Jennifer Darling
19th August 2026
When I was a student, my essays were a challenge for me, as they offer many for lots of reasons. My experience often was not that ideas were wrong. But the way I expressed them did not fit the expected form. I am dyslexic. I think in connections, in patterns, in several directions at once. And for most of my education, that was treated as a problem to be managed rather than a mind to be understood. Implicitly in marking and explicitly at times.
I went on to train as a mental health nurse, then as a cognitive behavioural therapist, and eventually became a lecturer in higher education. I have spent years sitting at the intersection of clinical practice and academic life. And the question that has followed me through all of it is this: who is higher education built for?
That question has become pressing again, not because of anything new about dyslexia or disability, but because of artificial intelligence. The conversation universities are currently having about AI is, I want to argue, is the wrong one. And getting it wrong has real consequences for the students who were already being failed before ChatGPT/Claude existed.
The inequality hiding in plain sight
Universities have made genuine progress on inclusion. Disability support services exist. Widening participation policies are real. The language of equity is everywhere. And yet has the fundamental architecture of academic life shifted? I would suggest very little.
Consider what a university still mostly asks students to do: produce fluent written /oral prose, under time pressure, in a single standardised format. Every part of that design disadvantages students whose cognitive style, educational background, or first language does not map onto this narrow template. Dyslexic students. Students from non-traditional educational backgrounds. Students for whom English is an additional language. Students managing mental health conditions.
The inequality is not just in who gets in. It is in who gets to demonstrate what they know once they are there. And it is compounded by something rarely named: the hidden labour of performing to the ‘status quo’. The extra time spent not developing ideas but translating them into an acceptable form. The self-monitoring required to approximate conventions that other students absorb without noticing. The energy spent managing how much of yourself to reveal. This labour is never acknowledged and never assessed. But it accumulates across years of study in ways that exhaust without educating.
The panic is pointing in the wrong direction
When AI writing tools became widely available, universities responded with some initial alarm. Detection software was deployed. Policies were written. The conversation became almost entirely about academic integrity, about catching students who might be using AI to complete their work.
What almost nobody asked was: what does it tell us that these assessment tasks could be completed by a machine? If an essay can be produced by AI, perhaps the essay was never really measuring what we thought it was measuring.
The detection tools themselves reveal the deeper problem. Research shows that AI detectors flag neurodivergent writers, students who write in non-standard ways, and non-native English speakers at disproportionate rates. (Eaton 2025) The tools built to protect academic standards are, in practice, penalising the students who were already most disadvantaged by those standards. The technology has changed. The inequality has not.
This is not an accident. As researchers studying machine learning have argued ( McQuillan 2022) AI systems are trained on dominant patterns and produce outputs that reflect and reinforce those patterns. A system trained on normative academic prose will treat deviation from that norm as suspicious. The student who has always written differently finds themselves, in the AI moment, newly legible as a potential cheat.
What AI could actually do if we let it
I want to make a case that the current conversation is missing. For students who have always struggled with the gap between what they think and what they can produce in the expected form, AI offers something genuinely significant: a way to close that gap.
When I use AI as a writing tool, I am not outsourcing my thinking. I am translating it. The ideas, the arguments, the clinical experience, the years of sitting with people and noticing what actually helps…all of that is mine. What AI helps me do is get it onto the page in a form that others can receive. I also use it to navigate the landscape of available knowledge, to find relevant research, identify useful reading, and map the territory of a topic more comprehensively than I could alone. This is what technology has always done at its best: expand what we can access in terms of information and express. Th leaps we have made forward from sitting in a library that only has access to its physical books and journals, to accessing the Internet and a much wider set of research, to AI potentially guiding this to more research and evidence base to all of the time enriching our contact with theory and information to help us learn better, is a point of hope.
For a dyslexic writer, this is not a shortcut. It is an accessibility tool. It is the removal of a barrier that was never measuring anything relevant. What universities are assessing, at their best, is not the ability to produce fluent prose unaided. It is the capacity to think critically, engage with complex ideas, form independent judgements, and contribute something genuinely new. AI cannot do any of that. But it can help people whose thinking outpaces their ability to express it in the approved form finally demonstrate what they know. The Universal Design for Learning framework developed by CAST has argued for years that inflexible curricula disable students who could otherwise demonstrate genuine capability. AI as a flexible medium of expression is exactly the kind of structural change UDL has been calling for not lowering expectations, but asking which expectations were ever really about learning.
The deeper question
The equity issue at the heart of this debate is not really about AI. It is about what higher education has always been measuring, and who that measurement has always served.
Students who have long known that the system was not designed for them, neurodivergent students, first-generation students, students from communities historically excluded from academic life, stand to gain the most from a genuine reckoning with these questions. They also stand to lose the most if universities respond to AI by adding a new layer of surveillance onto an unreformed system.
The real question is not how to stop students from using AI. It is what we were assessing in the first place, and whether we were doing it equitably. If we ask that question honestly, the answers will require changes that go well beyond AI policy .. changes to how we design assessment, how we think about academic writing, and how we decide whose ways of knowing count as knowledge. (Eaton 2025)
AI has not created the inequality in higher education. But it has made it harder to look away. That might be the most useful thing it has done.
References
Eaton, S. E. (2025). Neurodiversity and academic integrity: toward epistemic plurality in a post plagiarism era. Teaching in Higher Education, 1–20. https://doi.org/10.1080/13562517.2025.2583456
McQuillan, D. (2022). Resisting AI: An anti-fascist approach to artificial intelligence. Bristol University Press.
Jennifer Darling – Lecturer M.sc/PG Dip. / PG Cert Cognitive Behavioural Psychotherapy.
© 2025 University of Salford