{"id":"W2946581393","doi":"10.7202/1059172ar","title":"Approches psychométrique et didactique de la validité d’une évaluation externe en mathématiques : quelles complémentarités et quelles divergences ?","year":2019,"lang":"fr","type":"article","venue":"Mesure et évaluation en éducation","topic":"Mathematics Education and Teaching Techniques","field":"Social Sciences","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Humanities; Philosophy; Mathematics","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch","metaepi_narrow","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.0522205,0.0005032021,0.0005170489,0.0003601612,0.00035629,0.000538505,0.0006057205,0.0006673107,0.002230856],"category_scores_gemma":[0.002327248,0.0005349232,0.0002268379,0.0006725379,0.000269929,0.00134412,0.00009483664,0.000840072,0.0002617761],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009245395,"about_ca_system_score_gemma":0.002402121,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.006728735,"about_ca_topic_score_gemma":0.003211471,"domain_scores_codex":[0.9672081,0.02933801,0.0009841399,0.0007291493,0.001273121,0.0004674381],"domain_scores_gemma":[0.9928457,0.005053963,0.0008873327,0.0005909832,0.0004569091,0.0001651743],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.00001479624,0.0008424022,0.009773304,0.0002606436,0.00007520736,4.293059e-7,0.6117406,0.0002635608,0.001165324,0.363965,0.006941717,0.004956915],"study_design_scores_gemma":[0.0008249604,0.0002468129,0.2118912,0.001301238,0.0003030099,0.00003124382,0.05656321,0.007939586,0.003563969,0.6560282,0.06008429,0.001222248],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7636308,0.0009881413,0.1360153,0.05503225,0.002500104,0.002178724,0.00003253861,0.0004741655,0.03914794],"genre_scores_gemma":[0.9345382,0.004977186,0.05239928,0.001350873,0.0003286007,0.0003843127,0.000374143,0.00007416232,0.005573272],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.5551775,"threshold_uncertainty_score":0.9998856,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.061065612239627,"score_gpt":0.4275992748591815,"score_spread":0.3665336626195546,"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."}}