Choose training examples
Examples include measurable features such as temperature, debris, algae and water clarity, plus a ranger-reviewed label.
Machine learning for kids · Ages 9–13
Teach with examples.
Test what the model learned.
Machine learning becomes understandable when children can change the evidence and see the result. This beginner guide leads directly into a free classification game—no account or prior coding required.
TRANSPARENT MODEL
Nearest-neighbor classification
EXAMPLES→COMPARE→PREDICT→REVIEW
Every prediction names the closest training example, so the learner can inspect the model's evidence.
BEST FORAges 9–13Beginner AI learners
TIMEAbout 5–10 minutesGuide plus activity
CODINGNot requiredExamples before syntax
ACCESSFree in a browserNo child account
A simple mental model
What machine learning means
in this activity
The learner is not asked to trust a black box. Reef Rescue uses one inspectable comparison rule from start to finish.
Classify new cases
The model finds the most similar example and carries its safe-or-risk label to an unfamiliar reef zone.
Inspect the mistakes
A false alarm closes a healthy habitat; a missed risk leaves damage untreated. The kinds of error matter.
Improve the evidence
The learner adds missing edge cases, reruns the same test and explains why the second model performs differently.
Free machine-learning game
Play Reef Rescue.
Train, test and revise.
The first training set is intentionally incomplete. A hidden heat risk, natural seaweed and a harmless sand plume reveal why varied examples matter.
- Real deterministic classifier, not generated scoring
- Visible closest-example explanation
- False alarms and missed risks separated
- Human-review reflection before completion
ONE TEST CASE
Cloudy sand plume
Incomplete model: compares it with polluted runoff and raises a false alarm.
Improved model: sees a reviewed sand-current example and keeps the healthy zone open.
The world and test case stay fixed. Only the training evidence changes.From blocks to code
Where Scratch-style ideas
and Python fit next
A learner can understand examples and classification before writing code. Later, visual rules and Python make data structures, distance functions and evaluation steps more precise.
Machine learning, clearly explained
Questions from families and beginners
What is machine learning for kids?
Machine learning for kids means introducing how computers use examples to make predictions. A child can label examples, test unfamiliar cases, find mistakes and improve the evidence without needing advanced mathematics.
Can children try a machine-learning activity for free?
Yes. Wovi's Reef Rescue is a free browser activity for ages 9–13. It needs no account and guides learners through training examples, classification, false alarms, missed cases and human review.
Does a child need Scratch or Python first?
No. Example-based classification can be explored before formal coding. Scratch-style blocks and Python become useful later when a learner wants to express the same logic more precisely.
Is this a real machine-learning model?
Reef Rescue uses a transparent nearest-neighbor classifier. It is deliberately small so learners can inspect which training example influenced each prediction rather than treating the model as magic.
What should a child be able to explain afterward?
They should be able to explain that a model learns comparisons from examples, missing examples create blind spots, accuracy alone can hide different kinds of mistakes and some decisions need human review.
What machine-learning project can a child complete?
Reef Rescue is a short beginner project: choose labeled training examples, test four unfamiliar cases, compare false alarms with missed risks and decide when a human should review the model.
Founding families
Want the next free machine-learning mission?
Join the adult early-access list for new playable worlds and parent-readable learning updates.
Join early access
Free to join · Parent email only · Unsubscribe anytime