Addaly is in open beta. Things will change, and AI answers can be wrong — check anything that matters.

Explaining AI to a class that argues back

AI for Teachers · lesson 1 of 8 · 7 min

The one sentence to start with

"It is a machine that has read an enormous amount of writing by people, and it guesses what words come next."

That is the whole thing. Everything else is elaboration, and this sentence survives the questions children actually ask. "It's a robot brain" does not survive. "It's like a search engine" does not survive — the first follow-up is "then why is it wrong sometimes?" and you have no answer.

Make them do it before you explain it

Write on the board:

The old woman walked slowly into the ______

Every student writes one word. Silently, no discussion. Then take a tally by show of hands. Room. House. Shop. Kitchen. Somebody will write "sea" and the class will laugh.

Put the tally on the board. You have just built the thing you are teaching: twenty-eight people, four common answers, one rare one. The machine has that same shape of guess, except it built it from billions of sentences instead of one classroom.

Now the second stem:

The capital of Japan is ______

Nearly unanimous. Tokyo.

Ask them what changed. They will get there: for one question almost everyone's guess is the same, because the writing they have seen nearly always says the same thing. For the other, many guesses are fine. That is the entire reason AI is dependable on some questions and unreliable on others, and a ten-year-old can now explain it to their parents.

The questions they will ask, and honest answers

"Does it know it's me?" No. Usually it starts each conversation with nothing about you. Some products keep notes about you; that is a feature someone built on top, not the machine remembering.

"Is it alive? Does it have feelings?" It writes sentences about feelings because people wrote sentences about feelings and it learned from them. Whether anything is going on inside is genuinely argued about by the people who build these systems. You are allowed to say that. Try: "Nobody has shown it feels anything, and most researchers think it doesn't. But it is a real argument, and you are old enough to know it's an argument rather than a settled fact."

"Why does it lie?" Lying means knowing the truth and saying something else. It does not do that. It produces the sentence that fits, and a sentence can fit beautifully and be false. It has no way to check. Give them the phrase: it is not lying, it is making things up without knowing it.

"Where did it get all the words?" From writing by people — websites, books, articles, forums. That is also why it repeats what people say, including the unfair things people say.

"Can it do maths?" Sometimes, badly, unless a calculator is running underneath. Guessing the next word is a poor way to multiply.

Two analogies to avoid

"It's a robot brain." This invites every film they have seen into the room and you spend the period on robots instead of prediction.

"It's just autocomplete." Closer, and useful, but if you stop there a student will show you something it wrote that is genuinely good and conclude you were wrong. Say instead: "It works like autocomplete, and it turns out autocomplete over enough writing is far more powerful than anyone expected — including the people who built it."

The line worth ending on

"It is very good at sounding right. Sounding right and being right are different things, and telling them apart is now part of your job."

Eleven-year-olds understand that sentence immediately, and it carries the rest of the course.

Before you move on

A student tells the AI it made a mistake. It apologises and changes its answer. She says this proves it understood its error. What should you tell her?

Pick the one you would defend. Nobody sees your answer.

No ads. No data sale. No public scores on people. Ever.

© 2026 Addaly