{"id":"W4416814854","doi":"10.1177/00469580251399374","title":"A Pilot Study on Generative Artificial Intelligence’s Reliability in Qualitative Research Quality Appraisal Using CASP and JBI Checklists","year":2025,"lang":"en","type":"article","venue":"INQUIRY The Journal of Health Care Organization Provision and Financing","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Toronto Metropolitan University; Public Health Agency of Canada; University of Guelph","funders":"","keywords":"Checklist; Inter-rater reliability; Critical appraisal; CASP; Qualitative research; Reliability (semiconductor); Research design; Human research","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":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.01198397,0.0001331634,0.0004067846,0.0004538379,0.0006364072,0.00005112551,0.00008985557,0.00007992674,0.000006185699],"category_scores_gemma":[0.008875734,0.00009524504,0.00001844637,0.001246299,0.0002338583,0.0001580167,0.00006825115,0.0009256339,0.000001255557],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006133256,"about_ca_system_score_gemma":0.001812855,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002394291,"about_ca_topic_score_gemma":0.0005590756,"domain_scores_codex":[0.9948032,0.00274979,0.001472551,0.0002594063,0.0004604355,0.0002545555],"domain_scores_gemma":[0.9953065,0.001863719,0.0004627504,0.000219992,0.002040519,0.0001065378],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"qualitative","study_design_scores_codex":[0.001639925,0.001192356,0.06475151,0.0006647055,0.00002214506,0.000009125034,0.822551,0.0003424951,0.001645055,0.002815957,0.0002185014,0.1041473],"study_design_scores_gemma":[0.0002072436,0.006601045,0.1013263,0.0009745578,0.00002569952,0.00003116241,0.8829199,0.000495,0.00422335,0.003019494,0.00004825377,0.000127974],"study_design_candidate":"qualitative","study_design_consensus":"qualitative","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9734781,0.0009149146,0.01414125,0.009760873,0.000716719,0.0009665094,0.000001563332,0.000008143933,0.00001186974],"genre_scores_gemma":[0.9983934,0.0002001951,0.0005333034,0.0006670347,0.000176612,0.000004361574,0.000003254461,0.00001291978,0.000008963739],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.1040193,"threshold_uncertainty_score":0.9994729,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.4243587056817439,"score_gpt":0.6018586717368326,"score_spread":0.1774999660550887,"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."}}