Measure false-negative rates separately for each group: Which additional control most directly investigates
The identified fairness concern requires comparing the relevant error rate across groups.
The question
A university uses AI to prioritize research-administration applications. Overall approval accuracy is stable, but applicants in Group A have a substantially higher false-negative rate than applicants in Group B. Assume privacy, security, reliability, transparency, documentation, and human-review controls are adequate. Which additional control most directly investigates the identified concern?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Limit staff access to application recordsAccess limits support privacy and security, but do not examine why false negatives differ between the groups.
- Measure false-negative rates separately for each group ✓Disaggregated false-negative measurement directly quantifies the reported group disparity and supports investigation of harmful bias.
- Send applicants more detailed decision noticesDetailed notices improve communication and possibly explainability, but they do not measure or investigate unequal error rates.
- Optimize the single overall accuracy scoreImproving aggregate accuracy could leave the disparity unchanged or make it worse, so it does not directly investigate the identified concern.
The trap
When the evidence names unequal errors between groups, select group-level error analysis rather than an aggregate score. How to remember it
The identified fairness concern requires comparing the relevant error rate across groups.
How many of these would you get right?
One of 1581 AIGP questions on Certsqill. Take a free five-minute check and see your score per domain — not one number, but which section to open tonight.
Test your AIGP readiness — freeMore Understanding the Foundations of AI Governance questions
- Test hazards and block unsafe sequences before execution: Which additional control most directly addresses the →
- Give the customer a plain-language reason: Which additional control is most direct for the customer-facing →
- Model A is supervised: Which classification follows from these training methods? →
- All 337 Understanding the Foundations of AI Governance questions →
Part of the Certsqill AIGP question bank · Understanding the Foundations of AI Governance ·
Every answer, right and wrong, comes with its own explanation.
