Children meet systems labelled AI in search, recommendations, games, cameras, and writing tools. A useful lesson gives them questions to ask: What information went in? What did the system produce? How could we test it? What decision still belongs to a person?
Begin with a visible pattern
Students can build simple guessing rules before using a complex model. They choose a pattern, test new examples, and record where the rule fails. This establishes an important fact early: a plausible guess can be wrong for reasons we can investigate.
The AI Guessing Game preview uses that approach. Correct and incorrect results matter equally because the work is evaluation, not applause.
Give the activity a boundary
A first investigation needs one question, a defined tool, and a way to record mistakes. Students might test whether a fictional chatbot follows three character rules or compare how a classifier handles new examples.
Open-ended use makes both evaluation and safety harder. A narrow task lets the group decide what evidence counts and when to stop.
Treat privacy as part of the lesson
Personal details, private conversations, school records, and identifiable images do not belong in external tools used for a class exercise. Tool choice, accounts, retention, permission, and what is sent to a service require adult review in advance.
Families should hear the actual activity—not simply the word AI. A guessing experiment, image classifier, and chatbot test have different inputs and risks. Clear names support useful questions and meaningful consent.
Polish proves very little
A chatbot can write a confident falsehood. An image system can reproduce stereotypes. A prediction can fail because important examples were absent. Students need to collect these cases, verify claims with suitable sources, and record uncertainty.
Human judgment remains responsible for what is believed, published, or acted upon. Fluency is a style of output, not a certificate of truth.
Why AI is listed for later
School of Code marks AI & Smart Machines Coming later. Tool selection, privacy arrangements, classroom boundaries, and the teaching plan have to be settled before a class can open.
The aim is careful literacy: students who can state what a system did usefully, show where it failed, protect information, and keep responsibility with people.
A stop rule belongs in the plan
A written activity plan should name the input students will use, the output they will inspect, and the checks they will perform. It should also say what happens if the tool changes, requires an unexpected login, or starts requesting information outside the task. Stopping is a valid technical decision.
No classroom rule can make an external system perfectly predictable. The practical safeguard is layered: choose a limited service, avoid personal data, supervise use, keep the task narrow, and discuss problematic output rather than quietly discarding it. That gives students a procedure they can understand.
The same procedure needs review whenever a service changes its terms, retention policy, or classroom access.