Visualization clinic

Visualization designers Jörgen Abrahamsson and Johan Wastring met up with Jonas Löwgren of Medea to toss around ideas and constructive criticism on ongoing work. The general theme for the clinic was emergent glanceability.

Visualization Clinic 1

Jörgen introduced Rematrix, which is being done to give Malmö City better ways of understanding and working with welfare data for different parts of the city. His idea is to build what Bertin would have called a reorderable matrix, where the 40-some columns correspond to measured welfare variables and the rows are data per part of the city (broken down into women and men).

The demo contains several nifty tools for exploring the data; the discussion was mostly on how to develop it further into a useful system for city administration clerks and policy makers, as well as the general public. Some ideas included algorithmic support for showing which variables are most like a selected variable, a more organic focus+context approach and a multilayered interface for the public, where a set of predetermined views would be available in an outermost layer, with a possibility to dig deeper by deploying the more powerful general tools.

Visualization Clinic 2

Jonas talked about what could be emergently glanceable in social networks, in an effort to go beyond the visualizations of network topology that we have seen a million times in the last few years. We didn’t reach any definitive insights; a slightly promising idea was to focus on combinations of pure surface features (such as connections, activity, location, etc.) and intentionally disregard any attempts to process the contents of the online communication. Basically, this would mean to do the exact opposite of what we did in Pinpoint and try to imagine that we are looking at Facebook in Swahili, for instance.

Visualization Clinic 3

Johan shared some ongoing work on visualization of the state of large, complex enterprises. The whole notion of capturing large amounts of dynamic data in a form with strong Gestalt properties is highly intriguing, and we concluded tentatively that it needs to be approached by a combination of hands-on coding experiments and analytical reasoning on the level of abstracted visual primitives.

During the clinic, we found that adding the skeleton (red lines) to the more detailed, emergent shape (blue components) actually amounted to a visualization with two discernible perceptual layers, perhaps reducing the need for zooming or focus+context displays. This was seen as a useful step towards emergent glanceability.

A next step for Johan is to test the visualization algorithm with datasets for three or four corporations with known characteristics. We hope, of course, that the visual representations will come out different, and that the differences can be related to what we know about the underlying characteristics of the corporations.