AI0-001 AI Security, Ethics and Governance Practice Question
A financial institution uses an AI model to approve loan applications. The model was trained on historical data that included biased lending practices. The bank's ethics committee wants to mitigate bias without removing protected attributes. Which approach best balances fairness and model performance?
⚠ Common exam trap
CompTIA often tests the misconception that simply removing protected attributes (Option B) is sufficient to eliminate bias, when in reality proxy features and correlated variables can perpetuate discrimination.
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
✓
Apply adversarial debiasing during training
Adversarial debiasing is the best approach because it directly optimizes the model to reduce bias during training while preserving predictive accuracy. It uses an adversarial network that tries to predict the protected attribute from the model's predictions, forcing the main model to learn representations that are less correlated with that attribute. This allows the bank to keep protected attributes in the data (as required by the ethics committee) while actively mitigating bias.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Retrain the model using a balanced dataset
Why it's wrong here
Balancing the dataset alters outcome proportions but leaves the model free to learn residual correlations from proxy attributes, so bias persists without a fairness constraint. It is tempting because resampling is a standard preprocessing remedy, and would suit label imbalance rather than historical discrimination encoded in features.
- ✗
Remove all protected attributes from the training data
Why it's wrong here
Dropping protected attributes does not remove bias, because correlated proxies such as postcode or employment history let the model reconstruct the excluded characteristics. It is tempting as a quick anonymisation step, and it would suit scenarios where attributes are irrelevant to the target, not ones where historical lending bias must be actively corrected.
- ✗
Post-process model outputs to adjust for demographic parity
Why it's wrong here
Adjusting outputs after training cannot repair bias learned from historical lending data, and forcing demographic parity can lower accuracy for qualified applicants. It is tempting because post-processing needs no retraining, making it correct where a deployed model must be corrected quickly and parity is the agreed fairness definition.
- ✓
Apply adversarial debiasing during training
Why this is correct
Adversarial debiasing trains a predictor alongside an adversary that tries to infer protected attributes from predictions, penalising reliance on them. This reduces disparate impact while retaining protected attributes in the data, preserving predictive performance better than attribute removal.
About these practice questions
This AI0-001 question is part of Courseiva's 962-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 AI0-001 practice question is part of Courseiva's free CompTIA 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 AI0-001 exam.