Why Policing AI Glasses in Exams Won’t Fix Cheating: Lessons From South Korea
The video was already playing when I walked into homeroom. Four students were leaned over a desk, backpack straps still tangled around their arms, watching a news report from South Korea. Someone had turned the volume low, as if that made the clip less strange.
“Those are just glasses,” one girl said.
“Look,” said the boy holding the phone. “There’s a camera in the frame. He takes a picture of the test, somebody outside reads it, and the answer comes back through his ear.”
The room went quiet for a second. Then another student asked, “Couldn’t we do that?”
He did not say it like a threat. He said it like the obvious technical question it was.
The South Korean cases involved AI-powered glasses during the college entrance exam, a test so high-stakes that entire days are built around security. Phones are collected. Watches are checked. Proctors walk the rows. Yet ordinary-looking glasses could apparently photograph questions, send them to accomplices, and return answers through tiny earpieces. Dozens of people were investigated. Schools and parents were left asking the same question: how do you ban a device nobody recognizes?
I stood there with my attendance list and lukewarm coffee and thought, this is the wrong question.
Don’t get me wrong. Exam security matters. Students who follow the rules deserve a fair room. But a list of banned devices will always be one product cycle behind. One year it’s smartphones, then smartwatches, then earbuds, then glasses. Soon enough, someone will ask about rings, badges, or ordinary-looking pens. Schools can spend more time checking pockets and sleeves than actually watching how children work, and the most determined cheaters will still find the next gap.
The South Korean story is frightening because the stakes are so extreme. A single exam can shape university admission, career choices, and family expectations. When one score carries that much weight, the temptation to protect it does not stay small. The glasses were not really the problem. They were just the latest tool attached to a much older pressure: the belief that a result matters more than the learning behind it.
I saw a smaller version of this a few weeks earlier. A student in my class wore her smartwatch into a history test. It slipped halfway down her wrist under her sleeve. Her mother had been texting about a dentist appointment. We followed the school rule: the watch went to the office, and she finished the test in another room. She cried afterward, convinced she would get a zero and be labeled a cheater.
I have no way of knowing whether she intended to use the watch. What I know is that the whole conversation immediately became about punishment, not about what she understood. Nobody asked her to explain a feudalism timeline or talk through the essay question. The device told us nothing about her learning.
That same week, a parent sat across from me at conferences and asked what app she should install to stop her son from using AI on math homework. She looked exhausted. She had already checked his browser history, limited screen time, and moved the router to her bedroom.
I asked her what happened when he did math without help.
“He stares at the page,” she said. “Then he gets angry and says he’s stupid.”
The AI was not the cause. It was exposing the cause: he was behind on fractions, and every homework assignment felt like a reminder. Blocking every tool would have made the evenings quieter only in the way that locking a door makes a room quieter. The problem would still be inside.
We made a simpler plan. He would do twenty minutes of homework at the kitchen table, phone on the counter and laptop screen visible. He had to try two problems first and circle the exact step where he got stuck. If he used AI, it could only be after those attempts, and only to ask for one worked example similar to the circled problem. I cut his assignment from twelve problems to six and checked in with him for five minutes each Monday. The goal was not perfect surveillance. The goal was to see where his thinking stopped.
That distinction matters in classrooms too. If my only evidence of learning is a multiple-choice test taken in silence once a month, I am basically inviting the newest device to become a threat. If I also see drafts, wrong answers, in-class writing, and short explanations, the final score becomes harder to fake and less important by itself.
I now use shorter quizzes more often. Some are open-note, with one handwritten page allowed. Once per unit, each student records a thirty-second voice note explaining one answer: “I chose C because…” or “The part I still don’t understand is…” I collect essay notes, thesis statements, and messy drafts before the final version. AI can produce a polished paragraph, but it cannot invent a paper trail of confused edits, highlighted sources, and a student’s handwriting.
None of this eliminates cheating. It reduces the reward and makes the evidence more human.
At home, parents can do something similar without turning the kitchen into a screening station. A predictable work setup helps more than sporadic spying: laptop in a common area, phone charging across the room, paper notes within reach. Try saying, “Before you ask AI, try one and circle where you stop.” Then look at the circled spot together. If AI is allowed, ask your child to paste the prompt and the response into the document, not because every word must be policed, but because they should be able to tell you what they asked and what they did with the answer.
Schools also need rules written by function, not brand name. “No device that can record, photograph, send, receive, or display information during a test” is clearer than a list naming particular watches or glasses. But students should help update those rules. They know what devices are appearing in backpacks. Asking, “What loopholes should we fix before the next test?” gives them a stake in fairness instead of making them feel like suspects. Ordinary glasses, medical devices, and approved accommodations should never become reasons for suspicion.
The hardest part is accepting that no system will be perfect. Some families want total bans. Others want every test redesigned. Most teachers I know are somewhere in between, trying to protect honest students while still measuring real understanding.
Last week, a boy raised his hand during study hall and asked whether AI contact lenses were coming next. The class turned to look at me.
I said probably, somewhere.
He seemed disappointed that I did not promise a rule against them. Then he asked whether tomorrow’s quiz could be open-note. I told him one handwritten page, front side only. He nodded and went back to his review sheet.
That is where we are right now: smaller stakes, better evidence, and rules we revise together. The devices will keep changing. The tests will have to catch up in the other direction.
Please indicate: Thinking In Educating » Why Policing AI Glasses in Exams Won’t Fix Cheating: Lessons From South Korea