{"id":"W4392669787","doi":"10.18653/v1/2023.eval4nlp-1.9","title":"Can a Prediction’s Rank Offer a More Accurate Quantification of Bias? A Case Study Measuring Sexism in Debiased Language Models","year":2023,"lang":"en","type":"article","venue":"","topic":"Hate Speech and Cyberbullying Detection","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Debiasing; Computer science; Sentence; Rank (graph theory); Statistics; Benchmark (surveying); Artificial intelligence; Natural language processing; Econometrics; Psychology; Mathematics; Social psychology","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.00906174,0.001038935,0.0007787221,0.001100207,0.0005302072,0.001566081,0.0007424916,0.001149945,0.001665867],"category_scores_gemma":[0.04328873,0.00028379,0.000414228,0.0005963871,0.001384415,0.00231881,0.001263842,0.001615587,0.0008300305],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006312946,"about_ca_system_score_gemma":0.0008896624,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00277772,"about_ca_topic_score_gemma":0.00364792,"domain_scores_codex":[0.9951521,0.002196127,0.0002429322,0.0008118093,0.00129396,0.0003031248],"domain_scores_gemma":[0.9694336,0.02130906,0.00197885,0.004124464,0.002595624,0.0005583841],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.003645494,0.0007000208,0.266811,0.0006682673,0.0006550968,0.001087599,0.002529129,0.1910557,0.06122814,0.01223672,0.009626399,0.4497564],"study_design_scores_gemma":[0.00009156083,0.001572423,0.06536925,0.0001399267,0.0001959576,0.0008909776,0.001120696,0.8420256,0.05946199,0.02165706,0.00731146,0.000163082],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.887264,0.0008330991,0.1044021,0.001158308,0.0001271584,0.00007516155,0.0007123438,0.001387218,0.004040624],"genre_scores_gemma":[0.9738202,0.00009012756,0.02481107,0.0001250886,0.00003624084,0.0000218782,0.00039146,0.0001141217,0.0005897995],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.00906174,"threshold_uncertainty_score":0.04792362,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1003945971048337,"score_gpt":0.2963982998557292,"score_spread":0.1960037027508955,"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."}}