Align an AI Strategy with Microsoft Responsible AI Policies AB-731 Practice Questions (Page 9)

Questions 41–45

1. 41foundation · easy

   Which practice enhances transparency in AI systems?

   • AProviding clear documentation of the model's purpose, data, and limitations.
   • BAvoiding any communication about how the AI works.
   • CKeeping the model's algorithms proprietary and secret.
   • DUsing complex models that are difficult to interpret.
   Reveal Solution Discussion Select an answer first
2. 42foundation · easy

   What does inclusiveness in AI primarily aim to achieve?

   • AEnsuring AI systems are accessible and useful to people with diverse backgrounds and abilities.
   • BMaking AI systems as complex as possible to challenge users.
   • CLimiting AI use to only certain groups to reduce complexity.
   • DUsing only English language data to simplify development.
   Reveal Solution Discussion Select an answer first
4. 43expert · hard

   A university is developing an AI system to screen student applications. The admissions team wants to ensure fairness, but the historical data reflects past biases in admissions. The university also wants to maintain transparency for applicants. What is the best approach?

   • AUse fairness metrics to evaluate the model and adjust the decision threshold to reduce disparities, while documenting the process.
   • BKeep the model's decision process secret to avoid controversy.
   • CRemove all demographic data from the model and assume it is fair.
   • DUse the historical data as-is because it reflects past decisions.
   Reveal Solution Discussion Select an answer first
5. 44expert · hard

   A company is developing an AI model to screen job applicants. The model is highly accurate overall but shows a significant accuracy gap for applicants with non-traditional educational backgrounds. The company must decide whether to deploy the model now or delay. What should the company do?

   • ADelay deployment until the accuracy gap is investigated and mitigated, and document the decision.
   • BDeploy the model only for applicants with traditional backgrounds to avoid the gap.
   • CUse a different model that is less accurate overall but has no gap.
   • DDeploy the model now and monitor the gap over time, adjusting if needed.
   Reveal Solution Discussion Select an answer first
6. 45application · medium

   A hospital is deploying an AI system that recommends medication dosages for patients. The system must be reliable and safe. What should the hospital do before full deployment?

   • AUse the AI system only for non-critical patients to avoid risk.
   • BRely on the AI vendor's claims that the system is safe and skip internal testing.
   • CRun a pilot study with a small group of clinicians and compare AI recommendations against standard protocols.
   • DDeploy the system immediately to all departments to gather real-world data quickly.
   Reveal Solution Discussion Select an answer first