AI›Level 1 · AI Explorer›Lesson 16

L1 · 16

Challenge: Train-the-Teacher (Unplugged)

No screens today. You become the model: your teacher shows you labelled examples, you spot the hidden rule, then predict new ones. Can you learn it?

⏱ 1.5 hours🔥 Challenge · unplugged📚 After AI-L1-15✏️ Whiteboard + cards
01

Learning Goals

5 min

By the end of this lesson you can:

  • Act out the train → predict → test loop as a class.
  • Discover a hidden rule from labelled examples.
  • Explain what helped or hurt your "learning".
02

Warm-Up · Guess My Rule

8 min

Last lesson we read confidence scores. Today you are the learner.

Your teacher says: 4 is IN, 7 is OUT, 10 is IN, 3 is OUT. Quietly think — what might the secret rule be?

Reveal one possible rule

"Even numbers are IN." You just learned a rule from labelled examples — exactly what a model does.

03

How the Game Works

12 min

This is a challenge lesson, so here are the rules, then we play.

  1. The teacher has a secret rule (the pattern to learn).
  2. They show labelled examples — each one marked IN or OUT — one at a time. This is the training data.
  3. After a few, the class predicts whether a new example is IN or OUT.
  4. The teacher reveals the answer (the test). Right predictions mean the class learned the pattern.
You are the model

Your brain is doing the training. More varied examples make the hidden rule easier to spot — just like for a real AI.

04

Worked Round · One Played Aloud

13 min

Here is a sample round to show how it feels.

Training examples (shown one by one)

"cat"   -> IN
"dog"   -> OUT
"car"   -> IN
"sun"   -> OUT
"cup"   -> IN

Predict: is "can" IN or OUT? Is "box" IN or OUT?

Reveal the secret rule

The rule was "starts with the letter c". So "can" is IN, "box" is OUT. Notice the rule was about spelling, not meaning — easy to miss at first!

05

Play the Rounds

20 min

Now play three rounds as a class, getting harder each time. After each, talk about what helped you learn.

  1. Round 1 (easy): a number rule (e.g. "more than 5").
  2. Round 2 (medium): a word rule (e.g. "has the letter a").
  3. Round 3 (tricky): a two-part rule (e.g. "an animal AND short word").

After each round, discuss:

  • How many examples did you need before you could predict?
  • Did a confusing example (an exception) slow you down?
  • Were you ever confident but wrong?
Link it back

This is exactly an AI's life: more good examples → faster learning; a wrong label → confusion; high confidence is not the same as correct.

06

Mini-Challenge · Be the Teacher

12 min

Invent your own secret rule. Write five labelled examples (IN / OUT) that point clearly to it. Then "train" a partner and see if they can predict two new ones.

It works if your partner can learn your rule from your examples — which means your examples were clear and varied.

Show a sample puzzle

Rule: "words that rhyme with cat". Examples: hat → IN, dog → OUT, mat → IN, sun → OUT, bat → IN. New ones: rat (IN), pen (OUT).

07

Recap

5 min

You became the model. You learned hidden rules from labelled examples, predicted new ones, and tested your guesses. More clear, varied examples made learning easier — and confidence didn't always mean correct. That's machine learning, no computer needed.

Vocabulary Card

model
The learner — today, that was you and your class.
training
Seeing labelled examples until the rule becomes clear.
test
Predicting a new example to check the rule was really learned.
08

Homework · Make a Rule Puzzle

≤ 20 min

Design a "guess my rule" puzzle for the class. Write your five labelled examples (IN / OUT) and keep the secret rule on a separate line. Make it learnable — clear and varied.