An AI system used by a government agency to assess eligibility for public housing produces systematically lower scores for families with non-English surnames. The agency's AI vendor claims the model does not use ethnicity or national origin as input features. Which investigation approach would BEST determine whether discrimination is occurring?

B is correct because matched pair testing directly tests for proxy discrimination by isolating the effect of the surname (which carries national origin information) while controlling for all other variables. If scores differ systematically for matched profiles with different surnames, this is direct evidence of proxy discrimination, regardless of whether national origin is an explicit feature. Code review (A) only shows direct feature use, not indirect proxy effects. Surveys (C) provide anecdotal evidence but cannot establish statistical discrimination. Vendor certification (D) is an attestation with no evidentiary value for the specific discrimination question.