Tie the metric choice to the relevant access harms: What evidence should guide approval?
Competing fairness metrics require a context-based choice tied to the relevant harms.
The question
A medical-administration model prioritizes appointment requests without giving clinical advice. Overall error is similar across groups, but one fairness metric shows lower false-negative rates for Group A while another shows closer positive predictive values for Group B. What evidence should guide approval?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Select the metric with the highest overall value after reviewing its statistical confidence.A high aggregate value and confidence interval do not establish that the metric captures the most consequential access harm.
- Tie the metric choice to the relevant access harms. ✓Conflicting metrics should be assessed according to the administrative harms the system may cause, especially unequal access or missed prioritization.
- Choose the metric supported by the largest stakeholder group after collecting preferences.Stakeholder support is relevant input but does not by itself identify which error or access consequence the approval decision must control.
- Use both metrics as evidence of fairness after documenting their different group-level results.Different metrics showing different advantages demonstrate a tradeoff, not proof that the system is fair for the administrative purpose.
The trap
When fairness measures conflict, identify the affected harm before selecting or interpreting a metric. How to remember it
Competing fairness metrics require a context-based choice tied to the relevant harms.
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 security and rare-case safety: Before approval, which evidence best addresses both risks and the →
- Provide engineers feature attributions and leaders: Which evidence plan best serves both audiences? →
- Use supervised learning trained on verified category: Which approach best fits? →
- 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.
