AI can make math homework easy. These professors say we shouldn’t let it

The teenagers have never heard of Leonhard Euler, but on a July morning in a Dalhousie University classroom, they are using blankets and sticks to try to solve the great mathematician’s most famous, 300-year-old puzzle.

The challenge is part of the Halifax university’s summer math camp for Indigenous junior high school students – an attempt to rebrand the world’s most dreaded school subject as fun and interesting between city outings and dorm sleepovers.

Sara Faridi, the math professor leading the morning activity, explains that the 18th-century citizens of Königsberg were bedevilled by a simple riddle: how could they walk the German city’s seven bridges exactly once, and end up back in the same place?

Volunteers take turns testing different routes along the numbered stick “bridges” to each “blanket” island, trying not to cross the same bridge two times, and yet still return to their starting point. One attempt after another fails. Dr. Faridi finally shares the secret. The task is impossible.

On the board, she shows how Mr. Euler figured this out with dots and lines and created graph theory in the process. One of the most significant discoveries in modern mathematics emerged out of failure – a demonstration of one of the many important life lessons that math can teach.

In Depth: AI has changed the calculus for computer-science students, and schools that hope to teach them

Today, Mr. Euler would never have heard from the curious citizens of Königsberg; they would have simply asked the chatty math genius on their phones for the answer.

The temptation is irresistible: a computer can now solve millions of those painstaking quadratic equations – and a growing number of more complex math problems – in the time it takes a student to lift a pencil.

Math educators say a growing number of students are taking AI shortcuts on their homework, and at the cost of their test scores. But the long-term loss is more significant than grades, these professors say. They’re also missing out on the brain-boosting benefits that math can offer: the resilience and self-confidence and critical thinking that comes from persisting with tough problems.

Last spring, University of Victoria math professor Trefor Bazett realized that 42 per cent of his students in his linear algebra class had used AI to solve a particular homework question; they’d turned in a one-line answer using an obscure, highly specialized formula not found in the textbook or easily on the internet.

Dr. Bazett decided to watch and wait. In the midterm exam, that same group performed 22 percentage points worse than students who’d submitted a longer solution to that one question without using AI. On average, the group who used their brains, passed the test; the ones who outsourced their thinking to AI, failed.

Dr. Bazett’s concern – that his students are off-loading hard work and critical thinking to AI at the cost of their own deep understanding and cognitive development – is supported by other recent studies.

A 2025 paper published by University of Pennsylvania researchers found that when high school students had access to a ChatGPT-style math tutor, their grades improved 48 per cent. When they were tested without it, however, those same students performed 17 per cent worse than those who had not had access to the same tutor.

In the age of artificial intelligence, professors try to make cheating hard again

The machines, meanwhile, are only getting faster and smarter. In September, OpenAI announced that 10,000 autonomous agents running at the same time had partially cracked a 90-year-old stumper, the Navier–Stokes existence and smoothness problem, in 88 hours. (Those agents aren’t cheap labour, the BBC quickly noted; based on the company’s own pricing, the task cost as much as US$10-million.)

A month later, OpenAI released answers to 400 famous math problems, leaving the world’s best math minds to decipher and validate a massive stack of computer-generated solutions and proofs.

While AI mires mathematicians in an existential crisis, its arrival in classrooms has also hastened the discussion around adding more humanity and well-being to the only subject with a well-known psychological condition attached to its name.

Math anxiety increased over 10 years in almost every country where 15-year-old students take the Programme for International Student Assessment exam, according to findings in 2022, the last year the question was asked. That year, 39 per cent of Canadian students agreed they felt “very nervous” doing math problems, compared to 31 per cent in 2012.

In the 2025 Education Quality and Accountability Office exam, taken by Grade 6 and 9 students in Ontario’s public schools, only half said they liked math, or felt they were good at it. And math’s gender gap persists: girls experience math anxiety more than boys and pursue advanced math less often, closing the door as early as middle school to meaningful, well-paying STEM jobs.

