A team preparing a production rollout wants to avoid gaps in their governance coverage. Which pillar addresses whether the model performs consistently under unexpected or adversarial inputs?
Select an answer to reveal the explanation.
Short Explanation
You need the pillar that checks what happens when the world doesn't look like the training data. That's robustness, because it covers stability and resilience under unexpected or adversarial inputs.
Full Explanation
The correct answer is robustness, which covers the model's stability and resilience when conditions deviate from training assumptions. Robustness specifically concerns a model's ability to maintain acceptable performance under noisy, unusual, or adversarial inputs, which is distinct from the other four pillars in the governance framework.
The fairness option, which covers whether outcomes differ unjustifiably across demographic groups, is incorrect because fairness concerns equitable treatment across groups, not resilience to unusual inputs.
The transparency option, which covers whether stakeholders can see how the model was built and deployed, is incorrect because transparency concerns visibility into process and design, not runtime stability.
The accountability option, which covers who is responsible for the model's decisions, is incorrect because accountability concerns ownership of outcomes, not the model's technical resilience.
For the exam, keep the five pillars distinct rather than treating them as interchangeable synonyms for 'good governance'.