A.I. is here! Artificial intelligence (AI) is already making an impact on various aspects on our lives and is positioned to significantly change how health maintenance and healthcare delivery are handled (1, 2). We have seen time and again that AI can be a game-changer in many industries, and the field of medical education is no different. AI can be used to transform learning environments, improve psychological safety, and promote lifelong self-directed learning. Depending on the educational need, AI can help with approaches to professionalism, reflective practice, and teaching competencies, and can inspire the future of education. Not search engines, but open-access models such as ChatGPT 3.5 or Gemini are leading the way in natural language processing and multimodal interactions. ChatGPT 3.5 excels in text-based conversational interactions, while Gemini extends this capability to support images, audio, and video. Their dialogue format enables them to handle follow-up questions, acknowledge errors, challenge faulty assumptions, and decline unsuitable requests. These models are versatile, finding use in creating outlines, crafting NBME-style questions, developing vignettes, fulfilling requests, and performing editing tasks. The AI-powered platform Perplexity can provide informed academic answers to queries and delivers summaries of queries, including inline citations, ensuring the credibility of the sources. With multiple search modes like "Copilot" for refining queries, "Focus" for setting specific topics, and "Writing" for generating results without a web search, this engine adapts to your needs, offering a more personalized and efficient search experience. Finally, the AI research tool Consensus is able to find and summarize peer-reviewed scientific research papers. So, AI can transform health profession education in various ways, including individualized feedback, assessment enhancement, curriculum development, data mining, and educator support. Need support outlining a four-hour workshop? Ask ChatGPT, Gemini, or Microsoft Copilot! Need assistance developing a program for pre-clinical students starting with exploring other medical school programs? Start with Perplexity, interact with its built-in Copilot function for clarification, and appreciate the in-line citations. Starting a new research project and exploring your preliminary research question? Ask Consensus and receive a list of peer-reviewed articles, ready to download in an Excel file. In our recent professional development session with Dariano Joseph, Director of IT at Western Atlantic University School of Medicine, I focused on applications and on engineering the correct “prompt" for AI-enhanced medical education. A prompt is text written in natural language that describes the task you expect AI complete. For a text-to-text language model, a prompt can be a simple question or a directive (“write an NBME-style vignette"). It can also be a longer passage that provides context, instructions, or previous conversation history. Please find a blog post on prompts for healthcare professionals here: https://www.paubox.com/blog/100-chatgpt-prompts-for-healthcare-professionals.
A word of caution, while LLMs (large language models) provide text for prompts in polished English, they may hallucinate (generate text that is incorrect or nonsensical) and exhibit knowledge gaps. So – be critical, use responsibly, and edit thoughtfully (3). As a teaching tool, you could have students critique AI-generated text as part of their learning! The following highlights some applications for pre-clinical health professions educators. If you would like to try: Accounts with ChatGPT, Bard, Perplexity and Consensus:
Prompt ChatGPT; Gemini; and Microsoft Co-Pilot to generate multiple-choice questions, written in style. Use this prompt, “Please write an NBME-style question and vignette on [….add your concept…..] for first-year medical students with explanations”. If you refer to an image, Gemini might insert an image. Prompt ChatGPT, Gemini, and Microsoft Co-Pilot to develop clinical skill training cases. The following examples were kindly provided by Dr. Poh Sun Goh, National University of Singapore. Example: “Develop clinical skills training simulation to manage patient with suspected pneumothorax”. The following response from Bard (Gemini) was generated (30 December 2023) - https://g.co/bard/share/3fdefbcb772e. "Develop clinical case to manage breathless patient - For Training Medical Students". This response from Bard (Gemini) was generated (30 December 2023) - https://g.co/bard/share/2387f7ae53dc. Prompt Perplexity to support precision learning: “For a third-year medical student, please list [nosocomial infections] with resources, including etiology and microbial pathogens, diagnosis, complications, prevalent cases, treatments, prevention, and colloquial names. Add some low-stakes quiz questions (may skip the co-pilot)”. Enjoy the references you will receive! Try the prompt, "Look up leadership pathways and leadership electives for first- and second-year medical students.” Again, lots of great references. Prompt Consensus and inquire if your research question have been asked before. Example: “How do perceptions of medical student wellness affect academic outcomes?” Make sure to download the references as an Excel file: https://consensus.app/results/?q=How%20do%20perceptions%20of%20medical%20student%20wellness%20affect%20academic%20outcomes%3F%20chools%3F Give it a try! Tried them out yet? These AI-powered tools, language models, or search engines have the potential to significantly enhance medical education as we know it. By integrating them into the learning process, we can encourage critical thinking, stimulate creativity, and provide innovative learning opportunities. Furthermore, gaining a deeper understanding of these models prepares all of us – educators and students alike – for future roles in a healthcare landscape increasingly influenced by artificial intelligence. Let’s familiarize ourselves with these technologies and learn to utilize them responsibly and effectively for our ultimate goal: to improve patient care. Many Thanks to Linda, my awesome human co-pilot, for feedback and edits! References 1. Shankar, P. R. (2022). Artificial intelligence in health professions education. Archives of Medicine and Health Sciences, 10(2), 256-261. 2. Safranek, C. W., Sidamon-Eristoff, A. E., Gilson, A., & Chartash, D. (2023). The role of large language models in medical education: applications and implications. JMIR Medical Education, 9, e50945. 3. Pal, A., & Sankarasubbu, M. (2024). Gemini Goes to Med School: Exploring the Capabilities of Multimodal Large Language Models on Medical Challenge Problems & Hallucinations. arXiv preprint arXiv:2402.07023.
2 Comments
|