Ambitious Math Instruction in the Age of AI: Prioritizing Student Thinking and Productive Struggle

Summary

In this article, you will learn how:

• AI elevates the need for ambitious math instruction where students focus on reasoning, collaboration, and critical thinking rather than just getting answers.
• Well-designed AI deepens learning, supporting collaboration and making student thinking more visible through AI enhanced formative assessment and learning tools.
• Teachers remain essential, using the S.M.A.R.T. framework for human-centered AI integration in mathematics education to guide and refine the use of AI-generated content and emerging AI tools.

How AI Is Transforming Math Instruction, Collaboration, and Student Thinking

Artificial intelligence is quickly becoming part of the fabric of today’s classrooms. It’s showing up in how lessons are planned, how students interact, and even how mathematical thinking is captured and assessed. For many educators, that shift brings both excitement and uncertainty.

After all, when AI can generate correct answers (at least most of the time), step-by-step solutions, and even polished explanations in seconds, it’s natural to wonder: What does meaningful and successful math learning look like now?

For those committed to ambitious math instruction, teaching that prioritizes student thinking, reasoning, and collaboration moving beyond simply finding answers, the answer isn’t about resisting AI. It’s about reimagining what matters most in your math classroom.

Keeping Mathematics Meaningful in an AI-Driven World

Ambitious math instruction (Lampert et al., 2013) has long been grounded in a simple but powerful idea: students should be active sense-makers, not passive answer-getters. That vision becomes even more important in the age of AI.

When answers are easy to access, the work of learning shifts. What matters is no longer just arriving at the correct solution, but understanding the reasoning behind it, being able to explain, question, evaluate, and refine ideas.

In classrooms shaped by this approach, students aren’t simply completing problems. They are collaboratively problem solving, comparing strategies, critiquing arguments (including those generated by AI), and working together to make sense of complex ideas. Teachers, in turn, are not stepping back, they are stepping more fully into the role of facilitator, helping students decide when an AI-generated response is useful, incomplete, or even misleading.

This is where the real opportunity lies. AI may make it easier to get answers, but it cannot replace the kind of reasoning, justification, and collaboration that define deep mathematical understanding.

Preserving Productive Struggle in an Age of AI

AI is frequently praised for making tasks faster and more efficient. Ambitious math instruction, by contrast, depends on slowing down and giving students time to explore, struggle productively, and engage in meaningful discussion.

AI can help streamline certain aspects of teaching, but it also challenges educators to design learning experiences that preserve complexity also known as cognitive demand and productive struggle in mathematics education (NCTM, 2014) or “cognitive friction” in the current AI literature (Vendrell & Johnston, 2026). The goal isn’t to move faster through content, but to create space for deeper thinking through reasoning, pressing for evaluative judgement, promoting metacognition, and sparking intellectual curiosity within an AI-supported environment. 

For example, students might first work on a rich and complex real-world, mathematical-modeling problem, such as analyzing and determining how much food waste is in their school cafeteria (Garfunkel & Montgomery, 2019). After developing their own assumptions, identifying key variables, and constructing a model, they could compare their approach with an AI-generated model and explanation. Students might examine differences in assumptions, question which variables were included or excluded, and evaluate the strengths and limitations of each model. Through discussion and debate, they could refine their reasoning and determine which model is more mathematically sound and useful for decision making. In this way, AI, rather than simply serving as a source of answers, becomes a way to deepen students' thinking as they make sense of the problem, reflect on and critique various approaches and persevere in developing a solution.

AI as a Thought Partner in Responsive Task Design

Teachers are increasingly experimenting with AI as a partner in Responsive Task Design, an instructional approach where teachers create or select tasks that adapt to students' current understanding, cultural backgrounds, and mathematical thinking. Teachers are using AI to generate ideas, suggest contexts, or help connect mathematics to students’ lived experiences.

In practice, this often looks less like handing over control and more like starting a conversation. Teachers might use AI to draft a task or propose a real-world scenario, then refine it by adjusting the mathematics, reworking the context, and aligning it more closely to their students’ needs.

In one study, educators described AI as a helpful “first draft” tool that supported idea generation, differentiation, and culturally relevant task design (Suh & Vora, 2025). At the same time, they noticed its limitations, for example, occasional inaccuracies, shallow cultural references, or tasks that didn’t quite match the intended level of mathematical rigor. Addressing these limitations required teachers to apply their Mathematical Knowledge for Teaching (MKT, Ball et al., 2008) and Technology Pedagogical Content Knowledge around AI (TPACK, Mishra, & Koehler, 2006), using professional judgment to evaluate mathematical accuracy of AI outputs, to use AI to anticipate student responses, and adapt tasks for meaningful learning.

What emerged was not a replacement for teacher expertise, but a clearer picture of its importance. AI could generate possibilities, but it was the teacher who ensured quality, coherence, and relevance (Digital Promise, 2024; Nucci et al., 2024).

Supporting Student Collaboration Skills

While much of the conversation around AI in education has focused on personalization and tutoring, another shift is beginning to take shape, one that centers on collaboration.

Instead of positioning AI as something that delivers answers, researchers are exploring how it can function as a collaborative partner. In these environments, students interact with “Collaborative AI Peers” designed to think alongside them.

