- Artificial Intelligence (AI) has the potential to better train doctors via personalising learning experiences, simulating clinical scenarios and supporting research.
- Yet, there are human, financial and resource-based barriers to its adoption.
- Healthcare institutions, medical schools, industry partners and government bodies need to work together to develop responsible and effective solutions.
SINGAPORE, 13 NOVEMBER 2025—Imagine a medical student diagnosing a virtual patient or a junior doctor practicing procedural skills such as drawing blood in a metaverse classroom. These Artificial Intelligence (AI)-powered tools aren’t science fiction—they are emerging realities that could train more doctors, faster and better, to meet the world’s growing healthcare needs.
(From left) Researchers Assoc Prof Liu Nan, Dr Ning Yilin and Dr Jasmine Ong on how collaborations are key to unlocking the potential of AI in transforming medical education // Image credit: Duke-NUS Medical School
A new study published in The Lancet Digital Health shows how AI could transform medical education, while calling for stronger collaboration across schools, hospitals, and regulators to make it safe, responsible, and effective.
The study by researchers from Duke-NUS Medical School, Singapore General Hospital and Tsinghua University, also identifies key barriers to AI adoption, such as ethical considerations and resource constraints. To address these challenges, the researchers call for a tightly coordinated network spanning medical schools, healthcare and academic institutions, industry partners and regulators to develop AI-enabled medical education and physician training schemes.
This study comes at a time when health systems around the world face staffing shortages and escalating expectations to deliver high‑quality care. WHO forecasts an alarming shortfall of approximately 10 million healthcare workers by 2030[1].
AI supports more diverse and engaging learning
The scientists highlight how AI can help bridge this gap, especially with advancements in technologies and the growth of large language model (LLM) applications—systems such as ChatGPT trained on vast amounts of text data to perform language processing tasks, including generating human-like text.
Specifically, AI tools can be used to personalise learning experiences in medical education. AI-generated virtual patients can simulate more realistic and complex clinical scenarios with greater consistency and versatility, without logistical and financial constraints. The combination with augmented reality or virtual reality technology also offers more immersive learning experiences. AI-powered metaverse environments further innovate medical education, facilitating activities such as team-based learning and case discussions anytime, anywhere. AI is also increasingly supporting medical research by streamlining tasks such as literature reviews. The incorporation of such tools could allow medical students and residents to devote more time to critical thinking.
Dr Jasmine Ong from Duke-NUS AI + Medical Sciences Initiative and Principal Clinical Pharmacist at Singapore General Hospital, is a joint first author of the paper. She said:
“AI is not here to replace clinical educators and mentors, but to empower them. AI enables educators and mentors to focus on what matters most – fostering meaningful connections with their learners. Serving as a digital co-tutor, AI enhances the learning experience through personalised feedback and realistic clinical simulations, helping to shape the next generation of healthcare professionals.”
Challenges to realising the potential of AI
Despite the potential of AI, its use in medical education currently faces challenges in terms of insufficient qualified trainers and a lack of tested implementation strategies. Another major concern about LLMs is their accuracy and credibility, with hallucinations or fabricated information remaining a persistent issue.
LLMs have presented biases related to gender and race, among others. Such biases, particularly when embedded within the medical literature, risk perpetuating systemic disparities over time. In addition, privacy concerns have also emerged, with the risk of patient information being exposed.
Dr Ning Yilin, Senior Research Fellow at Duke-NUS’ Centre for Quantitative Medicine and joint first author of the paper, said:
“As AI becomes more deeply integrated in medical education and training, we need to address the ethical concerns it raises, such as ensuring appropriate use, maintaining learning integrity and preventing unintended harms. These challenges call for clear guidance and inclusive, responsible design.”
Call for collaboration: promoting responsible adoption of AI
Associate Professor Liu Nan from Duke-NUS’ Centre for Quantitative Medicine and director of the Duke-NUS AI + Medical Sciences Initiative, who’s also a senior author of the paper, added:
“AI is transforming medical education worldwide. By working towards a comprehensive, global strategy and partnering across sectors, we can deploy generative AI responsibly to create more interactive, accessible training and translate gains into better care for patients.”
The researchers also pointed out that sustainable AI adoption in medical education and training calls for close collaboration across sectors. Healthcare institutions, medical schools, industry partners and government bodies need to work together to develop responsible, scalable and evidence-based solutions.
The researchers hope such collaborations will bring about the development of practical frameworks to implement AI-integrated medical education and physician training. These partnerships are also key to establishing funding models and resource supports.
Duke-NUS is at the forefront of biomedical research and translational innovations. This new study is part of the School’s ongoing efforts to improve global health through systems research and scientific breakthroughs.
###
DOI: 10.1016/j.landig.2025.100900