Machine learning sounds complicated, but the core idea is simple: a computer sees many examples and learns patterns from them. In this activity, your child is the computer.

  • Ages: 7–12
  • Time: 25 minutes
  • Skill: How machine learning learns from examples

What you need

  • About 15 small objects (fruit, toys, stationery, cutlery)
  • Two bowls or boxes
  • Paper and a pencil

Step by step

  1. Choose a secret rule. The parent secretly picks a rule, such as "things that are round" or "things you can eat".
  2. Show examples. Put a few objects in the "yes" bowl and a few in the "no" bowl, without saying the rule.
  3. Let the child predict. Hand over a new object. Your child guesses which bowl it belongs in. That is a prediction.
  4. Reveal and repeat. Say if they were right. After several objects, ask them to guess the rule.
  5. Break it on purpose. Now use misleading examples, such as only red round things in "yes". Watch how the predictions go wrong.
  6. Talk about it. Ask: how is this like AI? What happens when the examples are unbalanced?

The big idea

AI learns from examples (training data), finds a pattern (a model) and makes predictions about new things. When the examples are unbalanced, the predictions go wrong, which is exactly how real AI ends up biased or mistaken.

Make it easier or harder

  • Easier: use very obvious rules, like colours.
  • Harder: use two-part rules, such as "round and you can eat it".

Read more in how machine learning works, explained for kids.

This activity is a taste of our free AI Explorers workshop (ages 7–10). Older children can take Artificial Intelligence for Kids.

Frequently asked questions

What does this activity teach about AI?
That AI learns patterns from examples and makes predictions, and that unbalanced examples lead to wrong or biased predictions.

Does this need a computer?
No. It is an unplugged activity using everyday objects.

More free activities for kids