eli,
a classroom
that feels less lonely.

Students in online classrooms were struggling. We thought it was access to resources. It wasn't.

Rural students — and everyone during COVID lockdowns — kept falling off academically in online classes. Our initial assumption was that the barrier was access to resources: bandwidth, devices, materials.

Research told a different story. The real driver was isolation and disconnection. Without the informal social fabric of a physical classroom — glances, side-chatter, shared reactions — relationships between students and teachers broke down, eroding motivation and, eventually, academic performance.

Interviews, affinity diagramming, empathy mapping, then Crazy 8s and a decision matrix tied directly to our research criteria.

Four core themes surfaced from the research. We ran generative exercises to develop a dozen concepts, then filtered through a decision matrix.

The winning direction treated engagement as its own dedicated entity — a physical device, deliberately distinct from the digitally-overloaded screens students already used for learning. We built early web and physical prototypes, tested them in real live classes, and iterated based on what broke.

A working end-to-end prototype — physical device, onboarding flow, teacher dashboard.

Students used eli to send real-time signals during class (mood, comprehension, reactions via custom emoji they created during onboarding). Teachers received this data in a low-density dashboard designed to give them a pulse on the class without disrupting the lesson.

Ended the project with a 3D-modelled and printed device, a student onboarding flow including custom emoji creation, and a Webflow-built teacher dashboard wired up via Zapier. The project received a High Distinction and was shortlisted for the USyd Architecture, Design and Planning Industry Award.

Assumptions early in a research process can be confidently wrong.

The pivot from "access to resources" to "social disconnection" completely changed the solution space. I also learned how much value there is in testing with real users in real contexts, not just controlled conditions — experience sampling was critical.