I have five papers this week, ranging from a Nature Medicine perspective to a small classroom study, and they circle a single question: does working with AI help students learn to think, or does it let them skip the thinking altogether? The most reassuring finding is that the answer seems to depend far less on the technology than on the conditions we build around it.

Ke et al. set the conceptual frame in a Nature Medicine perspective that gives the problem a name1. They separate three distinct risks: deskilling, the erosion of competence in experienced clinicians; mis-skilling, the uncritical absorption of flawed AI outputs; and never-skilling, the failure to develop foundational reasoning in the first place because AI has done the cognitive work during the formative years.

The sharpest idea is what they call “false proficiency”, where a trainee performs well in AI-rich clinical settings but collapses in AI-free, high-stakes assessment, because the underlying competence was never built. They are candid that direct evidence in medical training is still absent and present never-skilling as a risk model rather than an established fact. Their proposed answer is a three-phase, competency-protective framework: establish an AI-independent baseline, build critical calibration through structured teaching, then integrate AI under supervision. The practical message is about sequence. Protect the early years, and bring AI in once there is something for it to augment rather than replace.

Ong et al. then show how unsettled the evidence really is, in an editorial for npj Digital Medicine that reconciles two new studies pointing in opposite directions2:

1. In the first, a longitudinal study of 372 senior students using AI diagnostic tools under supervision across a year of rotations, greater AI engagement predicted higher AI literacy, which in turn predicted stronger critical thinking.

2. In the second, a randomised trial of 111 pre-clinical students, even accurate AI explanations produced no benefit, and plausible but wrong ones significantly reduced diagnostic accuracy while leaving students just as confident either way.

The authors argue these results are not contradictory but conditional: supervision and the mode of engagement decide the outcome. Active interrogation of AI within a supervised workflow builds skill, whereas fluent pre-digested answers in an unsupervised test invite what they call cognitive surrender. They also flag a “Matthew effect”, where students with prior technical experience and mastery goals gain the most, which raises real equity concerns.

The reframing I found most useful is their closing one: stop arguing about abstention and start designing the conditions of safe learning, treating this as harm reduction.

Hernández-Rincón et al. widen the lens with a scoping review in the Journal of Investigative Medicine that maps what the empirical literature actually shows3. Across 26 included studies, most reported a positive effect of AI tools on clinical education and decision-making, especially in working through clinical cases and interpreting images and test results.

The most interesting signal is distributional: AI seemed to help students with less clinical experience the most, potentially bringing them closer to the performance of more seasoned peers. The caveats are honest ones. Some studies found no significant difference, imprecise or unclear AI outputs could cause confusion, and one study found that clinical practice guidelines outperformed AI as a support resource.

The authors read this as promise rather than proof, and they call for structured integration into the curriculum rather than the ad hoc use that dominates now. It is a helpful counterweight to the more cautionary pieces, though the youth and inconsistency of the evidence base are impossible to miss.

Zarate makes a quieter but important methodological point in a letter to Medical Teacher4. Most studies, he notes, evaluate generative AI using accuracy, efficiency, and user satisfaction. Those measures are necessary, but they say little about how AI shapes the reasoning process itself: how a learner constructs an explanation, generates a differential, or weighs an ambiguous finding. He suggests treating generative AI as a semiotic-epistemic actor, an output that does not merely supply information but foregrounds particular interpretations and implicitly signals what counts as an acceptable clinical answer.

The tension he identifies is that a tool optimised to look right can subtly narrow how students learn to justify decisions and manage uncertainty. His prescription is modest and sensible: study the process, not just the performance, and keep educator guidance in the loop so that AI-assisted reasoning stays reasoning.

Embang et al. bring the theme down to a single classroom in Advances in Health Sciences Education, and their results are instructive precisely because they are mixed5. First-year students in a cardiovascular physiology module generated their own questions and used ChatGPT to answer them. Students using the chatbot reported higher autonomy, competence, and task value, although the differences did not reach statistical significance in this small sample.

