Transparency CRAGE Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A software company is developing an AI feature that summarizes customer reviews. The product manager wants to ensure transparency is considered throughout the AI lifecycle. Which action best exemplifies applying transparency during the design phase?

   • AAdding a disclaimer to the user interface after the feature is deployed
   • BDocumenting the intended use, data sources, and known limitations of the AI feature in a design document
   • CConducting a post-hoc audit of the model's decisions after release
   • DSetting up automated monitoring to detect when the model's performance degrades
   Reveal Solution Discussion Select an answer first
2. 12expert · hard

   A company operates an AI system in a jurisdiction that requires transparency but does not specify the exact method. The company must choose between using a highly interpretable model with lower performance or a complex model with higher performance and post-hoc explanations. Which consideration is most important when making this choice?

   • AThe regulatory requirement for meaningful explanations can be met by either approach, so the decision should weigh performance and the quality of explanations
   • BPost-hoc explanations are never accepted by regulators
   • CInterpretable models are always required by regulation
   • DPerformance is the only factor that matters because transparency is secondary
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   At which stage of the AI lifecycle should transparency considerations first be applied?

   • AOnly during the deployment phase, when the system is released to users
   • BOnly during the monitoring phase, when issues are detected
   • CDuring the design and development phase, and then maintained through deployment and monitoring
   • DTransparency is only relevant after the AI system is retired
   Reveal Solution Discussion Select an answer first
5. 14foundation · easy

   What is the primary ethical rationale for requiring transparency in AI systems?

   • ATo respect the autonomy of individuals by enabling them to understand and contest decisions that affect them
   • BTo simplify the process of obtaining patents for AI algorithms
   • CTo reduce the need for human oversight of AI systems
   • DTo increase the profitability of AI products
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   A bank uses a complex ensemble model to approve loan applications. A customer whose loan was denied asks for an explanation. The bank's data science team can provide the model's feature importance scores, but the customer wants to know what specific actions they could take to improve their chances. Which approach best addresses the customer's need?

   • AGenerate a counterfactual explanation showing the minimal changes to the customer's application that would have resulted in approval
   • BUse a simpler, inherently interpretable model for all loan decisions, even if it reduces accuracy
   • CProvide the feature importance scores and let the customer infer what to change
   • DExplain the model's internal weights and decision tree structure to the customer
   Reveal Solution Discussion Select an answer first