I once watched a student forget the formula for the area of a trapezoid.
That could have been the end of the problem. They did not have the formula, so they did not have the answer.
But instead of stopping, they began to rebuild the shape. They split it into pieces they understood — a rectangle, a triangle on one side, another triangle on the other — and used the areas they did know to work out the area they had forgotten.
It was slower than remembering the formula, but far more interesting.
That moment has stayed with me because it captures something important about mathematics. The student wasn’t simply retrieving information — they were reasoning, using structure, taking something unfamiliar and turning it into something they could work with.
That is the kind of thinking we need to protect as AI in math education becomes the norm, not the exception.
Why AI in Math Education Changes the Stakes
A student can now type a question into an AI tool and receive the answer, the method, a worked solution, an explanation and another example in seconds. In many ways that’s remarkable, and there are good uses for these tools — pretending they don’t exist will not help anyone.
But if mathematics becomes mostly about getting to the answer as quickly as possible, AI will always look like the better mathematician. That should make us pause — not because mathematics matters less now, but because the human work of mathematics matters more.
For a long time, correct answers have carried enormous weight in math education, and it’s easy to understand why: they’re visible, markable, and fit neatly into tests, reports and spreadsheets, giving teachers, students and parents a clear signal that something has been learned.
But a correct answer is not the same as mathematical thinking.
The purpose of mathematics education has never simply been to produce answers. It is to help students make sense of problems, notice structure, test ideas, reason carefully, recover from errors and decide whether an answer is reasonable. Those things cannot be outsourced without cost.
Is the Tool Helping Them Think, or Thinking For Them?
Of course, tools have always been part of mathematics — pencil and paper, diagrams, manipulatives, calculators. Good tools can reduce unnecessary load and help students think more clearly, freeing up attention for deeper reasoning and making invisible ideas visible.
So the key question is not whether students should use tools. The better question is this:
Is the tool helping the student think, or is it doing the thinking for them?
That distinction matters. When a student writes down working, they are offloading memory, but they are still making decisions. When a student draws a diagram, they are offloading abstraction, but they are still reasoning about structure. When a student uses a calculator in the right context, they may be offloading computation so they can focus on a larger mathematical idea.
But when a student hands the whole problem to AI before they have wrestled with it, something different is happening: the thinking itself may have been outsourced.
That’s where we need to be careful — not fearful, not reactive, not anti-technology. Just careful.
The Moments Before the Answer Arrives
As teachers, we know that some of the most important mathematical moments happen before the answer arrives. The pause. The false start. The messy page. The student who says, “Wait, that can’t be right.” The class discussion where one method suddenly connects to another. The moment a child sees that 5 × 44 can become 10 × 22 by doubling one factor and halving the other.
These moments are not inefficient interruptions on the way to the answer. Very often, they are the learning.
Mathematics develops through contact. Students need contact with uncertainty, structure, error, other people’s strategies and their own half-formed ideas. If we remove too much of that contact in the name of efficiency, students may still arrive at answers, but they may not develop mathematical judgement — and judgement is exactly what they will need in a world full of instant answers.
They’ll need to know when an answer is too large or too small, whether a graph is misleading, when a statistical claim sounds impressive but says very little. They’ll need to compare methods, question assumptions and make decisions when there is not a worked solution waiting at the back of the book.
They will also need confidence. This is sometimes missed in conversations about AI and education — the issue is not only academic integrity or assessment design, important as those are. It is also about the formation of the student.
What happens to a student who learns, quietly and repeatedly, that the machine is always better at thinking than they are? What happens when their first instinct is not to try, estimate, sketch, decompose or discuss, but to ask for the answer?
We should want students to use powerful tools. But we should also want them to experience themselves as capable mathematical thinkers, and that requires patience. I’m reminded of something mathematician Nalini Joshi once said about her own experience as a learner — that she often felt like she was a month behind everyone else because she wanted to understand everything properly. She needed to get underneath ideas, not just move past them.
From the outside, that can look like a struggling student falling behind. But sometimes the student who appears slow is not disengaged or incapable — sometimes they’re doing the deeper work, making connections, refusing to accept a shallow understanding, building something that may not show up on the next quiz but may matter enormously in the long run.
In many classrooms, that kind of patience is difficult — teachers are under pressure, curricula are crowded, lessons move quickly, and technology promises shortcuts. Students compare themselves to others, and mathematics, perhaps more than any other subject, can make a struggling student feel exposed.
Slow Thinking Is Not Failed Thinking
So let’s say this clearly: slow thinking is not failed thinking.
A student who takes time is not necessarily behind. A student who makes a mistake is not necessarily lost, and one who needs to draw, talk, test or restart is not necessarily weak — often, that is what genuine mathematical work looks like.
And sometimes the most helpful person in the room is the student brave enough to ask what everyone else is quietly wondering. We often call it the “dumb question,” but it rarely is — more often, it’s the honest question, the clarifying one, the question that reveals the gap in the room and gives everyone else permission to think more carefully. We should be very careful not to build classrooms, assessments or technologies that make students afraid to ask it.
This does not mean we abandon fluency — quite the opposite. Students need number facts, efficient methods, automaticity, enough knowledge in long-term memory to think with. But fluency should not be reduced to speed alone. Real fluency feels more like freedom. It gives students options, helps them move flexibly, lets them choose a strategy, change direction and make sense of what they are doing.
In an AI age, that kind of fluency matters deeply.
If students only learn mathematics as answer production, AI will seem like a replacement. But if they learn mathematics as reasoning, structure, creativity and judgement, AI becomes just one more tool that must be used wisely. That is where teachers remain irreplaceable.
Teachers do far more than deliver explanations. They notice hesitation. They choose when to step in and when to wait. They create the conditions for safe struggle. They ask, “Why?” and “How do you know?” and “Is there another way?” — and they know when a wrong answer reveals more learning than a correct one. No AI system can fully replace the relational, responsive and deeply human work of helping a child become a thinker.
So perhaps the challenge before us isn’t simply to redesign assessment or write better AI policies, though we will need to do both. Perhaps the deeper challenge is to recover a clearer sense of what mathematics education is for.
It is not merely for answer-getting. It is for helping young people become more careful, more courageous and more capable thinkers — for giving them the confidence to meet difficulty without immediately outsourcing it, and for helping them see that mathematics is not a machine for producing answers, but a human way of exploring pattern, structure and truth.
Three questions Worth Protecting
In the age of instant answers, I think there are some questions worth asking more, not less.
- Why does this work? This moves students beyond answer-getting and into structure.
- Can you find another way? This helps students see that mathematics is not a single narrow path.
- How do you know your answer makes sense? This builds judgement, estimation and responsibility.
These questions are simple, but they protect something important. They remind students that mathematics is not just about arriving — it is about seeing, testing, connecting and understanding.
The arrival of instant answers should not make us panic. But it should make us more intentional. We should protect time for thinking, value explanation, estimation, representation and strategy, and help students use tools without surrendering agency. We should remind them, often, that the struggle before the answer is not wasted time.
It is where the mathematics is happening.
In a world where answers are instant, thinking matters more.
Want to read more – check out -> “What We Deliberately Don’t Let AI Do”
