Microsoft Start Focus
Log In Create Account

Microsoft workspace

Detailed Explanations

A detailed answer review with side-by-side rationale, distractor analysis, related questions, and weak-topic practice.

Certification
Build Quiz

Microsoft catalog review

Describe considerations for fairness in an AI solution

Correct: D
Your answer Not answered Answer this in practice or the daily question to sync here.
Correct answer D Evaluate model outcomes and error rates across relevant applicant groups, then review tradeoffs with domain experts
Result Awaiting answer Synced from this browser when available.
Confidence Unset Use the practice page confidence controls to calibrate this item.

What is the best way to assess fairness before deploying a loan approval model?

A Remove all demographic data from evaluation so reviewers cannot see disparities
B Train on the largest dataset available and deploy if the overall approval rate increases
C Choose the threshold that maximizes profit and treat fairness as a post-release issue
D Evaluate model outcomes and error rates across relevant applicant groups, then review tradeoffs with domain experts
1. Answer captured 2. Key checked 3. Rationale review 4. Retry weak topic

Detailed explanation

Catalog rationale

Correct answer: D

Pre-release fairness assessment should combine metrics with context. Decision thresholds can change who benefits and who is harmed, so they should be reviewed deliberately.

Key concept Describe considerations for fairness in an AI solution

Microsoft Certified: Azure AI Fundamentals (AI-901)

Exam tip Map the requirement to the managed Microsoft capability.

Eliminate services that solve infrastructure, data movement, or routing when the stem asks for AI model access or governance.

Microsoft service references

02 Microsoft Certified- Azure AI Fundamentals (AI-901) 01 Identify AI concepts and capabilities 01 Describe principles of responsible AI 01 Describe considerations for fairness in an AI solution

Why the wrong answers are wrong

A

Incorrect. Evaluation often needs protected attributes to measure disparities responsibly.

B

Incorrect. Larger data and higher approval rates do not prove fair treatment.

C

Incorrect. Fairness should be addressed before deployment, not only after harm occurs.