What Is the Purpose of Our Writing?

When I was at the AMEE conference in Vienna this August (2026), I attended a symposium called “The Death of the Academic Author? Implications of Artificial Intelligence for Academic Writing and Publishing.” Jennifer Cleland moderated, and the panel was Lorelei Lingard, Erik Driessen, Yu-Che Chang, Ayelet Kuper, and Ken Masters. A title like that plus this line-up of speakers? For me, this was “Must See TV.”

It was an excellent conversation, and I think what happened in that session is a good example, and a good representation, of the problem I want to talk about in this post: if AI can do our writing for us, should we let it?

The argument in the room

It was clear that the panelists were all well acquainted and friendly. But the dynamic of the symposium had two clear sides. My interpretation of the two sides was Dr. Masters on one side arguing that AI can do our writing for us, if we want it to, and scholarship will not be diminished by this. The other four were arguing that offloading our writing to AI would be detrimental to us as scholars and our scholarship, thus impacting the health professions education (HPE) field in a negative way.

Lorelei Lingard is an expert in rhetoric. She went first. She named the idea she was arguing against, “We research, GenAI writes,” and called it “a seductive idea” and “a dangerous idea, based on a false assumption," that writing is “the least creative and most mechanical part of the entire process.” Part of her argument was that “writing is thinking.” From Dr. Lingard’s constructivist paradigm, this makes a lot of sense, and the origin of this idea is in her field of rhetoric. The claim is not that writing records thoughts you already have; it produces them. Putting an idea into explicit, connected, and logical sentences for a reader who is not in the room forces you to find the gaps between the ideas bouncing around inside your head. Then the text on the page becomes an object for inspection, revision, and clarification. She used a quote to summarize her thoughts:

Dr. Masters was very clear about what he thinks about this argument. He was pretty forceful when he spoke, and a little provocative. He said, “writing is not thinking; writing is writing.” His summary slide was more hedged: “Writing is not thinking; Writing supports, benefits, and benefits from thinking, but it should not be equated with thinking.” He further said (I don’t have direct quotes for these ideas; this is me paraphrasing) that some people like writing and some people hate writing, and there are lots of ways to think, so you don’t have to write to think. In the words of his slide, writing “does not define the academic teacher or researcher.”

He also made a point I agree with, as it speaks directly to the mission of MedEdMentor, about who gets to take part in our field’s conversations:

He then cited a body of literature that he said was the basis for the “writing is thinking” argument. The papers he showed all looked at typing vs. handwriting notes and their effects on recall. He said that this literature does not show that “writing is thinking,” and he is right.

However, he is conflating two different bodies of literature. Dr. Lingard’s slides provided many great quotes, but little data to back up the “writing is thinking” claim. There are some papers that support what she was arguing:

A 2020 meta-analysis of 56 studies in school-age students showed that writing about content improved learning, with an average effect size of 0.30 (caveat: 18% of studies found a negative effect).
David Galbraith and colleagues, across more than two decades, have studied and argued that content “is not pre-stored” but “is constituted as we write.” They have shown that two writing activities build understanding: drafting full sentences spontaneously and reorganizing the overall structure afterward (caveats: understanding was self-rated, the samples were small, and even the authors call the evidence “preliminary”). Papers [1], [2], and [3].

None of this literature proves Dr. Lingard’s claim that “writing is thinking,” but it is suggestive, and it is a more accurate set of studies to cite than the ones looking at typing vs. handwriting.

The paper I could not have handed off

I really appreciate and respect both of these people, and I want to share that. In fact, I think Dr. Lingard has done more for the craft of writing in our field than any other scholar. Her work, especially her book, “Story, Not Study,” with Chris Watling, has been massively influential to me in making the difficult task of writing more manageable. As a physician, my default has always been to come at problems from a cognitivist approach. But as my career has grown to include more and more education work, I have definitely been incorporating more constructivist ideas.

An example I come back to is the Abstracting-Contextualizing Model (ACM) of Mentoring for Theory Use in Medical Education, which was published in Academic Medicine this year. I wrote nearly all first drafts of that paper, and the whole process, from start to finish, took me five years. It is the hardest thing I have ever done academically, by far, harder than any single part of medical training.

If I had offloaded any piece of that thinking, writing, and attempting to communicate my ideas to AI, I don’t think I would have understood the model that we built. My co-authors, Gregory Ow, MD, and Bridget O’Brien, PhD, MS, mostly acted as editors who also knew the data. So, the act of writing this paper was mostly me trying to convince them of the model, and of the data that supported each piece of it, and them telling me when what I was communicating didn’t make any sense, or the data didn’t support what I was trying to claim. The process of writing that piece led to us co-designing, revising, and re-revising the model, and I don’t think it could be what it is without us as humans writing the piece together.

