One year of AI × MedEd
A year ago this week I started this newsletter. Here are some of my reflections and what comes next.
A year ago I published the first issue of this newsletter. I wasn’t sure anyone would read it: the premise, writing about what other people have written about AI in Medical Education(!), felt very niche. But it turns out that this was actually the point, and one year on, there is a growing community of thousands of educators who regularly turn up to see what the latest research says.
My motivation for writing this newsletter was not to build an audience. It was to curate the papers I was reading anyway in a kind of personal journal, so I could refer back to it later in my own scholarship and research. From that perspective this newsletter hasn’t been a huge amount of additional work; it has brought me value, and I’m pleased to see that it is bringing value to others too.
So, in the spirit of the milestone post, let me do three things: say thank you, share what I’ve learned, and tell you where this is heading.
Thank you
The most enjoyable part of the last twelve months has been watching who turned up. Early on, most readers were where I expected them to be: in the UK, where I am based. Since then, a wave of new readers has arrived from the US, India, Australia, Singapore and a ton of other countries from all over the world. This international audience has reframed how I think about this newsletter. AI in medical education is a global issue, and the questions of access, training, representation and equity that keep coming up in the research are being lived out in very different settings at once. This has inspired much of my own research into why genAI is often terrible at being demographically and visually representative.
To every one of you who subscribed, forwarded a post to a colleague or replied with a paper I’d missed: thank you.
A particular shout-out too, to Karim Hanna, MD and Yavuz Selim Kıyak, whose excellent newsletters has helped build an exciting ecosystem around this topic here on Substack. None of us is covering this alone, and the field is better for it.
Finally, I’d like to thank the people who support this newsletter with a paid subscription, which is the most direct way to keep this newsletter and its tools coming, and to help me keep pace with a field that moves faster than the literature can.
If you are not yet a paid subscriber, it would be great to have your support for year two and I’d love to share the Member Resources with as many of you as possible. So for the anniversary, I’m opening a 30-day free trial. Take a month, read everything, explore the curated lists of jobs/grants/conferences and play with SBA Studio. There’s no commitment, and you can cancel any time before the trial ends.
What I’ve learned
If there’s one theme that has run through the year, it’s the widening distance between what students, educators and clinicians are actually doing with these tools and what our institutions are ready for. Survey after survey says the same thing: enthusiasm is high, formal training is scarce, and people are self-educating in the gaps. That tension has become the recurring theme of this newsletter.
I’ve learned that educators’ views on AI are often “AI good” or “AI bad”. I’ve come to the opinion that this is a false dichotomy. Of course the temptation is to pick a side, but the research almost never supports either. I think we’re capable of finding a more nuanced position where we see the opportunity of AI but are also simultaneously deeply sceptical and worried about its potential harms.
I’ve also noticed that the research is gradually becoming more useful and generalisable. There are a lot of papers that describe one intervention/model/tool in one setting/context, as is typical in medical education. However, there is an emerging evidence base of multi-site studies and cross-model evaluations that are giving us more useful and transferrable knowledge about how AI works in our sector generally. I hope there is more of this to come.
A look at the numbers
In the spirit of the writers who share their dashboards, here is where things stand after year one:
3,000 subscribers/followers in total, counting everyone who reads by email or follows along on Substack.
Around 10,000 views every month.
A global readership. The top five countries this year were the United States, India, the United Kingdom, Australia and Singapore, which still surprises and delights me for a newsletter this specialised.
64 posts published, a mix of themed deep-dives and the weekly round-ups, including these ones which were the most popular:
What’s next
Well I suppose I have to keep writing now there’s thousands of people waiting for the next post!
Next year the core promise will stay the same: I’ll keep reading the literature, curating into themes and synthesising it into something digestible for educators. The free weekly round-up will continue, and I’d love to write some more practical guides and invite guest writers. Beyond that, a few things I want to build:
Migrating from Substack to aomalley.com. This might not be very exciting but it allows me to do more with the website, including:
Building tools educators need at tools.aomalley.com. This is the part I’m most excited about. Reading the research has made one thing obvious: medical educators are being asked to teach and assess in an AI world with tools that were not built for it. So I’m starting to build them. First out of the gate is SBA Studio, to take the pain out of writing high-quality single best answer questions. It is the first of several tools I want to put in your hands over the coming year.
Broadening the scope of my curated lists of jobs, grants, conferences and targeted journals, so they are relevant to all medical educators (not just ones interested in AI).
More of your voice. The best ideas this year came from your replies. I’d like to make room for that more formally, whether through the occasional reader question, a survey, or bringing in other people working on this. On the tools especially: if there is something you wish existed to make your teaching or assessment easier, hit reply and tell me. I’m building this list from what you actually need, not what I assume you do.
If this newsletter has been useful to you this year, the single most helpful thing you can do is share it with one colleague who’d get something from it. Word of mouth is how a niche newsletter like this grows.
Thank you for reading.
Andrew