Question 970
An applicant-scoring model was trained largely on one region yet is intended for nationwide decisions. The system owner has requested a decision on the reported gap. Which observation is the best evidence that the correction works?
Explanation
Correct Answer
The key evidence is the observed result: representative subgroup tests meet predefined acceptance criteria. It is relevant because sampling imbalance threatens generalization and fair treatment.
Incorrect Answers
That observation would verify a control associated with: Data governance was not applied to collection and model development. It does not establish the needed acceptance outcome. The direct evidence is “Representative subgroup tests meet predefined acceptance criteria”, testing: Sampling imbalance threatens generalization and fair treatment.
This checks a different technical dependency: Single-attribute summaries may hide intersectional harms. The case requires proof of “Representative subgroup tests meet predefined acceptance criteria”, which establishes the relevant condition: Sampling imbalance threatens generalization and fair treatment.
That observation would verify a control associated with: New evaluation data matches live conditions and decision requirements. It does not establish the needed acceptance outcome. The direct evidence is “Representative subgroup tests meet predefined acceptance criteria”, testing: Sampling imbalance threatens generalization and fair treatment.