{"id":"W4411627694","doi":"10.1073/pnas.2422633122","title":"Generative AI without guardrails can harm learning: Evidence from high school mathematics","year":2025,"lang":"en","type":"article","venue":"Proceedings of the National Academy of Sciences","topic":"Explainable Artificial Intelligence (XAI)","field":"Computer Science","cited_by":81,"is_retracted":false,"has_abstract":true,"ca_institutions":"Toronto Metropolitan University","funders":"","keywords":"Generative grammar; Harm; Computer science; Artificial intelligence; TUTOR; Productivity; Machine learning; Psychology; Social psychology","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01604744,0.0007773001,0.0009489639,0.001501402,0.002001273,0.003122056,0.003039845,0.002837542,0.008918816],"category_scores_gemma":[0.1068626,0.0006435878,0.0009527503,0.001154205,0.005855796,0.004804964,0.003795745,0.003202652,0.002996456],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001246098,"about_ca_system_score_gemma":0.001618819,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.006076618,"about_ca_topic_score_gemma":0.007514873,"domain_scores_codex":[0.9824533,0.009984167,0.0009709161,0.002172282,0.003703518,0.0007158761],"domain_scores_gemma":[0.7962126,0.1513574,0.02022743,0.0165256,0.00931411,0.006362918],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.01191499,0.04777903,0.3951317,0.006114253,0.002631711,0.001899348,0.07224753,0.00541976,0.006423021,0.01231198,0.02558804,0.4125386],"study_design_scores_gemma":[0.003394098,0.03252486,0.7867741,0.004668738,0.002122388,0.001465151,0.03104484,0.008177321,0.01333577,0.03127478,0.08470342,0.0005146238],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9819999,0.001999555,0.001753369,0.002764745,0.00007434851,0.0001647854,0.0003237787,0.0001227548,0.01079673],"genre_scores_gemma":[0.9919118,0.001497658,0.00196714,0.001313761,0.00005693472,0.0002425172,0.0002748235,0.00007063721,0.002664784],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01604744,"threshold_uncertainty_score":0.08486795,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06691070461142205,"score_gpt":0.3490396163586603,"score_spread":0.2821289117472382,"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."}}