What happened
The Hechinger Report describes two October studies led by University of Maryland researcher Jing Liu. In a Deer Valley middle-school rollout, frequent Goblins users scored higher in math, but the study did not establish causation. In a randomized course comparison, college classes assigned Maryland’s AI assistant had lower grades and less engagement than control classes.
Why it matters
The contrast cautions against treating AI as a single intervention: outcomes may depend on design, use and classroom context. Accelerate’s Goblins study linked at least two weekly problems with higher scores, but more motivated students may explain the association; only 6 percent used it for at least 30 minutes weekly. In Maryland’s study, only 15 percent used the assistant, mostly to seek answers, while outcomes were worse overall, with larger negative effects for first-generation students.
What to do next
Define an AI assistant’s tutoring role, preserve instructor control and state what use is permitted in coursework. Track participation, grades and whether students seek reasoning support or answers. Goblins supports testing teacher-assigned practice, not assuming it causes gains. Test locally before broad rollout; Maryland’s results do not establish why grades fell. Liu and Provost Jennifer King Rice said they could not determine whether the assistant encouraged outside chatbot use. Accelerate hopes to commission a randomized trial of Goblins, so its effects remain uncertain.
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