{"id":"W4412703947","doi":"10.1145/3696630.3731674","title":"An Empirical Study on the Impact of Gender Diversity on Code Quality in AI Systems","year":2025,"lang":"en","type":"article","venue":"","topic":"Software Engineering Research","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Saskatchewan","funders":"","keywords":"Diversity (politics); Computer science; Code (set theory); Quality (philosophy); Gender diversity; Empirical research; Programming language; Mathematics; Sociology; Statistics; Business; Anthropology; Epistemology; Philosophy","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":[0.01214342,0.0002383091,0.0002599655,0.002035018,0.001302842,0.001589357,0.0006891678,0.000534763,0.003833749],"category_scores_gemma":[0.1056759,0.0002377062,0.0003071938,0.002181708,0.001885677,0.002614357,0.001994651,0.0009207939,0.0004911431],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001550309,"about_ca_system_score_gemma":0.001958988,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003436978,"about_ca_topic_score_gemma":0.005768926,"domain_scores_codex":[0.9899429,0.00498781,0.0006269024,0.001036308,0.002669637,0.0007364046],"domain_scores_gemma":[0.7336745,0.187392,0.04612719,0.006075374,0.02002257,0.00670829],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0002844498,0.0002606465,0.9520483,0.0001714946,0.00004828658,0.0001725825,0.020378,0.0001988945,0.0004663597,0.0009591003,0.0006175615,0.02439441],"study_design_scores_gemma":[0.00002377882,0.0005132713,0.9683732,0.0001781049,0.00004171487,0.0002907925,0.02490345,0.001217852,0.000763879,0.0008600014,0.002809854,0.00002408934],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9969985,0.0001559844,0.0006319114,0.0001896939,0.000006920005,0.00002513068,0.00009444235,0.000004373859,0.001893091],"genre_scores_gemma":[0.9990636,0.00007261363,0.0003601371,0.00004591926,0.000008931625,0.00002998891,0.00005493761,0.000006008408,0.0003579256],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01214342,"threshold_uncertainty_score":0.06422132,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1336716054099922,"score_gpt":0.4461826592896935,"score_spread":0.3125110538797014,"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."}}