The revealing part came when the researchers coded the 31 student-generated questions against Bloom’s taxonomy: most sat at “Understand” (41.9%) and “Apply” (45.2%), with very few reaching “Analyse”. In other words, the tool lifted motivation and self-directed engagement, but it did not, by itself, push students toward higher-order thinking. That only happened with instructor mediation to turn information retrieval into genuine dialogic inquiry. The lesson is a useful corrective to the tutor hype: an AI chatbot can be a good scaffold for engagement and a poor substitute for teaching, and the analytical work still needs a human to prompt it.

I wrote about a very similar finding from educational researchers working in UK business schools earlier this year here:

The five papers disagree about plenty, but the throughline is consistent. AI’s effect on clinical reasoning is decided by the conditions we place around it, not by the model. Supervision, sequencing, active interrogation, and attention to process rather than output are what separate a tool that builds reasoning from one that bypasses it. The most useful shift is in the question we ask. We keep asking whether AI helps students learn, when the better question, and the one all five papers gesture toward, is under what conditions AI builds reasoning rather than replacing it. That is a question about curriculum design and teaching, which I find oddly reassuring, because it puts the outcome squarely back in our hands.

A note before we begin: I wanted to flag the 2026 Edward J. Stemmler M.D. Medical Education Research Grant Program, which supports the research and development of innovative assessment approaches with the potential to advance assessment in medical education. The grant ceiling is $150K. If you want to apply for this grant you have to submit your Letter of Intent today.

For advanced warning of more grants like this one, please consider subscribing to support this newsletter:

Subscribe now

This subscription is a professional development resource and may be tax-deductible or eligible for reimbursement by your employer.

Almost everything written about artificial intelligence in medical education is written from the front of the room. We argue about what to add to the curriculum, how to redesign assessment, and which competencies a graduate will need. The people sitting in the lecture theatre, the students who will actually inherit all of this, are too often treated as the objects of the debate rather than participants in it. So this week I want to turn the lens around and ask what the learners themselves think, how that thinking forms, and whether we can measure it well enough to act on.

I have selected four papers from 2026, and across the four the message is consistent: student appetite for AI is high and now well documented, but it runs well ahead of both student confidence and institutional provision. The interesting questions have moved on from whether students are interested to why their attitudes form the way they do, and what they are actually being offered in return.

Mastour et al. provide the headline numbers in a systematic review and meta-analysis in BMC Medical Education that pools 26 studies and 20,963 medical students1. The top-line finding is reassuring for anyone pushing for reform: 78% of students held positive attitudes toward integrating AI into the curriculum, and that support was consistent across geographic regions. The more important finding is the one they name the “optimism-competence gap”. While 83.1% agreed that AI training is necessary, only 36.4% felt confident actually applying AI in clinical practice. Enthusiasm, in other words, is not the bottleneck.

The authors are candid about the limits of their synthesis, reporting extreme statistical heterogeneity (an I² of 98.5%), so the pooled percentages are best read as a broad signal rather than a precise figure. They also surface the practical barriers that keep the gap open:

• curricular overcrowding (in 68% of the studies that discussed it)

• a lack of faculty expertise (52%)

• ethical concerns (41%).

Their conclusion, which I think is the right one, is that the task is no longer to persuade students but to build a standardised, competency-based AI curriculum that turns willingness into capability.

Xu et al. ask a subtler question in BMC Medical Education: not whether students accept AI, but through what psychological route, and whether that route travels across cultures2. They surveyed 1,232 medical students in China (681) and Pakistan (551) and used structural equation modelling to test how subjective norms, meaning the sense of what peers and respected others expect, feed into acceptance. Their model proposed that perceived usefulness and then attitude carry that social pressure through to acceptance, and that sequential path held in both countries.

What differed was the direct effect of social norms: in the Chinese sample subjective norms pushed acceptance directly (β = 0.231), while in the Pakistani sample the same direct effect was not significant (β = 0.113). The measurement model was invariant across the two groups, so this looks like a real cultural difference rather than a measurement artefact. The practical implication is that “students want AI” is too blunt a basis for strategy.

In some settings, signalling that AI use is expected and endorsed by faculty and peers will move the needle; in others, the case has to be made more directly on usefulness. Engagement strategies, in short, do not transplant cleanly from one system to another. I personally wonder how this finding intersects with differing cultural understandings of academic integrity.

