Beyond the Answer: Reclaiming Math in the Age of AI

by

Author: Denis Avetisyan

As artificial intelligence tools become increasingly capable of solving mathematical problems, educators must rethink how they foster genuine understanding and skill development.

This review argues for the use of open-ended mathematical tasks and explicit didactic regulation of AI to preserve students’ epistemic control and the formative value of mathematical activity in a post-AI educational landscape.

The increasing accessibility of generative artificial intelligence presents a paradox for mathematics education: routine tasks, traditionally formative for students, are now readily automated. This study, ‘Open Mathematical Tasks as a Didactic Response to Generative Artificial Intelligence in Post-AI Contexts’, investigates the implementation of open-ended mathematical tasks as a means of sustaining students’ mathematical agency in this evolving landscape. Findings reveal that, with explicit didactic regulation, students can maintain epistemic control over mathematical activity even when generative AI tools are present. How can carefully designed tasks and pedagogical approaches best leverage the potential of AI while preserving the core values of mathematical understanding and reasoning?

Unveiling the Shifting Landscape of Mathematical Education

The integration of generative artificial intelligence is fundamentally reshaping educational landscapes, prompting a critical need to reassess established teaching methods. Previously, educators often focused on delivering information and assessing recall; however, the capacity of AI to readily provide answers and complete tasks necessitates a shift towards cultivating higher-order thinking skills. This isn’t simply about adapting to new tools, but about redefining the very purpose of education in an era where information is abundant and readily accessible. Traditional pedagogical approaches, centered on memorization and procedural knowledge, are increasingly insufficient; instead, emphasis must be placed on critical analysis, problem-solving, creative application, and the development of metacognitive abilities – skills that empower students to navigate and leverage AI effectively, rather than being passively replaced by it. This evolution demands a proactive and thoughtful response from educators to ensure learning remains meaningful, engaging, and relevant in a rapidly changing world.

The accelerating advancement of artificial intelligence is fundamentally altering the landscape of mathematical education by automating traditionally laborious tasks. While calculations and procedural fluency – once central to mathematical training – are now increasingly handled by AI tools, this shift presents both unprecedented opportunities and significant pedagogical challenges. Rather than dismissing these tools, educators face the task of leveraging them to move beyond rote learning and cultivate deeper conceptual understanding. This requires a re-evaluation of curricula to prioritize problem-solving, critical thinking, and the ability to interpret and validate AI-generated results, ensuring students develop a robust understanding of why mathematical principles work, not merely how to compute them. The availability of instant calculation risks diminishing the development of number sense and estimation skills if not thoughtfully integrated into instruction, demanding a proactive approach to preserve core mathematical competencies alongside the embrace of new technologies.

As artificial intelligence increasingly handles procedural mathematical tasks, the emphasis in education must pivot towards cultivating students’ abilities to construct and justify mathematical knowledge – the epistemic dimension of learning. This isn’t merely about understanding how to solve a problem, but rather why a solution is valid, and how it connects to broader mathematical principles. Recent research demonstrates that a deliberate focus on this epistemic dimension is critical for preserving student agency; when students are empowered to reason, critique, and build their own understanding, they are less likely to passively accept AI-generated answers and more likely to develop robust, transferable mathematical reasoning skills. Without this shift, education risks producing individuals who can utilize AI tools but lack the foundational understanding to independently verify results, adapt to novel problems, or contribute meaningfully to the field.

Fostering Student Agency Through Open-Ended Mathematical Tasks

Open mathematical tasks, distinguished by a lack of immediately obvious solution paths, actively foster student agency by necessitating the formulation of underlying assumptions before problem-solving can commence. Unlike closed tasks with prescribed methods, these problems present initial indeterminacy, requiring students to define their own variables, constraints, and approaches. This process shifts the cognitive load from simply applying a known procedure to actively constructing a personalized problem space. Consequently, students are not merely executing instructions but are instead making deliberate choices about how to interpret the task and what mathematical tools are appropriate, thus promoting ownership and a sense of control over their learning process. The requirement for assumption formulation also encourages metacognitive awareness as students must explicitly identify and justify the basis of their solutions.

