Interview with Lim Hyun-sung, President of the Korean Optical Society

With less than a month until the College Scholastic Ability Test (CSAT), concerns are rising over exam management. The emergence of 'AI glasses,' which can relay test questions and answers through lens subtitles or bone conduction speakers, has made these devices popular as fashion items priced around 400,000 won. These new devices, which conceal cameras and circuits within a lightweight frame of about 49 grams, are difficult to distinguish from regular prescription glasses. However, exam supervisors are still relying on simple visual inspections and metal detectors, raising concerns about their effectiveness.

Lim Hyun-sung, President of the Korean Optical Society and a professor at Eulji University, emphasized the urgent need to establish a standard exam governance system that combines optical expertise with human rights protections. As the CSAT approaches, Aju Economy spoke with Professor Lim to discuss the technological level and popularity of AI glasses, as well as the challenges and practical solutions for exam management.

Transition from 'seeing glasses' to 'informing glasses'... collapsing defenses in exam rooms

Recently, smart glasses have shifted from being high-priced, specialized equipment for early adopters to everyday fashion items in the global wearable market. Major tech companies like Meta and Rokid have introduced affordable devices priced around 400,000 won, significantly increasing accessibility for students needing vision correction.

Professor Lim described this technological leap as a fundamental paradigm shift in visual media. While traditional glasses focused on correcting optical refraction, comfort, and aesthetics, the latest AI glasses integrate a multimodal AI platform with miniature camera modules, microphones, bone conduction audio speakers, and waveguide micro-displays that project information directly into the wearer’s field of vision.

According to Professor Lim, when a wearer turns their head to look at a challenging question on the test paper, the camera on the front of the glasses captures the text in real-time. On-device AI or connected networks then activate optical character recognition (OCR) and large language models (LLM), providing explanations and answers as subtitles on the lenses or through speakers at the temples within seconds. This technology allows students to write answers without moving a finger.

The bigger issue is that the control circuits and miniature batteries are precisely embedded within the 49-gram frame, making them nearly indistinguishable from regular glasses. It is nearly impossible for a non-expert supervisor to detect these devices during a brief visual inspection at the exam entrance. Metal detectors also fail to differentiate between the lightweight titanium hinges or nose pads and the signals from the tiny circuit boards. Professor Lim pointed out that requiring teachers to identify these devices without systematic training or equipment support is a form of administrative negligence.

Balancing fairness and human rights... three dilemmas of 'portrait rights, AI misjudgment, and psychological pressure'

As concerns about cheating rise, some have suggested extreme measures, such as forcibly removing glasses from suspicious students for inspection or installing cameras in exam rooms to record all seats continuously. However, Professor Lim warned that such actions could violate constitutional principles of due process and excessive prohibition, highlighting three dilemmas faced by the exam management system.

The first dilemma involves the potential violation of basic rights and treating students as potential cheaters without evidence. Forcing students to surrender their glasses or undergo physical inspections based solely on subjective suspicions about the thickness or unusual appearance of their frames violates the principle of minimal intrusion in administrative procedures. Additionally, continuously tracking students' identities through facial recognition or recording the entire exam room assumes all students are potential cheaters. Professor Lim stressed that the exam supervision system must adhere strictly to the principles of 'purpose limitation' and 'minimal intrusion,' ensuring that any AI monitoring and recording procedures are clearly communicated in advance to guarantee predictability.

The second dilemma concerns the risk of innocent victims due to misjudgments by AI-assisted systems. Natural personal habits, such as focusing on difficult questions, fidgeting with glasses, or mouthing words, could be misinterpreted as cheating signals. Professor Lim cautioned that if the system detects these as suspicious behaviors and immediately expels the student, it could lead to irreversible violations of rights. He emphasized that AI should serve only as a sensor for detecting anomalies, while human supervisors and exam authorities must confirm and make judgments based on the 'human in the loop' principle.

The third dilemma arises from the emotional pressure students may feel when supervisors wear AI monitoring glasses, as they may feel constantly surveilled. To alleviate this, Professor Lim suggested that supervisors should transparently inform students that their devices are not CCTV cameras recording everything but rather minimal objective safety devices that only record when anomalies occur. He also recommended establishing guidelines to prevent unnecessary prolonged scrutiny by supervisors and ensuring that collected logs are destroyed immediately after the exam unless there are disqualifying reasons.

