Teaching About AI or With AI? In-Service Teachers’ Experiences in a Korean AI Convergence Education Program

In classrooms across South Korea, a quiet transformation is underway. For decades, teachers have been the experts in their fields, guiding students through history, science, or literature. But recently, a new subject has arrived: artificial intelligence. The government decided that every student needs to understand how these smart machines work, and the logical next step was to train the teachers. This created a unique situation where experienced educators, who already know how to run a classroom, had to go back to school to learn a completely new technical language. The central question facing educators and policymakers is not just whether teachers can learn the facts about artificial intelligence, but whether they can take those facts and weave them into the lessons they already teach. Can a history teacher use artificial intelligence to help students understand the past, or does the new knowledge stay locked in a separate box, useful only for computer science classes?

This story comes from a specific corner of South Korea, where the government launched a massive program to retrain teachers. Since 2020, thirty-eight universities have offered special graduate degrees for working teachers, focusing on what is called "AI convergence education." The idea is to blend artificial intelligence with traditional subjects. To understand how this works in the real world, two researchers from Seoul and Hannam universities followed a small group of teachers and one instructor through this program. They did not just look at test scores or checklists; they sat down with the participants to hear their stories about what it felt like to learn this new material and try to use it back in their own schools.

The teachers who joined the program did not sign up because they were suddenly fascinated by robots or code. They signed up largely in response to practical pressures, such as the need to engage with AI in their schools, and because the government offered tuition subsidies that made the decision possible. One teacher, who works at a high school for science students, found her students building projects in computer languages she could not read. She felt left behind and needed to catch up. Another teacher, an elementary school instructor, admitted her motivation was "impure." She needed the tuition money the government offered, and she was also preparing for a promotion exam. Yet, beneath the practical reasons, there was a shared sense of urgency. They knew artificial intelligence was arriving in their schools, and they knew they would be expected to lead the way. They entered the graduate program hoping to find a clear path forward.

Inside the university classrooms, the teachers learned a great deal about the technology itself. They studied how machines learn, how to analyze data, and how to write code in a language called Python. The instructors, who were experts in computer science and data, did a good job teaching these technical skills. One teacher said she could finally understand what her students were doing when they built complex projects. However, a gap began to form between what they learned in the university and what they needed to do in their own classrooms. The teachers described the curriculum as "fragmented." They learned about data analysis in one class and programming in another, but the program never showed them how to connect these pieces to a history lesson or an ethics discussion. They were learning about artificial intelligence, but they were not learning how to teach with it.

The teachers found themselves stuck in a difficult spot. They understood the technology, but they could not figure out how to make it fit into their specific subjects. One teacher described the feeling as being "at a loss." She wanted to graft artificial intelligence onto her lessons, but she did not know how to make it happen without losing the focus of her subject. Another teacher tried to use a drawing app to help students illustrate a story, but she realized this was just using a tool, not truly blending the technology with the lesson. The program had given them the tools, but it had not given them the blueprint for how to build something new with them. The responsibility to figure out the connection was left entirely to the individual teacher, who had to solve this puzzle alone while managing a full-time job.

The researchers found that the difficulty was not just about a lack of skill, but about the nature of the subjects themselves. For a science teacher, data is everywhere, so finding a place for artificial intelligence is relatively easy. But for a teacher of language or literature, the connection is much harder to find. One instructor in the program, who teaches social studies, admitted that he could help a math teacher or a science teacher, but he felt lost when a language teacher asked for help. He could show them how to analyze text, but he could not find a direction for how to use the deeper algorithms of artificial intelligence in a way that made sense for literature. This suggests that the challenge is not that teachers are failing to learn, but that the link between artificial intelligence and certain subjects is not obvious, and the program did not provide enough support to build that bridge.

Another major hurdle was the structure of the schools themselves. Even when a teacher managed to create a great lesson plan that mixed artificial intelligence with their subject, the school system often made it impossible to carry out. In elementary schools, teachers are moved to different grades and subjects every year by the principal. A teacher who spent two years learning about artificial intelligence might be assigned to a different grade the next year, with no time to implement what she learned. In high schools, the credit system makes it difficult for teachers from different subjects to work together on a single project. The policy assumed that if you train a few specialists, they would spread the knowledge to everyone else. But the reality of daily school life meant that this knowledge often stayed with the individual, unable to take root in the wider school community.

The study also highlighted a confusion about what the goal actually was and a lack of practical support. The program was named "AI convergence education," which implies a deep blending of technology and subject matter. However, the teachers found that the program mostly taught them how to use artificial intelligence tools or understand the basic principles. They were rarely shown how to use the technology to solve problems specific to their own subjects. One teacher noted that the instructors seemed to disagree on the goal: some said you must understand the technology first, while others said the focus should be on the subject. This lack of a clear, unified vision left the teachers feeling like they were trying to assemble a puzzle with missing pieces. Furthermore, participants reported that practical guidance on AI ethics was insufficient. Instead of detailed support, policy materials were often merely "thrown" at teachers with instructions to teach the content, find suitable sites, and use the provided funds, turning the initiative into new work without the necessary guidance to make it meaningful. They were learning about the technology, but they were not learning how to make it meaningful for their students.

The researchers concluded that the missing piece in this entire process is what they call "pedagogical translation." This is the hard work of taking technical knowledge and turning it into a lesson that fits a specific subject. It is not just about knowing how a machine works; it is about knowing how to use that machine to help a student understand a poem, a historical event, or a scientific concept. The current programs are good at teaching the first part, but they are not good at teaching the second. The teachers are not failing; the system is simply asking them to do a job that no one has shown them how to do.

This study suggests that simply adding more technical training to teachers is not enough. To make artificial intelligence education work, the training needs to change. It needs to focus less on the technology in isolation and more on how that technology can serve the specific goals of history, language, or science classes. It also needs to recognize that teachers cannot do this work alone. They need support from their schools and from the programs that train them, with clear examples of how to blend these worlds. Until the gap between learning about artificial intelligence and learning how to teach with it is closed, the promise of this new education will remain just out of reach for many classrooms. The teachers are ready to learn, but they need a map that shows them how to cross the distance between the university lecture hall and their own school desks.