is intended to empower clinicians to know when to trust, question, or override AI recommendations.

However, these are hard-won skills developed through deliberate practice, and both studies suggest that without proper conditions, AI engagement may hinder such development. The longitudinal survey study’s moderation analysis reveals that students with prior technological experience and mastery-oriented goals were better equipped to develop AI literacy, illustrating the "Matthew effect", where early advantages compound over time.

However, AI literacy was measured using self-report scales, leaving open the question of whether high-scoring students would resist plausible misinformation. This has equity implications because unsupervised AI access may widen educational inequality, since democratizing tools does not necessarily democratize effective learning conditions.

Both studies, conducted in China, reflect a global trend in which AI tools like chatbots are being rapidly embedded in medical education and are widely accepted, often without institutional support. The technology acceptance model (TAM) explains this through perceived usefulness and ease of use. 

Rather than advocating for abstention, the authors argue for systematically designing safe conditions for AI engagement. The central question shifts from whether AI should be used to how educators can create frameworks that support harm reduction, enabling the next generation of clinicians to learn well and safely with AI assistance.

Is it ethical to do research in AI?

Toward a Balanced Framework for AI Integration

Collectively, these two studies offer a nuanced, empirically grounded perspective on AI's dual-edged role in medical education, moving beyond speculative fears to reveal that educational outcomes depend critically on implementation context, learner characteristics, and the quality of engagement.

The findings refute both uncritical enthusiasm and blanket prohibition, instead advocating for a middle path grounded in deliberate curricular design. Supervision emerges as pivotal: structured environments foster critical thinking, while unsupervised use invites cognitive surrender and the adoption of misinformation.

Rather than resisting AI's integration, educators must proactively architect frameworks that promote active appraisal, preserve foundational reasoning, and mitigate harms. Ultimately, the studies underscore that AI should augment, not replace, clinical pedagogy, with competence redefined to encompass both traditional reasoning and sophisticated AI literacy. This will ensure that tomorrow's physicians are neither deskilled nor mis-skilled, but thoughtfully empowered.

Journal Reference

Ong, A. Y., Sui, M., Rosen, K. L., and Kvedar, J. C. (2026). Reconciling how clinical reasoning is learned in the age of artificial intelligence. Npj Digital Medicine, 9(1). https://www.nature.com/articles/s41746-026-02873-2

Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.

Citations

Please use one of the following formats to cite this article in your essay, paper or report:

  • APA

    Nandi, Soham. (2026, June 22). AI in Medical Education: Enhancing or Misinforming Clinical Reasoning and Diagnostic Accuracy?. AZoRobotics. Retrieved on June 23, 2026 from https://www.azorobotics.com/News.aspx?newsID=16425.

  • MLA

    Nandi, Soham. "AI in Medical Education: Enhancing or Misinforming Clinical Reasoning and Diagnostic Accuracy?". AZoRobotics. 23 June 2026. <https://www.azorobotics.com/News.aspx?newsID=16425>.

  • Chicago

    Nandi, Soham. "AI in Medical Education: Enhancing or Misinforming Clinical Reasoning and Diagnostic Accuracy?". AZoRobotics. https://www.azorobotics.com/News.aspx?newsID=16425. (accessed June 23, 2026).

  • Harvard

    Nandi, Soham. 2026. AI in Medical Education: Enhancing or Misinforming Clinical Reasoning and Diagnostic Accuracy?. AZoRobotics, viewed 23 June 2026, https://www.azorobotics.com/News.aspx?newsID=16425.

Tell Us What You Think

Do you have a review, update or anything you would like to add to this news story?

Leave your feedback
Your comment type
Submit

Sign in to keep reading

We're committed to providing free access to quality science. By registering and providing insight into your preferences you're joining a community of over 1m science interested individuals and help us to provide you with insightful content whilst keeping our service free.

or

While we only use edited and approved content for Azthena answers, it may on occasions provide incorrect responses. Please confirm any data provided with the related suppliers or authors. We do not provide medical advice, if you search for medical information you must always consult a medical professional before acting on any information provided.

Your questions, but not your email details will be shared with OpenAI and retained for 30 days in accordance with their privacy principles.

Please do not ask questions that use sensitive or confidential information.

Read the full Terms & Conditions.