{"id":"W4413295144","doi":"10.1016/j.system.2025.103805","title":"ChatGPT-generated versus human direct corrective feedback on L2 writing","year":2025,"lang":"en","type":"article","venue":"System","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":13,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of Victoria","funders":"Fondo Nacional de Desarrollo Científico y Tecnológico; Agencia Nacional de Investigación y Desarrollo; Universidad del Bío-Bío","keywords":"Corrective feedback; Computer science; Second language writing; Mathematics education; Psychology; Linguistics; Second language; Philosophy","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"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":[0.001607047,0.001114755,0.0007896286,0.0005775619,0.0004426744,0.001133288,0.0009825705,0.001294975,0.0217847],"category_scores_gemma":[0.01673208,0.0002810559,0.0002572695,0.0002532146,0.0002869178,0.0007429343,0.001362894,0.0007274331,0.005381665],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003369801,"about_ca_system_score_gemma":0.0006319209,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0044447,"about_ca_topic_score_gemma":0.003954751,"domain_scores_codex":[0.9983423,0.0008258582,0.0001427478,0.0002867357,0.0002852155,0.0001171918],"domain_scores_gemma":[0.9894744,0.00801764,0.0002252057,0.0006766022,0.001214782,0.0003913672],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.02591166,0.001697784,0.008458371,0.002089716,0.0002455717,0.003004379,0.004794739,0.01342173,0.1673307,0.000638693,0.02805005,0.7443567],"study_design_scores_gemma":[0.003510085,0.0146388,0.07260079,0.0005797269,0.0007010379,0.003306313,0.004143912,0.4407049,0.4228161,0.002168808,0.03422559,0.0006039652],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9241564,0.0004054237,0.02932163,0.0003177713,0.0005431381,0.0003952175,0.001670732,0.02471637,0.01847346],"genre_scores_gemma":[0.9634866,0.0001143086,0.01864591,0.0003011971,0.00006893217,0.0002202771,0.001423055,0.001030942,0.01470871],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.0217847,"threshold_uncertainty_score":0.07287705,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1405332177187322,"score_gpt":0.4337121920250231,"score_spread":0.293178974306291,"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."}}