{"id":"W4385894685","doi":"10.18653/v1/2022.blackboxnlp-1.18","title":"Towards Procedural Fairness: Uncovering Biases in How a Toxic Language Classifier Uses Sentiment Information","year":2022,"lang":"en","type":"article","venue":"","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"National Research Council Canada","funders":"","keywords":"Computer science; Classifier (UML); Debiasing; Salient; Leverage (statistics); Artificial intelligence; Sentiment analysis; Machine learning; Natural language processing; Psychology; 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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.000375227,0.0001306881,0.0001350752,0.0002541191,0.00017911,0.0002624917,0.0006073793,0.0000269437,0.0001451024],"category_scores_gemma":[0.0002342287,0.0001242958,0.00004555929,0.0006736846,0.00001635114,0.002162752,0.001017777,0.0002692737,0.00001405198],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002871202,"about_ca_system_score_gemma":0.0001463141,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002641125,"about_ca_topic_score_gemma":0.00005213792,"domain_scores_codex":[0.9986993,0.0001083401,0.0002028617,0.0002134499,0.0004801511,0.0002959094],"domain_scores_gemma":[0.9994735,0.00006726423,0.0001102326,0.0002751397,0.00002620916,0.00004761906],"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":[0.00003752141,0.0001370596,0.00827456,0.00008710928,0.00003048931,0.00009255278,0.02283479,0.8674718,0.0009088136,0.0187966,0.002347041,0.07898169],"study_design_scores_gemma":[0.0009782668,0.0001189531,0.007390141,0.00003570591,0.000005670194,0.00004924001,0.01001535,0.9715084,0.00201122,0.0003745132,0.007070101,0.0004424498],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4531113,0.00003727503,0.5365633,0.003943474,0.0006590028,0.0003587181,0.000004264711,0.0004271469,0.004895465],"genre_scores_gemma":[0.9849371,0.000001237604,0.01371287,0.0007312734,0.00003331918,0.00006989273,0.00001208029,0.000006953336,0.0004952754],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.5318258,"threshold_uncertainty_score":0.5068637,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01775743042092704,"score_gpt":0.2606138412506929,"score_spread":0.2428564108297658,"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."}}