Ethical, Societal, and Security Concerns in AI AIE Practice Questions (Page 9)

Questions 41–42

1. 41expert · hard

   A city plans to deploy an AI system to allocate public housing. The system uses predictive analytics to identify families most in need. However, the training data is incomplete for certain neighborhoods, leading to underrepresentation. The city must ensure equitable allocation while also being transparent about how decisions are made. The city council is divided between using a simple, interpretable model and a more accurate but opaque model. What should the city do?

   • AUse an interpretable model with clear rules, and supplement it with community input to address data gaps.
   • BUse a random lottery to allocate housing, ensuring complete fairness.
   • CCollect more data from underrepresented neighborhoods before deploying any model.
   • DUse the opaque model because it is more accurate, and provide post-hoc explanations.
   Reveal Solution Discussion Select an answer first
2. 42application · medium

   A startup offers an AI-powered image recognition API. A competitor subscribes to the API and uses thousands of queries to reverse-engineer the model's decision boundaries, then creates a near-identical model for a fraction of the development cost. The startup wants to protect its intellectual property without degrading service for legitimate customers. Which mitigation is most appropriate?

   • ARate-limit API requests and monitor for unusual query patterns
   • BDisable the API and sell the model as a downloadable package
   • CAdd random noise to every prediction returned by the API
   • DRequire customers to sign a non-disclosure agreement before using the API
   Reveal Solution Discussion Select an answer first