Aggregate data can hide subgroup harm: What specific risk remains?
Aggregate monitoring does not establish effectiveness when subgroup impacts and changing context remain unmeasured.
The question
An employee-assistance chatbot’s mitigation plan tracks usage volume, aggregate satisfaction, and reported complaints after launch. The team claims these measures prove the mitigation works, but employee concerns are not categorized by subgroup and workplace guidance is changing. What specific risk remains?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Changing response volume can distort comparisons between monthly satisfaction results.Volume changes can affect interpretation, but this does not address the stated inability to detect subgroup-specific impacts under changing workplace conditions.
- The team may mistake positive sentiment for reliable performance evidence when complaints are not independently validated.Satisfaction is incomplete evidence, but this option omits the decisive subgroup and changing-context limitations in the monitoring plan.
- Sensitive complaint narratives may require restricted access and retention controls.Complaint privacy may require appropriate handling, but it is not the residual risk concerning whether the mitigation is effective across groups and conditions.
- Aggregate data can hide subgroup harm. ✓Aggregate usage, satisfaction, and complaint counts cannot show whether particular employee groups experience different harms or whether changing guidance alters impacts.
The trap
Check whether monitoring measures outcomes across affected groups and current conditions, rather than relying on aggregate sentiment. How to remember it
Aggregate monitoring does not establish effectiveness when subgroup impacts and changing context remain unmeasured.
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 Development questions
- The disclosure omits operational limitations and supplier: Which disclosure deficiency most directly threatens →
- The supplier update may perform differently: What remaining risk is most specific? →
- The integrated endpoint may expose applicant results: What specific risk remains? →
- All 426 Understanding How to Govern AI Development questions →
Part of the Certsqill AIGP question bank · Understanding How to Govern AI Development ·
Every answer, right and wrong, comes with its own explanation.
