Why AI makes mistakes
A kid’s map of AI mistakes
Mistakes come in different flavors. Naming the type helps you choose the right check.
1Try it yourself
Simulation game
Hallucination hunt
Stamp each claim: Trap or Trust. Confident voice ≠ true.
Quiz show
“Sydney is the capital of Australia.”
2Learn the idea
Read
Four mistake types you’ll meet
See it
Confidence is a tone — verify before you act
Use this map when something feels off:
- Invented facts — Names, titles, scores, quotes, or dates that never happened (fake book, fake citation).
- Wrong specifics — Real topic, wrong details (correct team, wrong score).
- Out-of-date info — Used to be true; isn’t now (old coach, old schedule).
- Wrong fit — True somewhere else, not for your context (another school’s rules pasted onto yours).
Sports stats are a perfect training ground because numbers feel official. AI can spit out a box score style line without ever “watching” the game. If the claim is measurable, demand a measurable check: league page, school athletic site, photo of the scoreboard, or a trusted reporter.
Read
Label before you argue
When friends say “AI is dumb” or “AI is always right,” pause and label:
- Was it inventing?
- Mixing two similar things?
- Sounding confident about something it can’t know (tonight’s private practice notes)?
Labeling stops you from overreacting. One bad score does not mean AI can’t help outline your paragraph. One good outline does not mean you can trust tomorrow’s injury report from the same chat.
Read
Build a personal map
Keep a tiny note on your phone titled “AI fails I’ve seen”:
- Invented homework source
- Wrong sports number
- Mixed-up character names from a book you’re reading
- Schedule that ignored a holiday
After a week, patterns appear. Your map becomes a checklist: for stats, open the official page; for books, search the catalog; for school rules, ask a teacher or check the handbook.
Read
How the map helps under time pressure
Sunday-night homework is when maps matter. You will not run a full investigation for every sentence. You will glance at a claim and pick a bucket: invented, wrong specific, out-of-date, or wrong fit. That glance tells you which check is cheapest.
Wrong sports stats usually land in “wrong specific” or “invented.” Wrong fit shows up when AI applies another country’s school rules to yours, or mixes two athletes with similar names. Out-of-date shows up when last season’s roster is treated like today’s.
Share the map with a friend who “trusts AI for everything.” You do not need to win an argument. You only need a shared vocabulary: “That looks invented — let’s check the athletics page.” Shared labels make verification social instead of lonely.
Read
Mini drill
Pick one AI answer from your week. Write the mistake type in the margin of your notebook. Write the check. If you cannot name a check, the claim does not belong in anything graded or public.
Go deeper
Before you start
Why this matters
For a sports blog post, AI says your school’s basketball team “won the regional final 78–61 last Saturday.” You were at the game. The score was 62–59, and it was a semifinal. Wrong number, wrong round — but the sentence sounded like a real recap.
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.
Related lessons
Check your understanding
Page assessment
Answer from memory. Completion is saved from this evidence, not from opening the next page.
All responses are required.