Loading…
Glossary term
The goal that an AI system treats people and groups equitably.
Fairness in AI means that a system does not treat people or groups unjustly. There are several mathematical definitions, such as equal error rates across groups, and they can conflict, so choices must be made for each use.
Fairness work includes better data, testing and adjusting models or thresholds.
There is no single definition, so fairness requires deciding what is right in context, then measuring it.