AI in Health Professions Education

Original Editor - Michael Rowe Top Contributors - Ewa Jaraczewska, Jess Bell and Kim Jackson

Introduction

Artificial intelligence (AI) has many potential applications in health education, from generating learning materials to providing student feedback and assisting with assessments. Incorporating AI tools into academic programmes creates opportunities for educators to develop new teaching strategies and deliver personalised, diversified learning experiences for students. It can be used to ensure equitable and inclusive access to education and facilitate appropriate learning paths for all students. While there are many possibilities, there are also specific considerations when adopting AI tools in an academic setting. This page provides an overview of some of the potential applications of AI in health education, including content generation, resource curation, and collaborative learning and AI policy considerations.

AI-Supported Teaching

Generating Learning Materials

• Summarising content and slides:
  • teachers can use AI to create summaries of learning materials for their students
  • students can generate personalised summaries in formats that best suit their learning needs (audio, visual, contextual)

• Brainstorming activities:
  • AI can suggest collaborative, peer-led, and self-regulated classroom activities based on lecture content and learning outcomes
  • prompt example: "I am going to give this lecture, here are the slides, this is the audience, this is what the content is about, here are some additional resources. Give me a few examples of some of the activities we can do in the classroom that will be collaborative, peer-led, and self-regulated. They should take about 20-30 minutes, and I want to discuss them at the end. Give me ten ideas."

• Course design:
  • AI can assist in creating course outlines and structuring material to meet specific learning outcomes
  • prompt example: "I am going to be presenting a course. The course is going to be about [...]. How should I structure the course material to meet these learning outcomes most effectively?"

• Building educational resources:
  • AI can help create websites or other digital resources for both teachers and students, even without coding experience
  • prompt example: "Build me a simple website on the topic of education for health professionals. I want it to be an introduction to Health and Physical Education for novice educators and researchers. Create an outline of whatever you think is the most important topic for novices to know. Create the framework for a course as well and include a quiz."

• Generating illness scripts or case studies:
  • AI can create accurate illness scripts and case studies for various conditions

Feedback on Student Writing

"Personalised, timely feedback matters a lot."

AI can provide detailed, personalised feedback on student assignments by:

• summarising student performance
• offering feedback on the task completion process
• creating a student feedback assistance system for draft review

Dai et al. found that ChatGPT "is capable of generating more detailed feedback that fluently and coherently summarizes students’ performance than human instructors"

AI Personas

AI can realistically portray patient personas for physiotherapy education (particularly Claude 3.5):

• effective prompts can establish persona attributes, medical history, conversation goals and sample dialogue
• prompt example to build an AI persona: "You are a 78-year-old lady with COPD. You have recently lost your husband. You live alone, and lately, you started feeling slightly confused and anxious, etc."
• a student can paste this prompt into an AI model (Claude) and have unique interactions and opportunities to reflect

Virtual patient simulations develop communication, empathy, and clinical thinking before entering practice.

AI Policies in Education

When implementing AI in education, it is important to consider AI policy at different levels:

• institutional-level policy: tends to include generic position statements supporting ethical and responsible AI use
• programme-level policy: usually includes broad statements about the position of the school and programme administrators on the use of generative AI (GenAI)
• classroom-level policy: should include specific details on when / how AI can be used to support learning. Examples of classroom-level policy:
  • "in this module, I encourage you to use generative AI to support your learning"
  • "learning to use AI tools is an emerging skill that employers may expect"
  • "you may use AI to ask questions you would typically ask a lecturer or peer, but not to write essays for you"
  • "for each assignment where you use an AI model, you should also submit a document with information that shows (1) what prompt was used, (2) what model was used, including name and version, and (3) the output the model gave."

AI-Supported Assessment

"Most of our assessment is prone to error because our inferences about what students know can only ever approximate what they really know."

"Cheating is a social problem, not a technology problem, and we cannot solve the social issue with technology."

Challenges in traditional assessment:

• assessment suffers from a threat to validity because of the errors inherent in the process
• AI detectors for cheating are often ineffective and can escalate problems

Goals when using AI in student assessment:

• use AI to solve assessment design problems
• build learning support systems that discourage cheating

Standard Assessment Paradigm vs AI-Supported Assessment

"The standard assessment paradigm defines how most teachers think about assessment. It is a predefined set of items, problems, or questions we use to infer claims about students' proficiency in one or more traits. The data used for these inferences are typically sparse, and student learning may not be the focus of the assessment."

