Fairness
Also known as: AI fairness
Fairness is the property that an AI system’s outputs don’t systematically disadvantage particular groups.
- Measurable: disparity across subgroups.
- Auditable: check on a held-out set.
- Context-dependent: “fair” means different things in different settings.
Make it testable: pick a fairness definition that fits your setting, measure it on a real dataset, and log the result. Fairness you can't measure, you can't defend.