Sustaining AI Transformation and Continuous Improvement CAIPM Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   An online media company uses a machine learning model to recommend articles. The model's click-through rate (CTR) has been stable, but the company's revenue per user has declined. The data science team suspects that the model is now recommending more clickbait articles that generate clicks but fewer subscriptions. The team wants to improve the model to align with business goals. What should they do?

   • AAdd a business metric, such as subscription conversion rate, to the model's optimization objective.
   • BSwitch to a completely different algorithm to see if it improves revenue.
   • CKeep the model as is because CTR is stable and the model is performing well.
   • DRetrain the model more frequently with the latest click data to improve CTR.
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A retail company deployed a demand-forecasting model six months ago. The data science team notices that the model's accuracy has declined over the last two weeks, but the business team reports no change in sales patterns. The data engineering team confirms that the data pipeline is still running without errors. What should the team do first to diagnose the issue?

   • ACompare the distribution of recent input features against the training data distribution
   • BRetrain the model immediately with the most recent data
   • CRoll back to the previous model version
   • DIncrease the frequency of model retraining to weekly
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   An e-commerce company has a recommendation model that is retrained monthly. The data science team wants to introduce a new feature that requires a different data schema. The compliance team requires that all model changes be traceable. Which process should the team follow to implement this change safely?

   • ASkip documentation because the change is internal to the data science team
   • BUpdate the production schema and retrain the model in the same release
   • CCreate a new model version with the new schema, test it in staging, and document the change before production
   • DRetrain the model with the new feature but keep the old schema in production
   Reveal Solution Discussion Select an answer first
5. 19foundation · easy

   Which element is essential for ensuring accountability in AI governance?

   • AAvoiding documentation to reduce administrative burden
   • BKeeping decision-making processes opaque to protect intellectual property
   • CClearly defined roles and responsibilities for AI decisions and outcomes
   • DAssigning responsibility to an external vendor only
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   What is a best practice for replicating a successful AI solution in other parts of an organization?

   • ARequiring each team to start from scratch
   • BKeeping the implementation details secret to maintain a competitive edge
   • CLimiting replication to teams with the same data
   • DDocumenting the solution's design, implementation, and lessons learned
   Reveal Solution Discussion Select an answer first