Objective 4.4 Address Business Risks, Ethical Concerns, and Related Concepts in Operationalizing the Model AIP-210 Practice Questions (Page 5)

Questions 21–25

1. 21application · medium

   A healthcare provider is deploying an ML model to prioritize patient appointments. The model uses electronic health records (EHR), which are subject to HIPAA. The provider wants to ensure compliance while using the model. What is the most important requirement?

   • AObtain patient consent for using their EHR data in the model.
   • BEnsure that the model and its data storage comply with HIPAA's security and privacy rules.
   • CAnonymize all EHR data before using it in the model.
   • DDeploy the model on-premises to avoid cloud compliance issues.
   Reveal Solution Discussion Select an answer first
2. 22expert · hard

   A bank has a credit-risk model that has been in production for two years. The model's performance has been stable, but a recent regulatory audit requires the bank to demonstrate that the model is not discriminating against protected groups. The bank has limited resources and needs to respond quickly. What is the most efficient approach?

   • ARun a comprehensive bias audit using the model's historical predictions and protected attributes, and document the results.
   • BHire an external consultant to conduct a full ethical review of the model.
   • CRetrain the model with a fairness constraint to ensure it is non-discriminatory.
   • DRemove all protected attributes from the model and document that the model does not use them.
   Reveal Solution Discussion Select an answer first
4. 23foundation · easy

   What is the role of a governance framework in AI and ML operations?

   • ATo automatically fix all bugs in the model code.
   • BTo guarantee that models are 100% accurate.
   • CTo replace the need for human oversight in model decisions.
   • DTo define policies and procedures for responsible model development and deployment.
   Reveal Solution Discussion Select an answer first
5. 24application · medium

   A company is about to deploy a model that predicts employee attrition. The model uses performance ratings and salary data. The HR director is concerned that the model might be perceived as punitive. The data science team needs to communicate the model's purpose and limitations to employees. What is the most effective communication strategy?

   • ASend a memo to managers only, instructing them to answer employee questions individually.
   • BDelay the deployment until the model is perfect and no longer raises any ethical concerns.
   • CPublish a detailed technical report on the company intranet explaining the model's algorithm and feature importance.
   • DHold a town hall meeting where the team explains the model's purpose, how it will be used, and the safeguards in place to prevent misuse.
   Reveal Solution Discussion Select an answer first
6. 25application · medium

   A retail company deploys a demand-forecasting model that automatically places replenishment orders. The model was trained on historical sales that included several promotional periods. After a major supply-chain disruption, the model's forecasts become unreliable, and the company faces potential overstocking costs. The data science team wants to keep the model in production. What should the operations team do first?

   • AImmediately retrain the model on the most recent data that includes the disruption period.
   • BDecommission the model and revert to manual ordering until the supply chain stabilizes.
   • CIncrease the model's confidence threshold so it only places orders when it is highly certain.
   • DPlace the model in a shadow mode and compare its forecasts against actuals before allowing it to place orders again.
   Reveal Solution Discussion Select an answer first