{"id":"W4413679715","doi":"10.1109/compsac65507.2025.00296","title":"AI Agents for Clinical Data Assessment: Enhancing Decision-Making with Human-AI Collaboration","year":2025,"lang":"en","type":"article","venue":"","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"Thompson Rivers University","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Computer science; Clinical decision making; Knowledge management; Data science; Medicine; Intensive care medicine","routes":{"ca_aff":true,"ca_fund":true,"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":[0.01751327,0.001674542,0.001068642,0.00200191,0.001083137,0.004151455,0.003095281,0.002056671,0.0043711],"category_scores_gemma":[0.05566356,0.0006537745,0.0008145304,0.0009631384,0.001928992,0.00420908,0.004945887,0.003239844,0.001145077],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001222923,"about_ca_system_score_gemma":0.004579745,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004220615,"about_ca_topic_score_gemma":0.004235417,"domain_scores_codex":[0.9857442,0.009209836,0.0009063246,0.001436545,0.002363812,0.0003392116],"domain_scores_gemma":[0.9553266,0.03176877,0.003132883,0.003564388,0.004573231,0.001634206],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.0008155789,0.001190067,0.01078718,0.0007581239,0.0004004992,0.0008039422,0.003605856,0.3073332,0.009929226,0.07741873,0.01188384,0.5750738],"study_design_scores_gemma":[0.0001400921,0.0001955556,0.0006749071,0.000152748,0.00006681083,0.0001527801,0.0003359905,0.8866866,0.005068177,0.09245111,0.01400463,0.00007054303],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.009277556,0.0004074849,0.9824754,0.001952431,0.00008509988,0.0003246622,0.00007688449,0.001498768,0.003901768],"genre_scores_gemma":[0.336486,0.0002823039,0.6600337,0.0006407428,0.0001036362,0.0004067208,0.0002066185,0.0001908771,0.00164945],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01751327,"threshold_uncertainty_score":0.09262007,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3503294715243584,"score_gpt":0.6491601273226334,"score_spread":0.298830655798275,"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."}}