This Week in MedEd
12 open funding calls, 70 upcoming events, 38 live jobs (5 new, 4 closing this week), including:
💰 NIH Research Education Program: Mentoring Activities & Networks (up to $500,000 from the US National Institutes of Health, for US institutions, standing deadline to May 2029)
🎓 AIMEC 2027, the 3rd International Conference on AI in Medical Education (Kuala Lumpur, 16-18 Apr 2027)
💼 Executive Director of Education and Training, Innovation and AI (Medical Council, Ireland, open now)
The full list, with deadlines, links and eligibility, is one tap away. Try everything free for 30 days.
Most of us have now watched a colleague type a one-line question into a chatbot, squint at the answer, and decide the tool is either magic or useless. What we see far less often is anyone asking whether the person typing was ever taught how to ask. We run workshops on how to give feedback and how to write a good multiple choice question, yet we tend to treat prompting as something people will simply pick up.
This week I am looking at three papers that treat prompting as a skill that can be designed, taught and tested. Two of them put students through structured instruction and compare them with students left to their own devices, and the third shows how an examiner can build a better prompt, layer by layer, when writing high-stakes exam questions.
If you want to learn more about prompting, I would recommend my paper on PDSA prompting1, and my previous post on Prompt Engineering:
Key points:
Students who were taught to prompt, verify and integrate AI output did better on knowledge and clinical reasoning tests than students given the same tool with minimal guidance.
Without instruction, students tended to fire the question at the AI and keep rephrasing it, whereas taught students identified their knowledge gaps and broke the task down first. That difference did not fade as students became more familiar with the tool.
Effects on test scores were real but modest, and both student studies were single-institution, so treat them as strong signals rather than settled evidence.
For examiners, prompt design behaves like item writing: context, structure, examples and a request for rationales all change what you get back, and expert review is still non-negotiable.