Objective 1.6 Identify Potential Ethical Concerns AIP-210 Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   A financial institution is building a credit-scoring model. The data science team discovers that the model relies heavily on a feature that correlates with an applicant's zip code, which is a proxy for race. The team must ensure the model is fair and compliant. What is the most appropriate step?

   • AKeep the feature because it improves predictive accuracy significantly
   • BAdd a note to the model documentation explaining why the feature is used
   • CRemove the zip code feature and retrain the model, then test for disparate impact across protected groups
   • DBlind the model to all geographic data but do not test the outcome
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   An insurance company uses an AI model to set premiums. The model uses a feature that is highly correlated with disability status, leading to higher premiums for disabled individuals. The company must comply with anti-discrimination laws. What is the most appropriate action?

   • AKeep the feature because it is a strong predictor of risk
   • BDocument the feature's use in a public transparency report
   • CAdd a surcharge to all customers to offset the higher premiums for disabled individuals
   • DRemove the correlated feature and retrain the model, then audit the new model for disparate impact
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A government agency is deploying an AI to detect fraudulent benefit claims. The model has a high false-positive rate for a specific ethnic group. The agency must address this bias while still catching fraud. What is the most appropriate action?

   • AIncrease the model's sensitivity to catch more fraud
   • BRemove the model and use random audits
   • CRe-balance the training data and adjust the model's decision threshold to reduce false positives for the affected group
   • DKeep the model because it catches more fraud overall
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A social media platform uses an AI to moderate content. The AI sometimes removes posts that are legitimate, and users have no way to appeal. The platform wants to improve accountability. Which action is most aligned with this goal?

   • APublish the AI's internal decision rules publicly
   • BMake the AI's moderation decisions final to reduce operational overhead
   • CIncrease the AI's sensitivity to catch more content
   • DProvide users with a transparent appeals process and explain why content was removed
   Reveal Solution Discussion Select an answer first
6. 10expert · hard

   A healthcare provider is deploying an AI to predict patient readmission risk. The model is accurate but uses sensitive data such as mental health history. The provider must comply with privacy regulations and maintain trust. Which approach best balances these concerns?

   • ARemove all sensitive features from the model
   • BUse differential privacy techniques to add noise to the training data, protecting individual patient information while preserving model utility
   • CShare the raw data with researchers to improve the model
   • DUse the model without informing patients that their data is used
   Reveal Solution Discussion Select an answer first