Something in this week’s Wonkhe discussion about AI and human learning caught my attention. The central premise, put fairly, is that artificial intelligence tilts educational value away from cognitive processing towards the things that only have genuine meaning when humans do them: ethical positioning, real connection, situated professional judgement. It is an important and largely persuasive argument. But I find myself pausing on an assumption that seems to underpin it.
Has higher education, as it has actually been practised and organised in recent decades, ever really prioritised those human dimensions? Or have we, quietly and incrementally, been building something rather different?
I ask this not as provocation but as a genuine question. Over the course of my career to date, I have been close to how universities actually organise learning. That architecture, if I reflect on it honestly, has generally been optimised for qualities that sit in some tension with the ‘human’ values now being described as education’s differentiator.
Standardisation as the operating system
Quality assurance in UK higher education depends on comparability and replicability. These are not unreasonable foundations. A system that awards equivalent classifications for very different standards of work is neither fair nor credible, and mass higher education requires common frameworks if it is to function at scale. The challenge is that comparability pulls steadily towards what can be measured, and replicability pulls towards what can be reproduced. Both tendencies favour the individual, the discrete, the assessable. They sit in real tension with learning that is collective, embodied or genuinely resistant to standardisation.
Mark Peace’s observation in this week’s Wonkhe briefing about the ‘myth of the lone student’ is striking precisely because it is accurate. The assessment architecture of most UK universities rests on the assumption that learning is something individuals do alone and then demonstrate alone. Academic integrity policies, individual submission requirements and the design of most summative assessment reflect this assumption consistently. It is a choice, albeit a largely unexamined one.
And it is not a choice that AI has created. The tension was there long before generative AI arrived. AI has simply made it harder to ignore.
The situated professional dimension
The Wonkhe briefing notes that AI cannot offer a situated professional perspective, even though it can produce something that looks disturbingly like one. This observation resonates strongly with me. Genuine professional expertise is not the application of general principles to specific situations. It involves tacit knowledge built through experience, ethical judgement developed in conditions of real consequence, and the capacity to read context in ways that draw on everything a practitioner has lived and learnt. This kind of expertise takes years. It cannot be acquired by processing text, however vast the corpus.
Yet if I reflect candidly on how higher education has actually approached the development of professional judgement, I am not certain we have been as deliberate as we might have been. Employability frameworks, graduate attributes and skills taxonomies have tended to treat professional capability as something that can be catalogued, measured and certified. That is not entirely wrong. But it may have given us a degree of false confidence – that we were developing situated professional judgement when we were, in some respects, developing something closer to a plausible approximation of it.
AI produces plausible approximations with considerable skill. The question is whether we have been sufficiently honest with ourselves, and with students, about what distinguishes the real thing from something that merely resembles it.
The pressure that AI creates
There is, perhaps, something useful in the discomfort that AI is generating for the sector. The challenge it poses is not primarily about plagiarism or academic integrity, significant as those concerns are. It is a deeper challenge to the value proposition of what universities have been offering. If the system was already tending towards replicable, comparable, individually demonstrable outputs, and AI can now produce those outputs at speed and scale, then the pressure AI creates may be exactly the pressure that was needed.
That pressure might encourage institutions to take seriously the things they have long claimed to value but found difficult to assess: genuine collaborative learning, ethical development as a process rather than a statement of intent, the slow and sometimes uncomfortable acquisition of professional judgement in conditions of real stakes. These things have featured prominently in mission statements and graduate attribute frameworks for years. Whether the curriculum and assessment architecture has consistently supported them is, I think, a more open question than is sometimes acknowledged.
A question worth sitting with
The most constructive thing AI may have done for higher education is not to threaten it but to force a harder look at what it has actually been doing. If the human dimensions of learning genuinely matter and I believe they do then the challenge is not only to protect them from AI. It is to ask whether they were sufficiently protected before AI arrived.
I do not think that question has a comfortable answer. But I suspect it is one that institutions, and those of us who have shaped their learning architectures, have a particular responsibility to sit with.