Which statement best describes fairness in bias testing?

• A. To quantify disparate impact across groups and guide mitigation.
• B. To maximize overall accuracy without regard to fairness.
• C. To reduce dataset size.
• D. To speed up training.

Answer: (A)

Fairness testing centers on identifying and measuring differences in model outcomes across different groups and then using what you find to reduce those unfair disparities. In practice, you look for disparities in error rates or favorable outcomes between protected groups, quantify how large those gaps are, and then apply mitigation steps to bring outcomes closer to equal opportunity or calibration across groups. This makes fairness something you actively measure and remediate, not just something you chase through overall accuracy. The other options miss the fairness piece: they focus on overall performance, data size, or speed, rather than on ensuring that different groups are treated equitably.