Overview
Students act out a small neural network on a paper board, using coins as the messages between neurons, and use it to play Coin Snap. The board shows that a network of very simple units, each following one rule, can produce behavior that looks smart. The activity is a two-page PDF from cs4fn with the board and a short guide.
How the network works
The board has seven neurons in three layers:
| Neurons | Count | Rule |
|---|---|---|
| Sensory | 4 | Each watches one coin and fires on heads or on tails (Coin 1 heads, Coin 1 tails, Coin 2 heads, Coin 2 tails). |
| Relay | 2 | Each fires when it gets 2 messages: one combines the two “heads” neurons, the other the two “tails” neurons. |
| Motor | 1 | Fires on 1 message, which means shouting “Snap!” |
The relay neurons work like AND (both heads, or both tails), and the motor neuron works like OR (either relay fired). Together the network fires only when the coins match.
| Coin 1 | Coin 2 | Heads relay | Tails relay | Snap? |
|---|---|---|---|---|
| Heads | Heads | Fires | — | Yes |
| Tails | Tails | — | Fires | Yes |
| Heads | Tails | — | — | No |
| Tails | Heads | — | — | No |
Running the activity
- Print page 1 (the board) for each player. The PDF is A4, so print it with “Fit to page” on Letter paper. Each player needs 6 small coins, and the referee needs 2 large ones.
- Have students first play a round of Coin Snap with their own brains, then play it with the board.
- In groups of three, two players race each other’s boards while the third is the referee.
Debrief
- The network plays perfectly, but nothing in it understands the game. Where does the “smartness” come from?
- The threshold is the number of messages a neuron needs before it fires. What happens to the network’s behavior if a relay neuron’s threshold drops to 1?
- This network was designed by people. Real neural networks learn their own connection strengths, called weights, from training data. How might this one learn?
Notes on the PDF
- The guide says “neutron” once where it means “neuron.”
- It says the yellow circles are the positions the “motor neurons” focus on. It means the sensory neurons.
Going further
- The Brain-in-a-bag activity from Teaching London Computing has students act as the neurons themselves to play Red-black Snap.
- cs4fn’s Build a brain to play snap page has more on the idea.
- Learning about Machine Learning from Teaching London Computing collects related activities.