Following on from the first wave of panic about generative AI in higher education, the mood has shifted.
The early scramble – cheating warnings and the rush to detection software – has given way to something quieter and harder to ignore.
Most of what AI has done is highlight the gaps already built into the system. That argument has been made in these pages more than once. The question now is whether higher education institutions will actually change anything.
Taking writing seriously
If we accept that academic writing is a teachable discipline that we have largely stopped teaching, the harder question is what taking it seriously again would involve. The sector has yet to find a way to answer this.
We should be specific about what has actually been lost. Academic writing was never absorbed by osmosis. For most of the modern history of British higher education, students were taught to write through structures that no longer exist at scale, principally the small tutorial in which a supervisor read student work carefully and discussed it with them. The essay was not simply a product to be marked. It was the occasion for a conversation in which a student’s thinking could be tested by someone who knew the field. The writing was the medium; the teaching happened around it.
Mass higher education has hollowed out those structures without putting anything in their place. Supervisors carry heavier loads and feedback on returned work shrinks to a kind of shorthand – “be more critical,” “develop your argument,” “needs more analysis,” “unclear” – phrases that assume the student already knows what to do with them.
From describing to analysing
Where structured writing teaching survives, it has usually been moved sideways: to writing centres on small budgets, or to “academic skills” units handling everything from referencing to time management. The result is that the actual work of writing is rarely taught to anyone in any sustained way: how an argument is built across paragraphs, how reasoning is signalled, how counter-evidence is handled, how a student moves from describing to analysing.
International and first-generation students are most visibly disadvantaged by this, but the problem is general. It is no longer unusual to meet doctoral students who have never been shown, line by line, how to build a paragraph that does anything more than describe.
Here is the kind of sentence that turns up week after week in the theses I edit:
Whilst new interventions have proved useful for framing our attitude to how pre-colonial West African populations formed, problems arise when we think about the functioning of societies whose longevity either was not a foregone conclusion or that, in some cases, outright collapsed or disbanded.
The grammar is intact and the vocabulary is appropriately academic; an over-stretched marker would let it through, and any editing tool would barely touch it. But the student’s thinking is buckling under its own weight.
There are at least two arguments in there, possibly three, fighting for the same syntax; the sentence cannot carry them all. That sentence is the most useful thing in the draft. It is where the teaching could happen, if anyone had time to sit with the student and ask which of the claims is the actual claim. In a tutorial, this would mean underlining the main verbs and asking the student to say in one plain sentence, which of those claims they actually wanted to make. In current conditions, almost nobody does.
A teachable skill
I taught professional writing at the University of Groningen for more than five years and have spent twenty years as an academic editor, working mostly through theses, journal articles and book manuscripts written by non-native English speakers. When you teach writing explicitly – when you sequence it properly and show students the choices sitting behind each sentence – they learn. They can do it. The reason most cannot do it on arrival is not that they are deficient or that the standards are unfair. It is that nobody has ever shown them. Once shown, they learn. There is nothing exceptional about this. It is how we teach everything else.
There is a counter-argument that deserves a direct response, because it is increasingly heard and rarely answered well. It runs roughly as follows: that insisting on academic writing standards is itself a form of cultural gatekeeping, and that AI usefully disrupts a tradition that was always more about exclusion than about thought.
Aristocratic standards
I understand why the argument has traction. It is right to be suspicious of the ways “good writing” has been used as a sorting mechanism. But on closer inspection the politics are inside out. The students who suffer most when academic conventions are left tacit – treated as something one is simply supposed to have – are precisely the working-class and international students the gatekeeping critique claims to defend. Tacit standards are, in practice, aristocratic standards. They reward students who arrive already fluent. Everyone else is quietly marked down for not having absorbed something nobody set out to teach them.
Making the conventions explicit and teachable is the democratising move. Refusing to teach them, on the grounds that the teaching is itself exclusionary, leaves the system as it is: a system in which only the already-equipped succeed.
No more workshops
What, in practice, would teaching writing again involve? Not more workshops. Not better detection software. It would mean treating writing as what it is – a discipline that can be taught with the same seriousness we apply to statistics or research methods, with real institutional standing rather than a corner of a skills unit operating in the margins of the curriculum.
A couple of questions for the people who actually make these decisions. Heads of department: when did you last appoint someone whose explicit remit was to teach writing? Programme directors: how many of your supervisors have had training in how to respond to a draft? Students do not need a corrective applied to their finished work. They need a properly sequenced education in which the decisions behind academic argument are made visible and practised on the page. Much of this has been done before, in places where the resources for tutorial teaching still exist. The challenge is to build something equivalent at scale.
What AI has forced into the open is not, in the end, about AI itself. It is that we have been running an academic writing system on the assumption that students would somehow acquire capacities we never actually taught them. The assumption was always shaky. AI has only made its consequences impossible to ignore.
What the sector needs now is not a cleverer shield against the technology. It is to teach the thing.
