Stereotypes Learned and Reinforced
刻板印象被學習並固化
Design & Development
Risk Description
When a generative system consistently produces a single ethnic or gender profile when depicting certain occupations; due to unmitigated control gaps, such content is used extensively in marketing and educational materials, and over time reinforces the very stereotypes the organization sought to reduce, drawing formal concern from external groups, 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設計與開發
MIT AI Risk RepositoryDomain 1
Risk Treatment & Implementation Guidance
Deploy stereotype detection examining occupation-demographic associations in generated content; Curate anti-bias training data balancing group representation across contexts; Require diversity checks for marketing and educational uses so homogeneous outputs are adjusted before adoption