Effective open task design necessitates careful consideration of both mathematical modeling and the validation process. Tasks should allow for multiple entry points and solution approaches, enabling students to represent real-world scenarios using mathematical concepts [latex] (e.g., equations, graphs, geometric constructions) [/latex]. Crucially, the task structure must then facilitate a robust validation phase, where students critically examine the reasonableness of their solutions, compare different approaches, and justify their reasoning with evidence. This includes encouraging students to identify assumptions made during the modeling process and to consider the limitations of their models, promoting a cyclical process of refinement and increasing the trustworthiness of the final result. A well-designed task will explicitly or implicitly demand this iterative process, moving beyond simply obtaining a numerical answer to evaluating the validity and applicability of the mathematical representation.

Open mathematical tasks directly engage the epistemic dimension of learning by necessitating that students actively construct their own understanding of concepts and mathematical relationships. This approach moves beyond rote application of procedures and instead prioritizes the justification of solutions through reasoning and evidence. Our research indicates that this emphasis on construction and justification is particularly critical in the context of increasingly accessible AI tools; by requiring students to independently formulate approaches and defend their reasoning, these tasks maintain a formative sense of mathematical activity and prevent reliance on externally generated solutions, thereby fostering deeper conceptual understanding and promoting agency.

Structuring Learning with Didactic Regulation and the COMPAS Framework

Effective didactic regulation is crucial for structuring student mathematical activity, particularly when integrating artificial intelligence tools. This regulatory process involves teachers providing intentional guidance and support to ensure students maintain agency and focus on core mathematical concepts rather than simply accepting AI-generated outputs. Without careful regulation, students may prioritize obtaining solutions over understanding the underlying mathematical reasoning and principles. This is because AI tools, while capable of performing complex calculations and offering potential solutions, do not inherently foster conceptual understanding; instead, they require educators to actively mediate the learning process and ensure students critically evaluate and interpret AI-provided information to build robust mathematical knowledge.

The COMPAS framework is designed to structure learning activities when generative AI tools are utilized, thereby maintaining focus on established pedagogical objectives. It provides a systematic approach to integrating AI, not as a replacement for mathematical thinking, but as a support for it. Specifically, COMPAS facilitates the creation of learning experiences that deliberately address both the ‘how’ and ‘why’ of mathematical problem-solving, ensuring that students engage with the underlying concepts and reasoning processes, rather than solely focusing on obtaining correct answers. This structured approach is crucial for preventing students from becoming overly reliant on AI-generated solutions without understanding the mathematical principles involved, and it allows educators to intentionally guide the learning process in the Post-IA context.

The COMPAS framework promotes a balanced approach to mathematical learning in the Post-IA context by simultaneously addressing pragmatic and epistemic dimensions. Pragmatic dimensions concern the how of problem-solving – the efficient use of tools and procedures – while epistemic dimensions focus on knowing – the development of mathematical understanding and justification. Research indicates that implementing COMPAS during AI-assisted learning preserves student control over these epistemic aspects of problem-solving; students retain agency in determining the validity of solutions and constructing their own mathematical knowledge, rather than passively accepting AI-generated outputs. This preservation of epistemic control is crucial for fostering genuine mathematical understanding and avoiding over-reliance on AI tools.

Envisioning a Future of Human-AI Collaboration in Mathematics

The COMPAS framework strategically integrates generative AI to foster a dynamic of Human-AI Complementarity, rather than simple automation. This isn’t about replacing human capabilities, but about intelligently distributing tasks based on strengths; AI excels at computationally intensive processes and pattern recognition, while students concentrate on conceptual understanding, critical analysis, and creative problem-solving. By asymmetrically assigning these different demands – AI handling calculations and procedural steps, and students focusing on reasoning and interpretation – the framework maximizes cognitive efficiency. This division allows learners to delve deeper into the ‘why’ behind the answers, building robust mathematical intuition and preparing them for future scenarios where collaborative intelligence with AI will be essential.

