Bias Detection CRAGE Practice Questions (Page 6)

Questions 26–30

1. 26application · medium

   A model predicts whether a patient will be readmitted to the hospital. The training data was collected from a hospital that uses a specific electronic health record system. The model later underperforms for patients from a hospital that uses a different system. Which source of bias is most directly responsible?

   • ASelection bias
   • BConfirmation bias
   • CAlgorithmic bias
   • DMeasurement bias
   Reveal Solution Discussion Select an answer first
2. 27expert · hard

   A model predicts whether a customer will default on a loan. The team finds that the false-negative rate is much higher for one demographic group. Which mitigation strategy is most likely to reduce this disparity?

   • ARemove the demographic attribute from the model
   • BUse a more complex algorithm to improve overall accuracy
   • CRaise the decision threshold for that group to reduce false positives
   • DLower the decision threshold for that group to increase true positives, while monitoring the impact on false positives
   Reveal Solution Discussion Select an answer first
4. 28foundation · easy

   Which of the following is a potential impact of bias on model performance?

   • AImproved accuracy for all groups
   • BElimination of variance in predictions
   • CReduced accuracy for underrepresented groups
   • DIncreased model interpretability
   Reveal Solution Discussion Select an answer first
5. 29application · medium

   A healthcare startup is building a model to predict patient readmission risk. The dataset was collected from a single hospital that primarily serves affluent patients. The team wants to detect potential bias before deployment. Which approach would be most effective?

   • AIncrease the model's complexity to capture more patterns
   • BCompare the distribution of key features and outcomes across demographic groups in the dataset
   • CUse a larger random sample from the same hospital
   • DRemove all demographic attributes from the training data
   Reveal Solution Discussion Select an answer first
6. 30application · medium

   A model predicts whether patients will develop a chronic condition. The team wants to ensure that the model does not disproportionately misclassify patients from different racial groups. Which fairness metric is most appropriate for this goal?

   • APredictive parity, because it requires equal positive predictive value across groups.
   • BDemographic parity, because it requires equal positive outcome rates across groups.
   • CCalibration, because it ensures predicted probabilities match actual outcomes.
   • DEqualized odds, because it requires equal false positive and false negative rates across groups.
   Reveal Solution Discussion Select an answer first