From the Speccie:
Why isn’t Durham University taking AI cheating seriously?
At Durham University, I have been Chair of the Board of Examiners for Philosophy since 2016. Last week I resigned, because I feel that it is my responsibility to raise a vital issue in higher education, one whose true significance is not understood. The existence of a crisis requiring immediate action is not generally recognised. I am not blaming the deans and pro-vice-chancellors. I want to hold the appropriate figures to account: the Vice-Chancellors.
Durham University is a beacon of excellence in the UK university sector. I wish to maintain standards in its top-rated Philosophy department. The issue I am addressing affects students past, present and future, in many leading British universities. It involves a crime that is not victimless. It is this: the lazy student cheats with professional-grade versions of AI chatbots such as ChatGPT or Claude, and gets a first-class result. The hard-working student uses no bots, or a non-professional bot honestly, thinking and writing for themselves, but gets a 2:1. That’s not fair. Yet Vice-Chancellors – to mix metaphors – are sitting on their lavishly-remunerated backsides and adopting the ostrich position.
Let me explain how the university marking system works, at least in the UK. Until Covid, there was a mix of assessment by sit-down exam and continuous assessment. During Covid, this was replaced, for understandable reasons, by “at home” exams done on computer – to the relief of academics who no longer had to decipher student handwriting. After Covid this continued, and worked satisfactorily. A system called Turnitin would check for plagiarism by scanning essays for text compiled from published sources, and it was hard to cheat.
That system has been subverted by ChatGPT and similar AI tools. These apps now work very well for producing university-style essays and essay components, especially in the professional version which many Durham students can afford. (You see that there is discrimination already against working-class students. What a surprise.)
A return to sit-down exams is the obvious solution. Universities say that these are financially impractical for various reasons. I doubt this, but in any case it is too late at Durham and elsewhere to implement sit-downs exams this year. (They should have been implemented a year ago.) So in response to the immediate crisis I must suggest sticking plaster solutions. I am old enough to recall the introduction of anonymous marking, which happened because research showed that female students were discriminated against. Now, when the alternative is what I call a chaotic system of guesswork, it may be our least-worst immediate alternative.
AI, in contrast, is now often impractical to detect with enough reliability to meet the high standards of proof required for accusations of cheating. Perhaps there is no ultimate alternative to sit-down exams – other options such as vivas are very time-consuming. My suggestion of abandoning anonymous marking seems a reasonable sticking plaster. We need an immediate response to the unfairness in the coming exam season. Yet most of my colleagues don’t seem to understand how it is fairer.
The buck stops with VCs, who should be exercising leadership. Obviously students who cheat are behaving badly – but the point is that the system allows them to do so with ease. Students have always cheated, but now they know they can do so with impunity.
I have been an academic since 1988, and worked at Durham University since 1991. I’ve been very lucky to get a job at Durham – and privileged. The students are excellent, the university is well-known and generally well-run. Any student who gets a place here will be both excellent academically, and well-set for future employment. Academic work is the only full-time employment I’ve had, after a series of temporary jobs as a student, and taking a PGCE course at Moray House Edinburgh (Primary). But now, shortly before retirement, I find that the skills I’ve developed as a marker – essential to my teaching role – aren’t able to be put to their proper use. Much effort is being expended by academics in working out how to identify the improper use of AI. But this vain pursuit is becoming very difficult and may soon be impossible. AI “tells” that remain are being aggressively stamped out by AI companies. One cannot mark essays under the Covid system, in the era of ChatGPT. Yet the university leadership is still in the ostrich position.
I was on research leave for the first term this academic year; Durham still operates a term system. Since then I have chaired one disciplinary panel for AI cheating, detected by the junior colleague who marked it. He noticed that the referencing contained tell-tale signs of AI “hallucinations”. This AI “tell” can prove misuse, as has happened in a number of high-profile cases. But AI makers have reduced its frequency, and cheaters can cover their tracks by minimally checking or omitting references. With existing tools, AI cheating is almost impossible to prove except by student confession. Stylistic cues, incongruous sophistication or maturity of writing for an undergrad cannot reliably show that AI has been misused, even when the marker is suspicious. Only with the last batch of essays which I started marking in April, did the true horrors of the situation become apparent to me. The system was broken and needed immediate emergency repair.
It should be clear that my complaint is not against Durham University. Few British universities have re-introduced sit-down exams extensively. Many departments have never had anonymisation; music, for obvious reasons. I know from bitter experience that one cannot play the piano with a paper bag over one’s head. Universities seem unable to respond in a timely manner to the current crisis. Academics are grumbling. Someone surely has to blow the whistle?
I guess that few whistle-blowers are enthusiastic. I have never engaged in a lawsuit, or been the victim of one, on any matter. A cursory Google search reveals that UK whistleblower laws protect my salary and pension, and I’ll get a hundred quid for this article. So I’m prepared to light the blue touchpaper and retire, in both senses.
It has been a very difficult decision to get this article published. Colleagues will be upset, and students disconcerted. But should they continue in ignorance? There needs to be quick action – something that large institutions such as universities have never been known for. Think of Air Chief Marshal Dowding in 1940, confronted with massive losses of planes in France. He goes to Winston Churchill and says that we cannot send any more Hurricanes and Spitfires there. But instead of reluctantly agreeing, Churchill sets up a cabinet committee to investigate, and committees in parliament. By the time they report, Britain is being invaded. I think I would be on Dowding’s side in that debate.
The imperative is the wellbeing of students who are suffering under the present system, though they don’t realise it. Staff in UK universities, and elsewhere, are not able to mark with the integrity the matter demands, because there is no sufficiently reliable way of detecting or preventing improper use of AI. Academics are doing their best, but as one young colleague said to me, “We are fighting with our hands tied behind our backs”.
Cromwell’s immortal words to the Rump Parliament apply to the current generation of VCs: “You have sat too long here for any good you have been doing lately…Depart, I say, and let us have done with you. In the name of God, go!” We need an equivalent of the mani pulite movement in Italian politics – one that does not end with a Berlusconi.
by Andy Hamilton, Professor of Philosophy and former Chair of Board of Examiners at Durham University
Full disclosure, my son is at Durham doing an humanities degree and I totally agree there should be a return to in person degrees for all exams. He doesn’t use AI to write essays tbf, but does use it to help plan, find sources and check his work acting on recommendations he agrees with.