The evolving relationship between students and artificial intelligence within educational settings increasingly prioritizes a division of cognitive labor. Rather than replacing human effort, generative AI tools are designed to handle computationally intensive tasks, freeing students to concentrate on the nuances of reasoning and problem-solving. This collaborative dynamic allows learners to move beyond rote calculation and focus on conceptual understanding, strategic thinking, and the critical evaluation of results. By offloading tedious processes to AI, the framework encourages students to engage in higher-order cognitive skills – formulating hypotheses, interpreting data, and creatively applying mathematical principles – ultimately fostering a more profound and adaptable skillset for future challenges.

The integration of generative artificial intelligence into educational frameworks extends beyond simply improving test scores; it actively cultivates a skillset essential for future success. This study demonstrates that by strategically pairing human cognitive strengths with AI’s computational power, students aren’t just learning how to solve mathematical problems, but also developing an intuitive understanding of mathematical thinking – a ‘formative sense of mathematical activity’ – even while leveraging AI tools. This approach suggests a shift in pedagogical focus, preparing students not as isolated problem-solvers, but as collaborators capable of effectively utilizing AI as an extension of their own reasoning abilities, a crucial competency in an increasingly automated world.

The research highlights a crucial shift in pedagogical approach, emphasizing that simply banning AI tools isn’t the answer. Instead, the focus must be on cultivating students’ epistemic control – their ability to critically evaluate information and construct their own understanding. This aligns with Niels Bohr’s assertion: “The opposite of a trivial truth is another trivial truth.” The article demonstrates that merely presenting a ‘correct’ answer, whether from a student or AI, doesn’t foster genuine mathematical understanding. Open-ended tasks, coupled with didactic regulation, provide the necessary framework for students to grapple with complexity, explore multiple solution pathways, and ultimately, retain agency over their learning process. The core idea isn’t about what students learn, but how they learn it – a point underscored by the need to move beyond seeking singular ‘right’ answers.

Beyond the Prompt: Charting a Course for Mathematical Agency

The preservation of epistemic control, as demonstrated through regulated open-ended tasks, offers a temporary reprieve, yet the landscape continues to shift. One observes a pattern: tools initially perceived as threats to cognitive skill often necessitate a refinement – not abandonment – of those very skills. The critical question now becomes not simply how to use AI in mathematical education, but which mathematical activities are fundamentally reshaped by its presence, and whether those shifts represent genuine progress or merely a different form of automation. Further investigation should explore the long-term impact on mathematical modeling itself; does readily available AI assistance subtly alter the types of models students choose to construct, or the level of conceptual understanding they deem necessary?

A compelling line of inquiry lies in the visualization of student agency. What visual patterns emerge when analyzing student work completed with, and without, AI assistance? Can one objectively measure the ‘depth’ of mathematical thinking, and does this depth correlate with the level of epistemic control exercised? The challenge is to move beyond simply assessing correctness and instead focus on the process of mathematical discovery – a process that, ironically, may require increasingly sophisticated analytical tools to fully reveal.

Ultimately, the success of this approach – or any approach – hinges on acknowledging a fundamental truth: education is not about predicting the future, but about cultivating the capacity to navigate it. The next iteration of research should therefore prioritize adaptability, encouraging students not merely to solve problems, but to critically evaluate the solutions offered by increasingly powerful algorithmic systems. The pattern suggests that the most valuable skill may not be mathematical proficiency itself, but the ability to discern when – and how – to apply it.

Original article: https://arxiv.org/pdf/2602.09242.pdf

Contact the author: https://www.linkedin.com/in/avetisyan/

See also:

2026-02-12 04:11