{"id":"W4408022558","doi":"10.2196/63058","title":"Advancing Clinical Chatbot Validation Using AI-Powered Evaluation With a New 3-Bot Evaluation System: Instrument Validation Study","year":2025,"lang":"en","type":"article","venue":"JMIR Nursing","topic":"AI in Service Interactions","field":"Computer Science","cited_by":10,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Chatbot; Health care; Test (biology); Computer science; Economic shortage; Automation; Artificial intelligence; Engineering","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.003612051,0.000266714,0.00032586,0.0004609781,0.0004542739,0.0005471839,0.0004965517,0.0001217728,0.00002695661],"category_scores_gemma":[0.0001924276,0.0002654252,0.00008459987,0.001262986,0.00003350119,0.002358576,0.00008197916,0.0003112306,0.00002531845],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00259463,"about_ca_system_score_gemma":0.00122393,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000191231,"about_ca_topic_score_gemma":0.00005659316,"domain_scores_codex":[0.9953349,0.001052106,0.0009593497,0.0008436834,0.001451344,0.0003585723],"domain_scores_gemma":[0.9971265,0.0001964739,0.0005633165,0.0009298376,0.001062012,0.0001218811],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0003167954,0.002463064,0.03782208,0.0001564455,0.0003423448,0.000005989477,0.01400456,0.09267485,0.003587399,0.004492442,0.0009552526,0.8431787],"study_design_scores_gemma":[0.00266044,0.0004283799,0.02743904,0.00181772,0.0004422999,0.00002636756,0.004550903,0.9589256,0.002079235,0.001244001,0.00007081265,0.0003151665],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5197759,0.0000372407,0.4720083,0.001404781,0.002473286,0.002876814,6.113344e-7,0.0002192841,0.001203759],"genre_scores_gemma":[0.9780503,8.227503e-7,0.02129196,0.0001333508,0.0002355027,0.0002032236,0.00002674527,0.00002147128,0.0000366554],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.8662508,"threshold_uncertainty_score":0.9999798,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08517371719177196,"score_gpt":0.4622386755596723,"score_spread":0.3770649583679003,"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."}}