Yet feeling happy and confident in math class is highly predictive of achievement, says Julia Hill, a math education researcher at RMIT University in Melbourne, Australia. Dr. Hill has developed a math well-being index that uses student surveys to measure how well math classes foster values such as meaning and perseverance.

Dr. Hill suggests that teachers explicitly tell students that struggling with math builds life skills. “If we want a better society where we have students who can solve social justice issues, we need math. We need math to understand other perspectives. To understand what fairness looks like.”

Besides, skipping math class isn’t a long-term solution; you’ll have to deal with numbers eventually, and the most scientifically supported way to reduce anxiety is to face the thing causing it.

Erin Maloney, an assistant professor of psychology at Carleton University, who studies the cognitive and emotional factors that affect math learning, sees studying the subject as “a fabulous training ground,” to build grit and composure when confronted by mistakes. “There’s lots of room to perform in situations that feel high pressure, but are actually pretty benign as far as actual pressures in life go.”

Sticking with math also helps with life down the road. Even if they never formally study math after high school, adults who know their way around numbers can more accurately assess risk, make smarter financial decisions and deduce whether politicians are lying about tariffs. At a macro level, math-literate populations are associated with more innovative societies.

Opinion: Universities should stop treating AI as a cheating problem

Taking the easy way out, as the University of Pennsylvania study found, made students overly optimistic about how much they’d learned. (By contrast, mathematicians say, wrestling with a tricky problem delivers a healthy dose of humility.) AI suffers from a similar overconfidence; unless prompted, it rarely expresses doubt. In the 2025 paper, researchers fed 57 problems multiple times to the fourth-generation AI: it gave the wrong answers 51 per cent of the time.

But that was then – today, Dr. Bazett estimates, AI gets the math right 99 per cent of the time. The current issue isn’t accuracy, he says, but the human knowledge we lose by not reaching the answers ourselves. The journey to the most significant math problems often leads to unexpected advancements or useful tangents along the way – just like Mr. Euler and graph theory.

Dr. Bazett advises his students to do their own hard thinking and use AI as a skilled tutor who offers hints when you’re stuck. Don’t leap to an easy AI answer without exercising your own memory and critical thinking, he tells them.

“Friction is where the magic of the learning actually happens,” he says. This is why, by fourth year, his students can write pages of calculations, with a logical, yet creative, focus that his first-year class can hardly fathom.

Nour Allam is a first-year master’s student in computer science at Dalhousie University, and a teaching assistant, including for the math camp. “Even if AI solves all these problems, it’s not really meaningful if a human doesn’t understand them,” she says.

For her own classes and research, she says she only uses AI for information, such as basic definitions that don’t involve skills such as critical thinking, or reading and interpreting proofs. But AI assistance isn’t always obvious on assignments, so Ms. Allam sees more professors weighting in-class exams more heavily than take-home work that often requires more creativity.

“Unfortunately, that has pushed students to just want the formula to do well on the test,” she says, “instead of really thinking.”

Opinion: The question isn’t whether AI can teach. It’s whether students will let it

In Francis Su’s puzzle-filled book, Mathematics for Human Flourishing, the math professor at California’s Harvey Mudd College makes an elegant case for the virtues of hard problems: the persistence sharpened by sticking with a challenge, the hopefulness of believing you’ll eventually solve it, the diligent search for truth, the openness to tackle problems from a different perspective.

In an interview, Dr. Su says he has recently observed students becoming more careful about how they use AI for classwork – a few even refusing to use it altogether. But fewer students are knocking on his door for extra help, even though he urges every class not to let AI replace one-on-one, in-person meetings.

“They’re missing out on the personal, human exchanges that help math feel more grounded,” he says. “It’s much more convenient to sit in your room and ask AI a question.”

But learning should be taxing, he says. Every year, on his final written exam, Dr. Su includes a bonus questions asking students to explain the value of a challenging experience they had in his class. As he writes in his book, “Math builds confidence in struggle.”

The AI world of instant answers may be easier to resist, he suggests, “if we remember the purpose of an education. It’s not just about getting answers. It’s not just about getting a good job. It’s about shaping who we are and how we live and move in the world.”