Objective 3.4 Identify and Mitigate Fairness and Non-Discrimination (bias) Risks CERTIFIED-ETHICAL-EMERGING-TECHNOLOGIST Practice Questions (Page 3)

Questions 11–15

1. 11foundation · easy

   An organization wants to systematically evaluate whether its AI system produces different outcomes for different demographic groups. Which method is specifically designed for this purpose?

   • AModel tuning
   • BData preprocessing
   • CFeature engineering
   • DFairness metrics
   Reveal Solution Discussion Select an answer first
2. 12application · medium

   A data science team is using AI Fairness 360 (AIF360) to audit a loan approval model. They want to measure whether the model treats different racial groups equally in terms of false-positive rates. Which AIF360 functionality should they use?

   • AThe Adversarial Debiasing algorithm to train a fair model.
   • BThe Fairness Metrics module to compute equalized odds.
   • CThe Preprocessing module to transform the dataset.
   • DThe Postprocessing module to adjust the model's outputs.
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A data science team is using AI Fairness 360 (AIF360) to audit a predictive policing model. They want to check if the model's false-positive rate is higher for certain racial groups. Which AIF360 metric should they use?

   • AEqualized odds difference
   • BDisparate impact
   • CCalibration
   • DStatistical parity difference
   Reveal Solution Discussion Select an answer first
5. 14foundation · easy

   A compliance team is required to document the potential societal and ethical impacts of a new AI system before deployment. Which bias identification method aligns with this requirement?

   • ACross-validation
   • BFeature selection
   • CHyperparameter optimization
   • DImpact assessment
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   A bank is developing a credit-scoring model. The training data includes a 'gender' column and a 'marital status' column. The model is intended to be used for loan approvals. During a bias audit, the data scientist finds that the model's false-positive rate is higher for married women than for any other group. The bank wants to reduce this disparity. Which action should the data scientist take first?

   • AIncrease the weight of the 'gender' and 'marital status' features in the model to ensure the model is aware of these groups.
   • BConduct a bias audit to determine whether the disparity stems from data, algorithmic, or human bias before making changes.
   • CApply a post-processing technique that adjusts the decision threshold for the affected group to equalize false-positive rates.
   • DRemove the 'gender' and 'marital status' columns from the training data and retrain the model.
   Reveal Solution Discussion Select an answer first