Practical guidelines to reduce confusion on-site... separating verification procedures post-exam

To address uncertainties on-site, Professor Lim proposed a six-step standard operating procedure to protect both supervisors and students: pre-exam device checks, familiarization with identification features, setting observation targets, recording and reporting suspicious signals, separating verification procedures during breaks and after the exam, and preserving and storing original data.

He suggested that supervisor training should shift from text-based guidelines to hands-on workshops where they can experience actual devices. Supervisors should familiarize themselves with the locations of pinhole camera holes, physical buttons, battery charging contacts, and the unique rainbow reflections of waveguide displays. Additionally, a '10-second quick identification infographic card' should be available at exam entrances to assist with initial screenings.

To ensure students' rights during exams, Professor Lim emphasized the importance of separating verification procedures post-exam. He advocated for strict prohibitions against disrupting students' concentration by requiring them to stand up or inspect their glasses during the exam. Instead, device checks should occur during breaks or after the exam, accompanied by two safety measures to prevent evidence destruction. Any detected anomalies should be timestamped and remotely frozen on a central security server, and glasses should be securely sealed in a special envelope immediately after the exam. If attempts to delete apps or reset devices are detected following a verification request, these should be treated as deliberate evidence destruction, warranting the highest penalties, including exam invalidation.

Responsibilities of the Ministry of Education and related agencies... establishing standard regulations and inter-agency governance

Professor Lim argued for the urgent need for the Ministry of Education to establish 'standard exam management regulations' that go beyond sporadic responses from individual schools or testing organizations. Key components of these regulations should include prohibiting the entry of AI smart glasses and requiring pre-registration and verification for low-vision students, banning forced confiscation during exams, separating verification procedures during breaks and after the exam, establishing a timestamp-based remote data freezing and evidence preservation system, and creating an independent 'appeals review committee' to guarantee students' rights to access their data.

He also suggested investing in security infrastructure for large-scale national exams like the CSAT or civil service exams, which involve hundreds of thousands of participants. Given the difficulty of disassembling devices on-site, a 'wireless frequency security communication network' should be established to control and block data packets transmitted to unauthorized Bluetooth, Wi-Fi, or wireless tethering from external generative AI servers. Most smart glasses, which operate in real-time with external networks, would become ineffective if their communication packets are blocked.

Furthermore, Professor Lim called for a collaborative framework that goes beyond the Ministry of Education's guidelines. He proposed developing a 'standard guideline for the safe use of AI glasses in educational settings' that integrates the privacy protection and minimal collection principles from the Personal Information Protection Commission, hardware security and network control standards from the Ministry of Science and ICT, and digital user rights and appeals standards from the Korea Communications Commission.

The perspective of optical science... transforming technology from a tool for cheating to a life-changing function

Professor Lim cautioned against framing smart glasses solely as tools for cheating. From an optical perspective, AI glasses serve as intelligent visual assistance platforms that read real-time text from signs, menus, and medication labels for visually impaired and low-vision individuals. They are also essential tools for remote medical care, allowing healthcare professionals to share first-person perspective videos with specialists at higher-level hospitals. Additionally, they are innovative tools in industrial settings, reducing errors and safety incidents by displaying complex diagrams.

According to Professor Lim, the optical science community is conducting unique research to control the side effects of this technology and enhance human compatibility. Research includes analyzing the 'accommodation-convergence conflict' and visual fatigue that occurs when viewing a virtual screen four meters away while simultaneously looking at a test paper 40 centimeters away, as well as measuring and correcting discrepancies between the pupil's line of sight and the camera module's axis. They are also measuring induced prisms that occur when the optical center and pupil center are misaligned by just 1mm, and analyzing '3D dynamic fitting' to evenly distribute facial load due to weight increases over 49 grams.

The Korean Optical Society's AI Smart Glass Task Force is working on creating a 'standard operating manual' for distribution to universities and testing sites, as well as a 'five-step clinical protocol' for optical clinics. They plan to expand simulation workshops for mock exam environments and submit policy proposals containing legislative amendments to the Ministry of Education and related agencies. Additionally, they aim to revise the curriculum for optical science programs at universities to establish specialized courses covering waveguide optics, 3D facial measurement, multimodal AI, and data ethics.

Professor Lim concluded, "We cannot confront a new paradigm by blindly blocking technological advancements. Only when we establish rigorous supervision procedures based on optical expertise and human rights-centered guidelines can we fully enjoy the benefits of advanced technology while preserving the integrity of exams and education."