Emerging research suggests that AI can be most powerful when it facilitates peer-to-peer dialogue, helping students articulate and defend ideas rather than replacing human interaction. In one example, middle school students worked with two AI “classmates,” each with a different role. One surfaced common misconceptions, while the other modeled precise reasoning. Together, they helped students engage in productive struggle, metacognition while critiquing the reasoning of others and refining their thinking (Suh et al., 2025; Yue et al., 2025). 

In another study (Lee et al., 2025), OKO, which is an AI-powered app named after the Polish word for "eye," was designed to facilitate small-group collaborative learning in elementary and middle school classrooms. Rather than tutoring individual students, OKO acts as a facilitating agent that encourages students to talk with one another, explain their thinking, build consensus, and engage in mathematical discourse.

When AI is designed to participate in the learning process, rather than shortcut it, it can support the kind of discourse and sense-making that ambitious math instruction values most.

Designing for Thinking, Not Just Support

One of the clearest lessons emerging from this work is that too much support can actually get in the way. Designing AI experiences that preserve productive struggle rather than remove it is essential.

When AI steps in too quickly or provides overly complete explanations, it can reduce the need for students to think, question, or persist. On the other hand, when it prompts students to reflect, nudges them to explain their reasoning, or encourages them to consider alternative approaches, it helps sustain cognitive demand (Lyu et al., 2026).

The difference comes down to design. AI tools for the classroom must be intentionally structured to support inquiry, leaving room for students to do the intellectual work that leads to understanding.

Making Student Thinking Visible Through AI

Perhaps one of the most promising developments is how AI is beginning to reshape assessment.

In many classrooms, especially during group work, much of students’ thinking remains hidden. Teachers do their best to notice, attend to, interpret and respond to student thinking, but it’s not always possible to capture everything that unfolds in real time (Jacobs et al., 2010).

Emerging multimodal systems can help capture patterns in student reasoning and collaboration that were previously difficult for teachers to observe in real time. Instead of working only with text, it can now interpret and respond to the many ways students communicate mathematical thinking, interpreting spoken language, written work, drawings, photos, manipulatives, and collaborative conversations. Because connecting multiple representations is a key process standard from the National Council of Teachers of Mathematics, these advances open exciting new possibilities for supporting mathematical discourse, formative assessment, and collaborative problem-solving processes. 

AI-powered teacher dashboards can analyze multiple representations of student thinking in real time, providing immediate insights into students’ reasoning, partial understanding, participation patterns, and problem-solving strategies. These dashboards can even offer purposeful questions to ask to promote deep classroom discussions. By making visible not only what students produce, but also how their understanding develops through questions, revisions, and collaborative interactions, AI creates new opportunities for more responsive and equitable formative assessment (Suh et al., 2025).

With these insights, instruction becomes more intentional, grounded in a richer understanding of student thinking.

Looking Ahead: Human-Centered AI Design for Math Teaching and Learning 

As AI becomes more integrated into instructional planning, a new kind of professional knowledge is taking shape: critical AI literacy. This literacy must be both critical and human-centered, enabling educators to thoughtfully evaluate, question, and shape AI use in ways that support meaningful teaching and learning.

Critical AI literacy isn’t just about knowing how to use AI tools but about being critical and judicious in its use and giving agency to teachers and students. For example, when integrating AI in designing math tasks, it is about how to recognize when something doesn’t quite add up, when a task has been oversimplified, or when an example is culturally biased or superficial. Teachers who develop this kind of literacy are better positioned to use AI thoughtfully. They can adapt what it produces, refine it, and ensure that it supports and not diminishes student thinking and agency (Suh & Vora, 2025).

In many ways, AI is making teacher expertise more visible. The difference between surface-level and meaningful instruction becomes clearer when educators are asked to interpret, evaluate, and improve what AI generates.

Across teaching, learning, and assessment, a consistent theme emerges. AI does not replace the work of teaching or the process of learning. Instead, it has the potential to expand both. When grounded in ambitious math instruction, with human-centered AI design in mind, AI can help create classrooms where reasoning is valued, discourse is central, and all students have opportunities to engage deeply with mathematics.

The question moving forward is not whether AI belongs in math classrooms. It’s how we choose to use it with smart human-centered design in mind. The S.M.A.R.T. framework (shown below) positions AI as a tool for amplifying, rather than replacing human thinking, teacher expertise, and meaningful mathematical learning. At its core, the framework emphasizes centering humans while supporting student agency, rigorous mathematics, ambitious instruction, reflective teaching, and transformative learning opportunities. Each core dimension is described with key reflective questions for educators to ask when integrating AI into their teaching and learning environments. 

The S.M.A.R.T. Framework for Human-Centered AI Integration in Mathematics Education

Educators are encouraged to consider these S.M.A.R.T. dimensions along with the reflective questions to ensure that we keep ambitious math at the core of instruction in the age of AI, prioritizing student thinking and productive struggle.

In addition, as schools adopt AI-enhanced tools, issues of data privacy, transparency, and ethical use remain essential considerations. Educators and technology developers must ensure that AI supports learning without compromising student trust, agency, or privacy.

If AI tools are adopted or integrated without a critical lens, it may undermine student agency, productive struggle, equitable participation, and teacher professional judgment. But if we use AI to amplify thinking, support collaboration, and make learning more visible, we can move toward classrooms that are not only more innovative, but more meaningful and ambitious. The future of AI in mathematics education depends not on the technology itself, but on how educators preserve the human work of teaching while leveraging AI's capabilities.

In the end, the goal remains the same: keep humans (educators and students) at the center of thinking and learning in the age of AI.