LARGE LANGUAGE MODELS (LLMs) like ChatGPT show promise as tools for enhancing medical ethics education, potentially improving virtue cultivation and moral knowledge acquisition in future doctors, according to a new study.
Researchers from Hiroshima University critically evaluated the integration of LLMs in medical ethics education, focusing on their potential to promote virtue and mimic nuanced human communication. The study positioned LLMs as exemplars and advisors in ethics education, highlighting their ability to address the lack of educational resources in this area. LLMs can serve as a “second-best” tool to supplement human instruction, effectively imparting moral knowledge and modeling virtue cultivation through tailored, interactive educational encounters.
The study advocates for including AI ethics in medical curricula and emphasizes LLMs’ utility as educational tools. While acknowledging LLMs’ capabilities, the researchers stress that these tools should supplement, not replace, human instructors. They also caution against deploying LLMs in actual clinical settings, as critical thinking in medical decision-making requires diverse moral perspectives that current LLMs may lack.
This research provides a framework for incorporating LLMs into medical ethics education, potentially addressing resource constraints while enhancing the learning experience for future medical professionals.
The integration of LLMs in medical ethics could significantly impact how virtue and moral principles are taught, offering a more personalized and engaging educational experience. However, future studies must focus on refining LLMs to better align with the complex ethical nuances required in clinical practice. As AI continues to evolve, its role in medical education will likely expand, but it must be carefully balanced with human oversight to ensure ethical integrity.
Reference
Okamoto S et al, AI-based medical ethics education: examining the potential of large language models as a tool for virtue cultivation, BMC Medical Education. 2025;DOI:10.1186/s12909-025-06801-y.
