{"id":"W4292583031","doi":"10.5539/elt.v15n9p82","title":"Examining the Dependability and Practicality of Analytic Rubric of Summary Writing Using Multivariate Generalizability Theory: Focusing on Japanese University Students with Lower-Intermediate Proficiency in English","year":2022,"lang":"en","type":"article","venue":"English Language Teaching","topic":"Educational Technology and Assessment","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Rubric; Generalizability theory; Psychology; Mathematics education; Context (archaeology); Constructive; Language proficiency; Dependability; Computer science","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.04628388,0.0007942724,0.0005820012,0.002678093,0.0008869899,0.001660634,0.0009917855,0.0006104364,0.001104289],"category_scores_gemma":[0.2426359,0.0004035433,0.00185776,0.001604268,0.001877202,0.002427834,0.002018036,0.001144328,0.0002087201],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001009163,"about_ca_system_score_gemma":0.001220185,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002439886,"about_ca_topic_score_gemma":0.003215651,"domain_scores_codex":[0.9577305,0.0201853,0.005304096,0.003709432,0.01241135,0.0006593627],"domain_scores_gemma":[0.7015567,0.201188,0.0279039,0.02893775,0.03882827,0.001585369],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0004831888,0.0004590897,0.8009669,0.0004704855,0.0007211551,0.0002184695,0.02428448,0.002216425,0.009691255,0.001848229,0.0007217758,0.1579185],"study_design_scores_gemma":[0.00006400684,0.001690377,0.9471823,0.0002179264,0.0004171626,0.0006437921,0.00879381,0.02524608,0.009744543,0.00303967,0.002828474,0.0001318822],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9297777,0.0003618245,0.06371032,0.0002347934,0.00005445978,0.0007933261,0.0001086239,0.0001316622,0.004827318],"genre_scores_gemma":[0.9863176,0.00008061628,0.01279922,0.00003144611,0.00002077303,0.0003307742,0.0001229962,0.00002856911,0.0002679348],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.04628388,"threshold_uncertainty_score":0.2447755,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01854952372686935,"score_gpt":0.2954321271611949,"score_spread":0.2768826034343255,"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."}}