We Had the Same Panic About Calculators

When calculators arrived in schools, the panic was immediate and familiar: children will stop learning mathematics. Why would they, if a machine can do it for them?

We know how that turned out. The students who used calculators best were the ones who understood the maths well enough to know when the calculator was wrong. The tool amplified the thinking that was already there. It exposed the gaps where thinking was absent.

AI is the same moment, with a much more convincing tool.

The trap we fell into

Most of us started using AI the way we started using search engines: as an answer machine. We typed the question, got the answer, and moved on. We used it to compress tasks, not to extend thinking.

Then we looked at children doing the same thing and assumed the worst. They're going to stop thinking for themselves. They'll just ask AI for the answer. The concern was real. The logic was sound. It was also, in large part, a projection.

We were using AI as a task-doer. We assumed they would too.

Here's what actually changed for me.

A school timetabling problem

I was working on a complex timetabling problem. I had dozens of variables that all had to resolve at once: subject combinations that either worked or didn't, depending on staffing, room availability, and curriculum requirements. Every configuration I tried seemed to satisfy some constraints while breaking others. I kept prompting the AI to solve it, and the output kept looking right while being wrong. Each time I pointed out an error, it would generate a new version, still wrong in a different way.

So I changed approach entirely. Instead of asking for the answer, I asked the AI to teach me the maths: the combinatorial logic, why certain configurations were impossible, what the underlying structure of the problem actually was. Once I understood that, everything changed. I came back with better questions. I asked for options and trade-offs rather than solutions. I made the decisions. The AI helped me think instead of thinking for me.

That shift, from answer machine to thinking partner, only worked because I had enough existing knowledge to catch the errors and redirect. My knowledge didn't become redundant when AI arrived. It became the thing that made AI usable.

The developer parallel

If you write code, you've done the same thing. You paste an error into an LLM and it generates a fix. Sometimes the fix looks right and misses something critical. You catch it because you understand what the code is supposed to do. Without that understanding, you'd accept the answer, ship the bug, and wonder why it broke.

Domain knowledge is the thing that makes AI use intelligent rather than passive, in any field. Curiosity, critical thinking, knowing when something doesn't add up. Those are the skills that determine whether someone uses AI well or just uses AI.

What the research shows

Nord Anglia Education ran a two-year study with Boston College, working across 27 schools, 20 countries, and more than 12,000 students. Teachers and leaders embedded structured thinking routines into everyday classroom practice. One example: See, Think, Wonder: What do I notice? What does that tell me? What questions do I still have?

The results after two years:

85% of students reported knowing what they're good at. 76% said they had become more independent. 72% said their understanding of how they learn had improved.

Up to 96% of teachers in the study agreed that this approach helps students succeed beyond school.

The World Economic Forum estimates that around 40% of core job skills will change by 2030. The skills identified as most essential: analytical thinking, creativity, adaptability. The same skills the thinking routines were building.

They were building something older and more fundamental than an AI curriculum: the capacity to think about your own thinking. To notice what you don't know. To stay curious in the face of a hard problem rather than shortcut to the nearest available answer.

You can read more about their research here: https://www.nordangliaeducation.com/metacognition/teaching-the-skills-ai-cant-replace#

What metacognition actually is

I believe metacognition is thinking about your own thinking. Knowing what you understand and what you don't. Noticing when your reasoning has gone off track and asking better questions of yourself before asking them of anyone else.

It sounds abstract. In practice, it looks like this: a student reads a passage, notices they didn't actually follow the argument, and goes back rather than moving on. A developer runs a test, sees a result that looks correct but feels wrong, and pauses to check rather than assuming.

It's the skill I was using, without naming it, when I changed my approach to the timetabling problem. It's the skill that made the difference between useless AI output and something I could actually work with.

In my opinion, schools that teach thinking routines aren't being anti-AI. They're building the exact foundation that makes AI use intelligent. A child who has learned to ask What do I notice? What does that tell me? What questions do I still have? is a child who will use AI as a thinking partner, not just a vending machine.

The children who will thrive

The children and adults who do best in an AI world won't be the ones who use it most. They'll be the ones who use it best. That requires something AI cannot provide: the ability to know when the output is right, when it's wrong, and when to push back.

That is not a technology skill. It is a human skill. Schools have been building it for a long time. The value of that work has just become much easier to see.

If this resonated, whether you're a teacher, a parent, or someone thinking about how AI is reshaping how we work and learn, I'd love to connect. Drop a comment or send me a message.