Page 8 of 8~116 min topic

Why AI makes mistakes

Your homework trust-with-evidence playbook

Trust is earned with checks. Build a short playbook you can reuse every assignment.

~13 min this pageMastery check

1Try it yourself

Simulation game

Hallucination hunt

Stamp each claim: Trap or Trust. Confident voice ≠ true.

Quiz show

SOUNDS SURE

Sydney is the capital of Australia.

2Learn the idea

Read

The playbook (keep it on a sticky note)

See it

Why fluent answers can still be wrong
01Predict ≠ lookupSounds like an answer
02Web is messyFacts + fanfic mix
03No embarrassmentCan sound sure
04Prompt trapAsked to invent detail

Confidence is a tone — verify before you act

Before you ask

  • Name the job: draft, explain, quiz me, or find search terms — not “be my source.”
  • Decide the risk level: low / medium / high.

While you read the answer

  • Highlight every outside-world claim (names, numbers, dates, quotes, titles).
  • Ask: fluency or evidence?
  • For math: audit steps; for citations: follow the trail; for class content: ground in notes.

Before you paste into homework

  • Run the smallest check that could prove it wrong.
  • Keep AI wording only if you can defend it out loud.
  • If you cannot find evidence, cut the claim or mark it as a question for your teacher.

After you’re done

  • Log one miss you caught (fake title, wrong score, bad step). Your personal mistake map grows.

Read

A Sunday-night run-through

  • English: AI suggests a book title → catalog search → discard if missing.
  • Math: confident steps → re-do hinge step yourself.
  • Science poster: bold statistic → textbook or assigned article only.
  • Caption for the poster photo: low risk → kindness skim.

Same tool, different bars. That is mastery: flexible trust with evidence, not blanket suspicion or blanket faith.

Read

Put the playbook where you will see it

Sticky note on the laptop bezel. Lock-screen reminder. First line in your AI chat saved prompt. The best playbook is the one that interrupts you before paste-into-Google-Docs.

Pair with a friend: swap one claim each night and challenge each other to find the smallest falsifier. Make evidence social. Celebrate catches (“fake title avoided”) the way gamers celebrate clutch plays.

Mastery is not never using AI. Mastery is knowing which jobs earn trust and which claims need receipts. When a teacher asks how you used tools, you should be able to describe your checks without sweating.

Read

Capstone self-test

Without looking, recite: before / while / before-paste / after. If you blank on a stage, rewrite that stage in younger words until it sticks. Your future self on a rushed weeknight will thank you.

Go deeper

Before you start

Why this matters

It’s Sunday night. You have an English paragraph, a math review set, and a science fact to confirm for a poster. AI is open in another tab. You need a personal system so “helpful” does not become “submitted fiction.”

In the wild

See how this idea shows up as a product and a company — then come back to the lesson. Skills transfer across vendors.

Check your understanding

Page assessment

Answer from memory. Completion is saved from this evidence, not from opening the next page.

1. What should you decide *before* you ask AI for homework help?
2. Which parts of an AI answer must be highlighted as claims?
3. What is the “defend it out loud” test?
4. How does logging caught mistakes make next week easier?
5. Give one stop rule you will actually follow on a school night.

All responses are required.