Majority of students deny using AI when forbidden, study finds

The majority of students using artificial intelligence (AI) as part of their studies claim to be honest about how they apply it, according to the results of a survey of more than 6,000 students at UK universities.

Of those who said they are incorporating the technology into their degree, 67 per cent said they would not use it to help with “part or all” of an assessment if they had been instructed not to, a study from Edinburgh Napier University has found.

Nonetheless, 5.2 per cent of all habitual users admitted to using generative AI “most of the time” or “always”, even when they knew they were not permitted to.

The results follow a report from plagiarism detection service Turnitin, which suggested American students were depending on AI far more than their Australian or British peers.

Generative AI's ability to produce entire essays has sparked fears of mass cheating and “cognitive offloading”, in which people may cede their own intellectual effort to the technology. Some universities have responded by returning to in-person exams.

The latest findings form part of a preprint research paper from independent researcher Stephen Gow and Edinburgh Napier academics Sam Illingworth, Khristin Fabian and Callum Goddard, who concluded that, “for those that are using it, the majority of students are honest the majority of the time and are trying to avoid submitting direct GenAI outputs”. 

Illingworth told Times Higher Education that students face an atmosphere of “uncertainty” when it comes to institutional perspectives on the fast-developing technology.

“Students want consistency, transparency and fairness,” he said. “At the moment they are not getting this. They understand that it is an evolving field but they need help in understanding what it means for their studies and their future careers. It is an amazing opportunity for universities and sadly one that we are failing [at].”

Inspired by a 2024 Australian study – Student Perspectives on AI in Higher Education – the researchers polled students at seven universities in England, Scotland and Northern Ireland on their trust in AI and motivations and capabilities when it comes to using it, among other questions.

The study highlights the widespread misconception that AI use is virtually universal, finding significant evidence of “conscientious objectors” – with 32 per cent of respondents reporting that they did not use it for their studies.

Some 59 per cent of men surveyed said they were “regular” users compared with 44 per cent of female respondents.

But the group most likely to rely on the technology was international students, with non-native English speakers often using it daily (32 per cent) or weekly (34 per cent).

Those who do use the technology are not totally trusting: more than half (51 per cent) had a “negative sentiment towards the accuracy of outputs”, although 76 per cent said they felt confident in their ability to prompt well enough to generate the desired result.

“The majority of students are using GenAI for a range of learning and assessment tasks, and it is increasingly embedded in everyday technology,” the report notes. It adds: “Conscious universal usage of the technology, however, cannot be assumed, and there are growing signs of an AI backlash, particularly by conscientious objectors who see there are concerns about the ethical foundations of the technology and see their futures threatened by its rise.”

“It’s a challenging time for higher education,” Gow told THE. “It’s also the opportunity to step up.”

Gow said it was time to “link up” approaches to how the technology is used across education, and called for schools and universities to work together on the matter.

He added: “It gave me profound hope, talking to students…the main [story] that people are drawing upon is that students are using it to cheat…but when I was talking to students I was seeing a very different picture.”

Illingworth added: “We need to involve students in our dialogues around AI.

“Not as focus groups after the fact, but as co-creators of meaningful policy that helps alleviate their fears and teach critical AI literacy they actually need.”

georgia.luckhurst@timeshighereducation.com