AI Assurance, Testing, and Auditing CRAGE Practice Questions (Page 10)

Questions 46–50

1. 46application · medium

   A social-media platform uses an AI content-moderation system. The system must accurately flag harmful content while minimizing false positives on legitimate posts. Which testing approach is most important to evaluate this trade-off?

   • AFunctional testing with a labeled dataset to measure precision and recall.
   • BPerformance testing to measure how many posts can be processed per second.
   • CExplainability testing to understand why posts are flagged.
   • DRobustness testing with adversarial examples.
   Reveal Solution Discussion Select an answer first
2. 47foundation · easy

   What is the main objective of robustness testing in AI systems?

   • ATo verify that the model's outputs are explainable to non-technical stakeholders.
   • BTo evaluate how well the model handles unexpected, noisy, or adversarial inputs without significant degradation in performance.
   • CTo check that the model's predictions are consistent with historical data patterns.
   • DTo measure the model's training time and computational efficiency.
   Reveal Solution Discussion Select an answer first
4. 48application · medium

   A hospital is deploying an AI system to recommend treatment plans. The hospital's ethics board requires that clinicians be able to understand why the system made a specific recommendation for each patient. The model is a deep neural network with high accuracy. Which action best satisfies the transparency requirement?

   • AUse a global surrogate model, such as a decision tree, to approximate the neural network's decisions across all patients.
   • BReplace the neural network with a logistic regression model to ensure full interpretability.
   • CPublish the model's architecture and training data documentation.
   • DGenerate a local explanation for each recommendation using SHAP or LIME.
   Reveal Solution Discussion Select an answer first
5. 49expert · hard

   A manufacturing company uses an AI system to predict equipment failures. The system is highly accurate, but the company has experienced a recent increase in false alarms, causing maintenance crews to ignore alerts. The company wants to reduce false alarms without significantly increasing missed failures. Which testing or monitoring approach would best address this trade-off?

   • ARetrain the model more frequently to adapt to changing equipment conditions.
   • BAdjust the model's decision threshold to balance precision and recall based on the cost of false alarms versus missed failures.
   • CIncrease the model's complexity to improve overall accuracy.
   • DMonitor the model's false alarm rate and alert the team when it exceeds a threshold.
   Reveal Solution Discussion Select an answer first
6. 50foundation · easy

   Which of the following is a widely recognized framework for AI risk management that provides guidance on governing and managing AI risks?

   • ANIST AI Risk Management Framework
   • BITIL Service Management Framework
   • CISO 9001 Quality Management Standard
   • DOWASP Top 10 for Web Applications
   Reveal Solution Discussion Select an answer first