AI is the lazy option for young people entering university. Log on. Send a prompt, and minutes later language and analysis appear, answering the tough professor's question in a usable form, ready to cut and paste. It is all so tempting and so simple. The language may be flat and lacking. The style may be anodyne. Delusional errors may result. Nuance is lost. Critical points may be lacking. Perspectives muddied or misconstrued. Yet students still think the risk is worth it. Most students are using AI (90 to 95 per cent in recent surveys), with foreign students particularly reliant on the AI-slop machine. Unfortunately, excessive use of AI is bad for young students.
Today, a battle is underway in universities. It is a struggle to ensure that the widespread reliance on AI by undergraduate students does not impair their intellectual development or cause a dumbing-down of their critical abilities and academic skillsets, right at the point when they ought to be flourishing under the guidance of their teachers.
This unstructured, ad hoc academic response is crucially important to us all. Generations of workers will use AI and must learn how to harness it to avoid ethical mistakes, misdirection and possible damage from a technology that is not designed to improve student performance but instead to enrich AI bros, regardless of the negative impact on undergraduates, workers and their success.
Research illuminates the scale of the dangers students face when AI becomes the all-too-easy option.
Read more:
- The AI revolution is already in Scotland's classrooms – but who's in charge?
- The £40bn case for maths: Why AI makes teachers more important than ever
- AI data centre giant to create 2,100 permanent on-site jobs in Scotland
- Scotland’s AI revolution: 18 key people shaping its future and defining its backlash
A key 2026 report by the MIT Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training warns of "cognitive surrender" by students who often fall back on AI at the first hint of difficulty in their work. The team found diminished critical thinking, weakened memory, eroded confidence and poorer mastery of subjects. All are indicators of easy regurgitation rather than education among AI addicts.
Students' brains do not benefit. A further MIT study of essays written by students reliant on ChatGPT found the number of neural connections to be lower, with students having less recall of their "own" text weeks later. The AI-fixated were not really learning, and their brains reflected this disappointing reality.
Research also finds that AI summaries of key texts are damaging, dumbing subjects down and making understanding harder, not easier. This rings true for any of us who benefited from great teachers. They taught us by engaging with us as we read the actual text, argued about it and internalised it.
Decades of studies show the importance of the gradual build-up of expertise and understanding through reading, writing, analysing and retaining critical knowledge, guided by talented professors.
Forty years on, I remember the names of my best professors, Mike Clarke and Phil Daniels. Why? Because they, and the lectures they delivered, enabled me to come to grips with and understand topics and approaches to thinking that were new to me. My writing gradually improved. My abilities evolved. My first essays would embarrass me today. But it is this cognitive friction, this struggle to understand, to form and defend theses, that makes us improve as engaged students. My final-year essays reflected this journey. They still surprise me positively when I read the tattered hard copies.
Student behaviour is signalling how AI reliance is an off-ramp from the university processes and practices that enhance the learning experience. Universities report that the number of students using study groups is down, and student-instructor interaction is down as well. This makes sense. If an AI large language model has already given you a bland, serviceable answer, there is no need to go to a study group. Nor is there a reason to take your questions to the professor. Indeed, because the student knows they have not really wrestled with the subject, exposure could be embarrassing. Better to keep your head down, submit the slop and move on.
A considerable proportion of students understand that easy answers from AI slop are not education, and they want direction and support. They are, after all, paying for their education. Twenty-four per cent, according to YouGov, want stricter rules. Yet 20 per cent admit cheating.
To fight back against this corrosive misuse and abuse of AI, its distortive anti-educational influence and its cognitive harms, universities and professors are going back to the future. What do I mean?
In-class tests are back, with no access to the web. Students who have prepared in advance must complete timed assignments during class. This high-pressure approach reveals who has learned and understood concepts rather than merely parroting AI summaries.
Seminars are back. Large lectures remain important for introductory coursework, but seminars matter because professor and student are face to face, helping the teacher judge the student's comprehension while improving the student's understanding and internalisation.
Oral examinations are back. These, like seminars, permit a judgement of the student's fluency and subject mastery, not their copying and plagiarism abilities.
Final written exams are back. Students do not like them, but they are a real test of what you know and how effectively you can articulate a case, cogently and compellingly, under tight time constraints.
Crucially, universities should require all students to take ethics courses on the use and abuse of AI large language models. Plagiarism was, and remains, disqualifying in academia. Universities need to re-teach and reinforce the importance of ethical behaviour. Morals matter. Shame is effective. Studies show that a student who has a clear personal moral compass on cheating is less likely to abuse AI and consequently performs better and scores higher at university. Ethical education on AI use is a must going forward.
Let me be clear. I believe misuse of AI by students is fixable. Universities can address it. We must learn how to use AI to support human intellectual development and evolution. We can do that by recognising the risks and addressing the dangers, instead of listening to techno-optimists who are selling an addictive, easy route to laziness and collective intellectual decline.
Stuart PM Mackintosh is a British economist and financial author who was formerly the Executive Director of the Group of Thirty, an international financial think tank.
Share