In AI governance, which indicators measure risk related to biased outcomes?

• A. Key Risk Indicators (KRIs) focusing on disparity and fairness metrics.
• B. Key Performance Indicators (KPIs) focusing on accuracy alone.
• C. Financial indicators.
• D. Network throughput indicators.

Answer: (A)

In AI governance, monitoring potential biased outcomes relies on Key Risk Indicators that track disparity and fairness metrics. These indicators are designed to signal when the risk of harming certain groups is rising, so governance teams can take corrective actions before harms escalate. They look beyond overall accuracy and examine how performance varies across protected attributes and subgroups, helping detect unfair effects.

By monitoring metrics such as demographic parity, equalized odds, and calibration within groups, organizations can spot biased outcomes, set actionable thresholds, and initiate audits or mitigations. This approach ties directly to risk management—identifying, measuring, and controlling the potential for harm or regulatory exposure.

Focusing on accuracy alone as a KPI can hide bias, because a model can be very accurate overall yet perform poorly for specific groups. Financial indicators assess monetary risk and don’t address fairness concerns, and network throughput indicators measure system capacity rather than bias in outcomes.