Australia’s first national plan for artificial intelligence aims to upskill workers to boost productivity, but will leave the tech largely unregulated and without its own legislation to operate under.
Please login below to view content or subscribe now.
There’s been endless discussion about what students do with generative AI tools, and what constitutes legitimate use of AI in assessment, but as the technology continues to improve there’s a whole conversation to be had about what educators do with AI tools.
We’re using the term “educators” to encompass both the academics leading modules and programmes and the professionals who support, enable and contribute to learning and teaching and student support.
Realising the potential of the technologies that an institution invests in to support student success requires educators to be willing and able to deploy it in ways that are appropriate for their context. It requires them to be active and creative users of that technology, not simply following a process or showing compliance with a policy.
So it was a bit worrying when in the course of exploring what effective preparation for digital learning futures could look like for our Capability for change report last year, it was noticeable how concerned digital and education leaders were about the variable digital capabilities of their staff.
Inevitably, when it comes to AI, some HE staff are enthusiastic early adopters and innovators; others are more cautious or less confident – and some are highly critical and/or just want it to go away. Some of this is about personal orientation towards particular technologies – there is a lively and important critical debate about how society comes into a relationship with AI technology and the implications for, well, the future of humanity.
Some of it is about the realities of the pressures that educators are under, and the lack of available time and headspace to engage with developmental activity. As one education leader put it:
Sometimes staff, they know that they need to change what they’re doing, but they get caught in the academic cycle. So every year it’s back to teaching again, really, really large groups of students; they haven’t had the time to go and think about how to do things differently.
But there’s also an institutional strategic challenge here about situating AI within the pedagogic environment – recognising that students will not only be using it habitually in their work and learning, but that they will expect to graduate with a level of competence in it in anticipation of using AI in the workplace. There’s an efficiency question about how using AI can reprofile educator working patterns and workflows. Even if the prospect of “freeing up” lots of time might feel a bit remote right now, educators are clearly going to be using AI in interesting ways to make some of their work a bit more efficient, to surface insight from large datasets that might not otherwise be accessible, or as a co-creator to help enhance their thinking and practice.
In the context of learning and teaching, educators need to be ready to go beyond asking “how do the tools work and what can I do with them?” and be prepared to ask and answer a larger question: “what does it mean for academic quality and pedagogy when I do?”
As Tom Chatfield has persuasively argued in his recent white paper on AI and the future of pedagogy, AI needs to have a clear educative purpose when it is deployed in learning and teaching, and should be about actively enhancing pedagogy. Reaching this halcyon state requires educators who are not only competent in the technical use of the tools that are available but prepared to work creatively to embed those tools to achieve particular learning objectives within the wider framework and structures of their academic discipline. Expertise of this nature is not cheaply won – it takes time and resource to think, experiment, test, and refine.
Educators have the power – and responsibility – to work out how best to harness AI in learning and teaching in their disciplines, but education leaders need to create the right environment for innovation to flourish. As one leader put it:
How do we create an environment where we’re allowing people to feel like they are the arbiters of their own day to day, that they’ve got more time, that they’re able to do the things that they want to do?…So that’s really an excitement for me. I think there’s real opportunity in digital to enable those things.
For our new project “Educating the AI generation” we want to explore how institutions are developing educator AI literacy and practice – what frameworks, interventions, and provisions are helpful and effective, and where the barriers and challenges lie. What sort of environment helps educators to develop not just the capability, but also the motivation and opportunity to become skilled and critical users of AI in learning and teaching? And what does that teach us about how the role of educators might change as the higher education learning environment evolves?
At the discussion session Rachel co-hosted alongside Kortext advisor Janice Kay at the Festival of Higher Education earlier this month there was a strong sense among attendees that educating the AI generation requires universities to take action on multiple fronts simultaneously if they are to keep up with the pace of change in AI technology.
Achieving this kind of agility means making space for risk-taking, and moving away from compliance-focused language to a more collaborative and exploratory approach, including with students, who are equally finding their feet with AI. For leaders, that could mean offering both reassurance that this approach is welcomed, and fostering spaces in which it can be deployed.
In a time of such fast-paced change, staying grounded in concepts of what it means to be a professional educator can help manage the potential sense of threat from AI in learning and teaching. Discussions focused on the “how” of effective use of AI, and the ways it can support student learning and educator practice, are always grounded in core knowledge of pedagogy and education.
On AI in assessment, it was instructive to hear student participants share a desire to be able to demonstrate learning and skills above and beyond what is captured in traditional assessment, and find different, authentic ways to engage with knowledge. Assessment is always a bit of a flashpoint in pedagogy, especially in constructing students’ understanding of their learning, and there is an open question on how AI technology can support educators in assessment design and execution. More prosaically, the risks to traditional assessment from large language models indicate that staff may need to spend proportionally more of their time on managing assessment going forward.
Participants drew upon the experiences of the Covid pivot to emergency remote teaching and taking the best lessons from trialling new ways of learning and teaching as a useful reminder that the sector can pivot quickly – and well – when required. Yet the feeling that AI is often something of a “talking point” rather than an “action point” led some to suggest that there may not yet be a sufficiently pressing sense of urgency to kickstart change in practice.
What is clear about the present moment is that the sector will make the most progress on these questions when there is sharing of thinking and practice and co-development of approaches. Over the next six months we’ll be building up our insight and we’d love to hear your views on what works to support educator development of AI in pedagogy. We’re not expecting any silver bullets, but if you have an example of practice to share, please get in touch.
This article is published in association with Kortext. Join Debbie, Rachel and a host of other speakers at Kortext LIVE on Wednesday 11 February in London, where we’ll be discussing some of our findings – find out more and book your place here.
I loved magic as a kid. Card tricks, disappearing coins, little felt rabbits in pretend top hats. “Now anyone can be a magician,” proclaimed the advert in the Argos catalogue. Ta da. Now that’s magic.
I’d make pretend tickets, rearrange the seating in the front room, and perform shows for the family – slowly learning the dark arts of misdirection and manipulation along the way. When I performed, I generated pride.
Over time I found that some of those skills could be used to influence people more generally – to make them feel better about themselves, to change their decisions, to trigger some kind of behaviour.
Sometimes, I’d rationalise, as long as I was doing it for the right reasons, it was better if they didn’t know it was a trick. The end justified the means. Or did it?
People love it when they know that magic is being performed as magic – the willing suspension of disbelief, the pleasure of being fooled by someone who’s earned the right to fool us. When they give permission to be illegitimately impressed, all is fine.
But what they can’t stand is being lied to. We don’t like being deceived. Most political news in this country centres on who lied and about what. We’re obsessed with it.
The cover-up is always worse than the crime, yet everyone still does it – they have to, they rationalise, to keep up, or to get permission. The gap between how things are and how we present them is the game.
Once they’re in, it won’t matter that the sector painted an unobtainable picture of student life for applicants. Once the funding is secured, universities can fess up that it isn’t as good as government thought it would be later. Once the rules are published, better to ask for forgiveness over the impact on net migration – not permission.
I think about that little magic set I got a lot, because so much of what AI does still sits for me in that “magic trick” space.
Ta da. Look what it can do. Generate an essay, write a play, create some code, produce an image of the Pope in a puffer jacket. But the line between magic and lies is a slippery slope, because its number one use case is pretence.
AI is used to lie – fake essays, fake expertise, fake competence. But mostly to make us look better, appear faster and seem wiser. The anxiety about being “found out” is the anxiety of the liar, not the audience at a magic show. Students worry they’ll be caught. Universities worry their degrees will be worthless.
Everyone worries that the whole edifice of qualifications and signals and “I know something you don’t” will collapse under the weight of its own pretending. But the pretending was already there – AI just makes the tricks cheaper, and much harder to sustain.
I’ve been in a particularly reflective mood recently – I turned 50 at the weekend (I can’t believe it either, it’s the moisturiser) and there’s something I can’t shake. When I look back upon my life, it’s always with a sense of shame.
When I got accepted to the University of the West of England in the mid-nineties, grandparents on both sides were thrilled that I had “got into Bristol”. A few extra Bonusprint copies of the sunken lawn at the St Matthias campus helped.
It hadn’t started as a deliberate lie – more a misunderstanding about where we had driven to on open days – but instead of correcting it, I doubled down.
Nobody in my family had been to university, and I doubt they would have discerned the difference. But on some level I thought I had to prove that their financial support was for something rare. Something… special.
