Biomedical Education in the Era of Large Language Models: A Paradigm Shift

A digest of Shen-Han Lee's commentary on how LLMs threaten critical thinking in biomedical training

Shen-Han Lee· Journal of Clinical Investigation· 2025· DOI 10.1172/JCI196863

The core problem

The integration of large language models (LLMs) into biomedical research and education has been hailed as a revolution. Yet, as Shen-Han Lee argues in the *Journal of Clinical Investigation*, this technological leap may inadvertently undermine the very cognitive skills that drive scientific progress. Lee opens with a stark vision of 2050: a world where scientists no longer ask "Why?" and where a PhD student, Jonathan, blindly accepts LLM-generated conclusions without understanding their biological relevance. This cautionary tale sets the stage for a critical examination of how LLMs, if adopted uncritically, could produce a generation of researchers incapable of independent thought. The commentary situates this concern within the broader context of AI's growing role in science, including the 2024 Nobel Prizes awarded for machine learning and protein structure prediction, which underscore both the promise and the peril of these tools.

The primary "result" of Lee's analysis is the identification of a dangerous feedback loop: as LLMs become more capable, students and researchers may increasingly delegate critical thinking to them, leading to atrophy of analytical skills. The narrative highlights specific failure modes: (1) uncritical acceptance of LLM outputs, as when Jonathan fails to question the model's conclusions; (2) inability to explain biological mechanisms, as shown by Jonathan's blank stare when challenged; and (3) loss of scientific curiosity, symbolized by the elder scientists' lament, "Remember when we used to think for ourselves?" These outcomes are not presented as inevitable but as likely if educational practices do not adapt. Lee implies that the current trajectory, if unchecked, will produce scientists who are technically proficient but intellectually dependent. The commentary does not offer a formal model, but the causal chain can be represented as: This reinforcing loop underscores the urgency of intervention.

The integration of large language models (LLMs) into biomedical research and education has been hailed as a revolution. Yet, as Shen-Han Lee argues in the *Journal of Clinical Investigation*, this technological leap may inadvertently undermine the very cognitive skills that drive scientific progress. Lee opens with a stark vision of 2050: a world where scientists no longer ask "Why?" and where a PhD student, Jonathan, blindly accepts LLM-generated conclusions without understanding their biological relevance. This cautionary tale sets the stage for a critical examination of how LLMs, if adopted uncritically, could produce a generation of researchers incapable of independent thought. The commentary situates this concern within the broader context of AI's growing role in science, including the 2024 Nobel Prizes awarded for machine learning and protein structure prediction, which underscore both the promise and the peril of these tools.

As a commentary piece, this work does not employ empirical methods but rather relies on narrative and rhetorical analysis. Lee constructs a hypothetical future scenario to illustrate the potential long-term consequences of current educational trends. The methodology is thus speculative and normative, drawing on anecdotal evidence and logical extrapolation. The central narrative device—a dialogue between a confused PhD student and a skeptical senior scientist—serves to dramatize the tension between technological convenience and intellectual rigor. Lee also references real-world events, such as the 2024 Nobel Prizes, to ground the fiction in plausible reality. This approach is effective for provoking reflection but does not provide quantitative data or systematic evidence. The commentary implicitly calls for empirical studies to assess the actual impact of LLMs on learning outcomes and research quality.

Lee's commentary raises profound questions about the purpose of biomedical education. If LLMs can generate hypotheses, analyze data, and even write papers, what role remains for human scientists? The author suggests that the answer lies in cultivating skills that LLMs cannot easily replicate: deep biological reasoning, ethical judgment, and creative problem-solving. However, the commentary stops short of proposing concrete curricular reforms. It implies that educators must redesign training to emphasize critical engagement with AI outputs, perhaps through Socratic questioning, hands-on experimentation, and explicit instruction in epistemology. The dystopian 2050 scenario serves as a warning, but also as a call to action. The piece aligns with broader concerns about AI's impact on society, including issues of bias, transparency, and accountability. In the context of biomedical research, where stakes are high, the need for human oversight is paramount. Lee's work is a timely reminder that technology should augment, not replace, human intellect. Future research should explore how to integrate LLMs in ways that enhance rather than erode scientific thinking.

CS practitioners and researchers

Opening member content…