AI Safety & Responsible AI
Fairness
The goal of avoiding unjustified differences in AI outcomes across people or groups.
Example
Approval rates are compared across groups to detect unequal outcomes.
Why it matters
It helps teams examine whether benefits and burdens are distributed justifiably.
You'll hear it
“Equal overall scores don't tell us how each group is treated.”
Can you control it?
Not directly
No, not directly. Fairness describes a learned property, observed result, or broader concept rather than a normal user setting. You may influence it indirectly by changing the tool, data, prompt, or workflow.
Common questions
- Does fairness have one universally accepted formula?
- No. Different ideas of fair treatment can lead to different measures.
- Can improving one fairness measure worsen another?
- Yes. Some aims conflict, so tradeoffs need to be understood rather than hidden.
- Does equal treatment always mean equal outcomes?
- No. Starting conditions and the meaning of appropriate treatment can differ.