Decades later I realised that the entire edifice of higher education runs on the same kind of slippage – the gap between what universities actually do, and the status they are assumed to have and confer.
Applicants and their families celebrate “getting in” as if admission itself were the achievement. Parents frame graduation photos, the ceremony mattering more than the three years that preceded it. Employers use degree classifications as sorting mechanisms while moaning that the sort has not delivered the graduates they wanted. There’s a graduate premium. And so on.
Those of us who write about higher education are no better. Our business model rests on “I know something you do not” – the insider knowledge, the things you haven’t noticed, the analysis you can’t get elsewhere. Scarcity of information, monetised. I’ve built a career on being the person in the room who has read the regulatory guidance.
But now, suddenly, a machine can summarise the guidance in seconds. Not as well as I can – not yet, not always – but well enough to make me wonder what I am actually for. What value I bring. How good I am at… pretending.
AI doesn’t create that anxiety. It exposes something that was always there – the fear that our value was never in what we knew, but in other people not knowing it. And that eventually, someone might find that out.
Back in 1995 my first (handwritten) university essay was about the way the internet lets you become someone you are not. Chatrooms were new and identity was suddenly fluid. You could lie about everything – your age, your appearance, your expertise – and checking was hard.
The internet has been flooded with exaggeration ever since. Wish.com tat that looks nothing like the picture. LinkedIn profiles that bear no relationship to actual jobs. Influencers selling lives they don’t live in places that barely exist.
But it has also liberated us. At UWE, I lived through the transition from index cards in libraries to DogPile, asking Jeeves and Google. The skill of navigating a card catalogue, of knowing which reference books to check – it felt essential, and then it was worthless. For one semester, we were told we weren’t allowed to use search engines. The faculty held on for a while, then let go.
In my first year, I chose a module involving audio editing on reel-to-reel tape. Splicing, cutting, winding, knives. At the end of the year, I got a job helping to put the equipment in a skip. The skills I’d learned were obsolete before I graduated.
Each time, there was a period of pretending that the old skills still mattered. Each time, the system eventually admitted they didn’t. Each time, something was revealed about what had actually been valuable all along. The card catalogue wasn’t the point – finding and evaluating information was. The handwriting wasn’t the point – thinking under pressure was. The reel-to-reel wasn’t the point – understanding how to shape a story with sound was.
Now the sector clings on to exams, essays, and the whole apparatus of assessment that assumes that producing a thing proves you learned something. The system holds on – but for what?
If I rummage through the AI pitches that land at [email protected], I can see a familiar pattern.
There are catastrophists. Students are cheating on an industrial scale. The essay is dead. Standards are collapsing and students are cognitively offloading while the great plagiarism machine works its magic.
There are tech evangelists. Productivity gains, personalised learning, democratised access and emancipation – just so long as you don’t ask who is selling the tools, who is buying the data, or what happens to students who can’t afford the premium tier.
Then there is the centrist-Dad middle. “It is neither all good nor all bad” – balance, nuance, thoughtful engagement, and very little about what any of this is actually for.
The catastrophists are wrong because they assume what’s being bypassed was valuable – that the essay-writing, the exam-sitting, the problem-set-completing were the point rather than proxies for something else. If the activities can be replaced by a machine, what were they measuring?
The evangelists are wrong because they assume more efficiency is always better – that if AI frees us from X, we’ll have more time to do Y. But they never say what Y is. Or whose time it becomes. In practice, we know – the efficiency dividend flows upward, and never shows up as an afternoon off.
The balanced view is just as bad, because it pretends there’s no choice to be made. It lets us sound reasonable while avoiding the harder question – what is higher education for?
At the high risk of becoming one of those bores at a conference whose “question” is a speech about that very issue, I do think there is a choice to be made. We ought at least to ask if universities exist to sort and qualify, or to form and transform. AI forces the question.
Let’s first admit a secret that would get me thrown out of the Magic Circle. The industrial model of education was built on scarcity, and scarcity made a certain kind of pretending possible.
Information was scarce – held in libraries, transmitted by experts, accessible only to those who got through the door. A degree meant three years in proximity to information others could not reach.
Attention was scarce – one lecturer, two hundred students, maybe a weekly seminar. The economics of mass higher education turned teaching into broadcast, not dialogue, but the scarcity, coupled with outcomes stats from the past, still conferred value.
Feedback was scarce – assignments returned weeks later with a grade and a short paragraph. The delay and brevity made the judgement feel weighty, even oracular.
In a scarcity system, hoarding makes sense. Knowledge is power precisely because others don’t have it. “I know something you do not” isn’t a bug – it’s the business model. But once something isn’t scarce any more, we have to search again for value.
We’ve been here before. Calculators didn’t destroy maths – they revealed that arithmetic wasn’t the point. Google didn’t destroy research – it revealed that finding information wasn’t really the hard bit. Each time the anxiety was the same – students will cheat, standards will collapse, the thing we valued will be lost. Each time the pretending got harder to sustain.
For me AI fits the pattern. Not because it knows everything – it obviously doesn’t. Its confident wrongness is one of its most dangerous features. But it makes a certain kind of information effectively free. Facts, frameworks, standard analyses are now available to anyone with an internet connection and the wit to ask.
And it hurts to carefully build and defend systems that confer status on things humans can do – only to have something come along and relieve humans from having to do them. It causes a confrontation – with value.
During the early days of Covid, I came across a Harvard Business School theory called Jobs To Be Done. People pay to get a job done, but organisations often misunderstand the real job they’re being paid to perform.
As a kid, the Sinclair ZX Spectrum in our house was marketed as an educational tool – an invitation to become a programmer. Some did. Most, like me, worked out how to make the screen say rude words and then played games.
Students have at least two jobs they want done. One is access to well-paid and meaningful work, made possible through obtaining a degree and supplied by academic programmes. The second is coming of age – the intoxicating combination of growing up and lifestyle. Becoming someone. Finding your people. Working out who you are when you’re not defined by your parents or your school.
Universities have always provided both, but only dare attribute value to the first. The second is treated as incidental – “the student experience”, something that happens around the edges. But for many students, perhaps most, the second job is why they came. The qualification is the price of admission to three years of transformation.
AI increasingly handles the first job – the information, the credentials, the sorting – more efficiently than universities ever could. If that were all universities offered, they’d already be obsolete. What AI can’t provide is the second job. It can’t help us become someone. It can’t introduce us to people who will change our lives. It can’t hold us accountable, or surprise us, or make us brave.
During Covid, I argued that universities should cancel as much face-to-face teaching as possible – because it wasn’t working anyway – but keep campuses open. Not for teaching – for being. For studying together, bonding, bridging, belonging.
I’ve not changed my view. AI just makes it more urgent. If the content delivery can be automated, the campus has to be for something else. That something else is formation.
A couple of years ago I came across Thomas Basbøll, resident writing consultant at Copenhagen Business School Library. He argues that when a human performs a cognitively sophisticated task – writes a compelling essay, analyses a complex case, synthesises disparate sources – we infer underlying competence. The performance becomes evidence of something deeper.
When a machine performs the same task, we can’t make the inference. The machine has processes that produce outputs. It doesn’t “know” anything – it predicts tokens. The output might resemble what a knowledgeable human would produce, but it proves nothing about understanding.
Education has always used performance as a proxy for competence. Higher education sets essays because it assumed that producing a good one required learning something. There was trust in the inference from output to understanding, and AI breaks it. The performance proves nothing.
For many students, the performance was already disconnected from competence. Dave Cormier, from the University of Prince Edward Island, described the experience of essay writing in the search era as:
have an argument, do a search for a quote that supports that position, pop the paper into Zotero to get the citation right, pop it in the paper. No reading for context. No real idea what the paper was even about.”
There was always pretending. AI just automated it.
Basbøll’s question still haunts me. What is it that we want students to be able to do on their own? Not “should we allow ChatGPT” – that battle is lost. What capacities, developed through practice and evidenced in assessment, do we actually care about?
If the answer is that appearing literate is enough, then we might as well hand the whole thing to the machines. If the answer is that we want students to actually develop capacities, then universities will need to watch students doing things – synchronous engagement, supervised practice, assessment that can’t be outsourced. A shift that feels too resource-intensive for the funding model.
What’s missing from both options is that neither is really about learning. One is about performing competence, the other is about proving competence under surveillance, but both still treat the output as the point. The system can’t ask what students actually learned, because it was never designed to find out. It was designed to sort.
