{"id":"W4417300428","doi":"10.48550/arxiv.2505.12181","title":"Reliable fairness auditing with semi-supervised inference","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Ethics and Social Impacts of AI","field":"Social Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; University of Toronto","keywords":"Imputation (statistics); Estimator; Audit; Missing data; Inference; Variance (accounting)","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.04402163,0.00108947,0.001912812,0.001640049,0.00187807,0.003352677,0.003569849,0.002295147,0.00232955],"category_scores_gemma":[0.1771876,0.0008411662,0.001132873,0.001717183,0.003851974,0.004826381,0.005570083,0.004474905,0.0008724831],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001931814,"about_ca_system_score_gemma":0.006535982,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00308894,"about_ca_topic_score_gemma":0.003501661,"domain_scores_codex":[0.9552125,0.03323645,0.001645143,0.00361193,0.005297449,0.000996471],"domain_scores_gemma":[0.8257441,0.1089796,0.017099,0.03455104,0.01198979,0.00163648],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0008785893,0.000527102,0.03810881,0.0005566163,0.0006219641,0.0003044324,0.001384622,0.3928207,0.003488468,0.2421393,0.01477359,0.304396],"study_design_scores_gemma":[0.00005244546,0.00006606765,0.001740951,0.00007626411,0.0000259662,0.000066884,0.0000858335,0.7830226,0.0023826,0.2106954,0.001745938,0.00003899322],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01901877,0.0002014923,0.9767693,0.001096264,0.00006310044,0.0001333469,0.0002351632,0.0005183924,0.0019642],"genre_scores_gemma":[0.704904,0.0002116601,0.2912193,0.0006924968,0.000247977,0.0004026754,0.0006433396,0.0001884791,0.001490003],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.04402163,"threshold_uncertainty_score":0.2328114,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.09059676199453812,"score_gpt":0.3768076159206245,"score_spread":0.2862108539260864,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}