{"id":"W7126397965","doi":"10.21428/594757db.e9f456d0","title":"Generative AI for Enhancing Active Learning in Education:A Comparative Study of GPT-3.5 and GPT-4 in CraftingCustomized Test Questions","year":2024,"lang":"en","type":"article","venue":"","topic":"Intelligent Tutoring Systems and Adaptive Learning","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Ontario Tech University","funders":"","keywords":"Active learning (machine learning); Test (biology); Generative grammar; Path (computing); Generative model","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":[0.0006021946,0.0001470928,0.0003164198,0.0003789526,0.0001145575,0.0001629545,0.0001587807,0.00004277278,0.000004972534],"category_scores_gemma":[0.0002530997,0.0001316482,0.00003413486,0.000533703,0.00002270284,0.000450195,0.00008795926,0.0003245813,0.000004290651],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001395283,"about_ca_system_score_gemma":0.0002588815,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0008192314,"about_ca_topic_score_gemma":0.001734481,"domain_scores_codex":[0.9986019,0.0002207863,0.0004161615,0.0004314071,0.0001343777,0.0001953362],"domain_scores_gemma":[0.9985194,0.001054038,0.00009206763,0.0001217222,0.0001749658,0.00003782806],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00009438465,0.002165559,0.03624891,0.0003116647,0.0001693271,0.00002280606,0.4629964,0.04474315,0.02431466,0.3875583,0.000226489,0.04114841],"study_design_scores_gemma":[0.003270552,0.002542961,0.0343406,0.002733184,0.00004229288,0.00002252031,0.1591828,0.755456,0.02843048,0.002470438,0.01047267,0.001035468],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6791901,0.0004209657,0.316902,0.0003796894,0.0004539001,0.001158209,0.000001398941,0.0001180325,0.001375728],"genre_scores_gemma":[0.9895812,0.000005321582,0.006178811,0.0000290961,0.00007466779,0.0001925142,0.000001301947,0.000008783839,0.003928257],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.7107129,"threshold_uncertainty_score":0.5368458,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02652257507694267,"score_gpt":0.3367431233355652,"score_spread":0.3102205482586225,"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."}}