How hard should education be? The “meritocracy of difficulty” ties academic value to how hard a course is to survive – dense content, heavy workloads, high-stakes assessment used to filter and sort rather than support students. Go too far in the other direction, and it’s a pointless prizes-for-all game in which nobody learns a thing.
Maybe the sorting and the signalling is the problem. The degree classification system was designed for an elite era where classification signalled that the graduate was better than other people. First class – exceptional. Third – joker. The whole apparatus assumes that the point of education is to prove that your Dad’s better than my Dad. See also the TEF.
Everyone pretends about the workload. The credit system assumes thirty-five to forty hours per week for a full-time student. Students aren’t studying for anything like that. The gap is vast, everyone knows it, and nobody says it out loud because saying it would expose the fiction.
AI intensifies it all. If students can automate the drudgery, they will – not because they’re lazy, but because they’re rational actors in a system that rewards outputs over process. If the system says “produce this essay” and the essay can be produced in ten minutes, why would anyone spend ten hours?
Mark Twain might have said that he would never let his schooling interfere with his education. Today’s undergraduates would more often lament that they don’t can’t their lectures and seminars interfere with their part-time job that pays the rent.
There’s a YouTube video about Czech railways that’s been stuck in my head for weeks now. They built a 200 km/h line between Prague and Budweis and held celebrations – the first domestic intercity service to break the 160 km/h barrier.
But only one train per day actually runs at that speed. It arrives late every time. Passengers spend the whole journey anxious about missing their ten-minute connection at the other end.
The Swiss do it differently. The Gotthard Base Tunnel was built for 230 km/h. Trains run at 200. The spare capacity isn’t wasted – it’s held in reserve. If a train enters the tunnel with a five-minute delay, it accelerates and emerges with only two. The tunnel eats delays. The result is connection punctuality of the kind where you almost always make your connection.
The Czech approach is speed fetishism – make the easily marketable number bigger, and assume that’s improvement. The Swiss approach is reliability – build in slack, prioritise the journey over the metric, make sure people get where they’re going.
It sometimes feels to me like UK universities have gone the Czech route. We’re the envy of the world on throughput – faster degrees, more students, tighter timetables, twelve-week modules with no room to fall behind.
But when anything goes wrong – and things always go wrong – students miss their connections. A bad week becomes a failed module. A failed module becomes a resit year. A mental health crisis becomes a dropout. Then we blame them for lacking resilience, as if the problem were their character rather than a system designed with no slack.
The formation model is the Swiss model. Slow down. Build in reserves. Let students recover from setbacks. Prioritise the journey over the metric. Accept that some things cannot be rushed.
Universities tell themselves similar lies about academics. It’s been obvious for a long time that the UK can’t sustain a system where researchers are also the teachers, the pastoral supporters, the markers and the administrators.
The all-rounder academic – brilliant at research, compelling in lectures, attentive in tutorials, wise in pastoral care, efficient at marking, engaged in knowledge exchange – was always a fantasy, tolerable only when student numbers were small enough to hide the gaps.
Massification stretched it. Every component became more complicated, with more onerous demands, while the mental model of what good looks like didn’t change. AI breaks it.
If students automate essay production, academics can automate feedback. We’re already seeing AI marking tools that claim to do in seconds what takes hours. If both sides are pretending – students pretending to write, academics pretending to read – what’s left?
The answer is – only the encounter. The tutorial where someone’s question makes you think again. The supervision where a half-formed idea gets taken seriously. The seminar where genuine disagreement produces genuine movement. The moments when people are present to each other, accountable to each other, and changed by each other.
They can’t be automated. They also can’t be scaled in the way the current model demands. You can’t have genuine encounters at a ratio of one to two hundred. Nor can you develop judgement in a twelve-week module delivered to students whose names you don’t know.
The alternative is differentiation – people who teach, people who research, people who coach, working in teams on longer-form problems rather than alone in offices marking scripts. But that requires admitting the all-rounder was always a lie, and restructuring everything around that admission.
If information is now abundant and feedback can be instant and personalised, then the scarcity model is dead. Good riddance. But abundance creates its own problems.
Without judgement, abundance is useless. Knowing that something is the case is increasingly cheap. Any idiot with ChatGPT can generate an account of the causes of the First World War or the principles of contract law. But knowing what to do about it, whether to trust it, how it connects to everything else, which bits matter and which are noise – these remain expensive, slow, human.
Judgement is not a skill you can look up. It’s a disposition you develop through practice – through getting things wrong and understanding why, through watching people who are better at it than you, through being held accountable by others who will tell you when you’re fooling yourself. AI can give us information. It can’t give us judgement.
Abundance makes it harder to know what we don’t know. When information was scarce, ignorance was obvious. Now, ignorance is invisible. We can generate confident-seeming text on any topic without understanding anything about it. The gap between performance and competence widens.
UCL’s Rose Luckin calls what’s needed “meta-intelligence” – not knowing things, but knowing how we know, knowing what we don’t know, and knowing how to find out. AI makes meta-intelligence more important, not less. If we can’t evaluate what the machine is giving us, we’re not using a tool. We’re being used by one.
That’s the equity issue that most AI boosterism ignores. If you went to a school that taught you to think, AI is a powerful amplifier. If you went to a school that taught you to comply, AI is a way of complying faster without ever developing the capacities that would let you do otherwise.
Cultivating judgement means designing curricula around problems that don’t have predetermined answers – not case studies where students are expected to reach the “right” conclusion, but genuine dilemmas where reasonable people disagree. It means assessment that rewards the quality of reasoning, not just the correctness of conclusions – teachers who model uncertainty, who think out loud, who change their minds in public.
Creating communities of inquiry means spaces where people think together, are accountable to each other, and learn to be wrong in public. They can’t be scaled, and can’t be automated. They require presence, continuity, and trust built over time. AI can prepare us for these spaces. It can’t be one of them.
Last week I was playing with a custom GPT with a group of student reps. We’d loaded it with Codes of Practice and housing law guidance, and for the first time they understood their rights as tenants – not deeply, not expertly, but enough to know what questions to ask and where to push back. They’d never have encountered this stuff otherwise.
The custom GPT wasn’t the point – the curiosity it sparked was. They left wanting to know more, not less. That’s what democratised information synthesis can do when it’s not about producing outputs faster, but about opening doors others didn’t know existed.
There’s always been an irony in the complaint that graduates lack “soft skills”. For decades, employers demanded production – write the report, analyse the data, build the model. Universities obliged, orienting curricula around outputs and assessing students on their capacity to produce. Now that machines produce faster and cheaper, employers discover they wanted something else all along.
They call it “soft skills” or “emotional intelligence” or “communication”. What they mean is the capacity to be present with other humans. To listen, to learn, to adapt – to work with people who are different from you, and to contribute to collective endeavours rather than produce outputs in isolation.
It’s always irked me that they’re described as soft. They are the hardest skills to develop and the hardest to fake. They are also exactly what universities could have been cultivating all along – if anyone had been willing to name them and pay for them.
Universities that grasp this can offer students, employers and society something they genuinely need – people who can think, who can learn, who can work with others, who can handle complexity and uncertainty. Employers will need to train them in their specific context, but they’ll be worth training. That’s a different value proposition than “job-ready graduates” – and a more honest one.
I remember visiting the Saltire Centre at Glasgow Caledonian and being amazed that a university was brave enough to notice that students like studying together. Not just being taught together – studying together. The spaces that fill up fastest are the ones where people can work alongside others, help each other, and belong to something.
It’s not a distraction from learning. It is learning. The same is true of SUs, societies, volunteering, representation – the “extracurricular” activities that universities tolerate but rarely celebrate. These are where students practise collective action, navigate difference, take responsibility for something beyond themselves. Formation happens in community, not just in classrooms.
Being brave enough to confront all this will be hard. The funding model rewards efficiency, the regulatory model rewards measurability, and the labour market wants qualifications. The incentive is to produce – people who can perform, not people who have developed.
Students – many, not all – have internalised this logic. They want the degree, the credential, the signal. They are strategic, instrumental, and focused on outcomes. It’s not a character flaw – it’s a rational response to the system they’re in. If the degree is the point, then anything that gets you the degree efficiently is sensible. AI is just the latest efficiency tool.
But while shame is a powerful disincentive to the fess up, the thing about pretending is that it’s exhausting. And it’s lonely.
For years at Christmas, I pretended UWE was Bristol because I was ashamed – ashamed of wanting to study the media, ashamed of coming from a family where going to any university was exceptional, ashamed of the gap between where I was and where people felt I should be. The pretending was a way of managing the shame.
