{"id":"W4415668195","doi":"10.2196/81936","title":"Integrating a Large Language Model Into a Socially Assistive Robot in a Hospital Geriatric Unit: Two-Wave Comparative Study on Performance, Engagement, and User Perceptions","year":2025,"lang":"en","type":"article","venue":"JMIR Human Factors","topic":"Social Robot Interaction and HRI","field":"Psychology","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Perception; Quality (philosophy); Robot; Human–robot interaction; Social robot; Social relation; Robotics","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":[0.0002317568,0.0002493119,0.0003363089,0.0004411238,0.0005535897,0.00009794268,0.0001258538,0.0001041292,0.000499423],"category_scores_gemma":[0.00003200293,0.0002315758,0.00008227022,0.0004233244,0.00006383515,0.0001627681,0.0001011731,0.0007073906,0.00004321949],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002229665,"about_ca_system_score_gemma":0.00006243375,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004151696,"about_ca_topic_score_gemma":0.003076592,"domain_scores_codex":[0.9984866,0.0003399579,0.0003276317,0.0003902778,0.0001616454,0.0002939159],"domain_scores_gemma":[0.9993851,0.0001590059,0.0001334156,0.000187881,0.00007610748,0.00005855898],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"observational","study_design_scores_codex":[0.0000908949,0.00263745,0.3486761,0.00001244138,0.0002056982,0.0000101292,0.6331145,0.00008255076,0.0003370205,0.01272827,0.001718497,0.0003864472],"study_design_scores_gemma":[0.001410346,0.0003820645,0.6934389,0.00003075841,0.00002653627,1.270166e-7,0.3038822,0.0005144114,0.000006837738,0.00006337727,0.00007530131,0.0001691263],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9753466,0.00001027384,0.0003007073,0.00006311807,0.0003002049,0.001000251,0.00001909004,0.00008409627,0.02287563],"genre_scores_gemma":[0.9952388,0.000001136877,0.00007712452,0.0001251807,0.00006564035,0.0002162317,0.00003578157,0.00001807618,0.004222033],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3447628,"threshold_uncertainty_score":0.9443386,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07888930352055695,"score_gpt":0.4376293429680049,"score_spread":0.3587400394474479,"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."}}