{"id":"W2100249939","doi":"10.5555/2019116.2019118","title":"Comparing Computer Versus Human Data Collection Methods for Public Usability Evaluations of a Tactile-Audio Display","year":2010,"lang":"en","type":"article","venue":"Journal of Usability Studies archive","topic":"Tactile and Sensory Interactions","field":"Neuroscience","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Toronto Metropolitan University","funders":"","keywords":"Usability; Human–computer interaction; Computer science; Context (archaeology); Multimedia; Data collection; Usability engineering; Perspective (graphical); Artificial intelligence","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":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.001624027,0.000183777,0.0006316663,0.0002052603,0.0007474565,0.0000650726,0.0006695147,0.00004942865,0.00003370769],"category_scores_gemma":[0.0172358,0.0001511483,0.0002745119,0.0003064141,0.0006629382,0.0007639618,0.0003985664,0.0006802542,0.000001771836],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001369593,"about_ca_system_score_gemma":0.0001632627,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00004051318,"about_ca_topic_score_gemma":0.001013002,"domain_scores_codex":[0.9970579,0.0008316269,0.001041144,0.0004407609,0.0003562645,0.0002723326],"domain_scores_gemma":[0.9843621,0.01291142,0.0009838182,0.0008759254,0.00073157,0.0001351583],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"observational","study_design_scores_codex":[0.002398149,0.003015112,0.03660955,0.0003700262,0.000985406,0.000003462923,0.008591907,0.0005962308,0.9054957,0.004582352,0.007873608,0.02947847],"study_design_scores_gemma":[0.01661506,0.009400459,0.3418031,0.0003711822,0.002214478,0.0005586494,0.01102811,0.2440847,0.2559539,0.04081115,0.07543384,0.001725336],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.941353,0.00001557707,0.05299955,0.001298201,0.003325882,0.0005554527,0.00009632812,0.00001858834,0.0003374458],"genre_scores_gemma":[0.9456607,0.00002462714,0.05387229,0.00004257175,0.000341542,0.00001965564,0.000006164136,0.00001171027,0.00002077447],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6495418,"threshold_uncertainty_score":0.9910424,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3805955447662984,"score_gpt":0.505505412120582,"score_spread":0.1249098673542836,"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."}}