Objective 3.4 Identify and Mitigate Fairness and Non-Discrimination (bias) Risks CERTIFIED-ETHICAL-EMERGING-TECHNOLOGIST Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   A developer wants to use a Python library that provides fairness metrics and mitigation algorithms, and is compatible with popular machine learning frameworks like scikit-learn. Which tool should they choose?

   • AGit
   • BNumPy
   • CDocker
   • DFairlearn
   Reveal Solution Discussion Select an answer first
2. 7expert · hard

   A government agency is deploying an AI system to determine eligibility for social benefits. The system uses a model trained on data that includes a 'neighborhood' feature, which is highly correlated with race due to historical segregation. The agency wants to ensure the system does not discriminate against any racial group. Which mitigation strategy should they use?

   • AUse a data preprocessing technique to remove the correlation between neighborhood and race, such as reweighting or transforming the feature.
   • BRemove the neighborhood feature from the model entirely.
   • CUse a post-processing technique to adjust the model's predictions to equalize approval rates across racial groups.
   • DUse an inprocessing technique that penalizes the model for using neighborhood as a feature.
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   After training a model, a team finds that its predictions are biased against a protected group. They decide to adjust the decision threshold so that the model's outcomes are more equitable across groups. Which mitigation strategy does this represent?

   • APost-processing
   • BData preprocessing
   • CFeature engineering
   • DAlgorithmic adjustment
   Reveal Solution Discussion Select an answer first
5. 9foundation · easy

   A data team notices that a training dataset underrepresents a certain demographic group. To reduce potential bias, they decide to add more samples from that group to balance the dataset. Which category of bias mitigation does this action belong to?

   • APost-processing
   • BModel validation
   • CData preprocessing
   • DAlgorithmic adjustment
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A retail company is using an AI system to predict customer churn. The model is found to be biased against customers from a certain region. The company wants to mitigate this bias using a tool like Fairlearn. Which Fairlearn mitigation algorithm should they use?

   • AReweighing
   • BExponentiated Gradient
   • CGrid Search
   • DAdversarial Debiasing
   Reveal Solution Discussion Select an answer first