Bias Detection CRAGE Practice Questions (Page 7)

Questions 31–35

1. 31expert · hard

   A credit card company is evaluating a fraud detection model. The model has high overall accuracy, but the false-positive rate is much higher for customers in a certain age group. The company wants to reduce this disparity without significantly increasing overall fraud losses. Which approach is most appropriate?

   • AAdjust the decision threshold separately for each age group to equalize false-positive rates while monitoring overall losses
   • BRetrain the model with a more complex algorithm to improve accuracy
   • CIgnore the disparity because overall accuracy is high
   • DRemove age from the model to eliminate the disparity
   Reveal Solution Discussion Select an answer first
2. 32foundation · easy

   During the model development phase, a data scientist chooses a model architecture that inherently favors the majority class. This is an example of which source of bias?

   • AHuman bias
   • BMeasurement bias
   • CAlgorithmic bias
   • DData bias
   Reveal Solution Discussion Select an answer first
4. 33expert · hard

   A model is being developed to predict housing loan defaults. The team has a dataset with many features, including race. They are concerned about bias. They decide to use a bias detection method that involves training the model with and without the race feature and comparing the outcomes. What is the primary limitation of this approach?

   • ARemoving the race feature does not eliminate bias because other features may be correlated with race.
   • BThe race feature is necessary for the model to be accurate.
   • CTraining with and without the race feature is computationally expensive.
   • DThe approach only detects data bias, not algorithm bias.
   Reveal Solution Discussion Select an answer first
5. 34foundation · easy

   A data collection tool consistently records income values that are rounded to the nearest $10,000, causing systematic inaccuracies. This is an example of which type of bias?

   • ARecall bias
   • BMeasurement bias
   • CSelection bias
   • DConfirmation bias
   Reveal Solution Discussion Select an answer first
6. 35foundation · easy

   In the context of AI, what is the most accurate definition of bias?

   • AThe inability of a model to capture complex patterns in the data
   • BThe random variation in model predictions due to differences in the training sample
   • CA systematic error that leads to consistently skewed outcomes for certain groups or inputs
   • DThe difference between the model's predicted value and the actual value for a single observation
   Reveal Solution Discussion Select an answer first