AI Models, Considerations, and Requirements AAIA Practice Questions (Page 3)

Questions 11–15

1. 11foundation · easy

   Which technique is commonly used to provide post-hoc explanations for individual predictions of a black-box model?

   • AGradient descent
   • BLIME (Local Interpretable Model-agnostic Explanations)
   • CCross-validation
   • DPrincipal Component Analysis (PCA)
   Reveal Solution Discussion Select an answer first
2. 12expert · hard

   A government agency is deploying an AI system to predict recidivism for parole decisions. The system uses a proprietary algorithm that is not transparent. Civil rights groups are concerned about bias and lack of accountability. The agency wants to address these ethical concerns while maintaining the system's effectiveness. Which action is most appropriate?

   • AReplace the system with a random decision-making process to ensure fairness.
   • BContinue using the system because it has high predictive accuracy.
   • CKeep the algorithm secret to prevent manipulation by inmates.
   • DRequire the vendor to provide explainability reports and conduct an independent bias audit.
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   What is the primary privacy requirement when using personal data to train an AI model?

   • AEncrypting the data at rest
   • BDeleting the data after training
   • CObtaining appropriate consent or legal basis for processing
   • DUsing only publicly available data
   Reveal Solution Discussion Select an answer first
5. 14expert · hard

   An internal audit team is reviewing an AI model used for employee performance evaluation. The model was developed by an external vendor and has been in production for a year. The audit team needs to assess whether the model complies with internal policies and regulations. Which step is most critical in the audit process?

   • ACheck the model's accuracy on a test set.
   • BInterview the vendor's development team.
   • CReview the model's documentation and governance records.
   • DTest the model's predictions on historical data.
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   Which validation technique involves splitting the dataset into a training set and a separate test set to evaluate the model's performance on unseen data?

   • AHoldout validation
   • BData augmentation
   • CBootstrapping
   • DCross-validation
   Reveal Solution Discussion Select an answer first