I suspect a lot of students feel something similar. The performance of knowledge, the strategic deployment of qualifications, the constant positioning and comparison – these are ways of managing the fear that you’re not good enough, that you’ll be found out, that the gap between who you are and who you’re supposed to be is too wide to bridge.
AI intensifies the fear for some – the terror that they’ll be caught, that the machine will be detected, that the pretending will be exposed. But it might offer a different possibility. If the pretending no longer works – if the performance can be automated and therefore has no value – then maybe the only thing left is to become someone who doesn’t need to pretend.
That is the democratic promise of abundant information. Not that everyone will know everything – that’s neither possible nor desirable. But that knowledge can stop being a marker of status, a way of putting others down, or a resource to be hoarded. “I know something you don’t” can give way to “we can figure this out together.”
The shift from knowledge as possession to knowledge as practice is a shift from “I have information you lack” to “I can work with you on problems that matter.” From education as credentialing to education as formation. From “I’m better than you” to “I can contribute.” From pretending to becoming.
We’d need assessment that rewards contribution over reproduction. If the essay can be generated by AI, then the essay is testing the wrong thing. Assessment that requires students to think in real time, in dialogue, in response to genuine challenge – this is harder to automate and more valuable to develop. The individual student writing the individual essay marked by the individual academic is game over if AI can play both roles.
We’ll need pedagogy that prioritises encounter over transmission. Small group teaching. Sustained relationships between students and teachers. Curricula designed around problems rather than content coverage. Something between a module and a course, run by teams, with long-form purpose over a year rather than twelve-week fragments. Time and space for the slow work of formation.
We’ll need recognition that learning is social. Common spaces where students can study together. Student organisations supported rather than tolerated. Credit for service learning, for contribution to community, for the “extracurricular” activities where formation actually happens.
We’ll need slack in the system. The Swiss model, not the Czech one. Space to fall behind and catch up. Multiple attempts at assessment. Pass/fail options that encourage risk-taking. Time built in for things to go wrong, because things always go wrong. A system that absorbs delays rather than compounding them.
None of this will happen quickly. The funding model, the regulatory model, the labour market, the expectations students bring with them – they are not going to transform overnight. We’ll all have to play along for a while yet, doing the best we can within systems that reward the wrong things.
But playing along is not the same as believing. And knowing what we’re playing along with – knowing what we’re compromising and why – is the beginning of something different.
The reason I came to work here at Wonkhe – and the whole point of my work with students’ unions over the years – has been about giving power away. Not hoarding insight, but spreading it. Not being the person who knows things – but helping other people act on what they now know.
The best email I got last week wasn’t someone telling me that I was impressive, or clever. I’ve learned how to get those emails. It was someone saying “really great notes and really great meeting – has got our brains whirring a lot.” Using what I offered to do something I couldn’t have done myself.
Maybe I’ve become one of those insufferable men who grab the mic to assert that what education is for is what it did for them. But the purpose of teaching is surely rousing curiosity and creating the conditions for people to become.
When I look back at the version of myself who told his family he was going to Bristol, I feel compassion more than embarrassment. He was doing the best he could in a system that made pretending rational.
Thirty years on, I’ve watched skills become obsolete, formats get put in the skip, pretences exposed. Each time we held on for a while. Each time we eventually let go. Each time something was revealed about what had actually mattered all along.
AI doesn’t end the system of pretending. But it does expose its contradictions in ways that might, eventually, make something better possible. If the performance of knowledge becomes worthless, then maybe actual formation – and the human encounters that produce it – can finally be valued.
The hopeful answer is that universities can be places where people become more fully human. Not because they acquire more information, or even because they become subject specialists – though many will – but because they develop the capacities for thought, action, connection and care that make a human life worth living.
They are capacities that can’t be downloaded, nor automated, nor faked. They can be developed only slowly, in relationship, through practice, with friction.
You came to university for skills and they turned out to be useless? That’s a trick. You came for skills and left ready to change the world? Now that’s magic.
If, like me, you grew up watching Looney Tunes cartoons, you may remember Yosemite Sam’s popular phrase, “There’s gold in them thar hills.”
In surveys, as in gold mining, the greatest riches are often hidden and difficult to extract. This principle is perhaps especially true when institutions are seeking to enhance the postgraduate taught (PGT) student experience.
PGT students are far more than an extension of the undergraduate community; they represent a crucial, diverse and financially significant segment of the student body. Yet, despite their growing numbers and increasing strategic importance, PGT students, as Kelly Edmunds and Kate Strudwick have recently pointed out on Wonkhe, remain largely invisible in both published research and core institutional strategy.
Advance HE’s Postgraduate Taught Experience Survey (PTES) is therefore one of the few critical insights we have about the PGT experience. But while the quantitative results offer a (usually fairly consistent) high-level view, the real intelligence required to drive meaningful enhancement inside higher education institutions is buried deep within the thousands of open-text comments collected. Faced with the sheer volume of data the choice is between eye-ball scanning and the inevitable introduction of human bias, or laborious and time-consuming manual coding. The challenge for the institutions participating in PTES this year isn’t the lack of data: it’s efficiently and reliably turning that dense, often contradictory, qualitative data into actionable, ethical, and equitable insights.
The application of machine learning AI technology to analysis of qualitative student survey data presents us with a generational opportunity to amplify the student voice. The critical question is not whether AI should be used, but how to ensure its use meets robust and ethical standards. For that you need the right process – and the right partner – to prioritise analytical substance, comprehensiveness, and sector-specific nuance.
UK HE training is non-negotiable. AI models must be deeply trained on a vast corpus of UK HE student comments. Without this sector-specific training, analysis will fail to accurately interpret the nuances of student language, sector jargon, and UK-specific feedback patterns.
Analysis must rely on a categorisation structure that has been developed and refined against multiple years of PTES data. This continuity ensures that the thematic framework reflects the nuances of the PGT experience.
To drive targeted enhancement, the model must break down feedback into highly granular sub-themes – moving far beyond simplistic buckets – ensuring staff can pinpoint the exact issue, whether it falls under learning resources, assessment feedback, or thesis supervision.
The analysis must be more than a static report. It must be delivered through integrated dashboard solutions that allow institutions to filter, drill down, and cross-reference the qualitative findings with demographic and discipline data. Only this level of flexibility enables staff to take equitable and targeted enhancement actions across their diverse PGT cohorts.
When these principles are prioritised, the result is an analytical framework specifically designed to meet the rigour and complexity required by the sector.
The partnership between Advance HE, evasys, and Student Voice AI, which analysed this year’s PTES data, demonstrates what is possible when these rigorous standards are prioritised. We have offered participating institutions a comprehensive service that analyses open comments alongside the detailed benchmarking reports that Advance HE already provides. This collaboration has successfully built an analytical framework that exemplifies how sector-trained AI can deliver high-confidence, actionable intelligence.
Jonathan Neves, Head of Research and Surveys, Advance HE calls our solution “customised, transparent and genuinely focused on improving the student experience, “ and adds, “We’re particularly impressed by how they present the data visually and look forward to seeing results from using these specialised tools in tandem.”
The commitment to analytical substance is paramount; without it, the risk to institutional resources and equity is severe. If institutions are to derive value, the analysis must be comprehensive. When the analysis lacks this depth institutional resources are wasted acting on partial or misleading evidence.
Rigorous analysis requires minimising what we call data leakage: the systematic failure to capture or categorise substantive feedback. Consider the alternative: when large percentages of feedback are ignored or left uncategorised, institutions are effectively muting a significant portion of the student voice. Or when a third of the remaining data is lumped into meaningless buckets like “other,” staff are left without actionable insight, forced to manually review thousands of comments to find the true issues.
This is the point where the qualitative data, intended to unlock enhancement, becomes unusable for quality assurance. The result is not just a flawed report, but the failure to deliver equitable enhancement for the cohorts whose voices were lost in the analytical noise.
Reliable, comprehensive processing is just the first step. The ultimate goal of AI analysis should be to deliver intelligence in a format that seamlessly integrates into strategic workflows. While impressive interfaces are visually appealing, genuine substance comes from the capacity to produce accurate, sector-relevant outputs. Institutions must be wary of solutions that offer a polished facade but deliver compromised analysis. Generic generative AI platforms, for example, offer the illusion of thematic analysis but are not robust.
