Help or cheating? Using AI honestly in school

The Class 7 lesson on AI in education opens with two children over the winter break:

Rahat's cousin types the essay question and the AI writes the essay: full marks. Hiba's app gives her a new question every time she's wrong. Only one of them learned.

Both used AI. Only one of them cheated — and the one who cheated mostly cheated himself.

The one test

The books give children a single question to ask, and it works for adults too:

Does it help me think, or think for me?

Helping you think looks like: the meaning of one word you didn't know; more practice after a mistake; speech-to-text when writing by hand is hard; an explanation of step 1 so you can do steps 2 to 10 yourself.

Thinking for you looks like: an AI essay handed in as yours; all twenty homework problems solved; answers you never checked.

The Class 9 book sharpens it into something you can check afterwards: AI helped you learn if you can explain the work unaided. It replaced the learning if the tool did the thinking that was the task. Or, as one card puts it: "Work you can't explain is work you don't have" — in a viva, an interview, or your own head next year.

Say what you used, precisely

When you do use AI, say so. But "I used AI" names nothing. The Class 9 book teaches students to write an AI-use declaration with four parts:

1. Which tool.
2. What it did.
3. What you did.
4. What you checked.

For example: "I asked Gemini to explain the difference between precision and recall. I wrote the answers to questions 1–5 myself. I checked its explanation against Lesson 10.3.4."

The book connects this to something older than computers. Hadith scholars would not let a narrator be vague about how he received a report — ḥaddathanā, "he told us directly", meant something different from balaghanī, "it reached me", and that one word told every reader how much to trust what followed. An AI-use line does the same job: it tells the reader exactly how the work came to be.

And it is not the same as citing a book. A model has no author to cite, so you describe the process instead.

Check it, because it can be confidently wrong

The second rule is just as important as the first. AI tools write in the confident voice of a textbook whether they are right or not. The Class 8 book calls this hallucination — "a confident, fluent statement that is simply false" — and explains that it comes from how the model works, so it can't be switched off. Using AI honestly means checking what it told you before you put your name on it. More on that in Why AI makes things up.

For teachers: design for thinking you can see

Banning AI doesn't work; children have it on the phone in their pocket. What works better is asking for things a machine can't hand in for them:

• Ask one question about their own page two. The Class 9 lesson opens with exactly this: "Rahat got the top mark. Then the teacher asked him one question about his own page two, and he couldn't answer."
• Grade the process, not just the product. Unit projects in the AmanahVale books are marked on a rubric the child sees before starting, so the work has to show how it was done.
• Make the context local. An AI knows little about your village, your school or your class's own survey data. Work built on those is work a child has to do.
• Ask for the declaration. Make the four-part AI-use line normal, and reward honesty rather than punishing it.

Why it matters

The Class 7 lesson ends with a card that says it best: learning is a trust (amanah). Honest work builds knowledge you keep for life. The mark is for today. What you actually learned is what you carry into every exam, interview and job after it — and nobody, human or machine, can do that part for you.