Unit 5 project: The fair-AI check-up

This unit you caught a biased leaf detector, named three kinds of bias, ran a decision desk, rewrote a consent form and signed a class charter. Today the school is thinking about an AI system — and nobody switches it on until your group has given it a check-up.

Work in groups of 3–4. One period. Everything in checkup-<group>.odt.

Each group gets one card — every system and every number invented for teaching:

• The talking library: a voice assistant that finds books when you ask. Its test table: understood 93 of 100 requests in the teacher's voice, 58 of 100 from Class 6 pupils.
• The homework checker: marks English homework automatically. Trained on essays from big-city schools; its test table shows it marks local place names as "spelling mistakes".
• The late-comer camera: counts who arrives after 9:00. Tested in summer; its winter test shows it misses a third of arrivals in fog and snow.
• The club picker: suggests which pupils join the science club, trained on ten years of club lists in which girls rarely appeared.

1. Predict: is this system fair enough to use? Write yes, no or yes-if at the top.
2. Find the bias. Which group does it fail? Which kind — missing-group, historical, measurement? Quote the number from the card that shows it. What's the fix?
3. Human in the loop. List what it decides. Mark each low or high stakes. For each high-stakes one: who checks it, and what might they know that the system doesn't?
4. The notice. Write the notice people would get, answering the six questions — what, why, who, how long, how to say no, and is an AI used, and what does it do? Under 100 words, and no personal data collected that it doesn't need.
5. Your verdict: yes, no or yes-if — with your conditions. Did it match your prediction?

Each person writes one promise for the class charter — "We will…" — that this system taught you. Sign with your roll number. This part is marked for you alone.

Each group reads its bias number and its verdict.