{"id":"W3154515585","doi":"10.1111/risa.13725","title":"What is a Good Calibration Question?","year":2021,"lang":"en","type":"article","venue":"Risk Analysis","topic":"Forecasting Techniques and Applications","field":"Decision Sciences","cited_by":13,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"","keywords":"Calibration; Set (abstract data type); Sample (material); Computer science; Reliability (semiconductor); Domain (mathematical analysis); Artificial intelligence; Machine learning; Statistics; Data mining; Mathematics","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":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.1017957,0.001338587,0.002235348,0.003509317,0.002084302,0.006653799,0.001832513,0.003888416,0.003857611],"category_scores_gemma":[0.3902308,0.0006777285,0.001808069,0.003102352,0.003404796,0.01175751,0.003620073,0.00331742,0.001890441],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002358096,"about_ca_system_score_gemma":0.003844483,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002054001,"about_ca_topic_score_gemma":0.001847951,"domain_scores_codex":[0.9126189,0.06168734,0.005509956,0.007178621,0.01175625,0.001248982],"domain_scores_gemma":[0.6408413,0.2569288,0.02626111,0.0279561,0.04381198,0.004200722],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0009385865,0.0009616961,0.1370743,0.003377647,0.001976459,0.000334634,0.01198373,0.02009716,0.01172841,0.07260974,0.01976907,0.7191485],"study_design_scores_gemma":[0.0003323802,0.001375421,0.1492954,0.004139827,0.001432365,0.001056138,0.01774606,0.1421821,0.02239683,0.5923547,0.06676765,0.000921206],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.2926228,0.005604108,0.6257746,0.03259689,0.001458355,0.001698356,0.001101928,0.001225737,0.03791723],"genre_scores_gemma":[0.8206488,0.001110261,0.1727328,0.002647817,0.0005867236,0.0006164163,0.0004860233,0.0002541782,0.0009170726],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.8982043,"threshold_uncertainty_score":0.5383538,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05990817643048678,"score_gpt":0.4026853487622595,"score_spread":0.3427771723317727,"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."}}