{"id":"W4287891037","doi":"10.18653/v1/2022.naacl-main.188","title":"How Gender Debiasing Affects Internal Model Representations, and Why It Matters","year":2022,"lang":"en","type":"article","venue":"Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies","topic":"Topic Modeling","field":"Computer Science","cited_by":15,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Azrieli Foundation","keywords":"Debiasing; Computer science; Data science; Linguistics; Cognitive science; Psychology; Philosophy","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"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":[0.004941644,0.000491545,0.0004187046,0.000750043,0.001001326,0.006259785,0.001076666,0.001409497,0.01538766],"category_scores_gemma":[0.0777735,0.0005870132,0.0005270726,0.0007897764,0.001602778,0.01252798,0.00205473,0.001957682,0.005289509],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001110306,"about_ca_system_score_gemma":0.0008299688,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008081241,"about_ca_topic_score_gemma":0.006203726,"domain_scores_codex":[0.9974852,0.00110001,0.0001071685,0.0006128866,0.0004651199,0.0002295875],"domain_scores_gemma":[0.9812226,0.01002257,0.001132279,0.003975575,0.003203602,0.0004434605],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.002690291,0.0003611422,0.1458813,0.000498537,0.0003381561,0.001011571,0.02342298,0.02130312,0.03443258,0.1473278,0.07095989,0.5517726],"study_design_scores_gemma":[0.0002781969,0.0004886402,0.0843522,0.0006534795,0.0007352931,0.00199125,0.0274493,0.18639,0.03893632,0.5896075,0.06871583,0.0004021059],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6859344,0.003031476,0.1758005,0.02112909,0.001186506,0.0001305001,0.003330485,0.004776014,0.104681],"genre_scores_gemma":[0.9789109,0.0003944482,0.01265957,0.000638677,0.00008514809,0.0000246644,0.001191156,0.001456798,0.004638623],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01538766,"threshold_uncertainty_score":0.05147684,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02677859955779483,"score_gpt":0.2586122176168469,"score_spread":0.2318336180590521,"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."}}