AI
Two-year school trial finds AI tutoring gains were limited by student engagement
A two-year cluster-randomized trial across 18 Tennessee middle schools found that students assigned to use Khan Academy with the Khanmigo AI tutor improved mathematics performance by about 0.06 to 0.08 standard deviations over a school year. The estimated full-year active-participation effect was 0.14 standard deviations, comparable to Khan Academy practice without AI assistance. Although 96% of students tried the tutor, typical use was infrequent and shallow. The study points to engagement, rather than access alone, as the central constraint.
Editorial summary · Feed excerpt
What matters
- Students used Khanmigo during existing daily remedial mathematics sessions.
- The median student messaged the tutor on only one-third of practice days.
- Students engaged the tutor in just 17% of sessions where they made a mistake.
- Most messages were bare answers or selections from suggested prompts.
Context & caveats
- The reported findings concern this trial’s configuration and participating Tennessee schools; they do not establish effects in every educational setting.
Why this made the edition
Provides unusually relevant experimental evidence from a two-year randomized school trial, including effect sizes, usage behavior and a concrete explanation for the limited AI-specific benefit.
Source material