Seven things students get wrong about AI, and how to fix each one

When a student says a timer street light is AI, marking it wrong is the least useful thing you can do. The useful question is why they thought so. Almost always, the answer is one of a small set of beliefs that feel completely reasonable from the outside. Fix the belief and dozens of future wrong answers disappear with it.

Here are the seven we see most, in roughly the order they show up. For each: why it sticks, a two-minute activity that fixes it, and a different question to ask a week later to check the fix held.

1. "Automatic means AI"

Why it sticks: AI is sold as the clever thing that does work for you, and so is every automatic door. The fix: ask a calculator and a maps app the same question three times each. The calculator gives the same answer every time; the travel time changes. A fixed rule versus a prediction from data. A week later, ask: is an automatic hand dryer AI?

2. "AI means robots"

Why it sticks: films. The fix: ask students to list the AI they used before reaching school today. Keyboard suggestions, maps, video recommendations, the spam filter, face unlock — not a robot among them. A week later, ask: a factory robot arm welds the same spot on every car, all day. Is that AI?

3. "Data AI just stores data"

Why it sticks: students meet "data" as tables and spreadsheets. The fix: play rock-paper-scissors against a program that keeps track of what you play after each move. Within a few rounds it starts predicting you. The data isn't the point; the pattern and the prediction are. A week later, ask: is a spreadsheet of everyone's exam marks AI?

4. "Computer Vision sees like our eyes"

Why it sticks: the word "vision". The fix: paint a smiley face on a grid of squares and show the same picture as a grid of numbers. That's all the AI ever gets. Then dim the light: every number changes, and the AI becomes less sure. A week later, ask: why might face unlock fail in a dark room?

5. "AI understands meaning like a person"

Why it sticks: chatbots sound fluent. The fix: make a mood-reading AI happy, then flip it with "not", then fool it with sarcasm. Fluent isn't the same as understanding. A week later, ask: how might it read "movie toh bilkul bakwaas thi, but songs were superb"?

6. "AI is always right, and fair"

Why it sticks: computers are supposed to be objective. The fix: a team-picker AI learns from old team lists, but for years only one section was allowed to try out. It keeps picking that section, however well the others play. The AI isn't spiteful; it learnt from one-sided data. A week later, ask: a voice assistant was trained mostly on adult voices. Who might it work worse for?

7. "More data always fixes it"

Why it sticks: it's half true. The fix: give the team-picker ten times more records — all from the same section. It becomes more confident, not fairer. Then switch to records from every section. Better data beats more data. A week later, ask: what kind of data would make the voice assistant fairer?

Making the fix stick

A misconception corrected once usually comes back, because the old idea was comfortable. Four habits make the correction last:

1. Track the idea, not just right or wrong. When a student picks a wrong option, note which misconception it reveals. That's what tells you what to teach next.
2. Re-ask the idea with a new question. Never the identical question, or students memorise the answer. Same idea, new example.
3. Space it out. Later the same day, then a few days later, then a week later. Each correct answer pushes the next review further out.
4. Ask why. On a review question, follow the answer with "why?" and three short reasons to choose from. Then show the student their old mistake next to their new right answer. Seeing yourself improve is wonderfully motivating.

This is exactly how our gamified lessons work: every wrong option is tagged with the misconception behind it, and it comes back as a new question until it's fixed. You can see it in action in the free AI Reflection lesson — get a few wrong on purpose and watch what happens.