Being Fair with AI

AI Can Be Biased — Here's Why

AI learns from data created by humans. And humans aren't always fair. So sometimes AI picks up unfair patterns from the data and repeats them. This is called bias — and fixing it is one of the most important challenges in AI.

If most photos of doctors in training data show men, AI may assume doctors are men
This can cause real harm: unfair job rejections, wrong medical diagnoses
Fixing it requires diverse data, careful testing, and human review
People who care about fairness AND understand AI are incredibly valuable

The Golden Retriever Judge

Here's a story that shows how unfairness sneaks in. Imagine a robot judge for a dog show, trained to spot dogs — but every single training photo was a golden retriever. Now a chihuahua trots in. "Not a dog," says the robot. A poodle? "Not a dog." The robot isn't mean. Nobody typed "only golden retrievers count." The unfairness snuck in through what was missing from the examples. That's how real AI bias usually happens — not because of a villain, but because of lopsided data. And it matters when the "dog show" is really loan approvals, face recognition, or medical scans. The fix starts with one question: who's missing from the examples?

Bias usually sneaks in through missing examples, not evil intentions
The AI can't be fair to dogs — or people — it never saw
Great question to ask about any AI: "who was missing from its training?"
Diverse examples + careful testing + human review = fairer AI

Prefer slides, quizzes, and saved progress? Read this lesson in the library — free, no sign-up.