S0086RS4-T16-S0086-Z · Full risk code

Amplified Discrimination at Feature Intersections

交叉特徵歧視放大

Verification & Validation
Risk Description

When an organization's fairness testing examines gender and age separately, with all metrics within tolerance; due to unmitigated control gaps, in reality, the group possessing both characteristics receives markedly worse treatment, an outcome entirely invisible in single-dimension testing, triggering compliance exposure and operational reputational costs.

Framework Mappings

EU AI ActArt.10
NIST AI 600-1Harmful Bias and Homogenization
NIST AI RMFMEASURE 2.11
ISO/IEC TR 24027
ISO/IEC 5338驗證與確效
MIT AI Risk RepositoryDomain 1

Risk Treatment & Implementation Guidance

Test intersectional fairness, examining combined protected-attribute groups; Use multidimensional bias analysis to reveal compounded disadvantage invisible to single-axis testing; Augment test data for sparse intersectional groups to preserve detection power