Limit adaptation and require review before suspensions: Which design choice best fits both constraints?
Limit adaptation and require review because novelty and autonomous adverse action jointly increase exposure.
The question
A public-benefit agency considers an adaptive system that can automatically suspend payments when new patterns appear. Evidence covers historical cases but not novel household circumstances. The agency requires uninterrupted human authority over adverse actions and cannot tolerate unexplained suspensions. Which design choice best fits both constraints?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Limit adaptation and require review before suspensions. ✓Constraining adaptation reduces novel behavior, while pre-action review preserves authority and enables case-specific scrutiny.
- Permit autonomous suspension with broader historical validation.More historical validation cannot resolve the stated absence of novel-case evidence or the requirement for human authority.
- Allow adaptation while publishing general system documentation.General documentation may improve transparency but does not prevent unexplained suspensions or preserve required human authority.
- Use fixed rules with periodic manual updates.Fixed rules preserve human control but may not address changing patterns, and periodic updates do not ensure case-level explanations.
The trap
Combine the evidence gap with the required control point; neither constraint can be evaluated alone. How to remember it
Limit adaptation and require review because novelty and autonomous adverse action jointly increase exposure.
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
- Require calibrated evaluation and human review: Which procurement condition best addresses the immediate →
- Address the individual harm while examining relevant group: Which response is most suitable? →
- Stability across conditions and the costs of errors: Which evidence should govern deployment? →
- 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.
