{"id":"W4318929966","doi":"10.1037/cep0000303","title":"Text validation: Overlooking discrepancies in question constructions.","year":2023,"lang":"en","type":"article","venue":"Canadian Journal of Experimental Psychology/Revue canadienne de psychologie expérimentale","topic":"Text Readability and Simplification","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Manitoba","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Interrogative; Interrogative word; Consistency (knowledge bases); Linguistics; Construct (python library); Verb; Psychology; Sentence; PsycINFO; Natural language processing; Computer science; Cognitive psychology; Artificial intelligence; Philosophy; MEDLINE","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.001292567,0.0002371516,0.0003222502,0.001077315,0.0002365456,0.0001378241,0.001130177,0.0001969732,0.0001144337],"category_scores_gemma":[0.0001649997,0.0002655936,0.0001414737,0.001207309,0.0003280461,0.0007773079,0.00004653487,0.0003888466,0.00006990162],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001270291,"about_ca_system_score_gemma":0.0004606947,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005792045,"about_ca_topic_score_gemma":0.0250726,"domain_scores_codex":[0.9974285,0.0002389931,0.000735354,0.0005579593,0.0001068154,0.0009324141],"domain_scores_gemma":[0.9980183,0.00009481623,0.0003274973,0.0007034017,0.00009853378,0.0007575137],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"observational","study_design_scores_codex":[0.0002005553,0.0008376741,0.223376,0.00006199806,0.0002273791,0.003813033,0.04575719,0.001064954,0.3943371,0.2611026,0.03339274,0.0358288],"study_design_scores_gemma":[0.007206031,0.002373897,0.7512699,0.0006149802,0.00004791196,0.0105662,0.03667096,0.00148848,0.0796769,0.05654851,0.05120215,0.002334042],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9856873,0.0006955048,0.00245445,0.004983955,0.003024925,0.0002516939,0.00002007074,0.00007671217,0.002805381],"genre_scores_gemma":[0.9947348,0.00003833253,0.004249317,0.0006372302,0.0001542951,0.00003545199,0.00001978917,0.00001972836,0.0001110299],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.5278939,"threshold_uncertainty_score":0.9999796,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03959266053740267,"score_gpt":0.3197643406229638,"score_spread":0.2801716800855611,"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."}}