But robust validation of any output is still required. This is the danger of smoke and mirrors – attractive dashboards that simply mask a high degree of data leakage, where large volumes of valuable feedback are ignored, miscategorised or rendered unusable by failing to assign sentiment.
When institutions choose rigour, the outcomes are fundamentally different, built on a foundation of confidence. Analysis ensures that virtually every substantive PGT comment is allocated to one or more UK-derived categories, providing a clear thematic structure for enhancement planning.
Every comment with substance is assigned both positive and negative sentiment, providing staff with the full, nuanced picture needed to build strategies that leverage strengths while addressing weaknesses.
This shift from raw data to actionable intelligence allows institutions to move quickly from insight to action. As Parama Chaudhury, Pro-Vice Provost (Education – Student Academic Experience) at UCL noted, the speed and quality of this approach “really helped us to get the qualitative results alongside the quantitative ones and encourage departmental colleagues to use the two in conjunction to start their work on quality enhancement.”
The capacity to produce accurate, sector-relevant outputs, driven by rigorous processing, is what truly unlocks strategic value. Converting complex data tables into readable narrative summaries for each theme allows academic and professional services leaders alike to immediately grasp the findings and move to action. The ability to access categorised data via flexible dashboards and in exportable formats ensures the analysis is useful for every level of institutional planning, from the department to the executive team. And providing sector benchmark reports allows institutions to understand their performance relative to peers, turning internal data into external intelligence.
The postgraduate taught experience is a critical pillar of UK higher education. The PTES data confirms the challenge, but the true opportunity lies in how institutions choose to interpret the wealth of student feedback they receive. The sheer volume of PGT feedback combined with the ethical imperative to deliver equitable enhancement for all students demands analytical rigour that is complete, nuanced, and sector-specific.
This means shifting the focus from simply collecting data to intelligently translating the student voice into strategic priorities. When institutions insist on this level of analytical integrity, they move past the risk of smoke and mirrors and gain the confidence to act fast and decisively.
It turns out Yosemite Sam was right all along: there’s gold in them thar hills. But finding it requires more than just a map; it requires the right analytical tools and rigour to finally extract that valuable resource and forge it into meaningful institutional change.
This article is published in association with evasys. evasys and Student Voice AI are offering no-cost advanced analysis of NSS open comments delivering comprehensive categorisation and sentiment analysis, secure dashboard to view results and a sector benchmark report. Click here to find out more and request your free analysis.
It is tempting to regard AI as a panacea for addressing our most urgent global challenges, from climate change to resource scarcity. Yet the truth is more complex: unless we pair innovation with responsibility, the very tools designed to accelerate sustainability may exacerbate its contradictions.
A transformative potential
Let us first acknowledge how AI is already reshaping sustainable development. By mapping patterns in vast datasets, AI enables us to anticipate environmental risks, optimise resource flows and strengthen supply chains. Evidence suggests that by 2030, AI systems will touch the lives of more than 8.5 billion people and influence the health of both human and natural ecosystems in ways we have never seen before. Research published in Nature indicates that AI could support progress towards 79% of the Sustainable Development Goals (SDGs), helping advance 134 specific targets. Yet the same research also cautions that AI may impede 59 of those targets if deployed without care or control.
In practice, this means smarter energy grids that balance load and demand, precision agriculture that reduces fertiliser waste and environmental monitoring systems that detect deforestation or pollution in real time. For a planet under pressure, these scenarios offer hope to do less harm and build more resilience.
The hidden costs
Even so, we must confront the shadows cast by AI’s advancements. An investigation published earlier this year warns that AI systems could account for nearly half of global data-centre power consumption before the decade’s end. Consider the sheer scale: vast server arrays, intensive cooling systems, rare-earth mining and water-consuming infrastructure all underpin generative AI’s ubiquity. Worse still, indirect carbon emissions tied to major AI-capable firms reportedly rose by 150% between 2020 and 2023. In short, innovation meant to serve sustainability imposes a growing ecological burden.
Navigating trade-offs
This tension presents an essential question: how can we reconcile AI’s promise with its cost? Scholars warn that we must move beyond the assumption that ‘AI for good’ is always good enough. The moment demands a new discipline of ‘sustainable AI’: a framework that treats resource use, algorithmic bias, lifecycle impact and societal equity as first-order concerns.
Practitioners must ask not only what AI can do, but how it is built, powered, governed and retired. Efficiency gains that drive consumption higher will not deliver sustainability; they may merely escalate resource demands in disguise.
A moral and strategic imperative
For educators, policymakers and business leaders, this is more than a technical issue; it is a moral and strategic one. To realise AI’s true potential in advancing sustainable development, we must commit to three priorities:
Energy and resource transparency: Organisations must measure and report the footprint of their AI models, including data-centre use, water cooling, e-waste and supply-chain impacts. Transparency is foundational to accountability.
Ethical alignment and fairness: AI must be trained and deployed with due regard to bias, social impact and inclusivity. Its benefits must not reinforce inequality or externalise environmental harms onto vulnerable communities.
Integrative education and collaboration: We need multidisciplinary expertise, engineers fluent in ecology, ethicists fluent in algorithms and managers fluent in sustainability. Institutions must upskill young learners and working professionals to orient AI within the broader context of planetary boundaries and human flourishing.
MLA College’s focus and contribution
At MLA College, we recognise our role in equipping professionals at this exact intersection. Our programs emphasise the interrelationship between technology, sustainability and leadership. Graduates of distance-learning and part-time formats engage with the complexities of AI, maritime operations, global sustainable development and marine engineering by bringing insight to sectors vital to the planet’s future.
When responsibly guided, AI becomes an amplifier of purpose rather than a contraption of risk. Our challenge is to ensure that every algorithm, model and deployment contributes to regenerative systems, not extractive ones.
The promise of AI is compelling: more accurate climate modelling, smarter cities, adaptive infrastructure and just-in-time supply chains. But the challenge is equally formidable: rising energy demands, resource-intensive infrastructures and ungoverned expansion.
When responsibly guided, AI becomes an amplifier of purpose rather than a contraption of risk
Our collective role, as educators and practitioners, is to shape the ethical architecture of this era. We must ask whether our technologies will serve humanity and the environment or simply accelerate old dynamics under new wrappers.
The verdict will not be written on lines of code or boardroom decisions alone. It will be inscribed in the fields that fail to regenerate, in the communities excluded from progress, in the data centres humming with waste and in the next generation seeking meaning in technology’s promise.
About the author: Professor Mohammad Dastbaz is the principal and CEO of MLA College, an international leader in distance and sustainability-focused higher education. With over three decades in academia, he has held senior positions including deputy vice-chancellor at the University of Suffolk and pro vice-chancellor at Leeds Beckett University.
A Fellow of the British Computer Society, the Higher Education Academy, and the Royal Society of Arts, Professor Dastbaz is a prominent researcher and author in the fields of sustainable development, smart cities, and digital innovation in education.
His latest publication, Decarbonization or Demise – Sustainable Solutions for Resilient Communities (Springer, 2025), brings together cutting-edge global research on sustainability, climate resilience, and the urgent need for decarbonisation. The book builds on his ongoing commitment to advancing the UN Sustainable Development Goals through education and research.
At MLA College, Professor Dastbaz continues to lead transformative learning initiatives that combine academic excellence with real-world impact, empowering students to shape a sustainable future.
Artificial intelligence is already reshaping higher education fast. For universities aiming to be AI-first institutions, leadership, governance, staff development, and institutional culture are critical.
How institutions respond now will determine whether AI enhances learning or simply reinforces existing inequalities, inefficiencies and, frankly, bad practices. This is not only an institutional or sector question but a matter of national policy: government has committed to supporting AI-skills at scale, and the UK has pledged an early ambition that a “fifth of the workforce will be supported with the AI skills they need to thrive in their jobs.” Strategic deployment of AI is therefore a pressing HE leadership question.
Universities will benefit from articulating a clear AI-first vision that aligns with their educational, research and civic missions. Leadership plays a central role in ensuring AI adoption supports educational quality, innovation and equity rather than focusing purely on operational efficiency or competitiveness. Cultivating a culture where AI is viewed as a collaborative partner helps staff become innovators shaping AI integration rather than passive users (as the jargon frames it, “makers” not “takers”). Strategic plans and performance indicators should reflect commitments to ethical, responsible, and impactful AI deployment, signalling to staff and students that innovation and integrity go hand in hand.
