Manage Accountability Documentation and Audit Trail PMI-CPMAI Practice Questions (Page 5)

Questions 21–25

1. 21foundation · easy

   What is the main benefit of using version control for models, data, and training processes?

   • AIt guarantees that the model will pass all regulatory audits
   • BIt eliminates the need for data governance policies
   • CIt automatically improves the accuracy of the AI model
   • DIt ensures that every change is tracked and that experiments can be reproduced
   Reveal Solution Discussion Select an answer first
2. 22expert · hard

   A technology company is developing an AI model for hiring. The team has documented the decision to use a particular algorithm, but the documentation is stored in a wiki that is not version-controlled. During an audit, the auditors find that the wiki page has been edited multiple times, and they cannot determine what the original decision was. What is the best way to handle this?

   • AUse the wiki's page history to reconstruct the original decision and then migrate the documentation to a version-controlled system.
   • BDelete the wiki page and create a new one with the current understanding.
   • CAsk the original decision-makers to recall what was decided and document it now.
   • DAccept the current version of the wiki page as the official decision record.
   Reveal Solution Discussion Select an answer first
4. 23application · medium

   A data science team is working on a model that uses a dataset that is updated monthly. The team needs to ensure that they can reproduce the model exactly as it was trained in a previous month. What is the best practice?

   • ARecord the date of the dataset update in the model documentation.
   • BKeep the latest version of the dataset and assume it is similar to the previous version.
   • CCreate a snapshot of the dataset at the time of training and store it with the model version.
   • DStore only the model code and not the data, as the data can be re-downloaded.
   Reveal Solution Discussion Select an answer first
5. 24application · medium

   A financial services firm is developing a credit scoring model. The data science team uses a shared network drive to store datasets and model files, and they manually rename files with version numbers (e.g., 'data_final_v3.csv'). During an internal audit, the auditors cannot determine which dataset version was used to train the deployed model. What should the firm implement to address this?

   • AA policy requiring all team members to use a standardized file naming convention.
   • BA manual logbook where team members record every file change.
   • CA formal version control system for datasets and model artifacts, with automated tracking of changes and metadata.
   • DA requirement to store all files in a single folder on the shared drive.
   Reveal Solution Discussion Select an answer first
6. 25application · medium

   A bank's AI governance committee must report to the board on the performance and risks of its AI models. The committee has collected data on model accuracy, fairness metrics, and incident reports. What is the best way to present this information to the board?

   • AA detailed technical report with all raw data and model equations.
   • BA verbal presentation with no written documentation.
   • CA dashboard that shows key performance indicators, risk levels, and trends over time, with the ability to drill down into details.
   • DAn email summarizing the highlights, with no supporting data.
   Reveal Solution Discussion Select an answer first