Bias CT-AI Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   How does bias in AI-based systems differ from general statistical bias?

   • AAI bias is random, while statistical bias is systematic
   • BAI bias is a systematic error that leads to unfair outcomes, while statistical bias is a mathematical property of an estimator
   • CAI bias only occurs in deep learning models, while statistical bias occurs in all models
   • DAI bias is always intentional, while statistical bias is unintentional
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   A recruitment firm is developing an AI system to screen job applications. The training data consists of resumes from successful hires over the past 10 years. The firm's leadership wants to ensure the system does not discriminate against candidates from non-traditional educational backgrounds. Which mitigation technique is most appropriate to apply during data collection?

   • AApply a fairness constraint during model training
   • BUse a more complex model to better capture the nuances of each resume
   • CRemove the 'university attended' feature from the dataset before training
   • DCollect additional resumes from candidates with non-traditional backgrounds to balance the dataset
   Reveal Solution Discussion Select an answer first
4. 3application · medium

   A credit card company is using an AI system to set credit limits for new customers. The system was trained on historical data that reflects past lending decisions. The company's compliance team is concerned that the system may be denying higher credit limits to customers from certain ethnic groups. Which potential impact of bias is the compliance team most directly concerned about?

   • AReduced model accuracy for all customers
   • BLegal and regulatory consequences due to discriminatory lending practices
   • CIncreased computational cost of running the model
   • DDecreased user trust among all customers
   Reveal Solution Discussion Select an answer first
5. 4expert · hard

   A financial institution is developing an AI system to approve personal loans. The training data includes a feature for 'gender'. The data scientists are considering two mitigation approaches: (A) remove the 'gender' feature and retrain, or (B) apply a fairness constraint that equalizes approval rates across genders. The institution's legal team warns that removing the 'gender' feature may not be sufficient to eliminate bias because other features, such as 'occupation' and 'income', are correlated with gender. Which approach is most appropriate given the legal team's warning?

   • ARemove the 'gender' feature and rely on the model to not use correlated features
   • BRemove the 'gender' feature and also remove all correlated features like 'occupation' and 'income'
   • CUse a more complex model to better capture the relationships between features
   • DApply the fairness constraint (B) because it directly addresses the disparate impact, even if correlated features remain
   Reveal Solution Discussion Select an answer first
6. 5foundation · easy

   Which of the following is an example of bias introduced through human interaction?

   • AA model that uses a large dataset with many features
   • BA model that uses a random forest algorithm
   • CA model trained on historical hiring decisions that favored certain demographics
   • DA model that has high accuracy on training data but low accuracy on test data
   Reveal Solution Discussion Select an answer first