{"id":"W2015457266","doi":"10.1016/j.system.2013.07.010","title":"Putting testing researchers to the test: An exploratory study on the TOEFL iBT","year":2013,"lang":"en","type":"article","venue":"System","topic":"Student Assessment and Feedback","field":"Social Sciences","cited_by":14,"is_retracted":false,"has_abstract":false,"ca_institutions":"Mount Saint Vincent University; Carleton University; Queen's University","funders":"","keywords":"Test of English as a Foreign Language; Computer science; Context (archaeology); Construct (python library); Test (biology); Exploratory research; Variance (accounting); Focus group; Psychology; Construct validity; Language assessment; Mathematics education; Psychometrics; Sociology; Developmental psychology; Programming language","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":"codex-gemma-dda1882f352a","candidate_categories":["sts","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.004578231,0.00009572165,0.0001052007,0.00003955551,0.001481876,0.000514094,0.0007694655,0.00003239389,0.00005110963],"category_scores_gemma":[0.001524988,0.00005108118,0.00002808023,0.0005843214,0.00006953815,0.0002130571,0.0001240843,0.0001768363,0.0009978964],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001601643,"about_ca_system_score_gemma":0.00009326054,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001870495,"about_ca_topic_score_gemma":0.001152889,"domain_scores_codex":[0.9973583,0.001001063,0.0001771759,0.0002287068,0.0008501651,0.0003846247],"domain_scores_gemma":[0.9974251,0.001822301,0.0000693697,0.0003723893,0.0001860394,0.000124821],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"qualitative","study_design_scores_codex":[0.000005300927,0.0003015884,0.4821432,0.00001862837,0.00004900682,0.000009699093,0.4830627,0.0000223833,0.0007291695,0.005692454,0.01629084,0.01167505],"study_design_scores_gemma":[0.0000954268,0.0002389631,0.08288997,0.00006102148,0.000006282732,1.593511e-7,0.9145845,0.0001134464,0.00001336156,0.0000650464,0.001844124,0.00008769867],"study_design_candidate":"qualitative","study_design_consensus":"qualitative","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9160377,0.00001295669,0.000006775448,0.004814158,0.0002625439,0.002026214,8.014131e-7,0.000138587,0.07670022],"genre_scores_gemma":[0.9973331,3.032636e-7,0.00007781402,0.0001908711,0.0006747076,0.0003687626,1.556999e-7,0.00001311586,0.001341146],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4315218,"threshold_uncertainty_score":0.9998181,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2468339053032725,"score_gpt":0.4152035437005829,"score_spread":0.1683696383973104,"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."}}