AI Discoverers · Ages 7–9
Teach the Robot
Can examples help a machine learn?
01 · Read the story
Meet Mina and Sami
A fruit helper sees only red apples and yellow bananas. Then Mina shows it a green apple. Is color enough to tell which fruit it is?
02 · Discover the idea
What can we learn?
A model can learn patterns from examples called training data. If all apples in the examples are red, a model might rely on color and make mistakes with green apples.
About this activity: The fruit experiment uses a small, transparent example-matching rule, not a trained neural network. The helper can be unsure or mistaken.
03 · Train, test, improve
Teach a fruit helper with examples.
Choose which clues the helper uses. Test a fruit it has not seen. Add examples and see what changes.
Training examples
These are the fruits the helper has seen.
- Red, round apple
- Yellow, long banana
2 examples
Test an example
Can the helper guess a different fruit correctly?
No prediction yet.
Choose a fruit to test.
Try changing the clues.
Begin with a green apple and color-only clues.
Important: This is a teaching simulation using an explicit example-matching rule. It does not train a real AI model. After adding an example, test a different example to see whether the rule works beyond those it has already seen.
04 · Explain and review
Choose, check, and explain.
There is no timer. If you make a mistake, learn from the explanation and try another question.
You explored all three questions! ✨
05 · Try it together
Family activity
Sort fruits or paper drawings by color and shape. Deliberately add an unusual example. Discuss what clues are useful.
06 · Reflect
Talk about it
How could you make a training collection more varied and fair?
No answers are collected or sent anywhere. An adult can guide the conversation.