Standard assessment:

• predefined items or questions
• sparse data for inferences
• discrete snapshots rather than continuous evaluation
• uniform and inauthentic

AI-supported assessment:

• connects peers
• scores open-ended responses
• continuously aware of student behaviours
• more accurate inferences
• adaptive based on individual ability
• authentic and simulates real-world tasks

Solution for assessment:

• adjust assessment difficulty based on student capabilities
• focus on how students apply information rather than memorisation
• evaluate how students and AI solve problems together

AI-Supported Learning

"The agency for directing learning should be situated in the student"

It is interesting to consider how learning might change if we use AI technologies as cognitive tools, as extensions of ourselves, as collaborators or creative partners.

AI Tutors

AI tutors are improving and many students value AI for its "availability, patience, and lack of perceived judgement". Examples of AI tutors include: LearnLM-Tutor (Google), ChatGPT Edu (chatGPT).

Sense-making

There are concerns about using GenAI as an "oracle" - i.e. the user poses questions and GenAI makes answers up. However, GenAI is very good at analysing, summarising and rewriting existing information. You can ask it questions like:

• "explain this to me as if I was a first-year student"
• "please translate this into my home language"

Multimedia Theory

Multimedia theory explains how people learn better from pictures and words than words alone. AI can transform information into different formats, including flowcharts, graphs, conversations, or podcasts. Students can use AI to transform information into formats that will assist their learning. Prompt examples:

• "turn this description of a process into a flowchart"
• "read the simplified version of this section back to me"
• "present this information in a graph and explain the variables to me"
• "turn this academic paper into a conversation between two novices"
• "transform these conference proceedings into a podcast series"

Ideal Learning Environment

Andy Matuschak provides an overview of what AI-supported learning might look like in "How Might We Learn?"

• tractable immersion: AI suggests small, relevant contributions students can make to their field
• guidance in action: AI provides context-specific support during learning activities
• dynamic media synthesis: AI generates tailored media to enhance understanding
• contextualised study: AI inserts relevant information into the learning environment
• dynamic practice: AI creates spaced repetition questions based on lecture content
• social connection: AI identifies local groups or communities for collaborative learning

Risks and Challenges of AI in Education

Integrating AI in health education poses several risks:

• potential loss of critical thinking skills
• limited development of problem-solving abilities
• oversimplification of complex topics
• widening gap between education and clinical practice

Strategies to Integrate AI into Curriculum

Some strategies to help implement AI into health education programmes include:

• ensure students understand AI limitations, capabilities, and ethics
• develop clear guidelines for AI use in education for both students and teachers
• establish policies on AI use in assessments, exams, and research projects
• promote ongoing dialogue between educators and AI developers

Conclusion

"Assessment reform is OK. Education reform is better."

AI has the potential to revolutionise health education. By understanding its applications, benefits, and risks, educators can effectively integrate AI into their teaching practices. This integration should focus on enhancing learning experiences, developing critical thinking skills, and preparing students for a future where AI will be an integral part of healthcare practice.

As we move forward, we must continually assess and adapt our approach to AI in education, ensuring that we are teaching skills and knowledge that will remain relevant in the evolving landscape of healthcare and technology.

Resources

• Campino J. Unleashing the transformers: NLP models detect AI writing in education. J. Comput. Educ. 2024.
• Mollick ER, Mollick L. Using AI to Implement Effective Teaching Strategies in Classrooms: Five Strategies, Including Prompts. The Wharton School Research Paper 2023.
• Jurenka I, Kunesch M, McKee KR, Gillick D, Zhu S, Wiltberger S, Phal SM, Hermann K, Kasenberg D et al. Towards Responsible Development of Generative AI for Education: An Evaluation-Driven Approach. arXiv:2407.12687 2024
• Laurillard D. Pedagogical forms of mobile learning: framing research questions.
• Dawson, P. (2024). no-you-dont-have-a-cheating-proof-or-ai-proof-activity-7235052878673944576-rPoT LinkedIn.
• OpenAI (2024). Introducing ChatGPT Edu. OpenAI blog.
• OpenAI (2024). Understanding the source of what we see and hear online. OpenAI blog.
• Rowe, M. (2023). Creating a classroom AI policy for my students. Michael Rowe blog.
• UNESCO (2022). The turning point: Why we must transform education now.
• Stanford Graduate School of Education (2023). What Do AI Chatbots Really Mean for Students and Cheating?