Task Statement 4.2: Recognize the Importance of Transparent and Explainable Models. AIF-C01 Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   What is a common tradeoff between model transparency and safety?

   • ASimpler, more interpretable models may have lower predictive performance, which could affect safety in high-stakes applications.
   • BTransparent models are always safer because they are easier to audit.
   • CTransparent models require more data to train, which can introduce privacy risks.
   • DTransparent models are more likely to be attacked by adversaries.
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A data scientist is explaining to a stakeholder why a random forest model is considered less transparent than a linear regression model. Which statement is accurate?

   • ALinear regression is opaque because it assumes a linear relationship.
   • BLinear regression is less transparent because it can handle many features.
   • CRandom forests are more transparent because they are based on decision trees.
   • DRandom forests are opaque because they combine many decision trees, making it hard to trace individual predictions.
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A data science team is using Amazon SageMaker Clarify to detect bias in a model. They want to understand how the model's predictions are influenced by different features. Which type of output from Clarify would be most useful?

   • ATraining loss curve
   • BSHAP-based feature importance values
   • CModel card
   • DConfusion matrix
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A bank is deploying an AI system to approve loan applications. To comply with regulations, they must provide applicants with a clear explanation of why their application was denied. Which approach best aligns with human-centered design for explainable AI?

   • AProvide a detailed technical report with feature importance values and SHAP plots.
   • BOffer a phone number to call for more information, but do not provide any written explanation.
   • CShow a simple, non-technical summary of the main factors that influenced the decision.
   • DProvide the full model code and training data to the applicant.
   Reveal Solution Discussion Select an answer first
6. 10expert · hard

   An e-commerce company uses a recommendation system to suggest products. They want to make the system more transparent to users by showing why a product is recommended. However, adding this feature increases the response time and reduces the number of recommendations that can be generated per second. What is the best way to handle this tradeoff?

   • AUse a simpler model that is faster but less accurate.
   • BShow explanations only for a subset of users to reduce the performance impact.
   • COptimize the explanation generation process to minimize the impact on response time.
   • DDo not add explanations to avoid any performance degradation.
   Reveal Solution Discussion Select an answer first