{"id":"W4406891154","doi":"10.55593/ej.28112r3","title":"Innovation in Learning-Oriented Language Assessment (Book Review)","year":2025,"lang":"en","type":"article","venue":"Teaching English as a Second or Foreign Language--TESL-EJ","topic":"Educational and Psychological Assessments","field":"Psychology","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Computer science; Cognitive science; Linguistics; Psychology; Philosophy","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":["metaepi_narrow","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.001855275,0.00040131,0.0005571131,0.000459513,0.000211122,0.00007031066,0.0004782755,0.0002604494,0.0618501],"category_scores_gemma":[0.001397452,0.0003283251,0.0001292583,0.001125523,0.00006941419,0.0002657771,0.0001235511,0.001751138,0.0002083262],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002418665,"about_ca_system_score_gemma":0.0001745501,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002102949,"about_ca_topic_score_gemma":0.00003134736,"domain_scores_codex":[0.9965005,0.0007378009,0.0009732093,0.00080706,0.0003601726,0.0006212812],"domain_scores_gemma":[0.9981367,0.0005870009,0.0003713023,0.0006233283,0.0001795403,0.0001021335],"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":[0.0003321354,0.002474924,0.02306577,0.001013153,0.0004239197,0.0004389817,0.04101494,0.00001492179,0.0007506351,0.2657602,0.601207,0.06350344],"study_design_scores_gemma":[0.003925974,0.0007065684,0.04731949,0.002011852,0.0001185873,0.00004490018,0.05786208,0.00006048761,0.0001462491,0.00209501,0.8846886,0.001020225],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"other","genre_gemma":"empirical","genre_scores_codex":[0.4312831,0.02524567,0.0002776276,0.0005519267,0.001202343,0.0009351704,0.00004258585,0.0003710536,0.5400905],"genre_scores_gemma":[0.724851,0.0002700695,0.0018853,0.05569213,0.000749328,0.0007089602,0.001371624,0.0000840187,0.2143876],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.3257028,"threshold_uncertainty_score":0.9999169,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0136532423756922,"score_gpt":0.3859894125051282,"score_spread":0.372336170129436,"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."}}