Bias Detection CRAGE Practice Questions (Page 8)

Questions 36–40

1. 36application · medium

   A ride-sharing company is building a model to predict driver earnings. The data was collected from drivers who opted into a telematics program. The model later performs poorly for drivers who did not opt in. Which source of bias is most directly responsible?

   • AAlgorithmic bias
   • BConfirmation bias
   • CSelection bias
   • DMeasurement bias
   Reveal Solution Discussion Select an answer first
2. 37application · medium

   A data science team is auditing a model for gender bias. They have access to the training data and the model's predictions. Which combination of methods would most effectively detect both data-level and model-level bias?

   • AOnly compute the model's overall accuracy on a test set.
   • BOnly interview the model developers about their assumptions.
   • COnly visualize the model's decision boundary.
   • DCompute demographic parity on predictions and analyze feature distributions across gender groups in the training data.
   Reveal Solution Discussion Select an answer first
4. 38foundation · easy

   Which of the following best distinguishes bias from noise in an AI model?

   • ABias is systematic and consistent, while noise is random and unpredictable
   • BBias only affects classification models, while noise only affects regression models
   • CBias is caused by the model architecture, while noise is caused by the optimizer
   • DBias is random and unpredictable, while noise is systematic and consistent
   Reveal Solution Discussion Select an answer first
5. 39application · medium

   A model is used to screen job applicants. The model was trained on historical hiring data that reflects past discriminatory practices. The model's predictions are highly accurate on historical data but perpetuate the same disparities. What is the most significant impact of this bias on model performance and fairness?

   • AThe model will have low variance because it is trained on a large dataset.
   • BThe model will have low predictive accuracy because biased data always reduces accuracy.
   • CThe model will have high fairness because it accurately reflects historical patterns.
   • DThe model may have high predictive accuracy but low fairness because it learns historical biases.
   Reveal Solution Discussion Select an answer first
6. 40foundation · easy

   Which of the following is an example of a qualitative method for detecting bias in model outputs?

   • AReviewing a sample of model predictions with domain experts to identify patterns of unfairness
   • BComputing the false positive rate for different demographic groups
   • CMeasuring the demographic parity of predictions
   • DRunning a chi-squared test on feature distributions
   Reveal Solution Discussion Select an answer first