Customizing LLM/AI for student learning in health professions education
Presentation on 12 November 2026 at 1300hrs
As part of Course - "From Data to Diagnosis: Teaching Clinical Reasoning with AI and EMRs"
DEPARTMENT OF MEDICAL EDUCATION AND BIOETHICS
and
DEPARTMENT OF HEALTH POLICY AND MANAGEMENT
FACULTY OF MEDICINE, PUBLIC HEALTH, AND NURSING
UNIVERSITAS GADJAH MADA, INDONESIA
Why is it important for masters students in health professions education to know about - Customizing LLM/AI for student learning in health professions education
For Master's students in Health Professions Education (HPE), understanding "Customizing LLMs and Generative AI for Student Learning" is becoming a core educational competency rather than simply a technology skill. As future educators, curriculum designers, assessment specialists, and academic leaders, they need to know not only how to use AI, but also how to adapt it effectively, safely, and ethically for health professions learners.
Why this topic matters
1. Healthcare education is moving toward personalized learning
Traditional teaching often follows a "one-size-fits-all" model. Learners differ in:
Prior knowledge
Clinical experience
Learning pace
Language proficiency
Learning preferences
Customized LLMs can provide:
Individualized explanations
Adaptive difficulty levels
Personalized feedback
Tailored learning pathways
Example: A first-year nursing student may require foundational physiology explanations, while a senior medical resident may need advanced clinical reasoning cases. The same AI system can be customized to support both learners appropriately.
2. Supports competency-based education
Most health professions programs are transitioning toward:
Competency-Based Medical Education (CBME)
Entrustable Professional Activities (EPAs)
Milestone-driven progression
LLMs can be customized to:
Generate practice cases aligned with competencies
Provide targeted feedback
Track learner progress
Create remediation activities
Students in HPE need to understand how AI can support competency development and assessment.
3. Enables scalable feedback
One of the biggest educational challenges is providing timely, high-quality feedback.
Customized AI systems can:
Review reflective writing
Analyze clinical documentation
Provide communication feedback
Support self-assessment
This is particularly important in large programs where faculty resources are limited.
4. Enhances clinical reasoning education
Clinical reasoning remains one of the most difficult competencies to teach.
Customized LLMs can:
Simulate patients
Present progressive disclosure cases
Challenge diagnostic thinking
Provide expert reasoning models
Students should understand both the strengths and limitations of using AI to teach diagnostic thinking.
5. Prepares graduates for AI-enabled healthcare systems
Healthcare is rapidly adopting:
Clinical decision support systems
Ambient AI documentation
Medical chatbots
AI-assisted diagnostics
Future educators must prepare learners to work effectively alongside AI.
The question is no longer:
"Should we teach AI?"
but rather:
"How do we teach healthcare professionals to use AI safely and effectively?"
6. Builds AI literacy and critical appraisal skills
HPE graduates need to understand:
Hallucinations
Bias in training data
Explainability
Reliability
Validation of AI outputs
Students should learn to evaluate AI recommendations with the same rigor used to evaluate research evidence.
7. Improves simulation and virtual patient learning
Customized LLMs can create:
Virtual patients
Standardized patient conversations
Interprofessional scenarios
Emergency simulations
Benefits include:
Lower cost
Greater accessibility
Unlimited practice opportunities
Immediate feedback
8. Supports self-directed and lifelong learning
Healthcare professionals must engage in lifelong learning.
AI can function as a:
Personal tutor
Learning coach
Reflection partner
Knowledge navigator
Understanding customization helps educators design AI systems that support continuous professional development.
9. Addresses educational equity
Properly designed AI systems can:
Support learners with varying abilities
Assist non-native English speakers
Increase access to learning resources
Provide 24/7 academic support
However, students must also understand risks such as algorithmic bias and the digital divide.
10. Future educational leadership requires AI competence
Many Master's students in HPE become:
Faculty members
Program directors
Curriculum leaders
Assessment experts
Educational researchers
These roles will increasingly require decisions about:
AI integration
Curriculum redesign
Assessment policies
Academic integrity
Faculty development
Understanding customization equips future leaders to make evidence-informed decisions.
Key Learning Outcomes for Master's Students
After studying this topic, HPE students should be able to:
Explain how LLMs function in educational contexts.
Design AI-supported learning activities.
Customize AI tools for different learner levels.
Critically evaluate AI-generated educational content.
Apply ethical and governance frameworks.
Develop AI-enhanced assessments.
Use prompt engineering effectively.
Evaluate educational outcomes of AI interventions.
Identify risks including bias, hallucinations, and privacy concerns.
Lead AI implementation within healthcare education programs.
Relevance to NUS Medicine / NUHS Context
For health professions educators in Singapore, customized LLMs have potential applications in:
Undergraduate medical education
Nursing and allied health training
Residency and postgraduate programs
Continuing professional development
OSCE preparation
Clinical reasoning teaching
Reflective practice
Entrustable Professional Activities assessment
As institutions increasingly adopt digital and AI-enabled learning environments, HPE graduates who understand customization of LLMs will be better prepared to design, implement, evaluate, and govern educational innovations.
A useful takeaway
For Master's students in Health Professions Education, learning about customizing LLMs and generative AI is not primarily about mastering a technology. It is about developing the capability to create personalized, scalable, evidence-informed, ethically governed learning experiences that improve learner outcomes and prepare future healthcare professionals for an AI-enabled healthcare system.
