The Living Lab Report
A multi-agent reader that watches how learners actually use a lab.
A pilot system that studies engagement patterns inside hands-on learning environments and translates them into narratives a program owner can actually use. It reads behaviour the way a good instructor would — with context, not just counts.
The problem
Learning analytics tend to be flat: completion rates, time-on-task, click totals. None of it answers the more important question — what did the learner do, and did it mean anything?
The approach
A composition of specialised agents, each doing a small interpretive job, coordinated to produce a single readable summary. Legal, privacy, and consent thinking sat inside the design from day one rather than being bolted on at the end.
The outcome
A pattern for turning raw signal into narrative — and a working example that responsible-AI review can happen alongside the build, not after.