Education has remained remarkably unchanged for centuries. Students gather in classrooms at fixed times to listen to teachers' lectures, review the content through homework, confirm their understanding through tests, and move on to the next unit. This familiar scene is essentially a transplant of the factory model from the industrial era. The problem is that we all already know this method is not "the most effective education." Yet inertia is strong. From schools to private academies in the after-school education market, Korea's educational landscape is still dominated by one-way lectures.
As someone who teaches at university while also volunteering to lead reading programs for high school students, I find a common expression on the faces of students I encounter in both settings. They listen but do not ask, memorize but do not contemplate. The responsibility cannot be placed on the students. The very structure of the classroom we have built is designed to operate that way.
The 'End of Homework' Has Already Begun
The emergence of artificial intelligence (AI) is creating a decisive crack in this old structure. The first thing to be shaken is "homework." The moment students ask AI for answers, the learning effect that homework was meant to deliver drops sharply. In fact, this phenomenon is not entirely new. Since the popularization of the internet in the 1990s, "copy-and-paste cheating" through search engines has already been widespread. AI has not created a problem that did not exist; it has merely accelerated a trend that had been underway for a long time.
However, the landscape after AI is quite different. Cheating is no longer a "special occurrence." Princeton University, a prestigious U.S. institution, abolished its 133-year tradition of unsupervised exams and revived the proctor system this year. At Korea's top universities, "AI cheating" has also become an open topic of discussion. The bigger problem is that the technical means to distinguish between AI-generated and human-written texts are virtually disappearing. Detectors are always one step behind, and models become more sophisticated each month.
Schools and teachers now stand before an unavoidable question: "To what extent will AI use be permitted?" The longer schools delay this question, the greater the confusion they will face.
The Calculator of the 1970s, the AI of the 2020s
Interestingly, we have already experienced a similar scene once before. In the early 1970s, when handheld calculators first appeared in classrooms, the educational community's initial response was close to rejection. Concerns about the deterioration of basic arithmetic skills were loud. But within just a few years, the atmosphere reversed. In 1980, the National Council of Teachers of Mathematics (NCTM) officially recommended that "calculators should be actively used at all grade levels." A debate that began in the early 1970s saw its official position reversed in less than a decade.
Just as the appearance of calculators did not eliminate the need to study mathematics, the emergence of AI does not eliminate the need for critical thinking and writing training. On the contrary. The more AI easily provides average answers, the more precious becomes the ability to question, verify, and reconstruct those answers in one's own language.
Bloom's '2 Sigma' Conundrum and AI as the Answer
It is worth revisiting a paper published 40 years ago. In 1984, educational psychologist Benjamin Bloom published "The 2 Sigma Problem." He empirically demonstrated that students who received one-on-one tutoring showed incomparably higher achievement than those in regular classrooms. It was a shocking result at the time: even ordinary students, when given personal tutoring, jumped to the top 2 percent level.
Bloom's discovery left a deep homework assignment for the field of education: "How can we achieve the effect of one-on-one tutoring even in group education?" Numerous studies were attempted, but the answer was always the same. A competent, universally applicable, and inexpensive personal tutor—producing such a being in reality was virtually impossible. Thus, the "2 Sigma Problem" remained an unsolved conundrum for nearly 40 years.
But now, AI has arrived in exactly that place. An AI tutor has emerged that remembers each student's learning history, diagnoses their weaknesses, and patiently repeats the same explanation without losing patience. "Khanmigo," the educational AI developed by Khan Academy, is the most advanced example showing that potential.
Yet Schools Will Not Disappear
No matter how much AI performance improves, it will be difficult to replace schools themselves. As the long history of schools shows, the classroom is a space that goes beyond "knowledge transfer." It is where students learn the art of cooperation by solving problems with friends, develop social skills through conflicts and reconciliations, and build self-efficacy through their teachers' attention and encouragement. Even with an excellent AI tutor by their side, the classroom adds unique value that can never be replaced.
There is, however, a clear change. AI helps teachers prepare lectures more systematically and transforms one-way lectures into classes where students actively participate. It reduces the time teachers spend on simple knowledge transfer, allowing them to spend more time on meaningful interaction with students—on discussion, feedback, and emotional exchange. What is in jeopardy, then, is not "schools" but the uniform lecture-style class itself.
Questions Students Will Ask, Questions Schools Must Answer
Before long, students will have AI by their side in almost everything they do. And they will ask their teachers new questions.
"Why must I do this assignment myself?"
"In an era when AI provides answers, what does learning mean?"
"How can using AI help my life?"
If schools cannot answer these questions, students will find answers outside the classroom. Fortunately, the clue to the answer is not far away. The essence of teaching does not lie in giving correct answers but in teaching how to ask questions. AI provides correct answers quickly, but it cannot ask questions. The ability to ask good questions—that is the core competency the classroom must cultivate going forward.
A Shift in Perception Comes First
In the end, what we must examine first may not be technology but our own perceptions. The long-held belief that lecture-style classes are efficient and inevitable is now due for reflection, and AI seems closer to a tool to be tamed step by step than a threat to be blocked. And to use that tool wisely, paradoxically, we must further cultivate uniquely human thinking and expressive abilities. This is the truth we must face again.
We now stand at the threshold of an era in which the methods of education are being fundamentally restructured. AI is simultaneously raising the capabilities of teachers and students, reconstructing the learning experience, and offering, for the first time, a realistic answer to the 2 Sigma Problem that no one had been able to solve. It is not something to fear. It is something to prepare for. Changing the landscape of the classroom is ultimately for the students, and what is for the students is the future of our society.
The classroom that once accepted the calculator now stands again before AI.
—The 2 Sigma Problem
A concept in education proposed in 1984 by U.S. educational psychologist Benjamin Bloom, referring to the phenomenon in which the average score of students who received one-on-one tutoring was two standard deviations (2σ) higher than the average of students in a regular classroom. Simply put, an ordinary student at the median level (50th out of 100) in a group class jumped to the top 2 percent (2nd out of 100) after receiving personal tutoring. Bloom did not celebrate this result but rather fell into deep contemplation. If personal tutoring is so effective, why can it not be provided to everyone? The answer was the limits of cost and manpower. "The dilemma of knowing the best educational method but being unable to give it to everyone"—that is why he called it a "Problem."
—Khanmigo
An educational AI personal tutor unveiled in 2023 by Khan Academy, the world's largest free online education platform, in partnership with OpenAI. The name combines "Khan" and the Spanish