Ethical and transparent leadership in AI-first institutions is vital. Decision-making, whether informed by student analytics like Kortext StREAM, enrolment forecasts, budgeting, or workforce planning, should model responsible AI use. The right governance structures need to be created. Far be it from us to suggest more committees, but there needs to be governance oversight through ethics and academic quality boards to oversee AI deployment across the education function.
Clear frameworks for managing data privacy, intellectual property, and algorithmic bias are essential, particularly when working with third-party providers. Maintaining dialogue with accreditation and quality assurance bodies including PSRBs and OfS ensures innovation aligns with regulatory expectations, avoiding clashes between ambition and oversight. This needs to be at individual institution, but also at sector and regulator level.
Staff capability underpins any AI-first strategy. This needs to be understood through taking a whole institution approach rather than just education-facing staff. Defining a framework of AI competencies will help to clarify the skills needed to use AI responsibly and effectively, and there are already institutional frameworks, including from Jisc, QAA, and Skills England, that do this. Embedding these competencies into recruitment, induction, appraisal, promotion and workload frameworks can ensure that innovation is rewarded, not sidelined.
Demonstrating AI literacy and ethical awareness could become a requirement for course leadership, or senior appointments. Adjusting workload models to account for experimentation, retraining, and curriculum redesign gives staff the space to explore AI responsibly. Continuous professional development – including AI learning pathways, ethics training, and peer learning communities – reinforces a culture of innovation while protecting academic quality.
Investment in AI-enabled infrastructure underpins an AI-first institution. We recognise the severe financial challenges faced by many institutions and this means that investments must be well targeted and implemented effectively. Secure data environments, analytics platforms, and licensed AI tools accessible to staff and students are essential to provide the foundation for innovation. Ethical procurement practices when partnering with edtech providers promote transparency, accessibility, and academic independence. Universities should also consider the benefits and risks of developing their own large language models alongside relying on external platforms, weighing in factors such as cost, privacy, and institutional control. See this partnership between Kortext, Said Business School, Microsoft and Instructure for an example of an innovative new education partnership.
Implementing AI responsibly requires trust. Leaders need to communicate openly about AI’s opportunities and limitations, critically addressing staff anxieties about displacement or loss of autonomy. Leadership development programmes for PVCs, deans, heads of school, and professional service directors can help manage AI-driven transformation effectively.
One of the most important things to get right is to ensure that cross-functional collaboration between IT, academic development, HR, and academic quality units supports coherent progress toward an AI-first culture. Adopting iterative change management – using pilot programs, consultation processes, and rapid feedback loops well – allows institutions to refine AI strategies continuously, balancing innovation with oversight.
AI interventions benefit from rigorous quantitative and qualitative evaluation. Indicators such as efficiency, student outcomes, creativity, engagement, and inclusion can offer a balanced picture of impact. Regular review cycles ensure responsiveness to emerging AI capabilities and evolving educational priorities. Publishing internal (and external) reports on AI impacts on education will be essential to promote transparency, sharing lessons learned and guiding future development. It almost goes without saying that institutions should share practice (what has worked and what hasn’t) not only within their organisations, but also across the sector and with accrediting bodies and regulators.
An AI-first university places human judgment, ethics, and pedagogy at the centre of all technological innovation. AI should augment rather than replace intellectual and creative capacities of educators and students. Every intervention must benefit from assessment against these principles, ensuring technology serves learning, rather than it becoming the master of human agency or ethical standards.
Being an AI-first institution is certainly not about chasing the latest tools or superficially focusing on staff and student “AI literacy.” It is about embedding AI thoughtfully in every part of the university. Leaders need to articulate vision, model ethical behaviour, build staff capacity and student ability to become next generation AI leaders. Staff and students need time, support and trust to experiment responsibly. Infrastructure and external partnerships must be strategic and principled. There must also be continuous evaluation to ensure that innovation aligns with strategy and values.
When implemented carefully, AI can become a collaborative partner in enhancing learning, facilitating creativity and reinforcing the academic mission rather than undermining it.
This article is published in association with Kortext. Join Janice and Rachel for Kortext LIVE on 11 February in London, on the theme of “Leading the next chapter of digital innovation” to continue the conversation on AI and data. Keynote speakers include Mark Bramwell, CDIO at Said Business School. Find out more and secure your spot here.
AI in all its multitudinous forms is here, it is here to stay, and its impacts are accelerating.
At a basic level, we see shifts in personal office practices with the tentative, steady adoption of large language models. We see AI being used alongside MS Teams or its equivalents to quickly produce summary transcripts of meetings or to generate starting places for documents which are then reworked.
As university educators and researchers, we also see debates regarding the ethics of AI adoption and a splintering ability of the collective and the individual to be able to discern fact from fiction. We are at the start of a long and unpredictable trajectory of impacts.
But, as we shape the skills, knowledge and abilities in our students that will see them thrive in an increasingly disrupted future world of work, where that track takes us is a subject of debate. What is consistently clear across various predictions is that the adoption of AI and increasing automation will deliver seismic changes to the world (of work).
86 per cent of employers surveyed for the World Economic Forum’s 2025 Future of Jobs Survey saw AI and information processing technologies as being the dominant technology driver for workplace change through to 2030, affecting workplaces across all sectors, not just those welcoming students from STEM disciplines. Similarly, the same survey indicates the greatest rise in demand in the workplace through to 2030 is for the ability to work with AI and big data.
Noting the dominance of AI in the WEF survey findings, we are reminded of the 1998 interview between Jeremy Paxman and David Bowie, happening just as the internet was forming. Paxman queries the internet as being anything more than a “different delivery system,” while Bowie asserts that it is an alien life form:
I don’t think we’ve even seen the tip of the iceberg – the things it will do, both good and bad are unimaginable right now. I actually think we’re on the cusp of something exhilarating and terrifying…
Looking back at what has happened to society in the quarter of a century since that interview, Bowie is unnervingly accurate in his foresight.
It seems that right now we are navigating similarly uncharted territories of an epoch-defining transition as the world starts to play in earnest with the next gen version of Bowie’s “alien lifeform.” Higher education is not immune – it is grappling with the challenges across its core activities.
However, what is of particular interest beyond the specific AI skills is the other in-demand skills that occupy the places immediately following the top three noted above. Fourth is creative thinking, followed by resilience, flexibility and agility, curiosity and lifelong learning and leadership and social influence. These skills are high value cognitive competencies inherently human in their nature – an equalising “soft” counterbalance to the “hard” technological literacies of the top three.
Reflecting on the duality between technological literacy and social, emotional and cognitive skills in this overall picture, it is clear that AI is not a replacement for the work of thought, deduction, critical reasoning and curiosity. Instead, it is a powerful augmentation to the already formidable arsenal of technological capability at our fingertips.
With education and the student experience in mind, we see two AI “swim lanes” forming out of the early stages of ubiquity ushered in by the popularisation of ChatGPT and other LLMs. These swim lanes should also acknowledge the broader mix of new and emergent technologies at play in tandem with AI – for instance AR/VR and data visualisation.
The first swim lane speaks to the need to optimise the complex wiring behind the institutional operations of higher education which provide our students with a world class experience. With efficiency, effectiveness and scale in mind, adoption of AI to underpin the crucial in-person experience with wider algorithmic personalisation becomes a highly desirable direction of travel. For instance, we can easily envisage a world in which AI is used to aid student navigation of module choice, tailoring the availability of elective courses and complementary extra-curricular and developmental activities.
The second swim lane is one of invention and co-creation, arguably pushing AI and the wider ecosystem of technological innovation to be the best it can be – far beyond the deployment of convenience or efficiency. At its best, AI can become a partner in creativity: an inspirer and a critical collaborator offering new perspectives. We are seeing promising points of innovation and departure in the early work at Loughborough as the range of technological capabilities within our DigiLabs continues to be adopted at pace.
However, to swim confidently in this lane we must dispel myths and fears with rigour and a critical navigation of AI as a co-creator. Scaffolding and skills development for staff and students are essential in order that we all might partner effectively with our new playmate.
Two points of skills development show themselves as a useful starting place towards consistency, innovation and collaboration in AI partnership. First, a good place to start would be recognition and development of prompt engineering as a fundamental digital skill and a shared structured practice. Second, it would be useful to focus on development of a consistent and structured means to better understand, interrogate and critically evaluate what the AI has generated in response to our prompting.
