What happened
Two October studies reported contrasting results, University of Maryland researcher Jing Liu told MindShift. In 2025-26, 6,000 middle schoolers in Deer Valley, Arizona, used Goblins, an AI math tutor. Students completing at least two weekly problems scored higher, though the study cannot establish causation. In a separate randomized study assigning professors’ classes to a university AI assistant or control, assistant-course students had lower grades and less course engagement.
Why it matters
The findings caution against treating AI as a single intervention. Goblins’ usage-linked math results may reflect student motivation, although half the students reached the two-problem weekly level. At Maryland, 15 percent of students in AI-assistant courses used it; those courses’ 1,100 students averaged lower grades than 1,200 controls and engaged less with course materials and instructors. Liu and Provost Jennifer King Rice suspect its presence may have encouraged outside AI use, but could not establish that.
What to do next
Pilot classroom AI before wider adoption, as Maryland did, and clarify whether it is for tutoring or answers. Keep educators involved in assigning practice; monitor use, learning measures, course-material interaction and help-seeking rather than assuming availability means engagement. In college courses, state AI expectations and watch grades, office-hour attendance and contact with instructors and peers, including for first-generation students, who had more pronounced negative effects in this study. Interpret findings cautiously: Accelerate hopes to commission a randomized Goblins trial, and Maryland researchers could not verify suspected off-platform use.
About this briefing
Reviewed by TutorFlow Editorial. We link the primary source, preserve its publication date, and distinguish reported claims from TutorFlow analysis. Our commentary focuses on practical decisions for educators and training teams.
Discussion (0)
Open to everyone