hardMultiple Select
Generative AI Leader Practice Question: A health-tech startup is fine-tuning a generative…
A health-tech startup is fine-tuning a generative AI model on electronic health records (EHR) to assist in clinical decision support. They need to ensure responsible AI practices. Which THREE measures should they implement? (Select three.)
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 model outputs for bias across different demographic groups
Option B is correct because systematically evaluating model outputs for bias across demographic groups (e.g., using disaggregated performance metrics like sensitivity, specificity, and equal opportunity difference by age, sex, race, and ethnicity) is essential in clinical AI to detect and mitigate health disparities before deployment. Option C is correct because requiring a human clinician to review all AI-generated recommendations preserves meaningful human oversight, keeping the clinician accountable for the final decision and preventing automation bias or unsafe autonomous actions in clinical decision support. Option D is correct because a Model Card documents the model's intended use, limitations, performance metrics, and ethical considerations, providing the transparency and traceability needed for responsible AI governance and regulatory review. Option A is not appropriate because fully automating all decisions removes human oversight and amplifies automation bias and accountability gaps, which is contrary to responsible AI in healthcare. Option E is not appropriate because training exclusively on data from a single hospital produces narrow, non-generalizable models and can encode site-specific biases, undermining fairness and robustness across diverse patient populations.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Automate all decisions to reduce human error
Why it's wrong here
Fully automating clinical decisions removes the clinician oversight that responsible AI in healthcare demands, and cannot mitigate bias or error in EHR-trained models. It is tempting because automation promises consistency and reduced workload, but human review is required for clinical decision support.
- ✓
Evaluate model outputs for bias across different demographic groups
Why this is correct
Bias evaluation across demographic groups detects disparate performance in the fine-tuned model, a core responsible AI control for clinical decision support. Because EHR data encodes historical inequities, measuring output disparities by subgroup satisfies the requirement to identify and mitigate harm before deployment.
- ✓
Require a human clinician to review all AI-generated recommendations before action
Why this is correct
Clinician review of every AI-generated recommendation provides human oversight for high-stakes clinical decisions, satisfying responsible AI requirements that automated output never directly drives patient care. The clinician retains authority to accept, modify, or reject, ensuring accountability sits with a qualified professional.
- ✓
Publish a Model Card that describes the model's intended use, limitations, and performance
Why this is correct
A Model Card documents intended use, limitations, and performance, giving clinicians and regulators the transparency needed to judge whether the model suits clinical decision support. Publishing it satisfies the responsible AI requirement for disclosed capabilities and constraints rather than opaque deployment.
- ✗
Train the model exclusively on data from a single hospital to ensure consistency
Why it's wrong here
Restricting training to one hospital's records narrows demographic and clinical diversity, embedding that site's coding and prescribing conventions and amplifying bias — the opposite of responsible practise. Single-site data suits a pilot validating local deployment, not a model intended for broad clinical decision support.
Go deeper
Related to this question
About these practice questions
This Generative AI Leader question is part of Courseiva's 1,008-question bank — original exam-style content with full explanations and wrong-answer analysis, never real exam questions or exam dumps. Learn why practice questions differ from exam dumps →
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
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.