Could AI make students more willing to think?

Imagine a Year 12 student who notices that rents near the school have risen much faster than in the next town. They want to know why, but the first table of housing data leaves them unsure where to start.

An AI tool could help them interpret the column headings and check whether the figures cover comparable properties over the same period. With that obstacle out of the way, the student can begin to investigate a question they might otherwise have abandoned.

As a former A level Economics teacher, that possibility interests me. Help at the right moment could make a difficult question feel worth pursuing. The educational challenge is to ensure that the student develops their understanding along the way.

How easier access could change ambition

In Cognitive Mercantilism, Venkatesh Rao compares our thinking about intelligence with an old view of wealth as a fixed stock of land and gold. If another country gained more, there was less left for yours. The priority was to guard what you had.

Drawing on the historian Joyce Appleby, he describes how access to new goods, including colourful cotton cloth, helped change ordinary people’s aspirations. Some were willing to work more because there were new things they wanted to buy.

His suggestion is that AI could change intellectual aspirations in a similar way. When help becomes easier to obtain, more projects may seem worth attempting. A question someone once set aside could become the start of an evening’s investigation.

For schools, that is an appealing possibility. It also raises a further question: what does the student come to understand through the work?

Students still need knowledge of their own

Return to the student investigating rents. Suppose the tool suggests that higher rents prove more people have moved into the area. A student who understands supply and demand has a reason to push back: fewer properties available to rent could also explain the rise. Both changes might be involved.

The rent figures alone cannot settle that question. To judge the explanation, the student needs some subject knowledge and an idea of what further evidence would help.

An AI tool could just as easily produce a plausible paragraph about the two towns. A finished paragraph, however, gives a teacher limited evidence of what the student understands. Access to an explanation is valuable, but part of the purpose of school is to develop the person who can judge it.

That is why concerns about students outsourcing their thinking deserve a proper response. Asking students to complete part of a task independently can create an opportunity to develop and demonstrate understanding. That can sit comfortably alongside generous access to help during other parts of the work.

What the research suggests

A 2025 study by Bastani and colleagues makes the distinction between performance and learning particularly clear. In a randomised study involving nearly 1,000 secondary mathematics students at one school in Turkey, access to a basic GPT-4 assistant improved practice performance. On subsequent tests without AI, however, that group scored about 17% lower than the control group.

A more carefully designed version provided hints and used teacher-supplied solutions and guidance about common mistakes. It largely avoided that harm, although it produced no statistically significant advantage over the control group on the unassisted tests.

There is encouraging evidence too. In a study involving 194 Harvard physics undergraduates, Kestin and colleagues found greater immediate learning gains from structured AI tutoring than from classroom lessons using active learning. The comparison covered two lessons, with students experiencing both approaches. That is an encouraging result in that setting; it cannot settle questions about long-term learning in secondary schools.

These studies examine learning under particular conditions. The possibility of greater intellectual curiosity remains a proposition to explore. What the findings give us is a reason to pay close attention to how help is designed and what students can subsequently do for themselves.

A classroom task worth trying

Consider an illustrative Year 12 activity on inflation.

Students first explain independently what it would mean for annual inflation to fall from 8% to 3%. Their answers give the teacher an opportunity to spot confusion between a slower increase in the overall price level and a fall in that level.

Students could then use an approved AI tool to question their explanation, request a worked example or explore a point they find confusing. There could also be room for a follow-up question of their own: how can some prices fall while the overall price level rises, for example?

The teacher would check the explanations and examples being used, particularly the distinction between the overall price level and individual prices.

With the tool closed, students would explain the same distinction in a different context and justify their answer. A later check would help establish whether their understanding had lasted beyond that session.

This is a suggestion for task design, whose value would need to be checked in practice. Its purpose is to make useful help available, give students room to explore and retain opportunities to see what they have learnt.

Access needs to include guidance

If we want more students to pursue ambitious questions, we also have to think about who gets useful help.

A login alone is a limited offer if some students have better tools, more reliable access or more guidance in using them. An ambitious approach would include suitable school provision and explicit teaching about how to question an explanation, check a source and recognise when further help is needed.

What I’d hope to see is the student returning to the rents question with an idea of their own: “We know the rents went up, but we don’t yet know what happened to the number of homes available to rent.”

Help was within reach, so the student could pursue a question they cared about. Knowing what to ask next is something they can take with them when the chat is closed.

Gary Roebuck is a former Head of Economics and A level Economics teacher, and founder of TeachEdge.