{"id":"W4378223607","doi":"10.1002/ev.20543","title":"Guest Editors’ notes","year":2023,"lang":"en","type":"article","venue":"New Directions for Evaluation","topic":"Evaluation and Performance Assessment","field":"Decision Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Timeline; Adventure; Ethos; Dream; Psychology; Sociology; Operations research; Library science; Computer science; Law; History; Political science; Artificial intelligence; Engineering","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch","insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.009514255,0.000105271,0.0001425645,0.0005823469,0.0004602176,0.0002291573,0.000236366,0.00007461989,0.00226025],"category_scores_gemma":[0.009121967,0.00008691567,0.0001268982,0.002037088,0.00002538977,0.0005328939,0.0000340383,0.00006901746,0.003745066],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00012568,"about_ca_system_score_gemma":0.0005420085,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0000584049,"about_ca_topic_score_gemma":0.0002585113,"domain_scores_codex":[0.9964512,0.0001950086,0.0004903719,0.0003762729,0.002273031,0.0002141057],"domain_scores_gemma":[0.9961861,0.001902264,0.0001787875,0.0004147493,0.001215424,0.0001026458],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.00001340269,0.00002584868,0.002491944,0.000001747237,0.00001180163,5.273941e-8,0.0004013477,0.003471426,0.0004200109,0.000567815,0.6949603,0.2976343],"study_design_scores_gemma":[0.0008332748,0.0000777477,0.03724575,0.000006810732,0.00004097519,9.617311e-7,0.0002740351,0.1610691,0.001229236,0.02600538,0.7730998,0.0001169129],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.390555,0.000508165,0.1009327,0.09566969,0.2654376,0.009821494,0.0002238294,0.003132968,0.1337186],"genre_scores_gemma":[0.9092156,0.00008943823,0.003582055,0.0004772673,0.0332639,0.001192392,0.0005073656,0.00004851216,0.05162346],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.5186606,"threshold_uncertainty_score":0.9992246,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3729067314618274,"score_gpt":0.5685543924994344,"score_spread":0.195647661037607,"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."}}