Academic dishonesty is becoming an increasing threat to UK universities. This concern is deeply connected to the emerging use of AI tools, particularly ChatGPT. Our recent survey found almost 7,000 documented instances of AI cheating during this current 2023-24 academic year. That comes to 5.1 positive cases per 1,000 students. These results point to a concerning trend that has developed as AI technology has gained more accessibility among students.
The problem of the nefarious uses of AI now has reached a level equal to that of regular plagiarism. This academic year, almost 27% of responding universities opted to go a different route. They pulled back from labeling AI misuse as a new category of misconduct, as in prior years’ reports. Without this formal recognition, it will be difficult for universities to tackle the endemic issue.
And these experts agree that even the scary misuse cases of AI that get reported on barely scratch the surface of what’s going on. In fact, most cases of misallocation go unchallenged. One unnamed top expert observed, “I would imagine those that get caught are just the tip of the iceberg. Compared to plagiarism, where you can verify the duplicated text in question, AI detection is extremely different. This feeling is representative of a deeper fear that has taken root in the academic community in response to the challenges that AI tools now present.
Dr. Thomas Lancaster, an academic expert in this field, noted, “When used well and by a student who knows how to edit the output, AI misuse is very hard to prove.” This begs the question of how well our detection methods are working right now. We don’t need gimmicks, we need more thoughtful, complex approaches to accurately measure and evaluate student work.
Scientists at the University of Reading decided to find out just how big a problem this is. To prove this, they tested their own assessment systems and found that students could submit AI-generated work without detection 94% of the time. The study, co-authored by Dr. Peter Scarfe and others, underscores the pressing need for educational institutions to adapt their assessment methods in response to emerging technologies.
The increase in AI-related misconduct is particularly striking given the regularity and consistency of a significant decrease in cases of traditional plagiarism. The share of confirmed instances of plagiarism has gone from 19 out of every 1,000 students in prior academic years down to 15.2 in 2023-24. Specialists anticipate this number to drop even more, down to an estimated 8.5 per 1,000 students by next year. It reflects a significant societal trend that the old school cheating is dead. At the same time, fresh threats from AI are quickly rising up to fill their stead.
The education sector recognizes that it needs to change to rise to these challenges. As Dr. Scarfe pointed out, “the education sector would have to adapt to AI, which posed a fundamentally different problem.” A lot of our colleagues are concerned that if we return to traditional assessment with in-person, proctored exams, we’ll avoid the pitfalls of AI misuse. They question if these approaches will be usable or successful.
Harvey*, a university student, shared his perspective on the integration of AI into academic life: “ChatGPT kind of came along when I first joined uni, and so it’s always been present for me.” To be clear, he doesn’t use AI tools to have students outright plagiarize. Rather, he uses them to ideate on next generation concepts. “I’m highly skeptical anyone would use AI to just lift its output verbatim,” he said. In fact, I think what they primarily use it for is to have blue sky type brainstorming to develop entirely new concepts.
Harvey’s remarks highlight how many students today perceive AI as an ally and collaborator in their work. They feel it’s very useful as a tool as opposed to just a shortcut. He emphasized the importance of focusing on skills that AI will not be able to easily replicate. Skills such as communication and interpersonal skills are increasingly fundamental to successful careers.
As the misuse of AI poses daunting challenges, there may be a silver lining, experts say. They think the educational system has the capacity to get students far more actively invested in their own learning journey. An alarming fact that Dr. Lancaster identified was that many students think that university-level assessment is a waste of time. He challenged educators to increase student interest by being transparent about the intent of assignments.
The worry goes deeper than just high-profile instances of officer malfeasance. So far the figures up to May are pointing to considerable backtracking. Proven misuses of AI are increasing and may soon be approaching 7.5 proven misuses per 1,000 students this school year. This projection sends shockwaves through the educational community as educators worry that we’re simply not doing enough to fight this rising tide.
The UK government is responding by investing more than £187 million in national skills programs. They have published new recommendations for the most responsible, effective uses of AI in teaching and learning environments. Science and Technology Secretary Peter Kyle emphasized that AI should be deployed to “level up” opportunities for dyslexic children, highlighting the technology’s potential benefits alongside its risks.