From navigation to contextual confidence
Visitors don't lack information. They lack context.
A companion that notices when a moment might be worth their attention — then gets out of the way.
This distinction changed everything downstream.
"The problem isn't that visitors can't find the penguin enclosure. The problem is that they don't know if it's worth the 12-minute walk right now — and there's no way to find out without walking there."
A companion that interprets the environment
and helps people notice what matters right now.
Each visit contains three moments where a small piece of context can change everything. Hover or click a signal point to explore each one.
A zoo visit isn't a linear experience. It's a series of small decisions made under conditions of incomplete information — each one shaped by uncertainty about what's worth it right now.
Three moments define the entire experience.
The system is designed for exactly those three.
The system doesn't recommend. It reads the environment and builds a picture of what's worth noticing — surfacing context at decision moments, then going quiet again.
The MVP defines confidence, not completeness. Every excluded feature was tested against one question: does this help visitors feel ready to choose, or does it add noise?
The interface is the smallest part of the system.
Three screens demonstrate the full design logic: how the system establishes context, how it surfaces a relevant moment at a crossroads, and — crucially — how it stays quiet when it should. Restraint is a feature, not the absence of one.
Navigate all three screens and test the quiet mode.
"Designing for uncertainty means accepting that the best outcome is not the most efficient one."
This project changed how I think about AI's role in physical environments — and about what it means to design for a place rather than a task.
Spatial thinking — the habit of reading environments as systems of movement, threshold, and territory — turns out to be exactly the right lens for designing AI experiences in places. Physical space has a grammar. Visitors read it instinctively. The design challenge is to speak that grammar fluently, not to replace it with a screen.
What this project clarified for me is the limit of optimization as a design goal. Optimization assumes we know what the visitor wants better than they do. Context assumes something different: that they already know what they want, but they're missing the information to act on it with confidence. One gives directions; the other gives ground to stand on.
There are things I didn't resolve. How does a system avoid becoming paternalistic when context becomes habit? I left them deliberately open — because the right answer depends on real visitors in a real zoo, not a design studio in Hamburg.
Designing for exploration means leaving some questions deliberately open.