ChatgPT recalciously writes work to students. “You are overwhelmed by a sense of anxiety”

Three years after CHATGPT's debut, generative artificial intelligence has become a common tool in the academic environment, and detecting its abuse remains a big problem – writes the Belgian daily de Morgen. Although the number of students officially caught using AI is still small, experts emphasize that these data are seriously underestimated.

“Sometimes while reading a student work you are a sense of anxiety. You know that something is wrong, but you can often not prove that the whole was written by generative artificial intelligence,” said the newspaper prof. Steven Vanderputten from the University of Ghent. He pointed out that lecturers do not have tools today to clearly confirm the authorship of the texts. If there is a lack of evidence, lecturers often have to withdraw from disciplinary activities.

According to the university in Ghent, in the academic year 2022/2023 artificial intelligence appeared in 19 out of 73 plagiarism proceedings. A year later, 13 such cases were recorded, including three in the master's thesis. Similar tendencies were demonstrated by the Catholic University of Leuven (Ku Leuven), where the number of cases increased from 18 to 23 during the year.

Professor of literature Pieter Vermeulen from Ku Leuven noted that the influence of AI can be seen even where there is no evidence of abuse – analyzes are becoming shallower and the work is more and more like each other. In turn, prof. Vincent Ginis from the Free University in Brussels assessed that AI “democratized” the possibility of commissioning external tasks to “helpers”.

Research cited by the journal indicates that in 2025 as much as 92 percent students in Great Britain declared the use of AI tools in the implementation of academic tasks, against 66 percent. a year earlier.

Universities are trying to react, ordering students more partial work and introducing mandatory oral presentations to make it easier to catch unfair practices. Experts admit, however, that the fight against fraud related to AI resembles a race between guards and poachers, in which none of the parties has a permanent advantage.

At the beginning of August, the 5.0 Chatgpt version appeared – the first one, which combines the advantages of earlier generations of the model and, according to the creators, is to work faster and more precisely.

Patryk Kulpok (PAP)

PMK/ MMS/