![\"\"](\"https:\/\/images.newscientist.com\/wp-content\/uploads\/2024\/06\/25114153\/SEI_209782933.jpg\")
<\/div>Exams taken in person make it harder for students to cheat using AI<\/p>\n
Trish Gant \/ Alamy<\/p>\n<\/div>\n<\/figcaption><\/figure>\n<\/p>\n
Ninety-four per cent of university exam submissions created using ChatGPT weren\u2019t detected as being generated by artificial intelligence, and these submissions tended to get higher scores than real students\u2019 work.<\/p>\n
Peter Scarfe<\/a> at the University of Reading, UK, and his colleagues used ChatGPT to produce answers to 63 assessment questions on five modules across the university\u2019s psychology undergraduate degrees. Students sat these exams at home, so they were allowed to look at notes and references, and they could potentially have used AI although this wasn\u2019t permitted.<\/p>\n The AI-generated answers were submitted alongside real students\u2019 work, and accounted for, on average, 5 per cent of the total scripts marked by academics. The markers weren\u2019t informed that they were checking the work of 33 fake students \u2013 whose names were themselves generated by ChatGPT.<\/p>\n The assessments included two types of questions: short answers and longer essays. The prompts given to ChatGPT began with the words \u201cIncluding references to academic literature but not a separate reference section\u201d, then copied the exam question.<\/p>\n <\/p>\n Across all modules, only 6 per cent of the AI submissions were flagged as potentially not being a student\u2019s own work \u2013 though in some modules, no AI-generated work was flagged as suspicious. \u201cOn average, the AI responses gained higher grades than our real student submissions,\u201d says Scarfe, though there was some variability across modules.<\/p>\n \u201cCurrent AI tends to struggle with more abstract reasoning and integration into information,\u201d he adds. But across all 63 AI submissions, there was an 83.4 per cent chance that the AI work outscored that of the students.<\/p>\n The researchers claim that their work is the largest and most robust study of its kind to date. Although the study only checked work on the University of Reading\u2019s psychology degree, Scarfe believes it is a concern for the whole academic sector. \u201cI have no reason to think that other subject areas wouldn\u2019t have just the same kind of issue,\u201d he says.<\/p>\n