{"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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0002946413,0.0001113227,0.000231578,0.0001465493,0.0002472993,0.00002479635,0.00005431381,0.0001046977,0.00003603911],"category_scores_gemma":[0.0001683888,0.0001014453,0.00005577611,0.0003679474,0.00002741149,0.00003213633,0.00001199745,0.0001675942,0.0002552084],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004984927,"about_ca_system_score_gemma":0.0001716423,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00100803,"about_ca_topic_score_gemma":0.0001281858,"domain_scores_codex":[0.9989505,0.00008835307,0.0003376076,0.0002543919,0.0001493269,0.000219861],"domain_scores_gemma":[0.9991772,0.0001842297,0.00007306748,0.0002435962,0.0002425474,0.00007939444],"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":[0.0072548,0.001583799,0.1223634,0.008051774,0.001334407,0.0002472162,0.03698068,0.0004888271,0.04488559,0.09537838,0.1071966,0.5742344],"study_design_scores_gemma":[0.001911884,0.004135836,0.04506458,0.01517679,0.0005602056,0.00005374133,0.15939,0.007302966,0.738857,0.0002446449,0.02615435,0.00114802],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8163508,0.0001909612,0.00005157258,0.0006926753,0.004036177,0.0004711077,0.000003138634,0.0001783729,0.1780252],"genre_scores_gemma":[0.9951841,0.00000554507,0.00002249798,0.0002192885,0.0006562104,0.0000578266,0.00002805063,0.00001300375,0.003813491],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6939714,"threshold_uncertainty_score":0.4136818,"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."}}