From Bus Stops to Treasure Caves: Using AI to Develop a Teaching Idea

I went down a bit of a maths rabbit hole after seeing someone on LinkedIn question the UK “bus stop” method for division.

I’ve taught the method plenty of times, and I’ve never really questioned the name. The written layout looks a bit like a bus shelter, so we call it a bus stop. Simple enough.

Except the more I thought about it, the less the actual analogy seemed to make sense.

Take 583 ÷ 7. We write 583 underneath the “shelter” and put 7 outside. But if I try to turn that into an actual bus-stop story, what is happening? Is 583 waiting under the shelter? Is 7 the bus? Are seven people waiting outside? What exactly is getting on?

More importantly, does any of that help a child understand division?

The shape might help them remember where to put the numbers, but the story itself doesn’t seem to connect particularly well with the mathematical understanding they have built beforehand.

From a disconnected procedure to a familiar structure

Before children encounter formal short division, they have already spent years developing ideas around equal groups, sharing, place value and exchanging. They know that one hundred can be exchanged for ten tens, and that one ten can be exchanged for ten ones. They may have represented this physically using Dienes, counters, place-value charts and other manipulatives.

So I started wondering: could the written method feel more like a continuation of that thinking, rather than a completely separate procedure that suddenly appears alongside a bus shelter?

Using AI to develop the thought

I used AI as a space to work through the idea.

At first, that mostly involved generating possible analogies. We tried a packing station, a shop, a resource table, party bags and a few others.

Most sounded reasonable for about thirty seconds. Then I started pulling them apart.

• If we use sweets, why would there be 583 sweets?
• If it is a resource table, why are the resources underneath the table rather than on top of it?
• If there are seven people, does it make sense for each person to receive 83 of whatever we have chosen?
• Does the visual position of each part actually correspond with the mathematics, or have I simply replaced one memorable-but-unhelpful analogy with another?

That back-and-forth was the most useful part of the process. AI could generate possibilities quickly, but I still had to decide whether any of them made mathematical and pedagogical sense.

A treasure cave as a working model

Eventually, I arrived at a working idea I quite like: a treasure cave.

The division symbol becomes the entrance to the cave. The total is inside, while the groups sharing it are outside.

For 583 ÷ 7, you could picture seven pirates outside the cave and, inside, five treasure chests, eight bags and three loose coins.

Open the treasure-cave diagram at full size.

The key is that these do not represent arbitrary objects.

A treasure chest represents 100 coins. Inside each chest are ten bags of ten. Each bag can then be opened into ten individual coins.

That gives the story the same underlying structure as place value: 1 hundred = 10 tens = 100 ones.

If five chests of 100 cannot be shared equally between seven groups, we open the chests. We now have 50 bags of ten, plus the eight bags that were already there: 58 tens.

Each pirate receives eight bags, representing 80 coins, with two bags left. We open those two remaining bags into 20 individual coins and add the three loose coins already in the cave. Each pirate then receives three more coins, with two coins left over. The result is 83 each, remainder 2.

The important mathematical idea has not changed. We are still sharing equally and exchanging between place-value units.

The treasure is simply a visual layer sitting on top of something children may already understand through Dienes. A hundred chest could be replaced by a hundred flat, a ten-bag by a ten rod, and a loose coin by a unit cube without changing the structure of the explanation.

What AI did—and did not—contribute

I am not suggesting that a pirate cave is suddenly the answer to teaching short division. It might turn out not to help at all. There may be much stronger models already being used elsewhere, and I am sure colleagues will immediately spot things I have not considered.

At this point, it is just a working idea. But it is now a much more developed working idea than the vague thought I started with.

That distinction has made me think about how I am increasingly finding AI useful professionally.

The value here was not that I asked AI how to teach division and accepted whatever it produced. In fact, most of the suggestions were discarded.

Instead, it gave me somewhere to externalise an unfinished thought, generate possibilities quickly, challenge them, make the problem more precise and keep iterating until I had something concrete enough to discuss properly.

The next stage is human

I would want to talk the idea through with colleagues, particularly people with different experiences of teaching maths. I would want them to challenge the assumptions behind it, point out misconceptions it could introduce and suggest other representations I have not considered.

Only then would I want to refine it further and try it with children, because ultimately they are the ones who will tell us whether the representation actually helps.

That is probably a more useful way for me to think about AI in education: not as something that replaces professional dialogue or produces the “best” teaching idea, but as something that can help me move from an observation to a proposition worth testing.

“I’ve noticed something that doesn’t quite make sense.”

“Here’s a reasonably developed idea. What do you think?”

Hopefully, that makes the conversation with colleagues more interesting too.

For now, the treasure cave stays firmly in the category of working idea. I’ll see where it goes next.