Taghiparast et al. take on the unglamorous but essential job of measurement in Medical Education Online, adapting and validating a Persian version of the Artificial Intelligence Learning Intention Scale across 800 medical sciences students at one Iranian university3. If we are going to track readiness and judge whether our interventions work, we need instruments that measure the right thing, and this one holds up well.

Factor analysis supported a coherent four-part structure covering epistemic capacity, psychological attitudes, facilitating environments, and behavioural outcomes, with reliability statistics that were consistently strong (Cronbach’s alpha from 0.854 to 0.904) and an acceptable model fit (CFI of 0.932). Importantly, the scale behaved the same way for men and women, which matters if it is to be used to compare groups. This is a single-institution validation and the authors are appropriately modest about it, but the value is in giving educators a standardised way to capture students’ intention to engage with AI, not merely their vague approval of it. That distinction, intention to learn versus general positivity, is exactly what the Mastour gap suggests we should be watching.

Enériz Janeiro et al. then deliver the uncomfortable counterpoint in JMIR Medical Education: a census of what is actually on offer4. They examined every one of the 52 Spanish universities awarding an official medical degree, reviewing published curricula for the 2025-2026 year with two independent reviewers and external validation.

The result is sobering against the backdrop of all that student appetite. Only 16 universities (30.8%) had incorporated AI in any form, and just 10 (19.2%) offered a course in which AI was the primary subject. Public and private institutions were almost identical (around 19% each), so this is not a resourcing story in the obvious sense. Where AI did appear it was lightweight and optional, averaging 1.17% of the 360-credit degree and mostly elective, with only the University of Jaén mandating a course with AI content. The regional picture was the starkest part: Andalusia had a specific AI course in 5 of its 9 universities (55.6%), while 10 of Spain’s autonomous communities offered nothing at all.

The authors frame their taxonomy as a reusable tool for monitoring over time, which is welcome, because the headline is that national provision is patchy, shallow, and geographically unequal.

Put the learner’s view next to the institutional reality and the shape of the problem is clear. Students are ready, in numbers we can now quantify, and we are getting better at measuring just how ready and why. What is missing is not motivation but a coherent, required, and evenly distributed response from the schools that teach them. The optimism-competence gap that Mastour and colleagues describe is not really a gap in students at all; it is the space left by a curriculum that has not yet been built. The encouraging news is that none of this asks us to manufacture enthusiasm. It asks us to meet it, with structured teaching, culturally aware engagement, and honest measurement of whether confidence is catching up with appetite.

Summer is well and truly here in Scotland, and I am looking forward to conference season. I will be attending AMEE in Vienna 🇦🇹 this year to present some exciting work on SimPatient. Please reply to this email if you’re going — it would be great to meet up.

I am also currently tracking 20 international Medical Education conferences and over £250K worth…

Read more

Read more

Hello everyone!

Many of you will remember that academics used to have something called ‘summer’, which was a break between academic years when we could planning, think and write about things that we didn’t have time to do during periods of busy teaching. Most of us don’t have much of a summer any more. In my University, assessments run later into June, a…

Read more

Hello everyone.

Before we get started on this week’s roundup, a couple of updates:

I wanted to take a moment to welcome a flurry of new subscribers to this newsletter, which is now routinely read by 7500 people per month from 70 different countries. If you have colleagues who would be interested in this newsletter please share it with them:

Share AI × MedEd

You may be int…

Read more

Read more

Last year I wrote about the AI tools that have become indispensable to research and teaching, and a number of readers wrote back asking the obvious follow-up question. What about everything else? The schedule, the inbox, the notes, the references, the slowly accumulating backlog of papers that need reading. The plumbing of academic work, in other words, rather than the AI work itself.

It is worth pausing on why this matters. The American Time Use Survey shows that people who work spend an average of 8.1 hours a day doing so, and those who pursue formal education add 5.4 hours on top. Most of us recognise these figures all too well. The point of this productivity toolkit, then, is not that it makes you more virtuous or harder-working. It is that the right tools let you finish the same work in less time, and the hours you reclaim are the ones that belong to your family, your friends, and the parts of yourself that have nothing to do with academia.

