{"id":"W6891697460","doi":"10.48448/c49p-n242","title":"Challenges in Applying Explainability Methods to Improve the Fairness of NLP Models","year":2022,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"National Research Council Canada","funders":"","keywords":"Training set; Interpretability; Key (lock); Business intelligence","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":[0.124914,0.001862007,0.002475193,0.00454636,0.003123487,0.01141124,0.005209887,0.003641117,0.006428355],"category_scores_gemma":[0.3626447,0.001161329,0.002924487,0.003746387,0.01056628,0.01873198,0.01059658,0.01275568,0.001441233],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.005583848,"about_ca_system_score_gemma":0.009164555,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.006357871,"about_ca_topic_score_gemma":0.005889237,"domain_scores_codex":[0.8982893,0.07407778,0.00452595,0.007601071,0.01437608,0.001129882],"domain_scores_gemma":[0.5172693,0.4165189,0.01023476,0.04070805,0.01361964,0.001649279],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0001095417,0.00009751529,0.006750866,0.0009797743,0.0005419628,0.0001361958,0.001836752,0.03020816,0.0006511901,0.7623498,0.00732114,0.1890171],"study_design_scores_gemma":[0.00002590764,0.00002301621,0.000515229,0.0003138357,0.0000430033,0.00004801979,0.0001852819,0.05934856,0.0005961166,0.928511,0.0103574,0.00003265811],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.004729259,0.004009106,0.9594262,0.02333545,0.0004642799,0.0001741001,0.0002990332,0.0006328005,0.006929774],"genre_scores_gemma":[0.2640761,0.004769868,0.7152852,0.00726833,0.002218425,0.0009960918,0.0008103445,0.001343976,0.00323172],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.124914,"threshold_uncertainty_score":0.6606162,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1271227836595832,"score_gpt":0.3973989403663739,"score_spread":0.2702761567067907,"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."}}