Amid all the doomsaying and soul-searching about what AI will do to academia, here is a positive story (and I know it’s a luxury position to have such small classes, and this cohort was particularly engaged and collaborative). I taught in a history of ideas class this year in our PPE master, and decided to go for essays as exams – but really loathed the prospects of having to read AI stuff. I get lots of AI-generated scam emails (suddenly, a book club in Australia really really wants to learn more about a volume on the reception of Hegel’s philosophy that I edited as a grad student…), and I really hate the bland, corporate, and sycophantic tone. For environmental reasons, I try to minimize AI use to where it might really have added value and save me some time – which, according to my non-representative sampling with n=1, is not very often. But back to the students.
As per our faculty policy, I requested an abstract to be sent to me for feedback beforehand, and planned a short oral checks with all students. I got 22 essay, none of which was AI written (or done so well that I didn’t notice at all). They were close to the texts we had read, asked intelligent questions about them, many brought in additional topics that they found interesting and applied the ideas of historical thinkers to them in a really thoughtful way. I got a sense from the students that they had made it a matter of honor not to use AI, but rather to take advantage of the opportunity to learn to write good essays.
I did not have to do any “checks”, in the end, but could have meaningful conversations with them about either the content, or about what it takes to write good philosophical texts. In the past, I had offered such feedback moments, together with written feedback, on a voluntary basis, but students hardly ever came. I realized that I should always have made them mandatory – the added value for the students seems quite high, and the time it takes for me is, in the grand scheme of things, negligible (again, the luxury of small classes…).
More generally speaking, my initial worries about how AI will undermine university teaching have changed in the last months. I remain convinced that what Dewey calls “the community of inquiry” retains its attraction (I here owe a lot to Éloise Soulier and Jason Branford, with whom I wrote a policy paper on AI in education and learned a lot about Dewey and education from them). It’s a social technique that is probably as old as human language: a few people sit together, in a room or in a cave or around a fire, and they discuss about things that matter to them. Those who already know more tell those who know less what they have learned over the years, they explain to them why certain assumptions might not hold, or why something is a logical fallacy. They use tools such as drawings, schemata, or written notes (once scripture was invented) to faciliate the learning process. They allow for discussion and questioning, because this is the best way for the human brain to learn and discover new things.*
What I no longer worry about – encouraged also by my fabulous students – is that this social technique would disappear. But what I do worry about is that it might become a privilege, even more than is currently already the case. Small-scale, maybe “electronics free” classes for the few, AI-based training and multiple choice tests for the many. Some governments might think: why pay for teachers for the hoi polloi if you can have personalized AI assistants?
But there is an interesting twist here. Those environments that are so resource poor that they cannot afford giving every student a digital device with internet access might, paradoxically, end up with better learning conditions than those where there is money (and political determination) to invest in AI for education.
It’s almost too painful to imagine a world in which even a fraction of what is currently being invested in AI was invested in human teachers. But maybe we could, for a second, imagine such a world? Or maybe we could imagine a world in which for every dollar invested in AI, there is a 20% tax that will be channelled directly into teacher training? (that’s an arbitrary number, it wouldn’t have to be a high %, at the moment, to have huge impact). Arguably, something like that would be a fair return to the teaching profession. It has, after all, been behind all the texts that Large Language Models are trained on, in the sense that it has taught someone, somewhere, to read and write!
* I’m not implying that there is no use at all for AI in education. But as we argue in the policy paper, AI often avoids precisely that (cognitive or interpersonal) friction that is most conducive to learning.
{ 0 comments… add one now }