A note on framing before we jump in to the list. Most academics no longer have the luxury of choosing their core infrastructure: institutional contracts have pulled almost everyone into the Microsoft 365 ecosystem, with Outlook, Teams, OneDrive, and the rest. That is not a complaint. Microsoft's offer has improved substantially in recent years and the data governance story for institutional users is, on balance, the right one. Where I name a Microsoft tool as the institutional default below, that is recognition of the constraint most readers actually face. Where I name something else as the stand-out, it is because there is a genuine case for paying for, or switching to, the better tool.

Contents

• 📋 Task and Project Management

• 🧠 Notes and Knowledge Management

• 📚 Reference Management

• 📖 Reading and Manual Note-Taking

• ✍️ Writing, Presenting, and Data

• 📅 Calendar and Scheduling

• 🎯 Focus and Deep Work

• 🎙️ Meeting Capture and Transcription

• 🤖 AI Assistants for Recurring Tasks

• ✈️ Travel for Conferences

📋 Task and Project Management

Where your week actually lives. Most academics now run three or more lists in parallel across institutional and personal tools.

Things 3

Best task manager for Apple users who don’t collaborate.

A focused, Apple-only task manager that turns a chaotic week into a calm review. Strong Today and Upcoming lists, lightweight projects, and a level of polish that makes capture genuinely frictionless.

This app uses the Getting Things Done (GTD) method and it has genuinely changed my life. No more panicking about forgotten tasks. No anxiety about future tasks, which always resurface when it’s time to work on them. No more missed deadlines.

If you need to stay in the Microsoft 365 ecosystem then I would recommend your institutional default, which is most likely Microsoft To Do. It’s not all bad, though! The visual design, on iOS at least, is clean and refreshing, and it is tied directly to Outlook flagged emails and Teams tasks, so it earns its place by being already wired in.

If Things 3 is too Apple and Microsoft To Do is too simple, I would suggest looking at Todoist. Cross-platform, with natural-language entry and powerful filters. The right pick for researchers working off Windows or Android, or collaborating across institutions

🧠 Notes and Knowledge Management

The second brain. The space where lecture prep, reading notes, half-formed ideas, and meeting minutes accumulate.

Notion

Best second brain for anyone who gets lost their notes.

What sets Notion apart for academics is not the document editor but the database layer. A linked database of reading items, grant applications, supervision tasks, or course materials behaves like a small research information system, with views, filters, and relationships across them. Once you start using it this way it is difficult to go back.

If you live on Apple devices, Apple Notes has quietly become the most reliable everyday tool. Quick capture, document scanning, handwriting recognition, smart folders, and instant sync across iPhone, iPad, and Mac. Free, fast, and good enough for most users.

If simplicity is your thing, Apple Stickies or Microsoft Sticky Notes is probably what you’re looking for. When the goal is “get this thought out of my head before the next meeting”, a sticky note pinned to the desktop is still the most honest tool. Microsoft’s version is slightly superior: it syncs across your devices via OneDrive.

Microsoft also make OneNote, but it is one of their weaker and clunkier Office products.

📚 Reference Management

Citations, PDFs, and the slow accumulation of “I’ll read this later”. Choice here usually lasts a career.

Read more

Read more

The focus of these conversations was ab…

Read more

Read more

This week’s roundup examines a collection of recent papers that explore the intersection …

Read more

Read more

Read more

Read more

Read more

Generative artificial intelligence is rapidly entering medical curricula, prompting widespread concern about learners uncritically accepting inaccurate outputs. However, a fascinating new study suggests that a student’s own personality might serve as a built-in safeguard against this phenomenon. This week, we explore the intersection of human psychology…

Read more

Read more

👋 Hello everyone. February is the peak time for new job adverts as institutions finalise their recruitment and funding for the upcoming academic year so I have created a new Jobs board for subscribers; there are currently nine active vacancies listed across AI research, clinical informatics, and medical education.

Subscribe now

As medical educators, our priority is t…

Read more

The four papers…

Read more