Extending Instrumental Orchestration for AI-Mediated Mathematics Classrooms: The AISIO Framework
This paper proposes the Artificial Intelligence-Supported Instrumental Orchestration (AISIO) framework as an extension of instrumental orchestration to address how teachers in AI-mediated mathematics classrooms must regulate AI-generated pedagogical participation while sustaining collective mathematical activity and epistemic authority.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In classrooms around the world, mathematics has long been a conversation between a teacher, a group of students, and the ideas they are trying to understand. For decades, researchers have studied how teachers organize this conversation, especially when digital tools are involved. They use a concept called "instrumental orchestration" to describe how a teacher arranges the room, selects the right software, and guides students so that their individual work with a computer turns into a shared, meaningful learning experience. The teacher acts as a conductor, ensuring that the technology serves the math, not the other way around. But a new force is entering the room: artificial intelligence. Unlike a standard calculator or a static textbook, modern AI can watch a student work, generate a unique explanation, offer a hint, or even grade a response in real time. It does not just wait to be used; it actively participates in the lesson. This shift raises a difficult question for educators: when the computer starts making pedagogical decisions on its own, how does the teacher keep the class moving together as a single group?
A team of researchers from Korea and France has set out to answer this question by observing mathematics classrooms where artificial intelligence is already in daily use. They did not simply ask teachers what they thought about the technology; they watched the lessons unfold, recording how teachers reacted when the AI stepped in. Their work, which draws on observations from ten lessons taught by two experienced teachers in Seoul, reveals that the old ways of thinking about classroom management are no longer enough. The researchers found that teachers are now facing three specific, new challenges that require a fresh approach. They are not replacing the old framework but expanding it to account for a classroom where the technology is a partner in teaching, not just a tool.
The first challenge the researchers identified is the problem of reassembling a scattered class. In a traditional lesson, a teacher might notice that several students are stuck on the same problem and decide to stop the class to explain it. With AI, every student moves at their own speed, and the system generates a unique score and feedback for each one. The teacher is presented with a dashboard full of individual results, but the AI does not tell the teacher which of those individual struggles represent a problem for the whole group. In one observed lesson, a teacher noticed that while the AI had given high scores to a few students, many others were silently struggling with a misunderstanding of the core question. The teacher had to ignore the individual scores and look for a pattern across the whole class to decide when to stop the individual work and bring everyone together. The AI provided the data, but the teacher had to interpret what that data meant for the collective group, deciding when the individual paths needed to be merged back into a shared discussion.
The second challenge involves sharing the role of the teacher. In these classrooms, the AI often speaks first. It might look at a student's answer and suggest a next step or point out a missing detail before the teacher has even seen the work. The researchers observed a moment where a student gave a plausible but incomplete answer to a statistics problem. The AI immediately prompted the student to look at the raw numbers more closely. The teacher then had to decide how to handle this suggestion. The AI had started the conversation, but the teacher had to finish it, guiding the student to understand not just what to look at, but why it mattered as mathematical evidence. The teacher had to take the AI's initiative and weave it into the lesson, ensuring that the student's reasoning remained deep and rigorous rather than just following a robotic prompt. The teacher was no longer the sole source of new ideas; they were now a coordinator, managing a flow of suggestions coming from both themselves and the machine.
The third and perhaps most subtle challenge is about trust and truth. Sometimes the AI gives a verdict on a student's work that is mathematically correct but explained in a way that is confusing or unhelpful. In one instance, a student provided a counterexample to a mathematical statement, and the AI marked it as wrong. The teacher had to pause and figure out why the system made that judgment. It turned out the AI was technically right about the math, but its explanation was so abstract that it would have confused the class. The teacher had to reconstruct the reasoning behind the AI's verdict, translating the machine's cold judgment into a clear, human explanation that the students could actually learn from. The teacher had to act as a filter, validating the machine's answer while rewriting its message so that it served the learning goals of the class.
Based on these observations, the researchers propose a new framework they call Artificial Intelligence-Supported Instrumental Orchestration. This is not a suggestion to replace the teacher with a machine, nor is it a call to ban the technology. Instead, it is a guide for how teachers can regulate the participation of AI. The framework suggests that teachers must now manage a complex ecosystem where the teacher, the students, the mathematical knowledge, the resources, and the AI all interact. The teacher's job is to ensure that the AI's dynamic, personalized interventions do not break the class apart. They must decide when to let the AI speak, when to step in and clarify, and how to turn individual, AI-generated feedback into a collective learning moment.
The researchers are careful to note that this is an exploratory proposal, born from specific classrooms in Korea where AI integration is advanced. They do not claim to have solved every problem, but they have identified the precise pressures that teachers are feeling. The study suggests that the future of mathematics education depends on teachers developing a new kind of professional judgment. They must learn to listen to the AI, interpret its hidden logic, and guide its contributions so that they strengthen, rather than weaken, the students' ability to think for themselves. The goal remains the same as it has always been: to sustain a meaningful, collective activity where students are truly doing the math, even when a powerful new partner is sitting at the desk with them.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.