Which of the following is typically reviewed during an AI model audit?

The marketing strategy for the model

Training data quality and provenance

A healthcare provider uses an AI model to predict patient readmission risk. The model was developed in-house, but the team is concerned about potential biases that could lead to unequal treatment. The provider wants to mitigate this risk before deployment. Which approach is most effective?

Rely on the model's overall accuracy as a proxy for fairness.

Deploy the model and monitor for complaints from patients.

Conduct a bias audit using fairness metrics across demographic subgroups.

Use a more complex model to capture subtle patterns in the data.

An e-commerce company is developing a recommendation model. The dataset is imbalanced, with a small number of users generating most of the purchases. The team wants to ensure the model performs well for all user segments. Which validation technique is most appropriate?

Use accuracy as the sole metric to judge model performance.

Use stratified cross-validation to preserve the class distribution in each fold.

Evaluate only on the majority class to simplify the analysis.

Use a simple train-test split without considering class distribution.

What is the primary purpose of an AI model audit?

To deploy the model to production

To write the model's documentation

To independently assess the model's compliance, effectiveness, and alignment with governance policies

To improve the model's accuracy by tuning hyperparameters

ISACAAdvanced in AI Audit

Domain 1Objective 1

AI Models, Considerations, and Requirements AAIA Practice Questions (Page 4)

Part of the AI Governance and Risk domain, which accounts for 33% of the AAIA exam.

38questions here

8free pages

10concepts

33%of the exam

Questions 16–20

16foundation · easy
Which of the following is typically reviewed during an AI model audit?
Select an answer first
17application · medium
A healthcare provider uses an AI model to predict patient readmission risk. The model was developed in-house, but the team is concerned about potential biases that could lead to unequal treatment. The provider wants to mitigate this risk before deployment. Which approach is most effective?
Select an answer first
18application · medium
An e-commerce company is developing a recommendation model. The dataset is imbalanced, with a small number of users generating most of the purchases. The team wants to ensure the model performs well for all user segments. Which validation technique is most appropriate?
Select an answer first
19foundation · easy
Which ethical principle is most closely associated with the ability to understand and explain how an AI model makes decisions?
Select an answer first
20foundation · easy
What is the primary purpose of an AI model audit?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to AAIA

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by ISACA. “AAIA” is a trademark of its owner, used for identification only.