Customized LLMs and Generative AI transform student learning in health professions education by shifting passive instruction into interactive, personalized coaching.
Thesen, T., & Park, S. H. (2025). A generative AI teaching assistant for personalized learning in medical education. NPJ digital medicine, 8(1), 627. https://doi.org/10.1038/s41746-025-02022-1
Zhui, L., Yhap, N., Liping, L., Zhengjie, W., Zhonghao, X., Xiaoshu, Y., Hong, C., Xuexiu, L., & Wei, R. (2024). Impact of Large Language Models on Medical Education and Teaching Adaptations. JMIR medical informatics, 12, e55933. https://doi.org/10.2196/55933
(following with assistance of Google AI)
Key Customization Strategies for Learning
• Retrieval-Augmented Generation (RAG): Restricts LLM responses strictly to instructor-curated textbooks, clinical guidelines, and lecture notes to eliminate hallucinations and maintain pedagogical accuracy.
Thesen, Thomas & Park, Soo Hwan. (2025). A generative AI teaching assistant for personalized learning in medical education. npj Digital Medicine. 8. 10.1038/s41746-025-02022-1.
• Virtual Patient Simulations: Configures LLMs to role-play diverse clinical cases, allowing students to practice history-taking, differential diagnosis, and bedside communication with immediate automated feedback.
Lai, N. M., Lim, Y. S., Win, M. T., Bhargava, P., Thomas, P., & Ong, Q. C. (2026). The Effectiveness of Artificial Intelligence in Undergraduate Health Professions Education: Systematic Review and Meta-Analysis of Randomized Controlled Trials. JMIR medical education, 12, e88933. https://doi.org/10.2196/88933
• Adaptive Study Plans & Content Generation: Automatically converts dense source material into tailored flashcards (e.g., Anki-ready formats), practice multiple-choice questions (MCQs), short-answer questions (SAQs), and customized outlines.
Zhang, Q., Huang, Z., Huang, Y., Wang, G., Zhang, R., Yang, J., Cheng, Y., Chen, B., Wang, H., Qiu, K., & Chen, H. (2025). Generative AI in medical education: feasibility and educational value of LLM-generated clinical cases with MCQs. BMC medical education, 25(1), 1502. https://doi.org/10.1186/s12909-025-08085-8
Sridharan, K., & Sivaramakrishnan, G. (2025). Large language models as educational collaborators: developing non-conventional teaching aids in pharmacology & therapeutics. BMC medical education, 25(1), 1525. https://doi.org/10.1186/s12909-025-08134-2
Lightning Demos: AI Tools for Medical Education | AI in Medical Education Symposium
• Passive Audio Tools: Transforms written medical summaries into conversational AI podcasts for review during commutes or clinical rotations.
Lightning Demos: AI Tools for Medical Education | AI in Medical Education Symposium
Benefits & Opportunities
• Scalable Personalization: Mitigates the scalability crisis in medical and nursing training by providing 24/7 one-on-one "tutoring" support when human faculty are unavailable.
• Targeted Remediation: Dynamically adjusts difficulty and focuses on individual student knowledge gaps identified during self-assessment.
• High-Stakes Prep: Intensifies active practice during exam periods and clerkship transitions.
Khakpaki A. (2025). Advancements in artificial intelligence transforming medical education: a comprehensive overview. Medical education online, 30(1), 2542807. https://doi.org/10.1080/10872981.2025.2542807
Thesen, Thomas & Park, Soo Hwan. (2025). A generative AI teaching assistant for personalized learning in medical education. npj Digital Medicine. 8. 10.1038/s41746-025-02022-1.
Thesen, T., & Park, S. H. (2025). A generative AI teaching assistant for personalized learning in medical education. NPJ digital medicine, 8(1), 627. https://doi.org/10.1038/s41746-025-02022-1
Challenges & Guardrails
• Accuracy & Verification: Over 70% of health professions students actively cross-check LLM outputs against primary literature or source material to guard against medical inaccuracies or "hallucinations".
• Overreliance & Critical Thinking: Risk of cognitive offloading where students passively accept AI-generated clinical reasoning rather than building foundational diagnostic skills.
Abd-Alrazaq, A., AlSaad, R., Alhuwail, D., Ahmed, A., Healy, P. M., Latifi, S., Aziz, S., Damseh, R., Alabed Alrazak, S., & Sheikh, J. (2023). Large Language Models in Medical Education: Opportunities, Challenges, and Future Directions. JMIR medical education, 9, e48291. https://doi.org/10.2196/48291
• Equity, Privacy, and Ethics: Concerns regarding data privacy, institutional governance, algorithmic bias, and academic integrity policies.
Abd-Alrazaq, A., AlSaad, R., Alhuwail, D., Ahmed, A., Healy, P. M., Latifi, S., Aziz, S., Damseh, R., Alabed Alrazak, S., & Sheikh, J. (2023). Large Language Models in Medical Education: Opportunities, Challenges, and Future Directions. JMIR medical education, 9, e48291. https://doi.org/10.2196/48291
Zhui, L., Yhap, N., Liping, L., Zhengjie, W., Zhonghao, X., Xiaoshu, Y., Hong, C., Xuexiu, L., & Wei, R. (2024). Impact of Large Language Models on Medical Education and Teaching Adaptations. JMIR medical informatics, 12, e55933. https://doi.org/10.2196/55933
