AI in Education #2: Five Things I Learned from Barbara Oakley about Why Practice Still Matters in the Age of AI

This is the second in a series of AI in Education specials, where I speak to the world’s leading experts about how AI is changing teaching and learning.

Barbara Oakley was terrible at maths as a kid. She sat in seventh-grade algebra, reading history books until the teacher confiscated them. Then she just pulled out another one.

She enlisted in the army, learned Russian, and did not touch maths again until she was 26. She started with remedial high school algebra. She is now a distinguished professor of engineering (and the co-creator of the most popular online course in the world)

It was an absolute pleasure to have Barb on the podcast, and I wanted to share five practical takeaways from our conversation that I think every classroom teacher and school leader will find interesting.

1. Students need knowledge in their heads, not just at their fingertips

Here is something Barb said that stopped me in my tracks. The most common fatal error in hospitals is someone typing an extra digit in a medication dosage. A patient gets ten times the dose they should have received.

Why does this matter for maths teachers?

Because if a student has internalised their multiplication tables, and you tell them 12 × 12 is 299, something inside them screams: no, it is not. But a student who has never committed those facts to memory? They shrug. Close enough.

Barb told me about her own daughter, who she put through ten years of Kumon maths – about 20 minutes a day of structured practice. Her daughter hated it. She went off and got a degree in studio art. Then she came back, got a master’s in statistics, and was handpicked by a professor who said he normally never selects students from the American education system because they do not have a feel for the numbers.

What was different about her daughter? Ten years of practice.

I regularly see the opposite in schools I visit. Students reach for a calculator for 6 × 7. They cannot estimate whether an answer is reasonable. They have no sense of whether 4.2 × 19 is closer to 80 or 800. And without that sense, no amount of access to AI or Google will save them from the medication-dosage problem.

The takeaway is simple. Students need certain knowledge in their heads. AI is not a substitute for that. It is a complement to it.

2. AI can make the boring-but-essential stuff less boring

So if practice matters, where does AI fit in?

Barb made a distinction I found really helpful. She talked about the two memory systems in the brain. One of them – the habitual, procedural system – thrives on repetition. That is how you form habits. That is how you get to the point where number facts feel as natural as your native language. And getting there takes a lot of practice, much of which is not pleasant.

This is where AI can help.

AI can keep practice at the right level. It can adapt in real time. It can add elements of gamification that a static worksheet cannot. Barb was clear: this is not about replacing the practice. It is about making the practice more targeted and, where possible, more engaging.

I think this is a much more useful framing than the one I often hear, which is that AI will replace the need for students to memorise things. Barb pushed back on that hard. As she put it: reform methodologies that say students do not need to memorise multiplication tables because they can always look them up – those approaches know nothing about how the brain works.

Strong words. But I think she is right.

3. Homework reliability is now zero – so test more in class

Regular readers will know I have form on this one. I wrote a newsletter called “Is Maths Homework Dead?” after my best friend’s daughter stopped messaging me for help with her Sparx homework. Not because she had suddenly understood fractions, but because she had discovered ChatGPT.

Barb sees the same problem from the university side. Her view is blunt: if you do not see students producing answers without access to AI, you have no idea whether they know anything. And if you are not careful, you will be widening the socioeconomic gap – because the motivated students will use AI to genuinely build knowledge, while everyone else will use it to avoid thinking.

Her solution? Test more. And before your alarm bells go off, she does not mean high-stakes, silence, grades-on-the-line testing. She means regular, low-stakes retrieval – weekly quizzes, maybe even daily quizzes – done in class, without access to AI.

This lines up with the approach I described in the homework newsletter, where Adam Boxer suggested that homework should prepare students for assessment in class, rather than being the assessment itself. Set the homework, then quiz them on the same material next lesson under controlled conditions. If they bomb the quiz, you know they did not put in the work. If they ace it, you know they did the homework and the material has stuck.

