Using ChatGPT to write an essay is a bit like using a forklift to lift weights

Irish Times

a day ago

A couple of weeks back, I did a public event in a bookshop, for which I and two other writers were each required to pick three beloved books, and to talk about each of them for five minutes or so. Choosing the books I wanted to talk about proved an interesting challenge, because although I can easily think of any number of books I have read and loved, it is considerably less easy to think of books I have not only read and loved but can also remember well enough to talk to an audience about for five minutes. A major criterion for the books I chose, I have to admit, was that I knew them well enough to not have to re-read them for the event. I felt that I had in some sense internalised these books, in a way that could not be said for very many others I had read.
In the few days afterwards, I began to wonder why it was that I remembered certain books so well, and others barely at all. It is not uncommon for me to read and greatly enjoy a book and then, within a year or two, remember next to nothing about it other than, perhaps, that I once read and greatly enjoyed it. But there are books that, many years after reading them, have remained a presence in my life. And these books – the ones that seem to belong to some small and select psychic library, whose volumes collectively account for my basic sense of myself as a literate person – are, I realised, the books that I have written about.
The Information,
Martin Amis
's comedy of thwarted literary ambition and writerly jealousy, features a protagonist whose career as a novelist has devolved into ceaseless book reviewing – a job in which he takes no real pride, but which he nonetheless does very well. 'When he reviewed a book,' writes Amis, 'it stayed reviewed.' It's a line that comes to mind when I think about this subject. It is, for me, the act of writing about a book that causes it to stay read.
And a disproportionate number of the books that have stayed read for me are ones that I wrote about at university. I read Ulysses
and wrote an essay about it; the essay, I can assure you, was not very good, but the book stayed more or less read. I read and wrote about the stories of Jorge Luis Borges, and those stories stayed read. I read and wrote about Edgar Allen Poe, and Poe has stayed read. The writing of those essays was in some sense inseparable from the depth and durability of the reading.
READ MORE
[
Zuckerberg saying AI will cure loneliness is like big tobacco suggesting cigarettes can treat cancer
Opens in new window
]
I am saying all of this now because of two facts that seem fairly self-evident: the fact that the writing of essays is a central aspect of an education in the humanities, and the fact that this centrality is increasingly threatened as a result of the widespread use of
ChatGPT
and other LLM (large language model) technologies.
The general feeling among university administrators, if not necessarily among academics as a whole, seems to be that it would be futile to try to stop students using AI in their work. The technology exists, and in the narrow sense of producing a functional piece of writing on a given topic, it's an effective tool, and it's not going anywhere. I've spoken to a few academics in the humanities recently who seem resigned to (though by no means at peace with) the idea that assessing students through essays may no longer be a viable pedagogical approach. No one seems quite sure what will replace essay writing, but it seems likely that something – a greater emphasis on written exams, perhaps, or some form of oral assessment – will have to.
I wasn't a particularly industrious student as an undergraduate. I half-assed a lot of my courses, and often didn't do nearly enough reading of supplementary material – works of academic literary criticism and other secondary sources – to give my essays a plausible veneer of academic credibility and rigour. The essays I wrote were not especially good, even by the standards of undergraduate essays. But I realise now that their being good or bad was of secondary importance to the writing of them.
The writing of essays seems to me to have two main uses as an educational tool. It is useful as a means of assessing a student – how much they know, how widely and deeply they have read in a subject, and how rigorous and original their thinking on that subject is. The other is both less measurable and more significant: in writing an essay, you find out what you think about a subject; you learn, in some sense,
how
to think about it.
[
AI is already a focus of endless delusion, magical thinking and plain old foolishness
]
In a recent study conducted by MIT's Media Lab, three groups of participants, aged 18-39, were asked to write essays using, respectively, ChatGPT, Google's search engine and no technology at all. The brain activity of the participants was measured using EEG. Of the three groups, the ChatGPT users consistently had the lowest level of brain engagement, and 'consistently underperformed at neural, linguistic and behavioural levels'. As the study progressed, the group using ChatGPT got steadily lazier with each subsequent essay, often simply copying and pasting the text produced by the LLM, rather than using it as a source for their own work.
One way of thinking about this, from an educational perspective, is that writing an essay is to one's intellect as lifting weights is to one's muscles. The point of going to the gym is not to get good at lifting weights; it's to train your body to become stronger. Using ChatGPT to write an essay is a bit like using a forklift to lift weights. The forklift might do a perfectly good job of moving around some heavy iron plates, but you'd be wasting your time. (You would also be causing a serious disruption, and would probably have your gym membership cancelled – and rightly so.)
Just as you can't get someone, or something, to work out for you, there is nobody and nothing that can think on your behalf. As with so many of the supposed benefits of so-called artificial intelligence, it's not clear what we're actually gaining. What we stand to lose is so large, and so fundamental, as to be incalculable.