Fair and accountable AI
When AI helps make big decisions about people, why fairness and accountability matter, a real example from 2020, and four questions to ask.
AI is starting to help make big decisions about people's lives. Those decisions must be fair. A I isn't only used for fun things like pictures and chat. It's also used to help make big decisions about people's lives. When that happens, being fair really matters. A mistake isn't just a wrong answer. It can change someone's future.
Big decisions AI can affect
A I can help sort job applications, decide who can borrow money, help with marking and school places, and help decide who needs medical care first.
A real example: exam grades in 2020
Here's a real example. In March 2020, because of COVID-19, school exams in England were cancelled. In August, a computer formula was used to decide students' A level grades, using things like each school's past results. Almost four in ten grades came out lower than teachers had predicted, and students at schools with weaker past results were hit hardest. After protests, the government scrapped the formula, and gave students their teachers' grades instead.
New word: Accountability
The lesson? A computer can't be blamed, or say sorry. People must be responsible. This is called accountability. Saying the computer decided is never a good enough answer. Someone must explain the decision, and fix it if it's wrong.
Four questions to ask about an AI decision
Here are four questions to ask about any A I decision. Is it fair to everyone? Can it explain why it made that decision? Does a person check it, especially for big decisions? And can people appeal, and ask for the decision to be looked at again?
Quick check
Quick check. An A I system turns down your loan application. What should you be able to do? Ask why, and ask a person to review the decision. That's what accountability means. Options: Ask why, and ask a person to review it; Nothing. AI is always right.; Delete your account; Shout at your phone.
Recap
A I can affect big decisions about people. In 2020, an unfair grading formula was scrapped after protests. People must be accountable, not computers. Good A I decisions are fair, explained, checked by people, and can be appealed.
Key points
- AI can help make big decisions about people, such as jobs, loans, school places and health care. These decisions must be fair.
- In 2020, when exams in England were cancelled because of COVID-19, a formula lowered almost 4 in 10 A level grades, hitting some schools hardest. After protests, it was scrapped.
- Accountability means a person or organisation is responsible for a decision, and must explain and fix it if it's wrong.
- Ask about any AI decision: is it fair to everyone, can it explain why, does a person check it, and can people appeal?
Try it: Fair or unfair?
Choose the fairest way to use AI in each situation.