Exam-style
A review board asks how the team will detect whether their assistant gives noticeably weaker answers to some customer groups than others, and whether it repeats stereotypes about them. Which responsible-AI dimension is that, and which control carries it?
Reveal the answer
D. Fairness, carried by prompt-stereotyping evaluation and accuracy scored separately for each customer group
Quality gaps between groups and stereotyped output belong to the fairness dimension. Stereotyping is measured for you: prompt stereotyping is one of the fmeval evaluation dimensions, and Builtin.Stereotyping is one of the built-in metrics in a judge-based Bedrock evaluation job. The quality gap is not, because AWS publishes no per-group fairness metric for a generative model, so it means scoring accuracy on your own dataset split by group and comparing the scores. The other dimensions cover different concerns: safety is harmful content, controllability is steering and human oversight, veracity is grounding answers in something true, and transparency is disclosing how the system works.
Q. AWS names its responsible-AI dimensions. Roughly, what are they?
A. Fairness, explainability, privacy and security, safety, controllability, veracity and robustness, governance, and transparency. Each has its own controls.
| Dimension | Controls |
|---|---|
| Fairness | Prompt stereotyping, in fmeval or a judge-based evaluation; accuracy per group |
| Explainability | Citations and the retrieved passages; Model Cards |
| Privacy and security | Guardrails sensitive information filters, blocking or masking; IAM, KMS, PrivateLink |
| Safety | Guardrails content filters and denied topics |
| Controllability | Human review before consequential output; feedback loops |
| Veracity and robustness | Contextual grounding checks; Knowledge Bases citations; evaluation jobs |
| Governance | Model Cards, Model Dashboard, invocation logging, CloudTrail |
| Transparency | AWS AI Service Cards; in-app disclosure |
Clarify and A2I once carried rows one and five. AWS put both into maintenance on 30 June 2026, closing them to new customers. Existing deployments keep running. Clarify’s evaluation engine is available on its own as the fmeval library.
An automatic Bedrock evaluation job scores accuracy, robustness and toxicity. A judge-based job scores Stereotyping and Harmfulness too.
Why? Naming the concern narrows what you have to build.