Responsible Practices and Governance Standards AIE Practice Questions (Page 5)

Questions 21–25

1. 21application · medium

   A bank uses a machine learning model to detect fraudulent credit card transactions. The model is a gradient boosting machine, which is not inherently interpretable. Regulators require the bank to explain to customers why a transaction was blocked. What is the most appropriate technique to achieve explainability?

   • AReplace the gradient boosting model with a linear regression model.
   • BProvide customers with a generic statement that the transaction was blocked due to suspicious activity.
   • CUse SHAP (SHapley Additive exPlanations) values to show which features contributed most to each decision.
   • DUse a global feature importance plot to explain all decisions.
   Reveal Solution Discussion Select an answer first
2. 22foundation · easy

   Why is stakeholder engagement important in AI governance?

   • AIt guarantees that the AI system will be profitable.
   • BIt helps incorporate diverse perspectives and values into AI decision-making.
   • CIt ensures that only experts make all AI decisions.
   • DIt eliminates the need for technical testing.
   Reveal Solution Discussion Select an answer first
4. 23foundation · easy

   Which of the following is a widely recognized framework for AI governance that provides guidance on ethical and responsible AI?

   • AISO 9001
   • BOECD AI Principles
   • CITIL
   • DTOGAF
   Reveal Solution Discussion Select an answer first
5. 24application · medium

   A credit union uses an AI model to decide credit limits for members. A member requests an explanation for why their credit limit was reduced. The credit union must provide a clear explanation. Which approach best meets this requirement?

   • AGenerate a personalized explanation that lists the key factors, such as payment history and debt-to-income ratio, that influenced the decision.
   • BShare the complete model code with the member to demonstrate transparency.
   • CRefuse to provide an explanation because the model is proprietary.
   • DProvide a general description of how the model works without referencing the member's specific data.
   Reveal Solution Discussion Select an answer first
6. 25expert · hard

   A large hospital network is deploying an AI system to triage emergency room patients. The system was developed with input from doctors and nurses, but not from patients or community representatives. The hospital's ethics committee is concerned about potential bias against minority groups. The system is already in pilot testing and shows promising accuracy overall, but subgroup analysis has not been performed. The hospital must decide whether to proceed to full deployment. What is the most responsible course of action?

   • AContinue the pilot indefinitely until more data is collected, without additional analysis.
   • BProceed with deployment but add a disclaimer that the system may have biases.
   • CProceed with deployment because overall accuracy is high and the system can save lives.
   • DPause deployment, conduct a subgroup fairness audit, and engage patients and community representatives before proceeding.
   Reveal Solution Discussion Select an answer first