This is notable because it speaks to Dr. Lingard’s point. The study used constructivist grounded theory, which, by nature, requires the investigators to be the tool through which the data is interpreted. Because Greg, Bridget, and I were the vehicle for the research, the writing had to be ours.

And then there are these posts

On the other hand, let’s talk about these blog posts, which are coming from me, but are definitely a hybrid between AI and my own writing. To write this one, I started by putting my ideas out into the world through verbal dictation, followed by an interview. I “waxed poetic” in my notes app for 15ish minutes about the symposium, what I remembered about what the panelists said, and what I thought about the different arguments. Then, my AI conducted a focused interview with me to fill in the gaps in my story.

In this case, I am taking more of a “dictation is thinking” than a “writing is thinking” approach, and AI is helping me to take those ideas, organize them, and communicate them clearly to others. However, I do find myself taking the first draft compiled by AI and rewriting nearly all of it, because it just does not communicate exactly the way I would. But it has laid out a clear blueprint for me to follow, based on the story I told.

Additionally, I am using AI to help me find the papers that I cited above. But I am reading them to find the pieces that are important to the various arguments here. Going back to Galbraith’s work and the two activities he ties to the development of understanding, while I am not writing spontaneous sentences, I am dictating them, then I do go through the draft, multiple times, in a highly critical way. In fact, the piece you are reading is my eighth draft, plus Greg’s edits and comments from the fifth draft.

Now, these blog posts have less at stake than a paper in Academic Medicine. These are scholarship-adjacent. There is no formal peer review, but I do care about what you, as my audience, think. A paper in a journal must get past multiple layers of scrutiny and is “set in stone” once it goes to print.

What I would hand off, and what I would not

Writing is a great way to learn (if you’re in a constructivist frame, creating knowledge is creating learning). But which pieces of writing are where the learning happens? I think that’s the question, because not all pieces of writing generate learning.

I do think AI can be incredibly helpful with writing, as I am demonstrating through these blog posts. In my own writing, I don’t ever hand off any first draft of any piece of an academic manuscript, and I do not hand off any idea generation. But, I have created a system here where I am still developing the ideas and the talking points through my own monologue and letting AI turn that into a blueprint that I can then flesh out. And I think it is relatively silly to not take advantage of AI for its editing capabilities. We don’t have to accept its suggestions, but we should at least run our writing through AI to see if we are missing anything, or if any grammar, wording, or structure can be improved. Dr. Lingard herself wrote in 2023 that “writers can harness [AI’s] power to make our processes more efficient and our products more robust.”

Where I land

So, where do I land? I agree with Dr. Lingard that writing is an incredibly powerful way to think. But I agree with Dr. Masters that there are multiple other ways to think. However, a productive reframing of this argument might focus on the question:

Like, why does our field publish journals in the first place? And why are publications in these journals often the “coin of the realm?”

I argue that writing in our field is a means of communication. We are trying to tell one another what we are thinking about and learning in our own contexts. To write well does require a tremendous amount of thought and is a very worthy skill to develop and hone, but it is not the only way to communicate. *Side note: depending on where you work, it might be the only mode of communication that your promotion and tenure (P&T) systems care about.

Now, if the purpose of writing is communication with each other, what happens when we hand this job off to AI? I think we are going to get a lot of AI talking to AI, and it is not going to get us very far as a field.

I think human-driven writing is the more direct route to this goal, but it would mean favoring quality over quantity. *Side note: again, something P&T systems struggle with.

When Greg reviewed this post, he asked what “AI doing our writing” actually looks like, and then he answered it:

"If you dictate all of your thoughts, or you write huge paragraphs into a prompt, and then the AI ‘writes’ the output… it’s hard to argue that the AI actually ‘wrote’ it. If you did that with a person, we would say that person was a co-author at best, or more likely an editor. This is why I like the phrase human-driven in your writing. If the original ideas and communication come from a person but the AI has just reshaped them, then, to me, that’s completely acceptable.

The issue (to me) is if [it means] you drop all your interview transcripts into the AI and out comes the manuscript. Well then, the AI did the thinking, not you. You’re not really able to stand behind the words on the page because you didn’t have the original thought. This is the problematic use of AI because it violates the contract where readers expect that the person writing it spent more time/energy/brainpower than the person reading."

So, I think it is vital that we maintain a high degree of human influence in our writing, because everything in HPE is contextual, and has a lot to do with our understanding and the experience of learning. The deeper I’ve gone into the field, the fewer areas of black-and-white I can still identify. This field is full of gray, and it requires our human experience to make sense of that gray and that context.

What I would love from you

Let me know your thoughts on writing as thinking versus writing as writing, and on these fascinating topics in our field.

Write to me at geoffstetson@mededmentor.org.

Thanks for reading! We are skipping next Monday, October 12, because of the holiday in the United States, so I’ll see you back here on October 19.