{"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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001769555,0.0001504981,0.0002409024,0.0002575573,0.0004522497,0.0002228561,0.003068981,0.00009334527,0.0000207437],"category_scores_gemma":[0.003661889,0.0001072664,0.00008080063,0.00182676,0.0007576513,0.001575279,0.0005914354,0.0003620836,0.000009357683],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001071044,"about_ca_system_score_gemma":0.0002304771,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001214011,"about_ca_topic_score_gemma":0.000001098361,"domain_scores_codex":[0.997315,0.00002277172,0.0005160801,0.0004719576,0.001424482,0.0002497434],"domain_scores_gemma":[0.9980874,0.0004641439,0.0005323322,0.00003078025,0.0008249636,0.00006035433],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.000008134942,0.00006825708,0.0129654,0.0000620997,0.00003867902,2.536378e-8,0.002206046,0.002095422,0.3386229,0.6413859,0.00167884,0.0008683595],"study_design_scores_gemma":[0.00004483434,0.00003354083,0.00723343,0.000318997,0.000009334774,0.000001133527,0.0004211312,0.03788547,0.6212502,0.3326238,0.00008407104,0.00009405284],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9272909,0.0003849521,0.006497722,0.05911487,0.0001438095,0.0005565066,0.00001000171,0.0001002067,0.005901074],"genre_scores_gemma":[0.9530157,0.00003746061,0.04488293,0.001127096,0.00006492039,0.00002064545,6.259768e-8,0.000003507052,0.0008476046],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.308762,"threshold_uncertainty_score":0.570298,"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."}}