University lecture hall. The photo is unrelated to the article content. Yonhap News. 

A collective cheating incident has been discovered during a midterm exam in a Yonsei University course, causing a stir on campus. Notably, it is known that a considerable number of students used AI, such as ChatGPT, in this process.

According to a comprehensive report by Yonhap News on the 9th, the professor of "Natural Language Processing (NLP) and ChatGPT," a course for third-year students at Yonsei University's Sinchon campus, recently announced, "Numerous instances of student cheating have been discovered. Students who confess will have their midterm exam scores marked as zero, while those who deny it will face academic probation in accordance with university regulations."

This course, which teaches generative AI such as natural language processing and large language models (LLMs), is reportedly attended by approximately 600 students. Due to the large enrollment, classes are conducted remotely. The midterm exam, held remotely on the 15th of last month, is where the incident occurred.

Measures were in place to prevent cheating. The exam involved answering multiple-choice questions by accessing an online site. Test-takers were required to submit videos of their computer screens, hands, and faces throughout the exam duration.

However, it is reported that some students engaged in cheating by adjusting their camera angles to create blind spots or by overlaying multiple programs on their computer screens. Some students also reportedly captured exam questions or continuously changed their screen windows and programs.

In response, the professor, after a thorough review of the videos with teaching assistants, identified these suspicious activities and encouraged students to "confess."

While the exact number of students who cheated is unknown, there are whispers among students that it could be more than half. One student posted a poll on the university community board "Everytime" titled "Let's vote honestly." As of that day, out of 387 respondents who identified as not being enrolled in the course, 211 voted "cheated" and 176 voted "solved it myself."

It is suspected that a significant number of students secretly used AI during the cheating process. Student A (25), a student in this course, told Yonhap News, "Most people use ChatGPT for the exam. It's a calculation that it would be difficult to get a good grade if I'm the only one not using it." Student B, who took the course last semester, also confessed, "Many of my friends, including myself, took the exam while searching with AI."

In relation to this, a Yonsei University official stated, "So far, about 40 students have confessed to cheating, and 10 students whose cheating is suspected have not yet confessed. If cheating is confirmed for students who have not confessed, disciplinary action may be considered."

Generative AI infiltrating various aspects of daily life [Yonhap News archival photo]

Although it has been three years since generative AI became widespread, the confusion in universities appears to be growing. As AI capabilities rapidly advance and surpass the level of mere learning aids, university policies and ethical discussions on AI usage are failing to keep pace with the speed of development.

Last year, 91.7% of 726 university students from four- and six-year programs reported having used AI for assignments or research (survey by the Korea Research Institute for Vocational Education and Training). However, 71.1% of 131 universities nationwide had not yet established guidelines for generative AI (Korean Council for University Education).

There are also concerns that students' increasing reliance on AI is diminishing their ability to think independently.

Professor Jeong Gi-in of the Department of Korean Language and Literature at Kyung Hee University stated, "It's like students who need to learn how to walk are riding motorcycles (with AI). We are emphasizing the risks of AI dependency, but it's not easy."

Arguments are also being made for redefining educational and assessment methods in the AI era. Suggestions include actively permitting AI use while requiring transparent disclosure of sources.

Kim Myung-joo, Director of the AI Safety Research Institute, said, "We need to educate students on critical thinking by having them submit not only AI-generated results but also their personal opinions." Choi Byung-ho, Director of the Korea University AI Research Institute, suggested, "We need to develop new educational methods such as in-person presentations and in-depth discussions."

Yonhap News