Barb also made a point that I think needs to be said more often: everyone loves to hate on tests, but try running schools without them. She volunteered in an inner-city school before national testing came in, and the picture she painted was grim. Paper aeroplanes. Students doing nothing. Teachers with no accountability. Testing is not perfect. But the alternative is worse.

4. AI is a brilliant thought partner for teachers – but it is not a curriculum

I asked Barb about AI’s role in lesson planning and resource creation. Her answer was nuanced, and I think she got the balance exactly right.

On the one hand, she thinks it is absurd that teachers are expected to create their own materials from scratch. She spent time in New Zealand, where there was a culture that said teachers had to make their own stuff up, otherwise you were dictating to them. Barb pointed to programmes like John Mighton’s JUMP Math, where the materials are so carefully designed that they give students visual insight alongside a step-by-step explanation. A typical classroom teacher, juggling five classes a day, cannot be expected to produce work of that quality every evening.

I have long wrestled with the idea of scripted lessons. For the first 12 years of my career, I would have hated the very notion of following to the letter something someone else had written. Then I read the research and thought: maybe I should follow to the letter something designed by experts, tested over years, and refined. I am still not fully there. But I am a lot closer than I used to be.

Barb’s position? Great teachers should have some freedom. But the more you loosen things up, the more that weaker teachers will struggle. And as she pointed out with characteristic directness: half of all teachers are below average.

So AI can help. Barb uses it herself – she feeds her notes into ChatGPT and asks it for a good hook, or for ideas for active learning activities to intersperse between her explanations. She described it as a tool for generating surprise, which I loved. As she explained, the brain is constantly predicting what will happen next, and when something different occurs, you get a mismatch that locks attention onto the new information. But coming up with those surprises is hard. AI can help you do it.

5. Schools of education need a shake-up – and AI might force one

This was the part of the conversation where Barb was at her most passionate. She argued that schools of education are doing teachers a disservice. They teach Dewey, Vygotsky, Piaget, and Montessori – theorists from 50 to 100 years ago – and the pedagogical approaches they promote are often at odds with what the science of learning actually tells us.

She drew an analogy with literacy. We know phonics works better than whole language. The research is clear. But schools of education keep reverting to whole language because it sounds so appealing: learning to read is just like learning to speak! It was only when governments stepped in – as in Mississippi, where reading scores vaulted forward – that things changed.

Barb sees the same pattern in maths and science education. Student-centred, inquiry-based, constructivist approaches that sound wonderful but leave students floundering. She told me she has travelled the world and heard primary teachers ask: “You mean it is okay to explain something to a student?” That is how extreme things have got in some places.

So where does AI come in? Barb thinks it could help create new pathways for teacher training that sit outside traditional schools of education – pathways grounded in how the brain actually works, not in century-old theories. She wants teachers to learn about memory systems, about the science of practice, about what cognitive science actually says about effective instruction.

I am more optimistic than Barb about progress in the UK. Things are improving here, particularly in maths, where the influence of cognitive science on curriculum design and teacher training is growing. But her broader point stands: if the people training teachers do not understand the science of learning, and if they are actively promoting approaches that make learning harder, then we have a serious problem. And AI – for all the disruption it is causing – might be the thing that forces the issue.

Over to you

This was a brilliant conversation, and I have only scratched the surface here. You can listen to the full episode on the Mr Barton Maths Podcast, and if you have not already, check out Barb’s book Uncommon Sense Teaching and her courses on Coursera – especially Learning How to Learn, which is free.

This is the second in a series of conversations I am having with leading thinkers about AI in education. The first, with Daisy Christodoulou, is here. I would love to know: Who do you want me to speak to next? What questions should I be asking? And how is your school handling all of this?

Let me know. And as always, I could be wrong about all of it.

Craig

🏃🏻‍♂️Before you go, have you… 🏃🏻‍♂️

… checked out my brand-new book series: The Tips for Teachers guides to…

And checked out my all-new, ad-free, Mr Barton Maths website