AI recommends the priority of insurance claims.
An error could cause significant harm to an applicant.
Which governance design is appropriate?
Keep a human decision maker, provide evidence and limitations, and offer logging, appeal, correction, and shutdown procedures.
Detailed explanation
Confidence does not eliminate error or disparate impact.
Without records, audit and appeal are impossible.
High-impact AI output should not be the sole final decision; users need recourse.
Transparency and informed recourse are lost.
Check official documentation
Compare the following points with the official documentation.
AWS公式AIF-C01ガイドのDomain 5.2で責任ある利用、監査、コンプライアンスを確認する。
Official references
What to confirm
AIを最終決定者にせず、人間の監督と救済手段を置く理由を説明できる。
Key points
- High impact
- Human oversight
- Recourse
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
Human-in-the-loop
Confidence from a model does not by itself replace accountability for a consequential decision.
User impact
The control design should account for how different people experience and contest an automated result.