A bank's loan-approval model is producing different acceptance rates for applicants of similar creditworthiness, and the team uses a tool to detect and measure this bias. Which responsible AI principle is being upheld, and what kind of tool helps?
- AFairness, using the Responsible AI Dashboard to detect bias
- BPrivacy and security, using disk encryption
- CReliability and safety, using load balancing
- DInclusiveness, using a content delivery network to cache pages
Detecting and measuring uneven outcomes for similar applicants is a fairness concern, and the Responsible AI Dashboard is the tooling that helps identify bias in data and improve model behavior. Encryption, load balancing, and content delivery networks address other concerns.
Why each option is right or wrong.
- A (correct). Fairness is the principle at stake, and the Responsible AI Dashboard helps identify and measure bias in datasets and models.
- B. Privacy and security and disk encryption address data protection, not biased acceptance rates.
- C. Reliability and safety and load balancing address uptime and safe operation, not bias.
- D. Inclusiveness and a content delivery network do not address measuring model bias.
Reference: fairness and the Responsible AI Dashboard; AI-901 Domain 1 (Identify AI concepts and capabilities).