{"id":"W4407036387","doi":"10.2139/ssrn.5118632","title":"Comparing Human-Only, AI-Assisted, and AI-Led Teams on Assessing Research Reproducibility in Quantitative Social Science","year":2025,"lang":"en","type":"article","venue":"SSRN Electronic Journal","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":6,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of Toronto; University of Ottawa","funders":"","keywords":"Reproducibility; Data science; Psychology; Computer science; Chromatography; Chemistry","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":["metaresearch"],"consensus_categories":["metaresearch"],"category_scores_codex":[0.3737993,0.0008779492,0.001908375,0.004042136,0.002731996,0.005081608,0.00311642,0.003912463,0.001107763],"category_scores_gemma":[0.6519526,0.0007953335,0.003236583,0.00244536,0.006018362,0.005774579,0.0089872,0.002373099,0.0003730736],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002251812,"about_ca_system_score_gemma":0.004497073,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004458024,"about_ca_topic_score_gemma":0.005280107,"domain_scores_codex":[0.5897287,0.3343875,0.02587427,0.01681167,0.02936861,0.003829207],"domain_scores_gemma":[0.0970774,0.8005458,0.03358825,0.03251975,0.03200995,0.004258855],"domain_codex":null,"domain_gemma":"reproducibility","domain_candidate":"reproducibility","domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.01343654,0.002566504,0.7943776,0.001525913,0.01235672,0.0003120912,0.0556062,0.008423992,0.002056747,0.003500203,0.002707249,0.1031304],"study_design_scores_gemma":[0.003017717,0.01107773,0.8677844,0.00145995,0.005403176,0.0005546988,0.03468301,0.04874437,0.005511576,0.015712,0.00538215,0.0006691921],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9835052,0.0006250754,0.009989063,0.0008725474,0.0002206634,0.0004947595,0.0001210672,0.00009599492,0.004075664],"genre_scores_gemma":[0.9945189,0.00008793395,0.004236016,0.0002254351,0.00006743577,0.0004213419,0.0001339171,0.00005745336,0.0002514629],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6262007,"threshold_uncertainty_score":0.772217,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2770211523412912,"score_gpt":0.5740021460993504,"score_spread":0.2969809937580592,"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."}}