Spatial Browser

Design Process

People naturally remember information spatially. We remember where things are, on a bookshelf, in a room, or in a city using spatial cues rather than symbolic labels. This ability is deeply ingrained and we rely on it daily to organize, navigate, and retrieve information from our physical environment.

When we need to find a book in a familiar library, we often recall not the book's title or author, but its physical location, its position on a specific shelf, its color, its proximity to other objects. This spatiotemporal memory is a critical human faculty for sense-making.

That raised a core question for the project:

How do people use spatial memory to find and recall information?

Goal: Design an XR browser that works with spatial memory, not against it.

Grounding the design in research

Before designing interactions, I studied existing research on spatial cognition, data organization, and immersive interfaces. This helped me move beyond intuition and define concrete design constraints.

From this research, I developed two key frameworks

Four spatial classification axes:

1. Visualization Size
   Smaller than human · Human-scale · Larger than human

2. Visualization Scale
   1:1 · Miniaturized · Magnified

3. Acknowledgement of Physical Environment
   None · Pre-computed · Real-time (occlusion, lighting, morphology)

4. View Alignment Method
   Marker-based · Morphology-based · Geolocation · Alignment-based

   
   

Five classical data organization modes for spatial systems:

• Hierarchical – Tree-like parent/child relationships rendered in 3D

• Linear – Sequential or temporal data with contextual distortion

• Spatial – Data mapped to architectural or real-world space

• Continuous – Smooth 3D surfaces representing gradients or fields

• Unstructured – Networks and grids for unordered datasets

Having these frameworks allowed us to reason clearly about trade-offs instead of designing ad hoc 3D layouts.

Defining the interaction flow

With this foundation in place, I collaborated with other designers to define a browsing flow that felt simple, predictable, and calm.

Instead of showing one folder view at a time, the Spatial Browser presents information as layers in space:

• An always-visible signifier indicates that the browser is available

• Users look to target elements

• A short gaze reveals previews

• A longer gaze + explicit confirmation selects an item

• The next layer opens in front of the user

• Previous layers remain visible, forming a clear spatial path

• Users can look back to return or close the browser

This approach was designed to reduce cognitive load, prevent accidental activation, and help users stay oriented at all times.

Together with other designers, we explored multiple ways people might move through information in XR and tested how different interaction choices felt in practice. Along the way, we built multiple small MVPs to explore how different interaction patterns behaved in XR and how people responded to them.

Prototyping with real content

To ensure the system worked at a realistic scale, we built the prototype using the ArtBench dataset, containing roughly 4,000 images. The data was structured into a familiar, library-like hierarchy:

Art Movement → Artist → Artwork → Single Item

This allowed us to test how well the spatial model supported real search tasks, not just small demos.

Evaluation

After building the prototype, I ran an early usability study to understand whether spatial browsing in XR actually helped users, or simply felt novel.

Participants completed the same search tasks using:

• Spatial Browser in XR

• A traditional laptop browser as a baseline

At first, performance between the two systems was comparable, which was an important baseline: the XR interface did not slow people down despite being unfamiliar.

What stood out was how behavior changed over time.

As participants became more familiar with the Spatial Browser, strong learning effects emerged. People began relying less on reading labels and more on spatial cues, remembering where things were, recognizing visual groupings, and navigating based on location rather than text.

Participants also consistently reported higher satisfaction with the XR experience. They described it as more engaging, more intuitive once learned, and easier to reason about visually. The persistent spatial layout helped them stay oriented and reduced the feeling of getting lost in a large dataset.

These findings reinforced the core idea behind the project:

Spatial interfaces may not always be instantly faster, but they support understanding, learning, and recall over time.