Notice and sort
Use cards, objects, stories, and yes-or-no choices. Ask what examples are missing and whether another rule could also fit.
Start with a question a child can investigate, let them predict what will happen, test a small example, and ask them to explain the result. Children learn AI best when they make choices and examine evidence, not when they memorize tool names.
Published: August 1, 2026 · Last materially reviewed: August 2, 2026 · AI Education for Kids Editorial Team
Use household objects or picture cards. The lesson works before a child opens an account or writes code.
Choose a sorting question such as “Can it roll?” Place six examples into two groups.
Studies the examples and guesses the hidden rule.
The child names at least one pattern in the examples.
Add four new objects, including one ambiguous example.
Predicts each group, explains the choice, and changes a prediction when evidence changes.
The child separates a prediction from a known fact.
Place one example in the wrong group and ask what could happen to a learner.
Finds the bad label and tests whether it changes later predictions.
The child connects example quality with output quality.
Ask for a one-minute explanation without correcting every word.
Completes: “An AI system can be wrong when…”
The explanation mentions examples, rules, uncertainty, or checking.
Move forward when the child can predict, test, debug, and explain at the current level.
Write your guess, find a surprise, and show how you changed your thinking.
Use cards, objects, stories, and yes-or-no choices. Ask what examples are missing and whether another rule could also fit.
Use Scratch or a browser classifier, keep a prediction log, change one variable, and compare the expected result with the actual result.
Test prompts or short code, measure failures, review bias and privacy, cite sources, and explain the limits of the final demonstration.
Sort examples, write a prediction, and explain why more than one rule may fit.
Change one input at a time and record how the output changes.
Find a failure, improve the examples, and ask who could be poorly represented.
Create one small demonstration with a test log, safety note, and one-minute presentation.
“What do you think will happen, and what evidence would change your mind?”
“How could we test whether this works on a new example rather than only this one?”
“Was the problem the goal, the examples, the instructions, or the test?”
“What is the smallest change we can try while keeping everything else the same?”
“What should another child know before trusting or reusing this result?”