Identify Guiding Principles for Responsible AI AI-900 Practice Questions (Page 5)

Questions 21–25

1. 21expert · hard

   A multinational retailer uses an AI system to recommend products. The system is trained on global sales data. In some countries, the system performs well, but in others it shows gender bias in recommendations. The retailer must fix this without losing overall accuracy. What should the retailer do?

   • AUse a single global model and accept the bias as a trade-off for simplicity.
   • BRemove gender-related features from the model to eliminate bias.
   • CCollect more representative data from the affected countries and retrain the model with fairness constraints.
   • DDeploy separate models for each country and tune them individually.
   Reveal Solution Discussion Select an answer first
2. 22application · medium

   A company deploys an AI system that automates hiring decisions. The system occasionally makes errors that could lead to unfair outcomes. The company wants to ensure accountability for the system's decisions. What should the company do?

   • ABlame the AI vendor if the system makes a mistake.
   • BDocument the system's design but do not assign a specific owner.
   • CLet the AI system operate without human oversight to avoid bias.
   • DAssign a human owner who is responsible for the system's decisions and establish an escalation process for disputes.
   Reveal Solution Discussion Select an answer first
4. 23expert · hard

   A credit union uses an AI model to approve loan applications. The model is complex and its decisions are not easily explainable. Regulators require the credit union to explain loan decisions to applicants. The credit union must balance model accuracy and explainability. What should the credit union do?

   • AKeep the complex model and use post-hoc explanation tools to generate explanations.
   • BUse a simpler, interpretable model even if it is slightly less accurate, and document its decision logic.
   • CHire a third-party auditor to explain the model's decisions to applicants.
   • DKeep the complex model and provide applicants with the model's confidence score.
   Reveal Solution Discussion Select an answer first
5. 24expert · hard

   A social media company uses AI to personalize content feeds. The AI uses user behavior data, including private messages. The company wants to improve personalization but must comply with privacy regulations and maintain user trust. What should the company do?

   • AUse all available data, including private messages, to maximize personalization.
   • BUse only aggregated and anonymized data for personalization and obtain user consent for any data use.
   • CStore all data securely and use it for personalization without informing users.
   • DAnonymize the data but do not obtain consent because anonymized data is not personal.
   Reveal Solution Discussion Select an answer first
6. 25application · medium

   A hospital deploys an AI system that predicts patient deterioration. The system was trained on data from one region and now is being used in a different region with different patient demographics and clinical practices. The hospital wants to ensure the system remains reliable. What should the hospital do?

   • ADeploy the model immediately and rely on clinicians to override incorrect predictions.
   • BRetrain the model from scratch using only local data.
   • CValidate the model's performance on local data before full deployment and set up ongoing monitoring for accuracy drift.
   • DUse the model only for patients who match the original training demographics.
   Reveal Solution Discussion Select an answer first