Objective 3.5 Address Business Risks, Ethical Concerns, and Related Concepts in Training and Tuning AIP-210 Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   Which explainability method provides a global view of which features are most influential in a model's predictions?

   • ASHAP
   • BLIME
   • CConfusion matrix
   • DFeature importance
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   A credit approval model is tuned to maximize overall approval rate. Which of the following is an ethical implication that should be assessed?

   • AThe model may use a complex ensemble method that is hard to debug.
   • BThe model may reject applicants from certain neighborhoods more often, even when their credit profiles are similar.
   • CThe model may require frequent retraining due to changing economic conditions.
   • DThe model may have a higher false positive rate for fraudulent transactions.
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   Which of the following is a data anonymization technique that removes or generalizes identifying information?

   • AOne-hot encoding
   • BPrincipal component analysis (PCA)
   • CBatch normalization
   • DK-anonymity
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A health insurance company is training a model to predict patient risk using electronic health records. The data includes rare genetic markers that could be used to identify individuals. The company wants to protect patient privacy while still training an effective model. Which technique should they use?

   • ACross-validation, which splits the data into training and validation sets.
   • BData augmentation, which creates synthetic variations of the original records.
   • CDifferential privacy, which adds noise to the training process to prevent the model from memorizing individual records.
   • DFeature scaling, which normalizes the range of features.
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   A university is using an AI system to screen student applications. The system was trained on historical admission data. The admissions office is concerned that the model may inadvertently disadvantage applicants from certain socioeconomic backgrounds. They want to evaluate the model's fairness. Which metric should they compute?

   • AOverall accuracy, which measures the percentage of correct predictions.
   • BAUC-ROC, which measures the model's ability to distinguish between classes.
   • CDemographic parity, which compares acceptance rates across different groups.
   • DF1 score, which balances precision and recall.
   Reveal Solution Discussion Select an answer first