At the Whiteboard and on the Table: Two Ways to Teach Neural Networks to Year 9s

Last Wednesday, fifth period, the classroom at the far end of Building B. The one with the view of the hills toward Kings Park — that's where I taught the same content twice. Once with chalk and words, once with a deck of cards and hands.

Both classes were Year 9, and for both it was a 30-minute "What is AI?" session I'd voluntarily wedged in after the landforms unit. It's not in the curriculum. I'm doing it because I want to. If anyone asks why a geography teacher is teaching neural networks, my answer is that AI is already inside geography — whether it's landform data analysis or satellite image classification.

---

Explaining Through Analogy: At the Whiteboard

With the first class (9B), I used the analogy I always reach for. When I explain a neural network, I always bring up the story of the signs at Perth Central Station.

"Say someone's visiting the station for the first time. They don't know whether it's Platform 1 or Platform 2, so they just start walking. There are signs, there's the direction people are moving, there are sounds. If they go the wrong way — they miss the train. And next time, they don't go that way again. That's exactly what it means for a neural network to learn. After getting it wrong, it shifts direction a little at a time."

On the board I drew three columns of circles. Input layer, hidden layer, output layer. Drew the arrows. Called the weights "how clear the signs are." Students nodded. But I didn't quite trust those nods. I couldn't tell whether they understood or were just going along with it.

At the end of the lesson I threw out two questions.

• "Explain what a weight is in one sentence."
• "How does the network know when it's wrong?"

Four out of six students answered the first question by giving my words straight back to me. "How clear the signs are." The second question hung in silence for a long time.

---

Doing It by Hand: At the Tables

For the second class (9C), I used cards. A pack of index cards from home, a few sheets of A4, and some coloured stickers. Total materials cost was probably under $2 — I already had the cards, and the stickers were left over from a sale at the stationery shop.

The structure of the activity was simple enough. I drew rough animal pictures on the cards — I'm a terrible artist, and the students laughed, saying they couldn't tell if it was a dog or a cat. That actually worked in my favour. I stuck small notes on each card listing features: things like "pointy ears," "has whiskers," "thick tail."

Three students became "nodes." Each received a rule card to place in front of them. The first student looked only at ear shape. The second looked only at whether there were whiskers. The third combined the judgements of the first two and called out "dog" or "cat."

It was a mess at first. Joel, who was in the third seat, kept getting it wrong. So I tweaked the rule card slightly. "Pointy ears = +2 points, whiskers = +1 point, total 2 or more = cat." After two more rounds, the wrong answers dropped.

That's when I stopped and asked. "What did I just change?"

Priya raised her hand. "The scores. Which one matters more."

"That's a weight."

The classroom felt different in that moment. Not a nod — Joel muttered to himself, almost under his breath: "Oh — so after you get it wrong, you adjust the scores."

---

How Different Were the Two Approaches in Practice

I put the same questions to 9C.

"Explain what a weight is in one sentence."

"How much importance I give to the ears." "The number you adjust when you're wrong." "How you distribute the points."

Not one student gave my words back to me. They were drawing from their own experience.

For the second question — "How does the network know when it's wrong?" — Joel answered straight away. "You compare it to the right answer. We did that, didn't we."

In 9B, that question had hovered in the air for nearly 30 seconds.

I'm not saying to give up on analogies. The Perth Central Station story is still useful — for grabbing attention in the first 30 seconds. But if you try to carry a full 30-minute lesson on a single analogy, students remember the analogy, not the content. When the 9B students answered "how clear the signs are," they hadn't understood the idea — they'd stored my words.

The card activity had its downsides too. The seating arrangement got complicated, and because 9C is such a quiet class, it felt awkward at first getting students to take on roles. Without someone like Joel, the activity might have dragged. And honestly, it takes more prep time — I spent about 20 minutes the evening before making the cards.

Even so, next time I'll reach for the cards first. The analogy can come after, as a way to fill in the gaps that the activity opens up.

I'm planning to go back to 9B next week with the cards. Understanding that arrives late is still understanding.