Question 32
IVBefore deploying an AI system to prioritize access to public housing, a city conducts community consultations, models potential disparate impacts, and documents mitigation plans. What does this process best represent?
Correct answer: B
Explanation
An AI impact assessment is a pre-deployment review of an AI system’s likely effects, including “social, ethical, and distributional consequences,” and the controls planned to reduce harm. Community consultations, disparate-impact modeling, and mitigation documentation fit that definition because they evaluate risks before the system is used to allocate public housing.
Why each option is right or wrong
A. Routine latency optimization.
Latency optimization improves response speed, not fairness, public impact, or predeployment governance review.
B. An AI impact assessment evaluating likely social, ethical, and distributional consequences, along with planned controls, prior to deployment.
The facts describe a pre-deployment review focused on whether the housing-prioritization model could create unequal effects, which aligns with the AIGP body of knowledge on governing AI deployment and use: evaluating key factors and risks before deciding to deploy, and performing key activities to assess the AI system. The community consultations, disparate-impact modeling, and mitigation documentation are the classic components of an AI impact assessment because they assess likely social, ethical, and distributional consequences and the controls planned to reduce harm before the system is put into operation.
C. A traditional penetration test focused on infrastructure vulnerabilities.
Penetration testing examines security weaknesses in systems, not social harms or disparate outcomes.
D. A marketing survey to promote adoption of the AI system.
Marketing surveys measure interest or adoption, not ethical risk, equity, or mitigation planning.