{"id":"W6981590369","doi":"","title":"ERIC ED464926: In Search of Soundness in Teacher Testing: Beyond Political Validity.","year":2002,"lang":"en","type":"other","venue":"Bulletin of Miscellaneous Information (Royal Gardens Kew)","topic":"Cultural Heritage Materials Analysis","field":"Arts and Humanities","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Soundness; Accountability; Politics; Test (biology); Test validity; Psychometrics; Predictive validity; Standardized test; Measure (data warehouse)","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01022344,0.00100122,0.0008615658,0.003233763,0.001579084,0.005193365,0.001970175,0.01307362,0.26733],"category_scores_gemma":[0.03991549,0.0007691951,0.0003003955,0.00365953,0.002234719,0.00490042,0.002385987,0.003820852,0.1974441],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002101627,"about_ca_system_score_gemma":0.003721836,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01216219,"about_ca_topic_score_gemma":0.02784406,"domain_scores_codex":[0.9954624,0.001811567,0.0005640263,0.0003088284,0.001627171,0.0002259782],"domain_scores_gemma":[0.9704298,0.01831721,0.001226302,0.001287641,0.006556097,0.002183035],"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":[0.00002409263,0.00001547096,0.0001607221,0.0001337859,0.000001400419,0.0000910161,0.00005908568,0.00002867799,0.00007203558,0.002647585,0.9474085,0.0493576],"study_design_scores_gemma":[0.00002563565,0.00003559612,0.001624806,0.000731237,0.000003157733,0.0002400964,0.0001652901,0.0000974108,0.0001570238,0.00248814,0.9944184,0.00001336356],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"other","genre_gemma":"other","genre_scores_codex":[0.0008092136,0.03029755,0.00435729,0.1019078,0.01071738,0.0005327692,0.002205908,0.001952502,0.8472197],"genre_scores_gemma":[0.003835263,0.009971805,0.002571989,0.02872091,0.004900406,0.0002990177,0.00101613,0.0006904797,0.9479939],"genre_candidate":"other","genre_consensus":"other","teacher_disagreement_score":0.26733,"threshold_uncertainty_score":0.894308,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03656837035646365,"score_gpt":0.2135587960209407,"score_spread":0.1769904256644771,"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."}}