Selecting Test Techniques for AI-Based Systems CT-AI Practice Questions (Page 5)

Questions 21–25

1. 21expert · hard

   A government agency is deploying an AI system for public benefits eligibility. The system must comply with strict fairness regulations. The team has a limited budget and must choose between testing techniques. Which approach best balances fairness testing with budget constraints?

   • AUse a comprehensive fairness testing suite with multiple metrics and subgroup analysis
   • BUse a single fairness metric on the overall population
   • CUse a combination of data bias detection and a few key fairness metrics on protected subgroups
   • DFocus on data validation to ensure the training data is unbiased
   Reveal Solution Discussion Select an answer first
2. 22expert · hard

   An autonomous vehicle company is testing its perception system. The system uses a deep learning model to detect pedestrians. The company has a high-quality labeled dataset, but the model occasionally fails in rare lighting conditions. They have limited time and budget. Which testing strategy is most effective?

   • AFocus on white-box testing to analyze the model's internal activations
   • BRely on differential testing against a simpler model
   • CUse metamorphic testing to generate variations of existing images
   • DCollect more data specifically for rare lighting conditions and use robustness testing
   Reveal Solution Discussion Select an answer first
4. 23foundation · easy

   Which characteristic of AI-based systems is best addressed by using data validation techniques?

   • AModel complexity
   • BUser interface design
   • CData dependency
   • DNon-determinism
   Reveal Solution Discussion Select an answer first
5. 24application · medium

   A team is testing an AI system that translates legal documents. They do not have a reliable oracle to verify the correctness of translations. They want to test the system's behavior by applying transformations to the input and checking if the output changes as expected. Which technique is most appropriate?

   • AAdversarial testing
   • BMutation testing
   • CDifferential testing
   • DMetamorphic testing
   Reveal Solution Discussion Select an answer first
6. 25expert · hard

   A team is testing a medical diagnosis AI system. The system is a black box, and the team does not have access to the model's internals. They need to test the system's robustness to variations in input data, such as different image resolutions. Which technique is most appropriate?

   • AMetamorphic testing with image transformations
   • BAdversarial testing with gradient-based attacks
   • CWhite-box testing with neuron coverage
   • DDifferential testing against a second model
   Reveal Solution Discussion Select an answer first