{"id":"W3208971449","doi":"10.1136/oem-2021-epi.84","title":"O-467 Evaluating the impact of sex and gender on the performance of machine learning for auto encoding of job titles","year":2021,"lang":"en","type":"article","venue":"Oral Presentations","topic":"Healthcare Systems and Public Health","field":"Medicine","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of New Brunswick","funders":"","keywords":"Computer science; Bespoke; Coding (social sciences); Artificial intelligence; Machine learning; Natural language processing","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":true,"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":[0.005072001,0.0004655839,0.0002885987,0.0006291095,0.0003641819,0.001007251,0.0003953134,0.0003604345,0.006198927],"category_scores_gemma":[0.02688629,0.0001452047,0.0004903544,0.0004541501,0.000365238,0.0006500388,0.0005350031,0.0004334592,0.001968638],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006670112,"about_ca_system_score_gemma":0.001091928,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0115263,"about_ca_topic_score_gemma":0.009578204,"domain_scores_codex":[0.9979334,0.001081508,0.0002235624,0.0002886126,0.0003406417,0.0001323313],"domain_scores_gemma":[0.9623969,0.02943156,0.001285466,0.001654694,0.004596679,0.0006347958],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00728072,0.001259103,0.5838836,0.0002971751,0.0002485355,0.0001785975,0.0009792034,0.02719851,0.01219411,0.000527131,0.005316779,0.3606365],"study_design_scores_gemma":[0.0002877408,0.005567061,0.4996387,0.0001691313,0.0002039164,0.0003924468,0.002884697,0.4195665,0.06581078,0.001334546,0.004014086,0.0001304286],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9839054,0.0001372347,0.009380596,0.0001942972,0.000061157,0.0001903993,0.001280933,0.0003448367,0.004505231],"genre_scores_gemma":[0.9828797,0.00004940457,0.01331064,0.00005646291,0.00001281915,0.00009298885,0.00177461,0.00004924252,0.00177429],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.0115263,"threshold_uncertainty_score":0.02682358,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2198778754194947,"score_gpt":0.4890386529592827,"score_spread":0.2691607775397881,"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."}}