{"id":"W4403847593","doi":"10.1111/jedm.12420","title":"Algorithmic Bias in BERT for Response Accuracy Prediction: A Case Study for Investigating Population Validity","year":2024,"lang":"en","type":"article","venue":"Journal of Educational Measurement","topic":"Imbalanced Data Classification Techniques","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"Concordia University of Edmonton; University of Alberta","funders":"","keywords":"Item response theory; Population; Psychology; Test validity; Predictive validity; Statistics; Computer science; Econometrics; Psychometrics; Mathematics; Clinical psychology; Demography","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":[],"consensus_categories":[],"category_scores_codex":[0.007525971,0.0001133221,0.0001611242,0.000433503,0.0001285067,0.0002213884,0.0003355138,0.00004041821,0.000004024763],"category_scores_gemma":[0.005787502,0.0001034885,0.00008798638,0.0004082581,0.00001425022,0.00096632,0.00003122394,0.0001519068,0.000001007347],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008594862,"about_ca_system_score_gemma":0.001116206,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00004275144,"about_ca_topic_score_gemma":0.00003149608,"domain_scores_codex":[0.9978583,0.0002880432,0.0008091516,0.0002516231,0.0006480083,0.0001448783],"domain_scores_gemma":[0.9967602,0.001588733,0.0003741854,0.0002554495,0.0009385707,0.00008279008],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.001949505,0.01338958,0.3256512,0.001999299,0.001208868,0.0004280433,0.04773556,0.003149619,0.06710986,0.06462865,0.1498084,0.3229414],"study_design_scores_gemma":[0.0048801,0.005222192,0.6825269,0.002294685,0.0002627446,0.005961016,0.003437563,0.1511058,0.009784258,0.11055,0.02295389,0.001020757],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4610288,0.0002685058,0.5216942,0.01327293,0.001923183,0.001709633,0.00004401473,0.00005221256,0.000006516774],"genre_scores_gemma":[0.8706062,0.000002870036,0.1285308,0.0000658822,0.0004168789,0.0003397839,0.000008872392,0.000009769979,0.00001886494],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.4095775,"threshold_uncertainty_score":0.6928598,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3090948321123259,"score_gpt":0.4135517798954063,"score_spread":0.1044569477830803,"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."}}