MANY MINDEDAI math review boosts retention in Tennessee students
A randomized experiment with 6,000 Tennessee middle school students found those using AI-assisted math software—requiring three consecutive correct answers after mistakes—to score 3 percentage points higher on a 15-minute retention test than peers. The intervention involved 50-minute sessions focused on fraction problems. Researchers from the University of Toronto and University of Pennsylvania’s Wharton School conducted the study, with a working paper scheduled for NBER circulation in August 2026.
The advantage was small, limited to simpler fraction problems, and did not extend to more complex tasks. Crucially, the study did not demonstrate deeper understanding or transferable skills, and the AI protocol’s long-term impact remains untested. The research was not peer-reviewed and evaluated only four learning approaches.
This offers a potential pathway for resource-constrained settings where teachers lack time for individualized feedback. If scaled, AI-driven review could help students retain foundational math concepts with less direct instruction—addressing a key barrier to educational access in low-budget schools. However, the narrow scope of benefits and lack of long-term data mean this is not a universal solution. What matters next is whether similar protocols improve learning in diverse contexts without sacrificing conceptual depth. The study’s limitations highlight that even modest gains require careful design to avoid reinforcing narrow skill sets.
Source: Hechinger Report