With frameworks for these two essentials of effective AI partnership in place, we can move beyond the cut-and-paste AI-as-shortcut, and beyond the simple fact checking of generated material. These two skills move us towards conversing and exchanging perspectives with AI, making content better together. The vantage point of having embedded these two AI partnership skills helps us then systematically inculcate the true value of AI by recognising the human skillset with which to strategically cocreate with it, rather than shortcut with it.
As our use of AI evolves, we should continually remind ourselves that understanding is not gained in the endpoint, but in travelling to that place (no student learns that much in the moment of a final assessment). AI becomes a meaningful companion on that journey, not a replacement for the experience of travelling. To shortcut the pleasure and frustration of our own creative and critical journeys by virtue of AI laziness is to deny ourselves the experience of our own essence – the struggle and the unknowing of what it means to question, to be alive and to be human.
When I ask apprentices to reflect on their learning in professional discussions, I often hear a similar story:
It wasn’t just about what I knew – it was how I connected it all. That’s when it clicked.
That’s the value of dialogic assessment. It surfaces hidden knowledge, creates space for reflection, and validates professional judgement in ways that traditional essays often cannot.
Dialogic assessment shifts the emphasis from static products – the essay, the exam – to dynamic, real-time engagement. These assessments include structured discussions, viva-style conversations, or portfolio presentations. What unites them is their reliance on interaction, reflection, and responsiveness in the moment.
Unlike “oral exams” of old, these conversations require learners to explain reasoning, apply knowledge, and reflect on lived experience. They capture the complex but authentic process of thinking – not just the polished outcome.
In Australia, “interactive orals” have been adopted at scale to promote integrity and authentic learning, with positive feedback from staff and students. Several UK universities have piloted viva-style alternatives to traditional coursework with similar results. What apprenticeships have long taken for granted is now being recognised more widely: dialogue is a powerful form of assessment.
In apprenticeships and work-based learning, dialogic assessment is not an add-on – it’s essential. Apprentices regularly take part in professional discussions (PDs) and portfolio presentations as part of both formative and end-point assessment.
What makes them so powerful? They are inclusive, as they allow different strengths to emerge. Written tasks may favour those fluent in academic conventions, while discussions reveal applied judgement and reflective thinking. They are authentic, in that they mirror real workplace activities such as interviews, stakeholder reviews, and project pitches. And they can be transformative – apprentices often describe PDs as moments when fragmented knowledge comes together through dialogue.
One apprentice told me:
It wasn’t until I talked it through that I realised I knew more than I thought – I just couldn’t get it down on paper.
For international students, dialogic assessment can also level the playing field by valuing applied reasoning over written fluency, reducing the barriers posed by rigid academic writing norms.
My doctoral research has shown that PDs not only assess knowledge but also co-create it. They push learners to prepare more deeply, reflect more critically, and engage more authentically. Tutors report richer opportunities for feedback in the process itself, while employers highlight their relevance to workplace practice.
And AI fits into this picture too. When ChatGPT and similar tools emerged in late 2022, many feared the end of traditional written assessment. Universities scrambled for answers – detection software, bans, or a return to the three-hour exam. The risk has been a slide towards high-surveillance, low-trust assessment cultures.
But dialogic assessment offers another path. Its strength is precisely that it asks students to do what AI cannot:
Scaling dialogic assessment isn’t without hurdles. Large cohorts and workload pressures can make universities hesitant. Online viva formats also raise equity issues for students without stable internet or quiet environments.
But these challenges can be mitigated: clear rubrics, tutor training, and reliable digital platforms make it possible to mainstream dialogic formats without compromising rigour or inclusivity. Apprenticeships show it can be done at scale – thousands of students sit PDs every year.
Crucially, dialogic assessment also aligns neatly with regulatory frameworks. The Office for Students requires that assessments be valid, reliable, and representative of authentic learning. The QAA Quality Code emphasises inclusivity and support for learning. Dialogic formats tick all these boxes.
The AI panic has created a rare opportunity. Universities can either double down on outdated methods – or embrace formats that are more authentic, equitable, and future-oriented.
This doesn’t mean abandoning essays or projects altogether. But it could mean ensuring every programme includes at least one dialogic assessment – whether a viva, professional discussion, or reflective dialogue.
Apprenticeships have demonstrated that dialogic assessments are effective. They are rigorous, scalable, and trusted. Now is the time for the wider higher education sector to recognise their value – not as a niche alternative, but as a core element of assessment in the AI era.
Imagine a student using a writing assistant powered by a generative AI chatbot. As the bot serves up practical suggestions and encouragement, insights come more easily, drafts polish up quickly and feedback loops feel immediate. It can be energizing. But when that AI support is removed, some students report feeling less confident or less willing to engage.
These outcomes raise the question: Can AI tools genuinely boost student motivation? And what conditions can make or break that boost?
As AI tools become more common in classroom settings, the answers to these questions matter a lot. While tools for general use such as ChatPGT or Claude remain popular, more and more students are encountering AI tools that are purpose-built to support learning, such as Khan Academy’s Khanmigo, which personalizes lessons. Others, such as ALEKS, provide adaptive feedback. Both tools adjust to a learner’s level and highlight progress over time, which helps students feel capable and see improvement. But there are still many unknowns about the long-term effects of these tools on learners’ progress, an issue I continue to study as an educational psychologist.
Recent studies indicate that AI can boost motivation, at least for certain groups, when deployed under the right conditions. A 2025 experiment with university students showed that when AI tools delivered a high-quality performance and allowed meaningful interaction, students’ motivation and their confidence in being able to complete a task – known as self-efficacy – increased.
For foreign language learners, a 2025 study found that university students using AI-driven personalized systems took more pleasure in learning and had less anxiety and more self-efficacy compared with those using traditional methods. A recent cross-cultural analysis with participants from Egypt, Saudi Arabia, Spain and Poland who were studying diverse majors suggested that positive motivational effects are strongest when tools prioritize autonomy, self-direction and critical thinking. These individual findings align with a broader, systematic review of generative AI tools that found positive effects on student motivation and engagement across cognitive, emotional and behavioral dimensions.
A forthcoming meta-analysis from my team at the University of Alabama, which synthesized 71 studies, echoed these patterns. We found that generative AI tools on average produce moderate positive effects on motivation and engagement. The impact is larger when tools are used consistently over time rather than in one-off trials. Positive effects were also seen when teachers provide scaffolding, when students maintain agency in how they use the tool, and when the output quality is reliable.
But there are caveats. More than 50 of the studies we reviewed did not draw on a clear theoretical framework of motivation, and some used methods that we found were weak or inappropriate. This raises concerns about the quality of the evidence and underscores how much more careful research is needed before one can say with confidence that AI nurtures students’ intrinsic motivation rather than just making tasks easier in the moment.
There is also research that paints a more sobering picture. A large study of more than 3,500 participants found that while human–AI collaboration improved task performance, it reduced intrinsic motivation once the AI was removed. Students reported more boredom and less satisfaction, suggesting that overreliance on AI can erode confidence in their own abilities.
Another study suggested that while learning achievement often rises with the use of AI tools, increases in motivation are smaller, inconsistent or short-lived. Quality matters as much as quantity. When AI delivers inaccurate results, or when students feel they have little control over how it is used, motivation quickly erodes. Confidence drops, engagement fades and students can begin to see the tool as a crutch rather than a support. And because there are not many long-term studies in this field, we still do not know whether AI can truly sustain motivation over time, or whether its benefits fade once the novelty wears off.
The impact of AI on student motivation is not one-size-fits-all. Our team’s meta-analysis shows that, on average, AI tools do have a positive effect, but the size of that effect depends on how and where they are used. When students work with AI regularly over time, when teachers guide them in using it thoughtfully, and when students feel in control of the process, the motivational benefits are much stronger.
We also saw differences across settings. College students seemed to gain more than younger learners, STEM and writing courses tended to benefit more than other subjects, and tools designed to give feedback or tutoring support outperformed those that simply generated content.
There is also evidence that general-use tools like ChatGPT or Claude do not reliably promote intrinsic motivation or deeper engagement with content, compared to learning-specific platforms such as ALEKS and Khanmigo, which are more effective at supporting persistence and self-efficacy. However, these tools often come with subscription or licensing costs. This raises questions of equity, since the students who could benefit most from motivational support may also be the least likely to afford it.
