The big picture
An AI learns from people's data and is built by people. Four ideas follow from that.
| Idea | Remember |
|---|---|
| Unfair picks | Bias is an unfair preference for or against someone. An AI can learn it from data, design or its makers. |
| Five kinds | Bias has five common kinds: data, historical, measurement, algorithmic and human. |
| Three reported mistakes | Check who is in the data, what it copies, what it measures. |
| A person in charge | For big decisions, a person checks the answer and decides. The makers and users of a program answer for what it does. |
Bias is not always planned. No one can promise it is gone. So the checking continues while a program is used.