What Is Bias Testing?
Bias Testing is the systematic evaluation of an AI system to detect whether it produces different, worse, or unfair outcomes for individuals in different demographic groups.
Bias Testing — the systematic evaluation of an AI system to detect whether it produces different, worse, or unfair outcomes for individuals in different demographic groups.
Bias testing evaluates AI systems for disparate impact — statistical differences in outcomes across groups defined by protected characteristics. Common metrics include: statistical parity (equal positive rate across groups), equalised odds (equal true positive and false positive rates), and calibration (equal predictive accuracy across groups). These metrics can conflict with each other, requiring explicit choices about which fairness definition to apply. NYC LL144 mandated annual bias audits for AEDTs; the EU AI Act's fundamental rights impact assessment process is also partly concerned with bias evaluation.
Source: Chouldechova (2017); NYC LL144; EU AI Act, Article 27
Plain-language explanation
Bias testing evaluates AI systems for disparate impact — statistical differences in outcomes across groups defined by protected characteristics. Common metrics include: statistical parity (equal positive rate across groups), equalised odds (equal true positive and false positive rates), and calibration (equal predictive accuracy across groups). These metrics can conflict with each other, requiring explicit choices about which fairness definition to apply. NYC LL144 mandated annual bias audits for AEDTs; the EU AI Act's fundamental rights impact assessment process is also partly concerned with bias evaluation.
See where you stand on AI governance
Take the free 7-question maturity assessment and get a personalised action plan.
Free assessment — 3 minutes →