{"id":"W4416074431","doi":"10.48550/arxiv.2506.03913","title":"When Fairness Isn't Statistical: The Limits of Machine Learning in Evaluating Legal Reasoning","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":"","keywords":"Normative; Cluster analysis; Refugee; Statistical inference; Philosophy of law; Statistical model","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":false,"about_ca":true,"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":[0.1564377,0.0009798029,0.001843358,0.004258904,0.004443433,0.01010507,0.002829672,0.003238965,0.002156585],"category_scores_gemma":[0.4894971,0.0005560022,0.00096521,0.006003334,0.02201903,0.01450782,0.007651618,0.007049792,0.0005527831],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.007088249,"about_ca_system_score_gemma":0.007648291,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.02277289,"about_ca_topic_score_gemma":0.02165519,"domain_scores_codex":[0.8673097,0.0970127,0.00510627,0.007665669,0.0212644,0.001641105],"domain_scores_gemma":[0.4455627,0.4741579,0.02701308,0.03282974,0.017497,0.002939505],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.001149197,0.0004055073,0.2358521,0.0006536869,0.0008366968,0.0001580948,0.008802114,0.05981495,0.0007097654,0.3945096,0.01317238,0.283936],"study_design_scores_gemma":[0.0001020356,0.000137764,0.03402566,0.0004716936,0.00008757708,0.0001074336,0.00206726,0.125484,0.001481223,0.827869,0.008042431,0.0001237579],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.4592023,0.005703599,0.4229816,0.05636718,0.0006264498,0.0004476891,0.00147525,0.000563352,0.0526326],"genre_scores_gemma":[0.9543254,0.0003871074,0.04178326,0.002103175,0.0002630572,0.0002025144,0.0003413195,0.0001140693,0.0004801209],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.1564377,"threshold_uncertainty_score":0.8273317,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1386095895008183,"score_gpt":0.4396049979263122,"score_spread":0.3009954084254939,"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."}}