{"id":"W3130937151","doi":"10.1038/s41746-021-00385-9","title":"Do as AI say: susceptibility in deployment of clinical decision-aids","year":2021,"lang":"en","type":"article","venue":"npj Digital Medicine","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":447,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute; University of Toronto; St. Michael's Hospital","funders":"Natural Sciences and Engineering Research Council of Canada; Government of Canada; Canadian Institute for Advanced Research; Vector Institute; Microsoft Research","keywords":"Advice (programming); Medical diagnosis; Decision aids; Software deployment; Task (project management); Quality (philosophy); Work (physics); Psychology; Medicine; Medical education; Computer science; Medical physics; Pathology; Engineering; Alternative medicine","routes":{"ca_aff":true,"ca_fund":true,"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","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.001043049,0.0001352863,0.0006364309,0.0001280242,0.00002548267,0.00001286908,0.00008983896,0.0001649767,0.001015551],"category_scores_gemma":[0.01192238,0.0001038093,0.0001171448,0.0005291419,0.0002609165,0.000147161,0.00005318708,0.0003653954,0.0001263351],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00009863031,"about_ca_system_score_gemma":0.0007429923,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004142906,"about_ca_topic_score_gemma":0.0002115021,"domain_scores_codex":[0.9971855,0.00006702465,0.001531282,0.0004180963,0.0005446801,0.0002533943],"domain_scores_gemma":[0.9972445,0.001203482,0.0001419473,0.0005525966,0.0005376634,0.0003198058],"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":[0.0002742773,0.0005370976,0.4632434,0.0000686374,0.0000153434,0.00007347792,0.000537292,0.000001627891,0.0001727195,0.0005112204,0.001886866,0.532678],"study_design_scores_gemma":[0.002348896,0.009886109,0.7865219,0.005364986,0.0001989339,0.0003541145,0.02181792,0.0002786978,0.01972598,0.09359109,0.05936166,0.0005497835],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9772666,0.0006927217,0.0002856614,0.009860774,0.001082187,0.0002875529,0.000004545521,0.00002717037,0.01049281],"genre_scores_gemma":[0.9955879,0.0002368009,0.0003141596,0.002633744,0.0006570057,0.00001027408,0.00004516705,0.00001557261,0.0004993826],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.5321282,"threshold_uncertainty_score":0.9998977,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2054665387131868,"score_gpt":0.527046304125125,"score_spread":0.3215797654119381,"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."}}