Understand
In supervised learning an example contains features, and its label is the outcome supplied for training. Suppose fictional fruit cards include colour and shape, with a fruit name as the label. Incorrect or inconsistent labels can teach an unreliable association. Labels describe the task we set; they are not automatically correct.
Original classroom scenario / sample claim
A fictional dataset labels all round red objects as apples, including a red ball.
Try it and check it
- Make eight paper cards: fruits and non-fruit objects.
- Label them using an explicit rule: fruit name or not a fruit.
- Find and correct the red ball label.
- Explain why colour alone is insufficient for identifying an apple.
Without a device: Print this lesson, use paper cards or the supplied fictional text, and discuss the checks in pairs. A student can later enter their own notebook record for teacher review when a device is available.
Notebook: write your corrected or unresolved conclusion, the evidence you actually checked, and a reflection. Do not include passwords, identity documents or private information. Original BrainCheck companion content, BCAI-v1.