Building Privacy, Trust, and Safety in AI Systems CRAGE Practice Questions (Page 6)

Questions 26–30

1. 26application · medium

   A ride-sharing company uses an AI system to set dynamic pricing based on demand. Customers have complained that prices seem unfair and unpredictable. The company wants to improve trustworthiness in the AI system. Which action most directly addresses the 'transparency' component of trustworthy AI?

   • AAssigning a customer service team to handle pricing complaints.
   • BPublishing a clear explanation of the factors that influence pricing, such as demand and time of day, in the app's help section.
   • COffering discounts to customers who complain about high prices.
   • DConducting regular audits of the pricing algorithm to ensure it is not biased.
   Reveal Solution Discussion Select an answer first
2. 27foundation · easy

   Which user data right allows an individual to request the deletion of their personal data from an AI system?

   • ARight to portability
   • BRight to rectification
   • CRight to erasure
   • DRight to access
   Reveal Solution Discussion Select an answer first
4. 28application · medium

   A government agency is deploying an AI system to automate the processing of visa applications. The system will make decisions that significantly affect individuals. The agency must conduct a risk assessment to identify and mitigate privacy, trust, and safety risks. Which risk mitigation strategy best addresses the high-stakes nature of the decisions?

   • APublishing the AI system's source code to the public for transparency.
   • BImplementing a mandatory human review process for all decisions that are adverse to the applicant.
   • CRelying solely on the AI system's accuracy metrics to ensure fairness.
   • DLimiting the number of visa applications processed per day to reduce system load.
   Reveal Solution Discussion Select an answer first
5. 29foundation · easy

   Which statement best describes the core idea of Privacy by Design in AI system development?

   • APrivacy is ensured by limiting the use of personal data only when a data breach occurs.
   • BPrivacy is embedded into the design and architecture of the AI system from the outset, proactively and by default.
   • CPrivacy is achieved by anonymizing all data after collection, regardless of the AI system's purpose.
   • DPrivacy is considered as an add-on feature after the AI system is fully developed.
   Reveal Solution Discussion Select an answer first
6. 30expert · hard

   An autonomous drone delivery company is deploying AI systems to navigate urban environments. The drones must operate safely in areas with pedestrians and vehicles. The safety team must decide on the level of human oversight. Full human oversight of every drone is expensive and may not be feasible at scale, but no oversight could lead to accidents. Which approach best balances safety and operational feasibility?

   • AUsing a single human supervisor to monitor all drones simultaneously, but with no ability to intervene.
   • BImplementing a tiered oversight model where drones operate autonomously in normal conditions, but a human supervisor is alerted and can intervene in high-risk situations or when the AI requests help.
   • CAllowing drones to operate fully autonomously with no human oversight, relying on the AI's robustness.
   • DRequiring a human operator to manually control every drone at all times.
   Reveal Solution Discussion Select an answer first