Artificial intelligence is both a threat and a necessary tool for academia, writes IAN RITCHIE
Our ‘ancient’ universities have been around for a long time. St Andrews was founded in 1413, Glasgow in 1451, and Edinburgh in 1582. Over the centuries they have changed their courses and delivery methods to survive and remain relevant. Now they are facing, arguably, their biggest ever challenge – coping with the AI revolution.
Universities are engaged in developing knowledge among their students, which are mostly measured by requiring students to write reports to demonstrate their understanding of the knowledge they have acquired.
The difficulty with this model is the recent arrival of powerful artificial intelligence (AI) systems, which are increasingly capable of automatically generating written reports which usually score highly when assessed.
It has been demonstrated, for example, that AI systems can successfully pass legal bar exams, receiving the same or equivalent evaluations as human candidates and scored against the same marking criteria.
A recent study at Edinburgh’s Napier University, based on a survey of more than 6,600 students at seven UK universities, found that 32% admitted using AI to create their submissions.
Universities Scotland reported in 2024 that more than 1,000 university students were caught using artificial intelligence to cheat in assessments including Abertay University in Dundee with 342 cases, Stirling University with 200, Robert Gordon University with 116, and Glasgow Caledonian University with 113.
What may, at first glance, seem to be the obvious solution – to ban the use of AI in preparing student submissions –would be completely inappropriate. After all, a university education is largely designed to prepare students for a future career after they graduate, and that career will undoubtedly call for individuals to make intelligent use of AI tools and learn how to get the best out of these powerful resources.
The Scottish government wants to grow these skills. It published in March this year a wide range of AI literacy educational actions. A new body, AI Scotland, will bring together government, academia, industry, and the third sector, supported by an independent Expert Advisory Board.
A graduate who can use AI to analyse a large dataset, write a first draft, generate software, search a literature base or explore alternative designs should be substantially more effective than someone who can’t.
For years, universities have taught skills in how to use computers, spreadsheets, databases, search tools and the internet and now AI literacy must be added to this list, knowing what it can do, how to instruct it, and when not to trust it.
One significant worry is that AI can still produce convincing but incorrect information, invent references, misunderstand a problem or reproduce biases. Universities therefore have a role in teaching students to critically evaluate AI output, rather than simply accept it.
Universities are having to reconsider traditional essays, take-home assignments and examinations because AI can convincingly generate them. Increasingly, the emphasis is shifting towards things that demonstrate understanding, reasoning, experimentation and application, rather than simply producing text.
The University of Glasgow has one of the clearest institutional approaches. It explicitly says its objective is to develop students who can use GenAI “effectively, critically, and transparently”, while preserving critical thinking, analysis and judgement.
Its student guidance encourages AI for things such as: exploring difficult concepts; brainstorming; structuring arguments; generating practice questions; checking language and grammar; and summarising material. And also tells students that AI should not substitute for independent thought, research, source evaluation or disciplinary knowledge.
The problem remains with some students tempted to ask the AI to write their reports for them and then submitting the result as their own work. University authorities are faced with the problem of detecting this unacceptable behaviour.
Software tools are available which claim to detect AI generated reports, but these have been widely discredited. There are several cases of students who have been disciplined for submitting AI-written reports which upon challenge and investigation were actually written by the author.
Instead, students are encouraged to use AI in preparing their submissions but are being asked to ‘show their workings’. They are asked to show their sources and demonstrate how their work has been assembled and may be asked to showing their intermediate research notes and various draft versions.
Ideally, they may participate in tutorial sessions and be subjected to oral challenge and testing of their understanding.
Of course, all of this requires that the evaluation and marking of student work will inevitably require more work for the lecturers doing it, at a time when university resources are under huge strain.
A few weeks ago, Sir Anton Muscatelli, the former Principal of the University of Glasgow published a report on the financial health of our university sector. The findings were stark – in 2023/24, there was a £200m gap between what universities spent and what they made.
And looking across all the data between 2016/17 and 2023/24, Muscatelli found the same “bleak” picture. Only in one of those years was there a surplus, caused by Covid.
If the government is serious about growing Scotland’s AI skills, it will have to square this circle.
Leave a Reply