Governance, Ethics, and Responsible AI in Adoption CAIPM Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   A bank is deploying an AI system to approve credit card applications. The system has been trained on historical data that may contain biases. The bank wants to ensure the system is fair and complies with regulations. What is the most effective way to incorporate human oversight?

   • AUse the system's decisions as final, but publish an annual report on its performance.
   • BHave a human reviewer approve every single application, regardless of the system's confidence.
   • CAllow the system to make all decisions automatically, with human review only after a complaint is filed.
   • DRequire a human reviewer to approve all applications that the system flags as borderline or uncertain.
   Reveal Solution Discussion Select an answer first
2. 7expert · hard

   A non-profit organization is deploying an AI system to allocate disaster relief resources. The system uses historical data that may underrepresent certain communities. The organization wants to ensure fairness and maintain public trust. The program manager must engage stakeholders. What is the most effective strategy?

   • AKeep the system's details confidential to avoid public confusion during the crisis.
   • BRely on the organization's existing relationships with local governments to represent community interests.
   • CUse a third-party auditor to validate the model after deployment and publish the results.
   • DEngage with community leaders from underrepresented groups during the design phase to understand their needs and concerns.
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   Which component is typically included in an AI governance framework?

   • AA mechanism for continuous model retraining
   • BA guarantee of zero bias in all AI outputs
   • CA set of ethical principles and accountability structures
   • DA requirement to use open-source AI models only
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A healthcare organization is deploying an AI model to assist radiologists in detecting tumors. The model was trained on a dataset that underrepresents certain demographic groups. The organization wants to ensure the model is fair and does not perpetuate bias. Which approach should the AI program manager take?

   • ADeploy the model as-is but add a disclaimer that it may not be accurate for all demographic groups.
   • BLimit the model's use to demographic groups that are well-represented in the training data.
   • CRe-train the model using a more diverse dataset and conduct a bias audit before deployment.
   • DIncrease the model's complexity to improve overall accuracy, assuming bias will be reduced.
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A university is developing an AI system to predict student dropout risk. Faculty members are skeptical because they fear the system will be used to evaluate their teaching performance. Students are concerned about privacy. The program manager needs to communicate the system's purpose. What is the best communication strategy?

   • AOnly communicate with the IT department and the administration to keep the project confidential.
   • BHold separate information sessions for faculty and students, clearly stating the system's purpose, data usage, and limitations.
   • CAvoid communicating until the system is fully deployed to prevent misinformation.
   • DSend a single email to all faculty and students explaining the system's technical specifications.
   Reveal Solution Discussion Select an answer first