{"id":"W4254924799","doi":"10.1080/0142159x.2020.1830052","title":"Performance assessment: Consensus statement and recommendations from the 2020 Ottawa Conference","year":2020,"lang":"en","type":"article","venue":"Medical Teacher","topic":"Innovations in Medical Education","field":"Medicine","cited_by":86,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Statement (logic); Consensus conference; Medical education; MEDLINE; Political science; Medicine; Library science; Computer science; Law","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.0007430675,0.000108153,0.0001747408,0.00001514249,0.0001379804,0.00002327292,0.0001316262,0.0000994168,0.01591992],"category_scores_gemma":[0.001975025,0.00007121474,0.00002170423,0.0002223758,0.0002734136,0.00003722747,0.00006826859,0.0007249651,0.00008321671],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005749988,"about_ca_system_score_gemma":0.0003763801,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0000463173,"about_ca_topic_score_gemma":0.000006100577,"domain_scores_codex":[0.9984831,0.0001164723,0.0004002608,0.0002383467,0.0005919116,0.0001699225],"domain_scores_gemma":[0.9992934,0.00009186428,0.00008591813,0.000199425,0.0001118943,0.0002174327],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.00002244933,0.0001592781,0.1831595,0.00003973077,0.00008172315,0.00001072526,0.002172423,3.131492e-7,0.00004990059,0.0007673671,0.6698421,0.1436945],"study_design_scores_gemma":[0.002141555,0.0002541523,0.1156663,0.0002432936,0.0001288127,0.00002622375,0.003865615,0.1761749,0.00002892077,0.0001642233,0.7011096,0.0001963546],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5444485,0.00005863347,0.0008262447,0.4468,0.0001999576,0.0002123043,0.000004359181,0.00003857363,0.007411459],"genre_scores_gemma":[0.9761168,0.0001594568,0.003389958,0.01869094,0.0006627005,0.00004629503,0.0001940241,0.00001271373,0.0007271607],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4316683,"threshold_uncertainty_score":0.9849797,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04581137911629056,"score_gpt":0.3664182182468435,"score_spread":0.3206068391305529,"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."}}