SafeTutors: Benchmarking Pedagogical Safety in AI Tutoring Systems
Тип публікації :
Препринт
Дата випуску :
18 березня 2026 р.
Автор(и) :
Rima Hazra
Bikram Ghuku
Ilona Marchenko
Yaroslava Tokarieva
Sayan Layek
Somnath Banerjee
Julia Stoyanovich
Mykola Pechenizkiy
eKNUTSHIR URL :
Журнал :
arXiv (Cornell University)
Цитування :
[APA 7] Rima, H., Bikram, G., Ilona, M., Yaroslava, T., Sayan, L., Somnath, B., Julia, S., & Mykola, P. (2026). SafeTutors: Benchmarking Pedagogical Safety in AI Tutoring Systems. arXiv (Cornell University),. https://ir.library.knu.ua/handle/15071834/34271
[ДСТУ] SafeTutors: Benchmarking Pedagogical Safety in AI Tutoring Systems / H. Rima та ін. arXiv (Cornell University). 2026. URL: https://ir.library.knu.ua/handle/15071834/34271 (дата звернення: 11.09.2026).
Large language models are rapidly being deployed as AI tutors, yet current evaluation paradigms assess problem-solving accuracy and generic safety in isolation, failing to capture whether a model is simultaneously pedagogically effective and safe across student-tutor interaction. We argue that tutoring safety is fundamentally different from conventional LLM safety: the primary risk is not toxic content but the quiet erosion of learning through answer over-disclosure, misconception reinforcement, and the abdication of scaffolding. To systematically study this failure mode, we introduce SafeTutors, a benchmark that jointly evaluates safety and pedagogy across mathematics, physics, and chemistry. SafeTutors is organized around a theoretically grounded risk taxonomy comprising 11 harm dimensions and 48 sub-risks drawn from learning-science literature. We uncover that all models show broad harm; scale doesn't reliably help; and multi-turn dialogue worsens behavior, with pedagogical failures rising from 17.7% to 77.8%. Harms also vary by subject, so mitigations must be discipline-aware, and single-turn "safe/helpful" results can mask systematic tutor failure over extended interaction.
Файл(и) :![Ескіз]()
Вантажиться...
Формат :
Adobe PDF
Розмір :
1.13 MB
Контрольна сума :
(MD5):eab6b17a9fc0b77c4d142f655a6acbfd