GCP Generative AI Leader Practice Question 100: Fairness Across User Groups
Google Cloud Generative AI Leader · Test 10 · question 10 of 10
A bank's gen AI loan-summary assistant is reviewed after launch. Analysts find that applications from one neighborhood are summarized with noticeably more hedging language and lower approval recommendations than similar applications from other neighborhoods, even after controlling for the applicants' financial details. Which responsible AI principle is most directly at stake?
Show answer and explanation
Answer:
The finding describes a system-wide disparity in outcomes for one group of applicants that is not explained by their financial details, which is precisely what the fairness principle addresses: a gen AI system should treat comparable inputs comparably across groups, and any disparity that correlates with a protected characteristic like neighborhood demographics needs investigation and mitigation, often by auditing training data and evaluating outputs by subgroup.
Option B, accountability, matters for governance but describes who owns the fix, not the specific harm found in the review. Option C, transparency, is about disclosing that AI was used at all, which is a separate obligation from whether its output is equitable. Option D, privacy, concerns protecting data from unauthorized access, not the pattern of biased outcomes described here.
Reference: Responsible AI practices, fairness