Start here

When an AI helps to decide, it should treat people fairly. How can we tell when it does not?

This is AI Ethics & Bias. It uses what you know about how an AI learns from examples.

Ethics means moral rules for choosing what is right and fair.

If an AI result seems unfair, tell a grown-up you trust.

What you will learn

• Unfair picks. What bias is and where it starts.
• Five kinds. Data, historical, measurement, algorithmic and human.
• Three reported mistakes. What happened and what each one taught.
• A person in charge. Who checks a big decision.

Need a quick recap?

• Training — a program learns a rule from labelled examples. A label is the right answer.
• What it learned from — the examples it was shown.
• Tracing a mistake — follow a wrong answer to its cause.