Measure local-language performance: What should governance address first?
Aggregate performance can mask local drift; governance needs measurements for the affected population and context.
The question
A travel-support assistant performs well overall after deployment, but support staff report worsening answers for a local-language population. The dashboard shows only aggregate accuracy. What should governance address first?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Increase overall accuracy targets for every travel queryHigher aggregate targets may leave local-language harms hidden and do not directly test the reported population-specific performance change.
- Measure local-language performance ✓Aggregate accuracy can conceal population-specific drift, so suitable local measurements are needed before judging mitigation effectiveness.
- Replace the assistant with a larger modelA model replacement may be premature because the organization lacks localized evidence identifying the cause and scale of degradation.
- Ask staff to report examples without tracking population contextExamples help investigation, but omitting population context prevents reliable comparison of local performance and drift over time.
The trap
When context changes or a subgroup is affected, aggregate metrics are not sufficient evidence. How to remember it
Aggregate performance can mask local drift; governance needs measurements for the affected population and context.
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 How to Govern AI Deployment and Use questions
- Reviewers may rubber-stamp outputs under deadline pressure: Which remaining risk is most specific? →
- The secondary use may create unassessed impacts: What remaining risk should governance prioritize? →
- Revoke record-editing permission and retain a manual: Which mitigation best fits? →
- All 424 Understanding How to Govern AI Deployment and Use questions →
Part of the Certsqill AIGP question bank · Understanding How to Govern AI Deployment and Use ·
Every answer, right and wrong, comes with its own explanation.