When AI Becomes a Resource: AI-Supported Teacher Design Capacity in Mathematics Teacher Education
This qualitative study proposes the framework of AI-supported Teacher Design Capacity (AI-TDC), characterized by five knowledge-in-action dimensions, to explain how pre-service mathematics teachers critically transform and integrate AI-generated resources into effective teaching documents through cycles of individual design, collective scrutiny, and enactment.
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 the classrooms of the future, a new kind of helper has arrived. It is an artificial intelligence that can write a full lesson plan in seconds, offering suggestions for how to explain a concept or what problems to give students. This technology promises to save teachers time, but it also raises a difficult question: if a machine can draft the work, how does a human teacher stay in charge of the most important parts? The answer lies not in how much the teacher uses the machine, but in how they think with it. To understand this, researchers look at two ideas. First, they view teaching materials not as static objects, but as living things that teachers shape and reshape to fit their specific goals and students. Second, they focus on "design capacity," which is simply the ability a teacher has to take available resources and turn them into a coherent learning experience. The central challenge is figuring out how to keep that human ability sharp when a powerful tool is doing so much of the heavy lifting.
A team of researchers in China set out to watch this process in real time. They followed twenty pre-service teachers—students training to become math teachers—through an eleven-week course. These future educators were asked to design lessons, teach them in short practice sessions, and then revise them. Crucially, they were allowed to use generative artificial intelligence, the kind of technology that can chat and create text, to help them. The researchers did not tell them exactly how to use the tool. Instead, they collected the teachers' chat logs, their lesson drafts, and their final videos to see how the AI actually fit into the teachers' thinking. The goal was to see if the AI became a master that dictated the lesson, a passive tool that just answered questions, or something more complex.
The study found that the relationship between the teacher and the AI was far more varied than a simple measure of usage would suggest. The researchers identified three distinct ways the teachers integrated the technology, and these ways had little to do with how many times they typed a question into the chat box. For one teacher, the AI became a central hub. She used it to connect different pieces of information—textbooks, online examples, and advice from classmates—into a single, organized flow. As she grew more confident, she actually used the AI less, asking fewer questions, but the tool became more important structurally because it helped her weave together a complex lesson plan around a specific mathematical idea. For another teacher, the AI was a specialized assistant. She used it only for specific tasks, like generating a list of practice problems or checking a derivation, while keeping the overall structure of the lesson firmly in her own hands. She used the tool more and more as the course went on, but it never took over the main design. A third teacher, who had more prior teaching experience, treated the AI as a reviewer. She would draft a lesson herself using her own experience and textbooks, then ask the AI to check it. She accepted only the suggestions that made sense to her and rejected the rest, keeping the AI on the periphery of her work.
What emerged from these different paths was a clear picture of what it means to be a teacher in the age of artificial intelligence. The researchers proposed a new way to look at this skill, which they call "AI-supported teacher design capacity." This is not just about knowing how to prompt a computer. It involves five specific actions. First, the teacher must orient the AI, deciding what the goal of the lesson is before asking for help. Second, they must verify the output, checking carefully to see if the math is correct and if the ideas fit the students' needs. Third, they must transform the suggestions, taking a raw idea from the AI and reshaping it into a lesson that makes sense. Fourth, they must regulate the tool, deciding when to let the AI lead and when to take control. Finally, they must collectivize, sharing the AI-generated ideas with peers to get feedback and improve the work together. The study showed that the most successful teachers were not the ones who used the AI the most, but those who could move fluidly between these five actions.
The most striking finding was that the AI-generated resources were never just copied and pasted into a final lesson. In every case, the teachers had to do significant work to turn the machine's output into a teachable experience. One teacher took a draft that the AI had organized and completely restructured it around a specific analogy between two types of number sequences. Another teacher took a list of mathematical steps generated by the AI and turned them into a sequence of questions that would guide students to discover the answer themselves. A third teacher looked at a generic example from the AI and decided it was too long for her class, so she replaced it with a shorter, more focused problem. In each instance, the human teacher acted as the architect, using the AI's bricks to build a house that the machine could not have designed on its own.
The course itself played a vital role in helping these teachers develop this skill. By forcing them to design a lesson, teach it, and then revise it based on feedback, the course created a cycle where the AI's suggestions could be tested. When a teacher used an AI suggestion in a practice teaching session, they could see immediately if it worked or if it confused the students. This real-world feedback, combined with discussions with their peers, helped them learn how to filter the AI's output. The study suggests that simply giving teachers access to artificial intelligence is not enough. Teacher education programs need to create environments where future teachers can practice these five actions—orienting, verifying, transforming, regulating, and collectivizing—so they learn to keep the responsibility for the lesson's direction firmly in their own hands.
Ultimately, the research suggests that the presence of artificial intelligence does not diminish the need for human expertise; it changes the nature of that expertise. The value of a teacher is no longer just in their ability to create a lesson plan from scratch, but in their ability to judge, reshape, and integrate the resources that technology provides. The study concludes that the future of teaching lies in a partnership where the machine offers speed and variety, but the human teacher provides the judgment, the context, and the moral responsibility for what the students learn. The technology is a resource, but the teacher remains the designer.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.