LAST UPDATE | 36 mins ago
IT IS A Sunday, and I am at the kitchen table with a stack of postgraduate assignments and a cooling cup of tea.
The references are clean, I’m happy the structure works, and even the prose is competent. Yet something is missing, call it the absence of struggle, or the absence of a person behind the page. I have marked enough of these over enough years to know what a mind working through difficulty actually looks like in writing, and this is not it.
Colleagues of mine, writing here recently, called this the front line of a war on human thought. It is a serious argument, made seriously, and parts of it I agree with. But the assignments worrying me most on a Sunday afternoon are not the ones that smell of ChatGPT. They are the ones that always read like this, polished, correct, hollow, long before any chatbot existed.
That is the part of this debate we are not having.
Let us concede the obvious. Writing is thinking, and students who use AI to skip the struggle are robbing themselves of something they cannot get back. Smartphones have done damage to attention spans that predate this whole debate by a decade. The instinct to protect the classroom as a place where something slower and harder can happen is a good one, and I share it.
But the harder question, the one underneath the AI panic, is this. Were the assignments we set actually measuring thinking? Or were they measuring something easier to measure that we hoped was a reasonable proxy? Because if AI can pass them in 30 seconds, that tells us something uncomfortable about the assignment, not just about the tool.
An analogue education
The classic university essay, written alone, over weeks, graded against a checklist, was not invented because it was the best way to know if someone could think. It was invented in an era when the library was the bottleneck, when only the diligent could pull a coherent argument from scattered sources, when the artefact on the page was the only practical evidence we had. It was always a rough proxy.
Anyone who has taught long enough has marked technically excellent essays that were intellectually dead. We have all met students whose minds raced miles ahead of anything they ever submitted on paper, and others whose written submissions flattered an understanding that fell apart the moment you asked a question out loud. AI has not created that gap between the artefact and the actual learning. It has just blown it wide open.
The uncomfortable bit is not that students can now produce essays without thinking. It is that we built a system in which they probably always could, if they were determined enough. We just did not have to admit it.
I teach mid-career professionals in procurement and supply chain, people who run things in the real world. The assessments that work in that room are the ones AI is structurally bad at.
These include a live case discussion where a student has to defend a procurement decision under questioning from peers who have made similar calls and know where the bodies are buried.
A presentation to an industry panel who will spot a hollow argument in 30 seconds and ask the question you hoped no one would ask. A reflective piece tied to a real problem in the student’s own organisation, where the only useful expertise in the room is theirs.
Watching a student change their mind mid-sentence, or admit they got something wrong and rebuild the answer in real time, that is learning happening in front of you. It does not show up in a polished essay. It cannot be faked by a chatbot. And, crucially, it can be assessed.
A new path
Now, imagine the graduates we will produce if we follow the wall-building instinct.
We give them four years of locked phones and blue-book exams. We keep AI out of the seminar room entirely.
We hand them their degrees and send them into a working world where AI is in every email client, every spreadsheet, every procurement platform, every legal review, every clinical decision-support tool, every supply chain dashboard.
They arrive at their first job having never been taught how to use these tools well, how to spot when they are confidently wrong, how to push back on a plausible-sounding hallucination, how to hold onto their own judgment when surrounded by a constant stream of machine-generated output.
We will have produced graduates who are pristine and unprepared. Their employers are already asking the opposite question. Not “did you keep AI out?” but “did you teach them to think for themselves while using it?”
That is a much harder pedagogical problem than banning the tool. It requires us to design work where AI is permitted but unhelpful, where the value lies in the judgment applied to the output rather than the output itself, where the student has to explain and defend choices a machine cannot make for them. It requires us to admit, finally, that the artefact was never the point.
Circling the wagons buys us a quiet decade in the lecture hall. It produces a generation who will be eaten alive in the workplace, and rightly so.
Time to get real
To my colleagues making the opposite case, course readers, phone-free classrooms, real attendance, the seminar room as a protected analogue space, there is real value in much of what they propose, and I am not pretending otherwise. Protect the conditions for thinking, by all means.
But protection is not the same as development. You can run the most analogue seminar in Ireland and still hand out an assessment regime that rewards polished reproduction over genuine engagement. The fight is not between pro-AI and anti-AI.
It is between assessments that measure thinking and assessments that measure something easier. AI is forcing that conversation whether we want to have it or not, and in an odd way, it is doing us a favour, though it does not feel like one at three in the morning with a hundred assignments still to mark.
The stack of papers is still there. The discomfort of marking them is a useful signal. It is not AI’s fault that polished output without thought used to slip through. It is ours for grading the polish.
The serious response is neither retreat to a pre-digital past nor surrender to an algorithmic future. It is to ask, plainly, what we want our graduates to be able to do when they walk out the door, and then to build the work they do with us so the answer is visible in the work itself.
A university is meant to be a place to think. That means a place where we actually assess thinking. We have not always been doing that. AI is the inconvenient mirror.
Dr Paul Davis is a lecturer at Dublin City University’s Business School. He specialises in supply chain management and procurement.