Objective 3.3 Identify and Mitigate Transparency and Explainability Risks CERTIFIED-ETHICAL-EMERGING-TECHNOLOGIST Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   An insurance company uses an AI system to set premiums. The system uses a wide range of data, including some that may be considered sensitive. The company wants to ensure that customers understand why their premiums are set at a certain level, but also wants to avoid revealing sensitive data patterns that could be exploited. Which mitigation strategy best balances transparency and privacy?

   • AProvide customers with the exact data values used in their premium calculation.
   • BProvide customers with explanations that describe the factors in general terms, without revealing specific sensitive data values.
   • CUse a simpler model that does not use sensitive data, even if it reduces accuracy.
   • DRefuse to provide any explanation, citing privacy concerns.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   Which tool is best suited for providing a standardized, human-readable summary of an AI model's purpose, performance, and limitations?

   • ASHAP
   • BModel cards
   • CLIME
   • DELI5
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A financial institution is deploying a machine learning model to approve loan applications. Before launch, the compliance team wants to identify potential explainability risks. They have a dataset of historical loan decisions and access to the model's predictions. Which technique would best help them detect if the model is relying on unintended patterns?

   • ARun SHAP analysis to determine feature contributions for individual predictions.
   • BConsult with external auditors to review the model's code.
   • CReview the model's overall accuracy on a held-out test set.
   • DAudit the training data for missing values and outliers.
   Reveal Solution Discussion Select an answer first
5. 19foundation · easy

   A company wants to ensure that all stakeholders receive consistent and clear information about an AI system's capabilities and limitations. Which mitigation strategy is most directly aligned with this goal?

   • AUsing a more complex model
   • BImplementing clear communication protocols
   • CConducting a data quality audit
   • DLimiting stakeholder access to the system
   Reveal Solution Discussion Select an answer first
6. 20expert · hard · select all that apply

   A government agency is deploying an AI system to assist with public benefits decisions. The agency must ensure transparency and explainability to the public. Which of the following actions would be appropriate to identify and mitigate transparency risks? Select all that apply.

   • AKeep the model's logic confidential to prevent gaming of the system.
   • BPublish a model card that documents the system's purpose, performance, and limitations.
   • CUse SHAP to generate explanations for individual decisions that can be provided to applicants.
   • DConduct a stakeholder impact assessment to identify affected groups and their information needs.
   • ELimit public access to the system's documentation to reduce confusion.
   Check answers Discussion Select an answer first