mediumMultiple Choice
Generative AI Leader Practice Question: A financial services company uses a generative AI…
A financial services company uses a generative AI model to summarize customer complaints. They notice that summaries for certain demographics consistently omit negative sentiment. Which responsible AI practice should they apply FIRST to address this bias?
⚠ Common exam trap
Generative AI Leader often tests the misconception that technical adjustments like temperature reduction can fix bias, but bias mitigation requires a systematic approach starting with evaluation.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Evaluate the model's outputs for bias using a diverse test set that represents all customer demographics
Evaluating the model's outputs for bias using a diverse test set is the first step because it directly identifies and quantifies the bias in the summaries across different demographics. This assessment provides the necessary data to understand the extent of the problem and informs subsequent mitigation strategies. Without this evaluation, any other action would be premature or ineffective.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Store all prompts and responses in Cloud Logging for auditing
Why it's wrong here
Cloud Logging captures prompts and responses for retrospective auditing, but logging alone neither identifies nor remediates the sentiment omission. It is tempting because audit trails underpin accountability and are the correct first step when the requirement is traceability of model inputs and outputs, not bias mitigation.
- ✗
Implement SynthID watermarking on all generated summaries
Why it's wrong here
SynthID watermarking marks AI-generated content for provenance detection; it cannot detect or correct demographic sentiment skew in summaries. It is tempting because watermarking supports transparency and content authenticity, which is the right control when the requirement is proving a summary originated from the model rather than addressing biased omissions.
- ✗
Reduce the temperature parameter of the LLM to 0.1 to make outputs more deterministic
Why it's wrong here
Lowering temperature to 0.1 reduces sampling randomness, yet the bias stems from training data or prompts, so deterministic decoding reproduces the same skewed omissions consistently. It is tempting because temperature tuning is the right control when outputs must be reproducible for testing or regulated decision consistency.
- ✓
Evaluate the model's outputs for bias using a diverse test set that represents all customer demographics
Why this is correct
Measuring outputs against a test set spanning all demographics surfaces where sentiment is systematically dropped, quantifying the disparity before any remediation. This satisfies the stem's first-step requirement because you cannot correct bias without evidence of which groups are affected and how severely.
Go deeper
Related to this question
About these practice questions
One of 1,008 original Generative AI Leader practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official Google Cloud exam blueprint
This Generative AI Leader practice question is part of Courseiva's free Google Cloud certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the Generative AI Leader exam.