The research, based on survey responses from 95,513 students across 20 major public research universities in the US, found that about two-thirds had used generative AI during the 2023-24 academic year, while 37 per cent used it at least monthly. About 9 per cent of students who used the tools said they had submitted AI-generated work despite knowing, or believing, that such use was not permitted.
The findings sharpen a dilemma confronting higher education: tools such as ChatGPT can support learning, coding, drafting and data analysis, but they also weaken the reliability of take-home essays, projects and other assignments that were designed to measure individual capability. The study’s authors argue that the answer cannot be a blanket ban, because AI competence is becoming relevant to many graduate jobs. Instead, they say universities must change what they assess and how they verify learning.
“Assessment reform is necessary and urgent,” said René F. Kizilcec, associate professor of information science at the Cornell Ann S. Bowers College of Computing and Information Science and director of the Future of Learning Lab. “The fact that students are misusing GenAI is a problem for assessment validity, and that’s a problem for the credibility of university credentials.”
The study was co-authored by Igor Chirikov of the University of California, Berkeley, Ivan Smirnov of the University of Technology Sydney and Complexity Science Hub Vienna, and Kizilcec. It used an indirect questioning method intended to improve reporting of sensitive behaviour such as academic misconduct, reducing the risk that students would understate cheating.
The data show wide variation across disciplines. Regular AI use was highest in computer science, at 62 per cent, followed by mathematics at 53 per cent and business at 51 per cent. Arts students reported much lower regular use, at 24 per cent. Misuse, however, was not concentrated only in technology-heavy fields. Estimated cheating rates were higher in some non-STEM disciplines, including economics at 17 per cent and journalism at 16 per cent, while biology stood at 5 per cent.
Frequency of use was a major warning sign. Among daily users, 26 per cent reported cheating with generative AI, compared with 7 per cent among monthly users. Researchers cautioned that the figures do not prove that frequent AI use causes misconduct, but the pattern suggests that repeated exposure may make it easier for students to cross boundaries, especially where course policies are unclear or assignments are easy to outsource to a model.
The study also identified equity concerns. Male students reported regular use at 45 per cent, compared with 33 per cent among female students. White and Asian students reported regular use at 39 per cent, compared with 29 per cent among students from underrepresented racial minority groups. The gap raises concerns that students with less access, confidence or training in AI may be disadvantaged as workplaces begin to expect familiarity with such tools.
Universities are already experimenting with responses, including oral examinations, supervised writing, in-class problem solving, practical demonstrations and assignments that require students to show drafts, reasoning, data choices and engagement with feedback. The researchers argue that disciplines should tailor reforms to their own learning goals rather than rely on generic detection software, which remains imperfect and can produce disputed findings.
The challenge is particularly acute because generative AI is now built into search engines, word processors, coding platforms and grammar tools. Students may encounter AI assistance without actively seeking to cheat, while one-click rewriting and summarisation features blur the line between acceptable support and unauthorised substitution of work.
Follow Arabian Post
Select Arabian Post as your preferred source on Google and MSN News for trusted business news and Arab politics and updates.
