Testing Techniques for AI Solutions AAIA Practice Questions (Page 3)

Questions 11–15

1. 11foundation · easy

   What is the purpose of regression testing for an AI solution?

   • ATo verify that changes or updates to the AI solution do not introduce new defects or degrade existing functionality
   • BTo measure the model's performance under peak load
   • CTo validate that the model meets business requirements
   • DTo test the model's ability to handle adversarial inputs
   Reveal Solution Discussion Select an answer first
2. 12foundation · easy

   What should be included in a test report for an AI solution?

   • AA summary of test results, including pass/fail status and any identified issues
   • BThe source code of the AI model
   • CThe names of all developers involved in the project
   • DOnly the final accuracy metric of the model
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A retail company is integrating an AI chatbot into its existing customer relationship management (CRM) system. The chatbot needs to retrieve customer order history and update records. The team wants to ensure the chatbot works correctly with the CRM and that any issues are documented. Which testing activity is most important?

   • ABias testing to ensure the chatbot treats all customers equally.
   • BIntegration testing to verify the chatbot correctly exchanges data with the CRM system.
   • CPerformance testing to measure the chatbot's response time.
   • DUnit testing of the chatbot's natural language processing components.
   Reveal Solution Discussion Select an answer first
5. 14expert · hard

   A streaming service has deployed an AI model that predicts user churn. The model's performance was excellent during testing, but the team is concerned that user behavior may change over time, affecting the model's accuracy. They want to detect any degradation early and take corrective action. Which approach is most effective?

   • AConduct a one-time robustness test with adversarial inputs.
   • BImplement continuous monitoring of model performance metrics and set up alerts for significant drift.
   • CRely on user feedback to identify when the model is performing poorly.
   • DRetrain the model monthly using the latest data without monitoring.
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   What does model validation and verification confirm?

   • AThat the model can handle peak loads
   • BThat the AI model is correctly implemented and its performance metrics are reliable and reproducible
   • CThat the model is free of all biases
   • DThat the model's decisions are explainable
   Reveal Solution Discussion Select an answer first