⇧ [VIDÉO] You might also like this partner content
Students see artificial intelligence as an unexpected boon. Some use it wisely, for example to summarize lessons, while others see it as a way to cheat, often successfully. Recently, a British study revealed that 94% of work written with AI escapes the vigilance of teachers.
In the past, plagiarism was one of the most common fraudulent academic practices. Today, with the advent of AI, the situation has changed dramatically. Technologies such as ChatGPT are increasingly used by students to complete their homework and take exams. These tools, which are ever more efficient, make it possible to obtain responses of often higher quality. Indeed, according to certain studies, the use of generative AI makes it possible to obtain higher grades than those that students would have obtained on their own.
Although these technologies have the potential to partially change the education sector for the better, they are beginning to pose a real danger due to their misuse. With such tools, assessments may no longer reflect actual mastery of a subject, which could diminish the value of degrees and other academic qualifications. Teachers sometimes seem helpless when faced with this technology. A recent study conducted at the University of Reading in the United Kingdom shows that professors struggle to identify work generated by AI, as detailed in the results published in the journal Plos
One.
Grades enhanced by AI
The experiment took place as part of an undergraduate psychology course at the University of Reading. The researchers created fictitious profiles of students, who submitted work written entirely by OpenAI’s GPT-4 language model, without informing teachers. During marking, 94% of these assignments escaped detection as AI-produced work. The remaining 6% were identified as cases of poor academic practice, without the correctors specifying the software origin of these texts. In addition, the majority of works received higher grades. Specifically, in 83.4% of cases, AI-written assignments received better evaluations.
According to the researchers, these high results are mainly explained by the performance of the technology. AI has learned fundamental skills valued in education, such as information integration and critical analysis. She is now able to provide analytical answers and detail the thought process behind her findings.
Online exams: fertile ground for cheating via AI
The rise of online sessions, particularly since the COVID-19 pandemic lockdowns, has significantly facilitated the use of AI to cheat on exams. In the absence of direct supervision, students have more freedom to exploit these tools. AI, easily accessible and often free, is widely used by students.
See also
To counter this growing threat, the authors of the study recommend a return to face-to-face exams. Failing that, they recommend increased monitoring of online exams. The researchers emphasize that AI detection tools are difficult to integrate due to their lack of reliability, especially since some AI tools are capable of “humanizing” content, i.e. to adjust them to resemble texts written by humans, in order to make them undetectable.
