This is a partial excerpt from a Maeil Business article.
https://v.daum.net/v/20260830174809497
https://www.mk.co.kr/news/society/12139925
According to Maeil Business reporting on the 30th, Seoul National University terminated its partnership contract with the AI detection service 'GPT Killer' on the 1st of last month. GPT Killer is a service that shows the probability that an inputted document was written by generative AI. Seoul National University is not providing GPT Killer usage rights to both students and faculty. A Seoul National University official explained, "Currently, AI detection services do not clearly determine whether AI was used, but only estimate the possibility that AI wrote it," and "We determined that using it in situations with low accuracy could create side effects."
Seoul National University's decision stands in contrast to the trend in the academic community, which has been strengthening the use of AI detection tools in response to the spread of AI academic fraud. According to Muhayou, the operator of GPT Killer, the usage of GPT Killer during the October midterm period last year increased 3.9 times compared to the same period the previous year. In particular, the usage for professor evaluation, which inspects assignments and such, surged 4.3 times.
The problem is that the reliability of AI detection tools has not been sufficiently verified. Choi Byung-ho, a research professor at Korea University's AI Research Institute, said, "Most AI detection tools analyze how much generative AI-specific patterns appear in text," and "Even if a person writes something, it can be misidentified as written by AI depending on frequently used expressions or patterns, and conversely, AI-written text can also evade detection." [Partial content omitted]
Major universities abroad are also not using AI detection tools in determining academic fraud due to low reliability and potential bias. About 10 major U.S. universities including Yale University, Johns Hopkins University, and New York University are restricting the use of the AI detection service 'Turnitin' due to low reliability and bias concerns. Japan's Waseda University also ended the provision of Turnitin functionality in December 2023 and stated as of September last year that "there are no plans for future implementation."
Experts emphasize that given generative AI has become an everyday learning tool, the existing evaluation method centered on reports and assignments should be changed. This is because if the existing evaluation method is maintained and only a detection tool is added, "whether AI was used" could become the focus rather than students' critical thinking.
[Remaining content omitted]
