Skip to content
Young Makers: AI for KidsLesson 6 of 11

6 · Fairness & bias (AI learns our mistakes too)

Remember that AI learns from examples — the training data. That's powerful, but it has a catch: AI learns the unfair stuff in the examples, too. If the examples are missing some people, or they show old unfair ideas, the AI can pick up those same problems. Grown-ups call an unfair pattern like this bias (say it: BYE-us). Bias just means leaning one way unfairly, like a seesaw that's stuck on one side.

Here are two ways bias can sneak in:

  • Missing examples. Imagine an AI learned to recognize "a doctor" from photos, but almost all the photos were of men. It might wrongly guess that a woman in a white coat is not a doctor — just because it didn't see enough examples. The AI isn't being mean on purpose; it only learned from what it was shown.
  • Unfair examples. People write all kinds of things on the internet, including unkind or untrue ideas about groups of people. If an AI learns from that, it can repeat those unfair ideas.

So when an AI says something about people — who's good at what, what someone "should" look like, who belongs — be extra careful. That's a moment to stop and think: Is that actually fair and true? Often a grown-up can help you talk it through.

The good news: knowing about bias is the first step to spotting it. UNICEF says AI made for kids should be fair and not leave anyone out, and the NIST safety rules list "fair — with harmful bias managed" as a goal for trustworthy AI. You can help by noticing unfairness and speaking up — because you know that everyone deserves to be treated fairly, even if a computer program hasn't learned that yet.

Think about it. If an AI only saw pictures of one kind of person doing a job, what unfair guess might it make? How could more, fairer examples help?

Sources

Sign in to track your progress