{"id":"W4200325561","doi":"10.1109/istas52410.2021.9629201","title":"Protecting marginalized communities by mitigating discrimination in toxic language detection","year":2021,"lang":"en","type":"article","venue":"","topic":"Hate Speech and Cyberbullying Detection","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Concordia University","funders":"","keywords":"Generalizability theory; Computer science; Language identification; Language model; Classifier (UML); Artificial intelligence; Machine learning; Identification (biology); Natural language processing; Natural language; 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":[0.0003997478,0.0000918967,0.00009911924,0.00009338515,0.0002181865,0.0002039313,0.0001887048,0.00005344247,0.00004451215],"category_scores_gemma":[0.00009951723,0.00009410753,0.00003367688,0.0004815126,0.00001529792,0.0004718436,0.0001044816,0.0002535319,0.00001758463],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00007744695,"about_ca_system_score_gemma":0.00002692651,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001278189,"about_ca_topic_score_gemma":0.002468836,"domain_scores_codex":[0.9989966,0.0002766386,0.0001812404,0.000175653,0.0001676748,0.0002022599],"domain_scores_gemma":[0.9995093,0.0000812342,0.00005811511,0.0002619769,0.00005507706,0.00003427978],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.000006274197,0.0000733802,0.000283077,0.0000551529,0.0000081915,0.00003632802,0.0112193,0.00004284746,0.6176621,0.001279737,0.00004891948,0.3692847],"study_design_scores_gemma":[0.0003933802,0.00004775136,0.0005745166,0.00007283765,0.000002416451,0.00006075171,0.01235305,0.03942607,0.9459288,0.0006921923,0.0002727656,0.0001754567],"study_design_candidate":"bench_or_experimental","study_design_consensus":"bench_or_experimental","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.653931,0.0001161296,0.3413319,0.000378561,0.0001327652,0.0001288506,6.015291e-7,0.0002144925,0.003765747],"genre_scores_gemma":[0.98676,0.000008026051,0.01217272,0.0001168184,0.00002165127,0.00004079982,0.000007714209,0.000007308931,0.0008649352],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3691093,"threshold_uncertainty_score":0.3837593,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01360007845269604,"score_gpt":0.2412180201285734,"score_spread":0.2276179416758773,"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."}}