AI Solution Development Methodologies and Lifecycle AAIA Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   A multinational bank is developing an AI system to automate loan approvals. The system must comply with fair lending laws, which require explainability of decisions. The bank's data science team has built a highly accurate gradient boosting model, but the model's decisions are difficult to explain. The compliance team insists on a simpler, interpretable model, while the business team wants to maximize approval accuracy to increase revenue. The project is in the problem definition phase. What should the project team do?

   • AProceed with the gradient boosting model and use post-hoc explainability tools to satisfy compliance
   • BUse a simpler, interpretable model and accept a potential reduction in accuracy
   • CDevelop both models and let the compliance team choose after seeing performance metrics
   • DDefine the trade-off between accuracy and explainability as a key requirement and seek stakeholder consensus
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   Why is effective communication with stakeholders important during AI solution development?

   • AIt guarantees the model will be accepted by all users.
   • BIt ensures the project is completed faster.
   • CIt eliminates the need for testing.
   • DIt helps align expectations, gather requirements, and build trust in the solution.
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A software company is developing an AI feature for its existing SaaS product. The company follows a DevOps culture and wants to continuously integrate model updates into the product. The team needs to ensure that model updates do not break the existing application. Which practice is most aligned with DevOps principles?

   • ASchedule model updates in a quarterly release cycle
   • BUse a separate model deployment process that bypasses the existing CI/CD
   • CImplement CI/CD pipelines that automatically test and deploy model updates
   • DManually deploy model updates after each training session
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A data science team is developing a churn prediction model. They have split the data into training, validation, and test sets. After training several models, they notice that the model with the best validation accuracy performs poorly on the test set. What is the most likely cause of this discrepancy?

   • AThe test set is too small to provide a reliable evaluation
   • BThe model is overfitting to the validation set due to hyperparameter tuning
   • CThe validation set is not representative of the test set
   • DThe training set contains data leakage from the test set
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   During feasibility analysis for an AI project, which of the following is a key consideration under organizational feasibility?

   • AWhether the existing IT infrastructure can support the model's computational needs.
   • BWhether the organization has the necessary skills and stakeholder support to adopt the solution.
   • CWhether the model can achieve the required accuracy on test data.
   • DWhether the required data is available in sufficient quantity and quality.
   Reveal Solution Discussion Select an answer first