Bias Detection CRAGE Practice Questions (Page 5)

Questions 21–25

1. 21foundation · easy

   What is a key fairness-related impact of bias in an AI system?

   • AIt improves model generalization to new populations
   • BIt reduces the need for human oversight
   • CIt can lead to discriminatory outcomes for protected groups
   • DIt ensures equal treatment across all groups
   Reveal Solution Discussion Select an answer first
2. 22application · medium

   A model is used to approve personal loans. The compliance team wants to detect whether the model is biased against applicants from a particular zip code. They have the model's predictions and the applicants' zip codes. Which quantitative method would most directly reveal this bias?

   • AInterview the model developers about their design choices.
   • BCompare the approval rates across different zip codes using a statistical test.
   • CCompute the model's overall accuracy on the entire dataset.
   • DVisualize the model's decision boundary in a two-dimensional feature space.
   Reveal Solution Discussion Select an answer first
4. 23foundation · easy

   After computing demographic parity, you find that the positive prediction rate for group A is 0.8 and for group B is 0.4. What does this result indicate?

   • AThe model is perfectly fair
   • BThe model has a potential bias favoring group A
   • CThe model is calibrated
   • DThe model has equalized odds
   Reveal Solution Discussion Select an answer first
5. 24foundation · easy

   A bias detection report shows that the equalized odds difference is 0.05, which is below the acceptable threshold. What is the most appropriate interpretation?

   • AThe model is definitely fair and ready for deployment
   • BThe model has a bias against the minority group
   • CThe model has no bias at all
   • DThe model meets the equalized odds criterion, but other fairness metrics should still be reviewed
   Reveal Solution Discussion Select an answer first
6. 25application · medium

   A social media company is developing a content moderation model. The training data was labeled by a team of annotators who were told that certain dialects are often used in spam. The model later flags those dialects as spam at a higher rate. Which source of bias is most directly responsible?

   • AAlgorithmic bias in model architecture
   • BSelection bias in sampling
   • CHuman bias in labeling
   • DMeasurement bias in feature extraction
   Reveal Solution Discussion Select an answer first