“I am a big believer in not letting robots dream for us”, Nicolas Cage.
“I want AI to do my laundry and dishes so that I can do art and writing, not for AI to do my art and writing so that I can do my laundry and dishes”, Joanna Maciejewska.
“The days of spending long hours reading through extensive academic content are over.” hix.ai.
In my early days online, there was a small blog that I loved. Every week or so, the blogger would share a new film review. The review would read as entirely sincere and serious but there was a catch – the review was only based on the film’s poster. The blogger would take a look at the poster and then make sometimes reasonable, sometimes wild assumptions about what the film would be about and how the whole thing would land. It was a funny read (and unfortunately one that apparently is long gone, I couldn’t find it to share now).
Those blog posts have been on my mind recently as those vague approximations, lucky guesses and massive misfires are characteristic of a more contemporary and less often amusing part of our online architecture – Generative AI.
GenAI is now seemingly embedded into everything (why is that when I go onto the search function on Whatsapp, I am now met with suggested Meta AI search terms ‘I want to write a birthday wish’, ‘I want to write an email’, ‘I need help with my school project’? Who was asking for this?). That ‘everything’ includes many aspects of contemporary higher education. This is a phenomenon that I feel genuinely uneasy about and it’s one where I think we need to be much more willing to speak with our students about.
There may be some genuinely useful uses of GenAI in education. However, the uses of GenAI in the classroom that I have seen and thought are interesting are mainly ones that are engaging critically with GenAI (for example, feeding questions into AI, counting the mistakes it makes, and asking students to try and identify that number of mistakes in work as part of a quiz, as practiced by Dr. Ronnie Olesker) rather than finding genuine use in it as an educational tool.
One of the concerns that I have with GenAI is that it puts an additional wall between the student and the education that they purportedly desire. This is something that was particularly well demonstrated in a practice detailed in Corbin and Walton’s recent paper in Higher Education Research & Development on the rise in use of GenAI “summarisers” – tools that students use to bypass reading a journal article or book chapter by getting the GenAI to generate a 300-400 summary of the piece.
Unsurprisingly, the ascent of GenAI has been accompanied with a sharp increase in concern over academic integrity (Kumar et al. 2024 have a good overview of this debate). I understand where these concerns come from but I also think that we can’t and probably shouldn’t want to be anti-GenAI cops. Whilst there are telltale signs of GenAI usage (with apologies to innocent users of the em dash and ‘delve’), a modicum of effort can be used to help a student cover their GenAI tracks. The detection software that we used for plagiarism checks in the past is simply not up to the task – the software that claims to be able to detect GenAI is also driven by AI and, shocker, it can also be completely incorrect in its diagnosis and produce false positives.
So, if we can’t prove GenAI usage, it also means we can’t ban GenAI usage (and again, we’re not the cops), so what options does that leave us with?
One of the key tools in our arsenal is to speak with students and make the case against GenAI usage. As Ted Chiang recently argued in The New Yorker: ‘Using ChatGPT to complete assignments is like bringing a forklift into the weight room; you will never improve your cognitive fitness that way.’ This is a framing that could be useful for those conversations with students (there is evidence that this framing is also correct – that trusting genAI lowers critical thinking). Afterall, they elected to come and study politics at university. There clearly was something that once really excited and motivated them in engaging with the subject. Why do they now want GenAI to do that engagement for them? If they are asking ChatGPT to do their work, they aren’t going to develop the skills that they wanted from their degree but are surely also depriving themselves of the potential for joy that is part of education.
Part of the solution to this is also signposting the struggle – that yes, like it or not, sometimes it’s going to be hard. There are very few people who are born natural journal article readers. However, to quote Robert Frost ‘the best way out is always through’, i.e. if students want to find it easier to read journal articles, asking GenAI to do it for them is not going to work, they need to embrace the struggle, read and read, and soon they will find that they are through that struggle and can actually get on quite easily with their reading. To go down a few lines in Frost’s A Servant to Servants: ‘We didn’t change without some sacrifice’.
We need to talk with students about why taking a shortcut with GenAI is not going to benefit them in the long run. It is our responsibility to ensure that they understand that assessment is not only a test of learning, it is also a site of learning. Students using GenAI to produce work are not only trying to cheat a test of learning, they are also excluding themselves from an important moment of learning. To me, this is the core of the argument against students using GenAI.
When students try to avoid that sacrifice or struggle they will have, ultimately, cheated themselves. It feels like such a twee point but it is also one that gets lost in other arguments and that, ultimately, might be one that is particularly resonant with students who did come to university with some hopes that it would benefit them.
Certainly, there are a litany of other arguments to make against the use of GenAI. We could, for example, discuss its disastrous environmental impacts with the energy intensity that is needed to generate responses. We could also discuss the fact that, despite all that energy, the responses it actually generates can often be absolutely useless – worse than useless really because people do use it – invented histories, hallucinated references, and some of the most bland, vapid writing you can find (as I sat writing this blog post in a campus cafe, I got chatting to a colleague who had a key summary about GenAI – “it’s a bit shit”).
Part of our response to GenAI in education needs to be squarely centred on a simple question – what do we want the benefits of our education to be? If it is “to be good at writing prompts for ChatGPT” then so be it but I would guess that it’s not. We want graduates that are skilled enough that should they actually find use in GenAI in the world of post-degree work, they will be the type of adaptable, capable people who can quickly pick up the tool and get going with it.
With rumours circulating that the Trump regime’s recent tariff calculations may have been the result of GenAI, there is a good context for speaking to our students about the flaws in generated responses and the risk of uncritical acceptance of those outputs. This is a risk of which politics students need to be especially aware. If GenAI usage is too readily normalised during the course of their education, I doubt they will have the right level of suspicion when considering their own or others’ usage of it (see the above point about trusting GenAI output and reduced critical thinking). As educators, we have a role to play in how students perceive GenAI usage.
But beyond the use case, we need to speak to students about putting the effort in to find pleasure in their work. Corbin and Walton’s article discusses a research agenda on GenAI summarisers – one thing I felt was missing was student enjoyment. Maybe there is pleasure in not taking the shortcut and actually spending long hours reading through extensive academic content. Maybe the use of GenAI summarisers and generated essays means students are being deprived of a challenging and enjoyable academic experience.
Whilst the blogger who wrote film reviews based only on that film’s posters was doing it for a laugh, I don’t think anyone would want to have posters actually replace sitting down to watch a film and getting the full experience. This might be stretching a point a bit (how many journal articles have you ever thought of as cinematic?) but still there is a core here of putting that bit more time in and getting a lot more out for that investment.
Maybe in a few months or years time I’ll look back on this post and think that I was being a total Luddite for my anti-AI resistance. For now, I do know that I enjoyed reading the entirety of Corbin and Walton’s paper and that come the release of the new Mission Impossible film in cinemas later this month, I won’t be asking ChatGPT to summarise the film for me.