Lessons From Horses

AI is a performance enhancing drug. How are universities responding and what does this tell us about where their priorities lie? 

I have always imagined university to be the place where I’d learn how to think. Somewhere that would give me the space, time, and tutelage to explore my creativity and interests: to discover not only my passions, but also critically and intelligently engage with them.

A slightly idealised image perhaps, but now, in the age of AI, a heavily threatened one. The use of AI seems paradoxical to everything that university stands for. Curiosity, creativity, and individuality are all watered down and flattened out.

As a first year History and Philosophy student, this seems obvious to me, and I know many of my peers share my concerns. The cognitive outsourcing that accompanies the use of AI for ideas generation, essay planning, problem solving has huge implications for the quality of our education.

However, many universities are doing little to apprehend the invasion of AI into education. Students are left feeling as if they have only two options, both of which are undesirable: use AI, potentially to the detriment of their intellectual and creative integrity, or don’t and settle for producing work that will appear at best average in comparison to work enhanced by AI. 

Perhaps predictably, considering the novelty of the technology, many universities seem uncertain– and conflicted– in their approach to AI. I have found that the personal opinion of academics often contradicts official university policy on permitted uses of AI. This leaves students feeling unsure about whether or not they should use AI to help complete their work, and, if they do, how extremely. The vagueness of both academics and university policy causes students to self-regulate their AI use, for fear of becoming overly reliant on it. The result of self-regulation, unsurprisingly, is huge discrepancy between students. 

Before setting off on a piece of coursework I have been told by an academic ‘don’t use AI, you’re better than that’. Along with being a slightly meaningless, totally unenforceable request, it also doesn’t align with university policy, which does permit some uses of AI- begging the question, better how? If using AI is not considered cheating by university policy, can you imply I will lose integrity if it helps me complete coursework? 

Admittedly, this question is an oversimplification because not all possible uses of AI are permitted. The Warwick History Department gives the advice “you should not use AI for any part of your assessments that you would not ask another human to do.” (Almost as vague as ‘don’t use AI’ is unenforceable!)

Although they do expand to include argument generation and planning/writing as inappropriate uses of AI and testing ideas, explaining concepts, or organising your thoughts as appropriate uses, I think the vagueness remains. On examination, there is very little distinction between argument generation (inappropriate) and testing ideas (appropriate). If I enter an idea into a chatbot, have a dialogue about that idea and then leave with an idea that, although might be descended from my original idea, has been deeply evolved by AI input, does that count as argument generation or testing ideas? 

Another takeaway from the original contradiction: some academics don’t actually know what the university guidelines on AI use are. It is very difficult for students to know how to approach AI when even their teachers do not. 

In comparison to the guidelines of other universities, Warwick sits mid-spectrum, neither particularly liberal or rigid. Durham, for example, permits many uses of AI such as “stylistic improvements to existing text”, “suggesting an outline for an essay”, or “suggesting avenues for research”. Although you are required to declare these uses (Warwick also requires a declaration, although I only discovered this through research rather than being told about it by academics), it is a much more lenient approach than the one taken by Warwick. Essay planning, I would argue, is an essential skill because of the condensing and organisation of vague, raw ideas that it requires. It is surprising that such a highly ranked Russell group university would be comfortable allowing this skill to remain undeveloped in their students. 

Oxford, on the other hand, has a less permissive policy than both Durham and Warwick. Their guideline states, “Any use of AI in work submitted for assessment is unauthorised unless you are specifically told differently in writing, in advance of the assessment, in instructions from your faculty’s or department for the specific assessment in question.” Similar to Warwick, and dissimilar to Durham, Oxford leaves decisions on AI allowance to the discretion of individual departments. 

Although it isn’t surprising that different universities have different approaches to AI since they are independent institutions, a comparison of these approaches is useful for more than just entertainment. After getting a degree, we all enter the same job market and the comparison of universities can become the difference between a dream job or persistent unemployment. If universities have very different policies on permitted uses of AI, the value of a degree from each of them could change. A university with a very lenient policy might gain a reputation for providing less impressive degrees than a university with a stricter approach. At a time when universities are being plagued by financial difficulties, such a reputation could be very harmful. 

