Machine learning projects for kids: 8 free ideas by age

New to the idea? Read machine learning for kids, explained simply first.

Projects at a glance

Unplugged projects (ages 7+)

1. Sort it like a computer. One person plays the "model" and learns to sort objects only from examples the others show. Then the group gives it a few confusing examples and watches the sorting go wrong. Follow our free sort it like a computer activity.

2. Decision tree guessing game. Build a yes/no tree that guesses an animal, then test it on animals it has never seen. Follow our free decision tree guessing game.

Browser projects (ages 7 to 12)

3. Quick, Draw! challenge. Quick, Draw! is Google's free game where an AI guesses your doodle in 20 seconds. Keep a score sheet: which drawings did it get, and which fooled it? Then ask why: it learned from millions of drawings by other people, so unusual drawings confuse it.

4. Rock, paper, scissors detector. In Teachable Machine, create an Image Project with three classes (rock, paper, scissors) and record about 30 webcam pictures of each hand shape. Train it, then test. Try it with a different person's hand: does it still work?

5. Clap or whistle sound sorter. Teachable Machine also has an Audio Project. Record background noise, claps and whistles as three classes, train, then test in a quiet and a noisy room.

6. Recycling sorter. Train a model to tell plastic bottles from paper using photos of only one brand of bottle. Then show it a different bottle. This is the clearest way to show biased training data: a model only knows the examples it was given.

A parent should be nearby for webcam projects. Teachable Machine runs in the browser, and children do not need an account.

Scratch project (ages 10+)

7. A smart Scratch game. The free website Machine Learning for Kids lets children train a model (for example, to recognise kind or unkind messages, or pictures) and then use it inside a Scratch game. It is one of the best bridges from playing with AI to building with it.

Python project (ages 13+)

8. Flower classifier. Teenagers who know some Python can use the classic iris flower dataset and the scikit-learn library to train a model that identifies a flower from four measurements. They split the data into training and testing sets and check the accuracy: exactly what real data scientists do, at small scale. Start with Python for kids: what to learn first.

How to talk about each project

After every project, ask the same three questions:

1. What did the model learn from? (the examples)
2. When did it get things wrong, and why? (missing or biased examples)
3. How could we make it better? (more, better and more varied examples)

These three questions are the heart of AI literacy for kids.

Our technology courses for kids: Fun with Coding (ages 10 to 16), Learn Python (ages 12 to 18) and Artificial Intelligence for Kids (ages 12 to 16).

Frequently asked questions

What is a good first machine learning project for kids?
For ages 7 to 9, an unplugged sorting game or Google's Quick, Draw!. For ages 9 to 12, a rock, paper, scissors detector in Teachable Machine is the best first project because it takes about 30 minutes and needs no coding.

Do kids need to know coding for machine learning projects?
No. Unplugged games, Quick, Draw! and Teachable Machine need no coding. Scratch and Python projects come later, once the ideas make sense.

Is Teachable Machine free?
Yes. Teachable Machine is a free Google tool that runs in a web browser. Children can use it without an account, ideally with a parent nearby for webcam projects.

What is the Machine Learning for Kids website?
Machine Learning for Kids (machinelearningforkids.co.uk) is a free website that lets children train models and use them in Scratch games and projects. It is a separate site from Learning Bubble.

What age can kids learn machine learning?
The core idea, learning from examples, can be taught from about age 7 with games. Training real models in Teachable Machine suits ages 9+, and Python machine learning suits ages 13+.

Written by the Learning Bubble teaching team. If something here does not match your situation, ask us. We would rather give you a straight answer than have you guess.