AI and Being Fair to Everyone

AI learns from huge piles of info from people. If that info is unfair, AI can be unfair too. Knowing this helps you spot it.

Some examples

• AI makes art with only one type of person.
• AI guesses jobs by gender — that's unfair.
• AI does worse with some accents.
• AI sometimes leaves out languages.

Try it!

Ask an AI to draw 'a doctor'. Look at who shows up. Talk about it with a grown-up.

Fairness is something AI has to learn — and so do we

If your class voted on a pizza topping but only asked five of the thirty students, that vote wouldn't be fair — it wouldn't represent everyone. AI can have the same problem. It learns from huge piles of writing and images, but if those piles mostly included one kind of person or one kind of story, the AI might act like that's 'normal' and leave others out. This is called bias. You might notice bias when AI only shows doctors as men, or when voice recognition works better for some accents than others, or when AI makes art that ignores whole groups of people. Bias in AI can feel small, but it adds up. When some people are left out or shown unfairly over and over, it can affect how they see themselves and how others see them. Noticing unfairness in AI — and saying so — is something YOU can do, even as a kid.

• Ask an AI image tool to draw 'a teacher' and see who shows up — then think about what that says.
• If AI doesn't understand your accent or a friend's, that's a fairness problem worth knowing about.
• If you notice AI leaving out a group of people, point it out to a trusted adult.
• Reporting unfairness — even in technology — is a form of standing up for others.