Lumilo — Carnegie Mellon's Teacher Smart Glasses Show Real-Time AI-Tutor Analytics Raise Learning in a 286-Student Classroom Experiment

Lumilo is a HoloLens-based teacher-support app, designed with teachers by Kenneth Holstein and Vincent Aleven at Carnegie Mellon University. It works alongside an AI tutor and overlays icons above students' heads showing whether each is on track, struggling or disengaged, with detail screens on where they are stuck.
A field study covered 343 middle-school students in 18 classrooms, 8 teachers (each with 5+ years teaching maths and prior AI-tutor experience) and 4 schools. Classrooms were randomly assigned, stratified within teacher, to three conditions: no glasses (business as usual with an AI tutor), glasses without analytics, and glasses with analytics (full Lumilo). Students did a 15-minute pre-test, two class sessions with the 'Lynnette' linear-equation tutor, and a 15-minute post-test. After excluding 57 students absent on one or more days, 286 were analysed with hierarchical linear modelling.
Results: Lumilo use improved learning relative to business-as-usual, and the analytics added benefit beyond the glasses alone (part of the gain came from the glasses without analytics, so a novelty or monitoring effect cannot be ruled out). Teachers using Lumilo spent more time with low-prior-knowledge students; those students gained most, reducing the usual 'rich get richer' pattern. The authors present it as the first experimental demonstration that teacher–AI partnership can improve outcomes.
Caveats: only two class sessions on one topic (linear equations), a small number of teachers and schools, no long-term follow-up reported, and in 2018 press coverage the glasses were described as too heavy for extended wear and not ready for market. The transferable lesson is the design pattern — surfacing AI-tutor data to the teacher in real time — rather than the specific hardware.

Read the full analysis: https://www.cmu.edu/ambassadors/february-2019/super-glasses.html