Transparency CRAGE Practice Questions (Page 6)

Questions 26–30

1. 26foundation · easy

   What is the key difference between explainability and interpretability in AI?

   • AExplainability and interpretability are synonymous terms that can be used interchangeably
   • BExplainability refers to providing understandable reasons for a model's outputs, while interpretability refers to the degree to which a human can understand the model's internal workings
   • CExplainability is only relevant for deep learning models, while interpretability applies only to traditional machine learning models
   • DExplainability refers to the ability to understand the internal logic of a model, while interpretability refers to providing post-hoc explanations for individual predictions
   Reveal Solution Discussion Select an answer first
2. 27expert · hard

   A hospital uses an AI system to predict patient readmission risk. The hospital must be transparent with patients and regulators, but also protect patient privacy under HIPAA. What is the best way to balance transparency and privacy?

   • AProvide patients with a general explanation of the system's purpose and how it uses their data, without revealing specific model details that could identify individuals
   • BPublish the full model and training data to be fully transparent
   • CShare detailed model explanations only with regulators, not with patients
   • DProvide no information to patients to protect their privacy
   Reveal Solution Discussion Select an answer first
4. 28application · medium

   A non-profit organization uses an AI system to allocate resources during disaster response. The organization wants to build trust with the communities it serves. Which action most directly demonstrates a commitment to transparency?

   • APublishing the raw sensor data used by the AI system
   • BKeeping the AI system's decision process confidential to avoid public confusion
   • CPublishing a plain-language explanation of how the AI system prioritizes resource allocation
   • DHiring a public relations firm to promote the AI system's successes
   Reveal Solution Discussion Select an answer first
5. 29application · medium

   A city government deploys an AI system to prioritize road maintenance. The system uses sensor data and historical traffic patterns. The city must communicate this system to residents, city council members, and road crews. What is the most appropriate approach to ensure transparency across these diverse stakeholders?

   • AShare the raw sensor data and code repository publicly
   • BProvide role-specific communications, such as plain-language summaries for residents, decision briefs for council, and operational manuals for crews
   • CPublish a single technical report with detailed model architecture and training data
   • DHold a single public meeting and answer questions from attendees
   Reveal Solution Discussion Select an answer first
6. 30application · medium

   A company is deploying an AI system that will be used in a jurisdiction with a new AI transparency law. The law requires that users be able to obtain an explanation of any decision that significantly affects them. What should the company implement to comply?

   • AAn internal audit of the AI system's performance
   • BA process for users to request and receive an explanation of decisions that affect them
   • CA privacy policy that states the company uses AI
   • DA public blog post describing the AI system's general capabilities
   Reveal Solution Discussion Select an answer first