{"id":"W4412568059","doi":"10.1109/ichi64645.2025.00069","title":"Assessing AI Explainability: A Usability Study Using a Novel Framework Involving Clinicians","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":"University of Toronto","funders":"National Center for Advancing Translational Sciences","keywords":"Usability; Computer science; Human–computer interaction; Pluralistic walkthrough; 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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.09740332,0.001319137,0.001592285,0.003980419,0.002061207,0.003802955,0.002342525,0.002218816,0.00161708],"category_scores_gemma":[0.1568028,0.000975871,0.001880464,0.002221128,0.00223424,0.003490632,0.003570901,0.001905391,0.0005152571],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.003619215,"about_ca_system_score_gemma":0.004165016,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00174524,"about_ca_topic_score_gemma":0.002361487,"domain_scores_codex":[0.942457,0.04191422,0.005507551,0.002994262,0.005409244,0.001717596],"domain_scores_gemma":[0.794539,0.1587239,0.005661867,0.01086384,0.02786866,0.002342765],"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":[0.00285228,0.01463872,0.08355729,0.01451364,0.0006267186,0.002779507,0.3006938,0.006181216,0.0342702,0.003700611,0.006358557,0.5298275],"study_design_scores_gemma":[0.00985045,0.09592783,0.2907578,0.01069756,0.002802254,0.005946992,0.3124048,0.07389963,0.06094561,0.01350697,0.120932,0.00232798],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9023808,0.0004293201,0.07250368,0.000891811,0.0001190559,0.02020731,0.0002506976,0.0005678782,0.002649369],"genre_scores_gemma":[0.7577769,0.0005609383,0.2037454,0.0007142049,0.00008877627,0.03559682,0.0004327446,0.0001237501,0.0009604567],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.09740332,"threshold_uncertainty_score":0.5151241,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3676552132695584,"score_gpt":0.5666245115505536,"score_spread":0.1989692982809952,"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."}}