When I asked the history admissions team at Oxford about how they prevent applicants using AI to enhance their submitted piece of work they said, “unsurprisingly there have been many discussions about AI, but there’s always been an issue with written work not being produced in the same circumstances by all applicants”. In other words, AI is just another way of enhancing work by means other than your own intellect or individual capability. It is not a uniquely threatening tool to intellectual integrity in education. A similar view was voiced by a Warwick History academic, who said “we are relying on past practices and then updating them for the age of AI”. While, theoretically, this approach may be sensible (although that is debatable- for starters, AI use is more difficult to detect than other forms of academic misconduct), it omits the issue of the free, easy availability of AI.

An academic from another Russell Group University said the difficulty of using past academic misconduct systems to deal with AI is that “the systems and processes we had in place before to detect academic problems [plagiarism or essay mills] are not big enough to deal with the number of referred AI cases”. Therefore, even ignoring the fact that AI is much more sophisticated than previous tools used for academic misconduct, it poses a greater threat simply because it’s so readily available. 

The enforcement of AI policies certainly isn’t helped by the fact that many academics still don’t seem to have a clear idea about what exactly AI is, and how it can be used. They are, essentially, behind most students in their understanding.

One of my seminar tutors referred, with slightly comically naive confidence, to ‘ChatGTP’ all year much to the silent, but collective, amusement of the students. Not that I would criticise such removal from the intrusion of AI into education as to be unaware you are mispronouncing the name of one of the most famous models- I think we would all benefit from the cat being put back in the bag. It doesn’t, however, fill me with confidence that we are being knowingly guided through how to cope with AI at university. 

It is not novel for younger generations to be more adept at understanding new technologies than older ones. The implications of this lag on understanding among academics is, however, unfortunate for the students who come before they catch up. An academic who conducted research into how students use AI was equally unaware of how pervasive an impact it was having on students, saying, “Many of us had quite a skewed understanding that… a few students use it [AI] in quite an extreme way but most students don’t use it”. Following the research it was discovered that “all the students we interviewed….. used it quite significantly….. in a more complicated way than we had understood”. As a student who was first exposed to AI in lower sixth, this isn’t surprising; experimentation will quickly teach you that the range of possible AI use is vast and complex. 

The same academic found that the greatest limiting factor on AI use was students’ self regulation, not any preventative measures put in place by the university. It was found that “All the students we interviewed had these lines they had drawn for themselves [on AI use]”. This implicitly reiterates the concern among students that excessive use of AI is damaging to the quality of their education, in tension with the concern that not using it will cause them to be left behind. One history student said, “we can’t use Chat GPT as an intellectual and creative crutch as we do ourselves a disservice”. 

Although I have been assured by academics that they don’t think avoiding AI will put someone at a disadvantage, this is difficult to believe. Academics with hundreds of essays to mark are bound to favour those that are clearer, more succinct, and easily readable, even if there is a higher likelihood that these are the essays that have been enhanced by AI. The takeaway from many academics’ combination of ignorance on AI and blasé confidence that AI is not uniquely threatening to academic integrity: students are abandoned to decide for themselves the appropriate level of use that will best satisfy both the concerns of cognitive outsourcing and being left behind.  

Finally, I wonder if education quality is not the only lens through which universities are viewing the AI conundrum. What if producing independent thinkers is not the only, or even the first, priority for universities? How big a role do financial concerns play? And, along with that, the pressure, particularly for Russell Group Universities, to maintain high grades that will attract prospective students.

A crackdown on AI use risks plummeting grades, especially if the Warwick History Academic is right in saying that “my suspicion is that the majority of students use AI in one way or another”. At a time when universities are seeking to grow their student bodies in an effort to achieve greater financial security, trying to seriously limit AI use, particularly now it’s been freely available for 6 years, might not be beneficial for grade statistics. The academic who carried out research into AI use said “the growth of universities is happening to the detriment of pedagogical quality”. Can students trust that the quality of their education is at the forefront of decisions made by the university regarding AI? 

I think it is clear that almost any use of AI, at least in the humanities, is not going to promote intellectual development within the human mind. The fact that universities are not taking a stricter approach suggests that either they have not fully grasped this or that we- their students- are not their priority.