An AI auditor is testing an AI model's robustness as part of an assurance engagement. Which test BEST evaluates the model's stability under minor input perturbations?
Select an answer to reveal the explanation.
Short Explanation
Here's the deal — sensitivity analysis is the best fit because you're checking whether tiny input changes make the output jump around. If the model wobbles under minor perturbations, that's a robustness problem. The other checks measure accuracy, fit, or dimensionality, not stability.
Full Explanation
Sensitivity analysis is correct because it directly tests model stability by observing how small, controlled changes in inputs affect outputs — revealing whether the model produces erratic or inconsistent predictions from minor input variations, which is a robustness risk. Confusion matrix analysis is wrong because it measures classification performance across classes, not stability under perturbed inputs. Learning curve analysis is wrong because it diagnoses overfitting versus underfitting by comparing training and validation error across dataset sizes. Principal component analysis is wrong because it is a dimensionality reduction technique, not a robustness test.