麻豆淫院

In one recent mathematics lesson at a state secondary school in the south of England, a student used generative artificial intelligence (AI) to explore whether it is valid to divide by an algebraic expression that could be equal to zero when solving a trigonometric equation. The concern was not how to reach an answer, but whether this step might exclude valid solutions. The AI responded quickly, offering a method that led to an answer, but shifted focus away from validity towards procedural completion.

Much of the conversation about AI in education focuses on outputs: correctness, efficiency and explanation quality (Létourneau et al., 2025; Luo et al., 2025). This blog post focuses instead on interaction: the sequence of prompts, responses and interpretations through which a student makes sense of mathematics.

It draws on early observations from an ongoing 麻豆淫院-funded practitioner research study examining how students with special educational needs and disabilities (SEND) engage with AI, and suggests that what happens within this interaction may matter more than the final answer.

The study observed two students with differing attainment taught by the researcher. Early observations suggest that, for both students, access is not the main issue; the challenge is whether students can use the interaction itself – not just the answer, but the sequence of prompts, responses and interpretations – in ways that support understanding.

What we are seeing in classrooms

Across early observations, two patterns emerge. In some instances, the students use AI as reassurance: confirming a method or identifying a missing step. At other times, the AI may offer dense explanations, move too quickly to a solution, or assume reasoning that does not match the student’s approach, causing interaction to drift. What matters for pupil engagement and learning is alignment: whether the AI’s response remains connected to the learner’s question and understanding.

One of the two students, a 15-year-old student with low prior mathematical attainment, worked with a linear cost model C = 4 + 2n, where C represents total cost, £4 is a fixed charge, and 2n represents an additional £2 per hour. The AI correctly explained the algebra required to answer how many hours were worked if the final cost was £22. However, the student struggled to connect this to meaning, wanting instead to interpret 2n as repeated addition. The researcher intervened by slowing the interaction, asking the student to reflect on the expression and focus on the meaning of 2n. By isolating this term and linking it to prior knowledge, the student reconstructed the model in a way that made sense to them.

The second student, a high-attaining A-level student, asked when it is valid to divide by an expression, specifically cos θ, which may be zero. This concern arose within a specific problem where dividing did not, in fact, exclude any solutions. While the student’s question was conceptually valid, the AI instead redirected to an efficient solution, bypassing the underlying issue the student was investigating.

Despite differences, a common issue emerges: AI can produce correct procedures while remaining misaligned with the ideas the learner is trying to explore.

‘AI can produce correct procedures while remaining misaligned with the ideas the learner is trying to explore.’

Why this is an inclusion issue

AI can appear inclusive because it is widely available and generates explanations on demand. Yet access alone does not guarantee meaningful participation: students may use AI without making sense of its output. This matters in mathematics, where learners vary in how they interpret symbols, process language and draw on prior knowledge (Sfard, 2008). An explanation clear to one may be too dense, too fast or misdirected for another. Inclusion is therefore not simply whether AI provides support, but whether the interaction enables learners to build usable meaning.

What teachers are doing that AI often is not

A clear finding in this study is that teachers are still required for interactional work that AI struggles to replicate:

• slowing down the pace of explanation
• drawing attention to key ideas
• linking explanations to prior knowledge
• allowing students to reason in ways that make sense to them
• reopening conceptual issues when dialogue moves too quickly to closure.

In other words, teachers are not only explaining mathematics – they are making explanations usable.

An emerging issue worth attending to

These are early findings from a small study, but they point to something important. The central issue may not be whether AI can generate correct answers, but whether learners can remain in a productive relationship with the interaction long enough to develop understanding. As such, the more important question may not be whether AI works, but for whom it works, under what conditions, and what kinds of learning it makes possible.

References

Létourneau, A., Deslandes Martineau, M., Charland, P., Karran, J. A., Boasen, J., & Léger, P. M. (2025). A systematic review of AI-driven intelligent tutoring systems in K-12 education. npj Science of Learning, 10(1), 29.

Luo, J., Zheng, C., Yin, J., & Teo, H. H. (2025). Design and assessment of AI-based learning tools in higher education: A systematic review. International Journal of Educational Technology in Higher Education, 22(1), 42.

Sfard, A. (2008). Thinking as communicating: Human development, the growth of discourses, and mathematizing (1st ed.). Cambridge University Press.