These and other recent findings should be seen as only a starting point. Because AI is so new and is changing so quickly, what we know today may not hold true tomorrow. In a paper titled The Death and Rebirth of Research in Education in the Age of AI, the authors argue that the speed of technological change makes traditional studies outdated before they are even published. At the same time, AI opens the door to new ways of studying learning that are more participatory, flexible and imaginative. Taken together, the data and the critiques point to the same lesson: Context, quality and agency matter just as much as the technology itself.
The lessons from this growing body of research are straightforward. The presence of AI does not guarantee higher motivation, but it can make a difference if tools are designed and used with care and understanding of students’ needs. When it is used thoughtfully, in ways that strengthen students’ sense of competence, autonomy and connection to others, it can be a powerful ally in learning.
But without those safeguards, the short-term boost in performance could come at a steep cost. Over time, there is the risk of weakening the very qualities that matter most – motivation, persistence, critical thinking and the uniquely human capacities that no machine can replace.
For teachers, this means that while AI may prove a useful partner in learning, it should never serve as a stand-in for genuine instruction. For parents, it means paying attention to how children use AI at home, noticing whether they are exploring, practicing and building skills or simply leaning on it to finish tasks. For policymakers and technology developers, it means creating systems that support student agency, provide reliable feedback and avoid encouraging overreliance. And for students themselves, it is a reminder that AI can be a tool for growth, but only when paired with their own effort and curiosity.
Regardless of technology, students need to feel capable, autonomous and connected. Without these basic psychological needs in place, their sense of motivation will falter – with or without AI.
This article is republished from The Conversation under a Creative Commons license. Read the original article.
by Mehreen Ashraf, Eimear Nolan, Manual F Ramirez, Gazi Islam and Dirk Lindebaum
Walk into almost any university today, and you can be sure to encounter the topic of AI and how it affects higher education (HE). AI applications, especially large language models (LLM), have become part of everyday academic life, being used for drafting outlines, summarising readings, and even helping students to ‘think’. For some, the emergence of LLMs is a revolution that makes learning more efficient and accessible. For others, it signals something far more unsettling: a shift in how and by whom knowledge is controlled. This latter point is the focus of our new article published in Organization Studies.
At the heart of our article is a shift in what is referred to epistemic (or knowledge) governance: the way in which knowledge is created, organised, and legitimised in HE. In plain terms, epistemic governance is about who gets to decide what counts as credible, whose voices are heard, and how the rules of knowing are set. Universities have historically been central to epistemic governance through peer review, academic freedom, teaching, and the public mission of scholarship. But as AI tools become deeply embedded in teaching and research, those rules are being rewritten not by educators or policymakers, but by the companies that own the technology.
From epistemic agents to epistemic consumers
Universities, academics, and students have traditionally been epistemic agents: active producers and interpreters of knowledge. They ask questions, test ideas, and challenge assumptions. But when we rely on AI systems to generate or validate content, we risk shifting from being agents of knowledge to consumers of knowledge. Technology takes on the heavy cognitive work: it finds sources, summarises arguments, and even produces prose that sounds academic. However, this efficiency comes at the cost of profound changes in the nature of intellectual work.
Students who rely on AI to tidy up their essays, or generate references, will learn less about the process of critically evaluating sources, connecting ideas and constructing arguments, which are essential for reasoning through complex problems. Academics who let AI draft research sections, or feed decision letters and reviewer reports into AI with the request that AI produces a ‘revision strategy’, might save time but lose the slow, reflective process that leads to original thought, while undercutting their own agency in the process. And institutions that embed AI into learning systems hand part of their epistemic governance – their authority to define what knowledge is and how it is judged – to private corporations.
This is not about individual laziness; it is structural. As Shoshana Zuboff argued in The age of surveillance capitalism, digital infrastructures do not just collect information, they reorganise how we value and act upon it. When universities become dependent on tools owned by big tech, they enter an ecosystem where the incentives are commercial, not educational.
Big tech and the politics of knowing
The idea that universities might lose control of knowledge sounds abstract, but it is already visible. Jisc’s 2024 framework on AI in tertiary education warns that institutions must not ‘outsource their intellectual labour to unaccountable systems,’ yet that outsourcing is happening quietly. Many UK universities, including the University of Oxford, have signed up to corporate AI platforms to be used by staff and students alike. This, in turn, facilitates the collection of data on learning behaviours that can be fed back into proprietary models.
This data loop gives big tech enormous influence over what is known and how it is known. A company’s algorithm can shape how research is accessed, which papers surface first, or which ‘learning outcomes’ appear most efficient to achieve. That’s epistemic governance in action: the invisible scaffolding that structures knowledge behind the scenes. At the same time, it is easy to see why AI technologies appeal to universities under pressure. AI tools promise speed, standardisation, lower costs, and measurable performance, all seductive in a sector struggling with staff shortages and audit culture. But those same features risk hollowing out the human side of scholarship: interpretation, dissent, and moral reasoning. The risk is not that AI will replace academics but that it will change them, turning universities from communities of inquiry into systems of verification.
The Humboldtian ideal and why it is still relevant
The modern research university was shaped by the 19th-century thinker Wilhelm von Humboldt, who imagined higher education as a public good, a space where teaching and research were united in the pursuit of understanding. The goal was not efficiency: it was freedom. Freedom to think, to question, to fail, and to imagine differently.
That ideal has never been perfectly achieved, but it remains a vital counterweight to market-driven logics that render AI a natural way forward in HE. When HE serves as a place of critical inquiry, it nourishes democracy itself. When it becomes a service industry optimised by algorithms, it risks producing what Žižek once called ‘humans who talk like chatbots’: fluent, but shallow.
The drift toward organised immaturity
Scholars like Andreas Scherer and colleagues describe this shift as organised immaturity: a condition where sociotechnical systems prompt us to stop thinking for ourselves. While AI tools appear to liberate us from labour, what is happening is that they are actually narrowing the space for judgment and doubt.
In HE, that immaturity shows up when students skip the reading because ‘ChatGPT can summarise it’, or when lecturers rely on AI slides rather than designing lessons for their own cohort. Each act seems harmless; but collectively, they erode our epistemic agency. The more we delegate cognition to systems optimised for efficiency, the less we cultivate the messy, reflective habits that sustain democratic thinking. Immanuel Kant once defined immaturity as ‘the inability to use one’s understanding without guidance from another.’ In the age of AI, that ‘other’ may well be an algorithm trained on millions of data points, but answerable to no one.
Reclaiming epistemic agency
So how can higher education reclaim its epistemic agency? The answer lies not only in rejecting AI but also in rethinking our possible relationships with it. Universities need to treat generative tools as objects of inquiry, not an invisible infrastructure. That means embedding critical digital literacy across curricula: not simply training students to use AI responsibly, but teaching them to question how it works, whose knowledge it privileges, and whose it leaves out.
In classrooms, educators could experiment with comparative exercises: have students write an essay on their own, then analyse an AI version of the same task. What’s missing? What assumptions are built in? How were students changed when the AI wrote the essay for them and when they wrote them themselves? As the Russell Group’s 2024 AI principles note, ‘critical engagement must remain at the heart of learning.’
In research, academics too must realise that their unique perspectives, disciplinary judgement, and interpretive voices matter, perhaps now more than ever, in a system where AI’s homogenisation of knowledge looms. We need to understand that the more we subscribe to values of optimisation and efficiency as preferred ways of doing academic work, the more natural the penetration of AI into HE will unfold.
Institutionally, universities might consider building open, transparent AI systems through consortia, rather than depending entirely on proprietary tools. This isn’t just about ethics; it’s about governance and ensuring that epistemic authority remains a public, democratic responsibility.
Why this matters to you
Epistemic governance and epistemic agency may sound like abstract academic terms, but they refer to something fundamental: the ability of societies and citizens (not just ‘workers’) to think for themselves when/if universities lose control over how knowledge is created, validated and shared. When that happens, we risk not just changing education but weakening democracy. As journalist George Monbiot recently wrote, ‘you cannot speak truth to power if power controls your words.’ The same is true for HE. We cannot speak truth to power if power now writes our essays, marks our assignments, and curates our reading lists.
Mehreen Ashraf is an Assistant Professor at Cardiff Business School, Cardiff University, United Kingdom.
Eimear Nolan is an Associate Professor in International Business at Trinity Business School, Trinity College Dublin, Ireland.
Manuel F Ramirez is Lecturer in Organisation Studies at the University of Liverpool Management School, UK.
Gazi Islam is Professor of People, Organizations and Society at Grenoble Ecole de Management, France.
Dirk Lindebaum is Professor of Management and Organisation at the School of Management, University of Bath.
