Privacy, Ethical, Trust and Safety Controls AAISM Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   Which privacy control is most directly applied to ensure that an AI system only collects personal data that is adequate, relevant, and limited to what is necessary for its intended purpose?

   • AData minimization
   • BEncryption
   • CAudit logging
   • DPseudonymization
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A financial services firm uses an AI system to detect fraudulent transactions. The system flags transactions as suspicious, but the firm's compliance team cannot understand why certain transactions are flagged. Regulators require the firm to explain its decisions. What should the firm do?

   • AReplace the AI system with a rules-based system that is inherently explainable.
   • BProvide only the final decision to regulators without explanation, citing proprietary algorithms.
   • CUse an explainable AI technique, such as LIME or SHAP, to provide feature attributions for each flagged transaction.
   • DHire more compliance staff to manually review every transaction.
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   Which technique is commonly used to assess whether an AI model exhibits bias by comparing its performance across different demographic groups?

   • ARegularization
   • BFairness metrics
   • CCross-validation
   • DFeature scaling
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A government agency is deploying an AI system to determine eligibility for social benefits. The system is designed to be fully automated, but there have been concerns about errors and lack of recourse for citizens. The agency wants to integrate human oversight to ensure ethical and safe operation. What is the most effective way to do this?

   • AUse a second AI model to double-check the first model's decisions.
   • BProvide a mechanism for citizens to request a human review of any AI decision.
   • CPublish the AI's decision criteria to the public.
   • DHave a human review every single decision before it is communicated.
   Reveal Solution Discussion Select an answer first
6. 10expert · hard

   A defense contractor is developing an AI system for target identification in military operations. The system has a low false-positive rate but occasionally misidentifies civilian objects as military targets. The contractor wants to integrate human oversight to prevent accidental strikes. Which approach is most effective?

   • ARequire a human operator to confirm all target identifications before any action is taken.
   • BAllow the AI to act autonomously but with a 'kill switch' that a human can activate.
   • CUse a second AI system to cross-check the first system's identifications.
   • DLimit the AI system to identifying targets but not taking any action.
   Reveal Solution Discussion Select an answer first