{"id":"W4392622771","doi":"10.2196/53019","title":"Assessing the Quality of ChatGPT Responses to Dementia Caregivers’ Questions: Qualitative Analysis","year":2024,"lang":"en","type":"article","venue":"JMIR Aging","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":22,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"National Institute on Aging; National Institutes of Health; University of Texas at Austin","keywords":"Preprint; Dementia; Quality (philosophy); Psychology; Gerontology; Medicine; Computer science; World Wide Web; Epistemology","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":[],"consensus_categories":[],"category_scores_codex":[0.001729814,0.00007374521,0.0001889865,0.0003337662,0.0001383185,0.00007744244,0.00006453307,0.00003561358,0.0001086866],"category_scores_gemma":[0.0005848742,0.00005371924,0.0001245879,0.001230448,0.00006240392,0.0001603021,0.00002248855,0.0001372883,0.00002713804],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001042648,"about_ca_system_score_gemma":0.0002491746,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003473348,"about_ca_topic_score_gemma":0.0002009146,"domain_scores_codex":[0.9986157,0.0004025499,0.0003946743,0.0001994375,0.0002415663,0.0001460689],"domain_scores_gemma":[0.9982609,0.001132476,0.00007177582,0.0002418953,0.0002245428,0.00006836702],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"qualitative","study_design_scores_codex":[0.0001244017,0.0001389884,0.1022782,0.0008167456,0.001529899,0.00001414821,0.7760206,0.0001512862,0.006908644,0.007558601,0.002236899,0.1022216],"study_design_scores_gemma":[0.00004097967,0.0001593825,0.2988586,0.001174178,0.001451591,0.000005461473,0.6764992,0.002098838,0.01108578,0.002948462,0.005425989,0.0002514708],"study_design_candidate":"qualitative","study_design_consensus":"qualitative","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9726037,0.00139209,0.01140134,0.0134626,0.00024928,0.0003129691,0.000006020169,0.00005816526,0.000513817],"genre_scores_gemma":[0.998339,0.00002257142,0.0007389479,0.0002994839,0.0001319827,0.00006062536,0.00001214586,0.000007703877,0.0003875244],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.1965805,"threshold_uncertainty_score":0.5250686,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3026698485249815,"score_gpt":0.6061262897221323,"score_spread":0.3034564411971509,"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."}}