Ensure Responsible AI Use CY0-001 Practice Questions (Page 3)

Questions 11–15

1. 11expert · hard

   A financial regulator is reviewing an AI system used by a bank to detect fraudulent transactions. The system is a deep learning model that is highly accurate but operates as a 'black box'. The regulator demands that the bank provide a clear explanation for why a specific transaction was flagged as fraudulent. The bank's data science team argues that the model's complexity is necessary for accuracy. What is the most appropriate way for the bank to satisfy the regulator?

   • AProvide the regulator with the model's source code and training data.
   • BReplace the deep learning model with a logistic regression model, which is inherently interpretable.
   • CRefuse the request, citing the model is proprietary and its internal logic is a trade secret.
   • DUse a model-agnostic explanation method (e.g., SHAP) to provide feature importance scores for the specific flagged transaction.
   Reveal Solution Discussion Select an answer first
2. 12expert · hard

   A healthcare research institute is using an AI model to analyze patient records to identify risk factors for a rare disease. The data is highly sensitive and subject to strict privacy regulations. The institute wants to share the model's insights with other researchers without sharing the raw patient data. Which approach best achieves this goal?

   • APublish the model's weights and architecture on a public repository.
   • BAnonymize the patient records by removing names and social security numbers, then share the dataset.
   • CTrain the model using federated learning, where the model is trained across multiple decentralized datasets without exchanging the data itself.
   • DShare the model's predictions for a new, synthetic dataset that mimics the original data's distribution.
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A social media company is launching a new AI feature that automatically summarizes news articles for users. Before launch, the company wants to evaluate the social impact of this feature. Which evaluation is most important to conduct?

   • AVerify that the AI model can process articles faster than the previous version.
   • BCalculate the cost savings from reduced human editorial review of news articles.
   • CMeasure the increase in user engagement time on the platform after the feature is released.
   • DAssess whether the summaries could amplify misinformation or create filter bubbles by presenting a single, biased perspective.
   Reveal Solution Discussion Select an answer first
5. 14foundation · easy

   Which of the following is an example of an accountability mechanism for an AI system?

   • ADeploying the model on more powerful hardware
   • BMaintaining detailed audit logs of model decisions and actions
   • CUsing a larger training dataset to improve accuracy
   • DAutomatically deleting all user data after processing
   Reveal Solution Discussion Select an answer first
6. 15expert · hard

   A government agency is deploying an AI system to screen visa applications. The system is intended to reduce processing time. During a risk assessment, the team identifies that the model has a higher false-positive rate for applicants from certain countries. The agency has a legal obligation to ensure non-discrimination. What is the most critical next step?

   • AImmediately halt the deployment and discard the model.
   • BConduct a disparate impact analysis to quantify the difference in outcomes across protected groups and determine if it is legally justifiable.
   • CAdd more data from the affected countries to the training set to improve accuracy.
   • DRequire a human to review every decision made by the AI system.
   Reveal Solution Discussion Select an answer first