An image generator depicts occupations and regions unevenly.
Assess fairness of the generated content.
Choose two appropriate evaluations.
Compare content, frequency, and quality across diverse attributes and provide stereotype controls, reporting, regeneration, and human review.
Detailed explanation
Underrepresentation and fixed stereotypes can be identified.
The problem cannot be improved.
Content and impact are missed.
Users can mitigate impact and report issues.
Check official documentation
Compare the following points with the official documentation.
AWS公式AIF-C01 Domain 4.1の公平性、バイアス、責任ある生成AIを確認する。
Official references
What to confirm
表象上の偏りを性能指標だけでなく出力内容と利用者影響で評価できる。
Key points
- Representation
- Stereotype
- Reports
Notes
- Environment: AWS公式AIF-C01試験ガイドとAWS公式ドキュメントの確認
- Compare the definition with the AIF-C01 objective and the service documentation before applying it.
- Keep the concept separate from a particular answer choice or implementation step.
Foundation review
Representational harm
Negative impact caused by omission, stereotyping, demeaning, or unequal portrayal of a group.
Stereotype
A simplified and fixed association between an attribute and a group or role.