{"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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0008407178,0.0001221201,0.0001711599,0.0004348597,0.00008830683,0.00007611267,0.000264053,0.00005388834,0.000004914553],"category_scores_gemma":[0.00007665656,0.0001098325,0.00004755944,0.001376877,0.00001507824,0.000383764,0.00008904968,0.0001203141,0.00001310802],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005774361,"about_ca_system_score_gemma":0.00006992579,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008354714,"about_ca_topic_score_gemma":0.003362659,"domain_scores_codex":[0.9985508,0.0001588289,0.0003516668,0.0003780015,0.0003283675,0.0002323152],"domain_scores_gemma":[0.9991493,0.00007795567,0.00009158344,0.0005247982,0.00008927043,0.00006712539],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0002744048,0.00289166,0.04364605,0.0004006299,0.0004094427,0.0165121,0.3710262,0.2551894,0.1192717,0.004988462,0.0009551466,0.1844348],"study_design_scores_gemma":[0.001046219,0.000125216,0.007077516,0.0000357149,0.00001123105,0.0002173202,0.01367525,0.9684946,0.008898872,0.0002521866,0.000004954317,0.0001609533],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9247079,0.00001388606,0.0738852,0.0002047834,0.0001408934,0.0004926009,0.000006738393,0.0004046847,0.0001433077],"genre_scores_gemma":[0.9987843,0.000004549093,0.0007879892,0.00002640118,0.00001784005,0.00008488657,0.000005063115,0.00001076029,0.0002781896],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.7133052,"threshold_uncertainty_score":0.9982488,"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."}}