AI Concerns, Ethical Principles, and Responsible AI CRAGE Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A healthcare AI startup develops a model that predicts patient readmission risk. The model uses sensitive health data. The startup wants to ensure compliance with privacy regulations while maintaining model accuracy. Which approach best balances privacy and utility?

   • ARemove all patient identifiers from the dataset and assume the model cannot re-identify individuals.
   • BApply differential privacy during training to limit how much the model reveals about any individual patient.
   • CStore the data in an encrypted database and use the model only internally.
   • DObtain broad patient consent for all future uses of their data.
   Reveal Solution Discussion Select an answer first
2. 12expert · hard

   A hospital is deploying an AI system to predict patient deterioration. The system is very accurate, but it occasionally makes errors that could be life-threatening. The hospital's ethics committee wants to ensure accountability. Which action is most important for accountability?

   • ADefine a clear process for who is responsible when the AI makes an error, including a review and remediation plan.
   • BEnsure the AI is transparent by providing a detailed explanation for every prediction.
   • CUse the AI only for patients who sign a waiver acknowledging the AI's limitations.
   • DMonitor the AI's performance and retrain it if accuracy drops below 99%.
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A financial services firm is developing a credit-scoring model. The product owner wants to launch quickly, but the legal team insists on documenting the model's decision logic and providing customers with a reason for adverse actions. Which practice best embodies responsible AI in this situation?

   • AOutsource the explanation to a third-party vendor that provides a confidence score instead of a reason.
   • BImplement a model documentation and explainability process that generates human-readable reasons for each credit decision.
   • CLaunch the model and provide a generic notice that 'the model uses multiple factors' without specifics.
   • DDelay the launch until a fully interpretable model (e.g., logistic regression) is built, even if it reduces predictive accuracy.
   Reveal Solution Discussion Select an answer first
5. 14expert · hard

   A credit union wants to deploy an AI loan approval system. The system must be fair to all applicants, but the credit union also has a legal obligation to prevent money laundering. The data available includes some sensitive attributes such as ethnicity. The compliance team is concerned that using these attributes could be discriminatory, but the risk team says they help detect fraud. Which approach best balances fairness and regulatory compliance?

   • ARemove sensitive attributes from all models and rely on other features for fraud detection.
   • BUse the same model for both loan approval and fraud detection to ensure consistency.
   • CUse the sensitive attributes only in a separate fraud-detection model with strict access controls and not in the loan approval model.
   • DInclude sensitive attributes in the loan approval model to improve accuracy for all groups.
   Reveal Solution Discussion Select an answer first
6. 15expert · hard

   A ride-sharing company uses an AI to set surge pricing. The AI raises prices during high demand, which often occurs in low-income neighborhoods after large events. Consumer advocates claim this is exploitative. The company wants to maintain profitability but also address ethical concerns. Which approach best balances these interests?

   • AUse a different AI model that predicts demand more accurately.
   • BCap surge pricing at a reasonable multiplier and provide transparency about how prices are set.
   • CKeep surge pricing but offer discounts to riders in low-income neighborhoods.
   • DEliminate surge pricing entirely to ensure fairness.
   Reveal Solution Discussion Select an answer first