A demonstration of cheating using smart glasses is conducted during a regular briefing by the Ministry of Education held at the Sejong Government Complex on the morning of the 16th. Despite controversy over cheating using ultra‑miniature cameras and artificial intelligence (AI), the Ministry of Education announced that it would not postpone the 2027 College Scholastic Ability Test (CSAT), scheduled for the 19th. However, specific measures to detect cheating using ultra‑miniature cameras have not yet been finalized. The ministry plans to review possible actions over the remaining three days. News1
Cases of cheating using artificial intelligence (AI) glasses (smart glasses) are being uncovered one after another, ranging from national qualification examinations to middle and high school academic assessments and university essay exams.
The Human Resources Development Service of Korea detected six cases of cheating involving smart glasses this year alone in national qualification examinations it administered. In the regular TOEIC tests, candidates wearing AI glasses were caught twice in May. Schools are no exception. A full survey of 5,679 middle and high schools nationwide by the Ministry of Education confirmed four cases of cheating in internal exams.
On the ground, there are complaints that there are inherent limits to proctoring itself. In a survey conducted in October by the Korean Teachers and Education Workers Union of 2,716 middle and high school teachers, 82.3% responded that it is difficult to identify AI glasses under current invigilation procedures.
As cheating using AI glasses increases, technologies for detecting and monitoring AI glasses are also emerging in quick succession. These include analyzing glasses worn by candidates to distinguish smart glasses, or detecting abnormal behavior to determine suspicion.
GREPP, an AI assessment solutions company, added a “glasses precision verification” function in August to its online test proctoring service “Monito.” This function photographs the candidate’s glasses in four areas before the test begins for verification. During the exam, it detects abnormal behavior such as a candidate’s face moving out of frame, leaving the seat, or another person appearing, assigns a risk level, and allows proctors to determine whether cheating has occurred based on this information. A vision-language model (VLM) is used for detection, recording the candidate’s behavior and situation for each short video segment in sentences and analyzing the context of actions before and after.
At universities, there have also been cases in which professors used self‑developed programs to detect AI‑enabled cheating. At Yonsei University, during an online midterm exam for the course “Natural Language Processing and ChatGPT,” which had around 600 students enrolled last year, the professor in charge analyzed the exam footage using a self‑developed program and identified suspicious students. Subsequently, 40 students voluntarily reported themselves, and some received disciplinary suspensions of two to three weeks this year.
The government is also moving to secure detection technologies. The Human Resources Development Service of Korea is preparing new detection technology in cooperation with the Korea Communications Agency. Lim Young‑mi, Director General for Policy at the Ministry of Employment and Labor, stated at a National Assembly audit in October, “We will prepare countermeasures together not only with test‑operating institutions but also with institutions that possess AI cheating detection technology.”
ⓒ dongA.com. All rights reserved. Reproduction, redistribution, or use for AI training prohibited.
Popular News