AI Models, Considerations, and Requirements AAIA Practice Questions (Page 5)

Questions 21–25

1. 21foundation · easy

   In the AI model lifecycle, which activity is typically performed during the post-deployment monitoring stage?

   • ARetraining the model on new data
   • BSelecting the final algorithm
   • CTracking model performance metrics over time
   • DLabeling training data
   Reveal Solution Discussion Select an answer first
2. 22foundation · easy

   Which ethical principle requires that AI models treat all individuals or groups fairly and without discrimination?

   • ATransparency
   • BPrivacy
   • CFairness
   • DAccountability
   Reveal Solution Discussion Select an answer first
4. 23application · medium

   A European bank is deploying an AI model that processes customer data to offer personalized loans. The bank operates in multiple EU countries. Which regulatory requirement is most directly applicable to this model?

   • AGDPR's requirements for data protection and automated decision-making.
   • BCCPA's consumer privacy rights for California residents.
   • CHIPAA's privacy rules for health information.
   • DSOX's financial reporting requirements.
   Reveal Solution Discussion Select an answer first
5. 24expert · hard

   A credit union is building a model to approve small business loans. The dataset is small and contains many missing values. The credit union wants to ensure the model is reliable and fair. They are considering using a complex gradient boosting model. Which approach is most appropriate?

   • ARemove all records with missing values to create a clean dataset.
   • BUse a complex model and rely on its high training accuracy.
   • CUse the complex gradient boosting model without imputation to capture all patterns.
   • DUse a simpler model with robust data imputation and cross-validation.
   Reveal Solution Discussion Select an answer first
6. 25expert · hard

   A financial regulator requires a bank to explain its AI credit model's decisions to customers. The model is a deep neural network. The bank wants to provide meaningful explanations without revealing proprietary information. Which approach is most appropriate?

   • AState that the decision is based on a complex algorithm without further explanation.
   • BProvide the model's global feature importance rankings.
   • CUse LIME to generate local explanations for individual decisions.
   • DShare the model's architecture and weights with the customer.
   Reveal Solution Discussion Select an answer first