AI in Medical Education: From Foundations to Practice

Audience: Medical educators, clinical faculty, simulation educators, healthcare trainers, instructional designers, curriculum leaders, and education technology teams

Delivery format: Live Online, four 2-hour sessions (8 total contact hours). All activities and deliverables are completed in class.

Artificial intelligence is rapidly changing how health professionals learn, teach, practice clinical reasoning, receive feedback, and demonstrate competence. This course provides a practical introduction to AI in medical education for faculty, clinicians, simulation educators, and education teams. Beginning with core AI concepts and a brief history of the field, the course moves into educational frameworks that help instructors decide when AI is pedagogically useful and when it is not, then into hands-on work with AI tutors, AI-powered simulation, and AI-supported assessment and feedback. 

Participants leave with four working deliverables built entirely during class time, along with strategies for evaluating AI outputs for clinical accuracy, bias, privacy, and learner safety. The course emphasizes that AI can support learning and exploration, but should not replace the development of foundational clinical knowledge, clinical reasoning, professional judgment, and human oversight.

1. Explore core AI concepts, key milestones in AI history, and the types of AI systems most relevant to medical education, including generative AI, large language models, multimodal AI, AI agents, and AI tutors.
2. Work with educational frameworks to determine when and how AI should be used in teaching, learning, simulation, and assessment.
3. Design an AI-supported teaching or tutoring activity aligned with a specific learner level, clinical domain, and learning outcome, including a prototype AI tutor prompt that supports learner reasoning without simply giving away answers.
4. Design an AI-enhanced simulation or virtual patient activity for clinical reasoning practice.
5. Draft an AI-supported feedback or assessment strategy using rubrics, performance criteria, and human oversight.
6. Evaluate AI-generated educational outputs for accuracy, bias, clarity, safety, alignment with learning goals, and appropriateness for the intended audience.
7. Create a responsible implementation plan that addresses privacy, data protection, academic integrity, learner safety, and faculty oversight.

Session 1: What Is AI? Foundations, History, and Types of AI

This opening session gives participants a shared foundation in AI. Participants learn what AI is, how the field developed, and why recent generative AI systems have created new opportunities and risks for medical education. The session distinguishes between different types of AI and connects each type to possible use cases in health professions education.

Topics:

• What artificial intelligence means in practical terms
• Brief history of AI: symbolic AI, expert systems, machine learning, deep learning, foundation models, and generative AI
• Key terms: algorithm, model, training data, inference, prompt, token, embedding, context window, hallucination
• Types of AI relevant to medical education: rule-based and expert systems, predictive models, generative AI and large language models, multimodal AI, AI agents, AI tutors, and AI-powered simulation
• What AI can and cannot do in clinical education
• Why AI outputs require human review
• Introductory prompting strategies for educators

In-class activity: Participants compare AI outputs across several educational use cases, including explanation, case generation, feedback generation, quiz creation, and clinical reasoning support. They identify where the AI is useful, where it is unreliable, and where human expertise is required.

In-class deliverable: AI Use Case Map covering learner group, teaching problem, AI opportunity, AI type, risks, safeguards, and human oversight plan.

Session 2: AI, Learning Design, and AI Tutors

This session focuses on how to connect AI to education rather than using AI as a novelty. Participants evaluate AI tools through educational frameworks and learning sciences principles. The session introduces AI tutors, Socratic prompting, scaffolding, deliberate practice, feedback loops, and learner-centered design.

Topics:

• Moving from tool-first thinking to learning-problem-first design
• Backward design: outcomes, activities, and assessment
• ADDIE: analyze, design, develop, implement, evaluate
• TPACK: technology, pedagogy, and content knowledge
• SAMR: substitution, augmentation, modification, redefinition
• Deliberate practice and feedback loops
• Cognitive apprenticeship and scaffolding
• Clinical reasoning frameworks: illness scripts, hypothesis generation, problem representation, and diagnostic justification
• AI tutors as explainers, Socratic coaches, practice partners, feedback generators, case challengers, and reflection guides
• Designing tutor boundaries: when to hint, explain, challenge, or escalate to a human instructor
• Prompt engineering for teaching and tutoring

In-class activity: Participants design an AI tutor workflow for a teaching scenario such as pre-clerkship physiology, differential diagnosis practice, OSCE preparation, clinical reasoning coaching, communication skills practice, board-style review, or feedback on a case presentation.

In-class deliverable: AI Tutor Design Template covering learner level, learning objective, tutor role, tutor boundaries, prompt sequence, feedback style, escalation rules, risks, and safeguards.

Session 3: AI in Simulation and Clinical Reasoning

This session focuses on AI-powered simulation, virtual patients, and clinical reasoning practice. Participants explore how AI can support realistic, adaptive, and scalable simulation experiences while preserving educator oversight and learner safety. An approved simulation platform will be used as the main demonstration environment.

Topics:

• What makes simulation educationally effective
• AI-powered virtual patients and adaptive case progression
• Designing simulations for clinical reasoning rather than simple recall
• Case complexity, learner level, specialty, and authenticity
• Building cases around chief complaint, history-taking, physical exam findings, diagnostic uncertainty, differential diagnosis, data interpretation, management decisions, communication, empathy, and reflection
• AI as standardized patient, examiner, coach, or debriefing assistant
• Simulation prompts and guardrails
• Using transcripts and analytics to support feedback
• Risks: unrealistic cases, inaccurate clinical content, premature closure, hidden bias, false confidence, and poor debriefing
• Integration into clerkships, residency training, faculty development, and continuing education

In-class activity: Participants design an AI-enhanced virtual patient case or simulation scenario. They specify the learner level, learning objectives, clinical reasoning targets, expected actions, feedback points, and debriefing structure.

In-class deliverable: AI Simulation Case Blueprint covering case title, target learners, learning objectives, patient profile, opening prompt, expected learner actions, clinical reasoning checkpoints, branching logic, debriefing questions, feedback rubric, and safety checks.

Session 4: AI for Assessment, Feedback, and Responsible Implementation

The final session focuses on AI-supported assessment and feedback. Participants examine where AI may help with formative feedback, rubric drafting, performance review, case analysis, learner reflection, and assessment design. The session also addresses where AI should not be used, especially in high-stakes decisions without human review.

Topics:

• Formative versus summative assessment with AI
• AI for rubric drafting and refinement
• AI-assisted feedback on written explanations, case presentations, clinical reasoning notes, and reflection assignments
• AI and OSCE preparation
• AI-generated quizzes and item review
• AI for assessment blueprinting
• Evaluating learner reasoning, not just final answers
• Human-in-the-loop assessment models
• Validity, reliability, fairness, transparency, and explainability
• Academic integrity and appropriate disclosure
• Data privacy and PHI/PII restrictions
• Bias and equity considerations
• Building local guidelines for AI use in assignments
• Small-scale pilot planning
• Participant share-out and peer feedback

In-class activity: Participants design an assessment or feedback workflow for one of their own teaching contexts. They define what AI will do, what the human instructor will do, what learners will see, and how quality will be checked. Class time is reserved for peer review and revision.

In-class deliverable: AI Assessment and Implementation Plan covering assessment purpose, learner task, AI role, instructor role, rubric or scoring criteria, feedback process, disclosure language, data restrictions, bias and accuracy checks, and pilot evaluation plan.