Compare group-specific error harms first: Which conclusion best reflects governance principles?
Conflicting fairness metrics require examining affected groups and error harms; no single metric determines trustworthy treatment.
The question
An employer’s screening model has equal selection rates across two groups, but validation shows materially different false-negative rates. Before deployment, the employer must choose a review focus; no legal fairness metric has been specified. Which conclusion best reflects governance principles?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Optimize whichever fairness metric produces the highest overall accuracy, then document the resulting tradeoff.Overall accuracy does not determine which group harms matter in employment screening and cannot substitute for context-specific fairness analysis.
- Discard the model immediately because any material difference in error rates is unacceptable.The disparity is important evidence for investigation, but the facts do not determine whether redesign, mitigation, or replacement is required.
- Use equal selection rates as the primary fairness basis because they are already comparable.Equal selection rates do not resolve materially different false-negative impacts or establish fairness across this employment context.
- Compare group-specific error harms first. ✓Competing metrics require examining which errors affect which groups and how consequential those harms are before selecting mitigation.
The trap
When metrics conflict, identify which errors harm which groups and why they matter in context. How to remember it
Conflicting fairness metrics require examining affected groups and error harms; no single metric determines trustworthy treatment.
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 safe behavior in unusual conditions as well as: Which evaluation is most appropriate? →
- Provide contextual interpretation to the adjuster: Which information pairing is most appropriate? →
- A labeled validation sample comparing cluster assignments: Before approval, which evidence would most directly →
- 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.
