Technology + design + art

Terra EnCognito: The Nature of Online Discussions


Terra EnCognito, 2007

This project is featured on the companion site to the Princeton Architectural Press book Participate: Designing with User Generated Content by Helen Armstrong and Zvezdana Stojmirovic.

Terra EnCognito is a realtime data visualization of online discussions as they unfold on the social news site The project’s primary goal is to enable quick, accurate, comparative, and qualitative assessments of individual story discussions across the entire site. By exploiting our instinctual response to organic form, movement, and behavior, I increased the number of data axes and overall complexity in the visualization without increasing the amount of learning or abstraction required by the viewer to understand it. The result is a visualization that embodies unique, dynamic, on-line discussions as a population of unique, dynamic, living creatures—with corresponding form and traits.

Discussions are scored according to a set of metrics designed to reveal more of the discussion’s nature rather than its quantitative attributes. Those scores are normalized and mapped to variables in Johan Gielis’ Superformula—creating organic forms, with added behavioral traits and lifecycles. The entire visualization is rendered in realtime using the Unity game engine.

There are nine metrics used to derive discussion qualities. For example, the Polarization Index describes the level of accord or discord within a discussion by looking at the spread and distribution of comment ratings and the ratio of uppercase to lowercase characters; Dialogical Balance describes how well the discussion is fostering dialogue by looking at the spread and distribution of comment nesting depth, occurrences of multiple comments by individual users, and the ratio between the number of first level comments and nested comments.

The rendered color palate is purposely muted to emphasize spatial, shape, and behavior attributes. Juveniles, or stories not yet promoted to the front page, are smaller, shapeless, flocking black orbs. Once promoted, they morph into adults and move to a category/topic based region defined by a treemap. At death, objects fall to the ground where their accumulated numbers provide a historical record. Shapes and behaviors are complex, and never identical, but are intuitive and consistent in their expression of the metrics–number of arms, thin/fat, long/short, docile/aggressive, slow/fast, spiked/rounded–so that one can immediately grasp the nature of a discussion, and over time can begin to recognize species.


Stamen, the studio in San Francisco, loomed large in innovative data visualization at the time. I wanted to use their Digg API (remember Digg?) to do something a little different. And since I found Digg comments entertaining, I wanted to build a practical way of visualizing all of those discussions.

I also wanted to experiment with formalizing complex information into something easy to grasp. People respond to aspects of the natural world instinctively. I wanted to exploit biophilic design to convey complex information in a way that a chart or graph couldn't.


Unity Pro. I had never heard of Unity before I began researching easy-to-use game engines for this project. I loved it the moment I started using it. Within a week, I was at the very first Unite Conference (2007, San Francisco) and, with CEO David Helgason's help, had working code to dynamically generate Superformula meshes. There was also some MySQL, PHP, Visual Studio, C#, and SQL Server.