Manifesto
The app that's farthest along is the Subway Graph Deformation app -- it's not totally representative of where I'd like to go but has some interesting points.
If I've done my job right, we've taken a hierarchical dataset covering (9 station dimensions) x (80ish destinations per such station, each with 5 more variables) for each station, and turned it into an exploration my Mom can enjoy using (Hi Mom!).
It can also serve as a jumping-off point for the(suprising) number of interesting mathematical topics touched on at the link above. I'm working on developing a variety of similar ways to share, display and investigate data online.
It's part of a larger project to build a series of standalone components to see, communicate and comprehend highly dimensional datasets. (In the words of Edward Tufte, to present information in a way that is "documentary, comparative, causal and explanatory, quantified, multivariate, exploratory, and skeptical")
My strategy for doing this is to choose among the many rich, freely available datasets with inherent appeal that have proliferated, build a tool to explore it, then adapt the frameworks I build in doing so into generalized standalone widgets. The goal is to have a flexible widget that, once set next to a data pool and an XML gloss of that data, would let any interested party flexibly and intuitively navigate the dataset, even if they don't know XML from DMX.
Climbing on my soapbox: almost every means of communicting data among people is
- Divorced from its source,
- Final, and
- Static;
and this no matter how sophisticated the participants or broad the channel. Scientific papers are presented either as ink smushed onto dead trees or pixels smushed onto static bitmaps or PDFs. Web pages containing quarterly reports usually feature clumsy charts and bar graphs with no connection to the legally-mandated-to-be-publically-available rich data set they present. None of these let you investigate, interact with, or animate the data. Most of them involve a laborious process in which data is graphed in Excel or R or somesuch, exported to a bitmap, then modified to suitably embed in the final product. A good tool for data exchange should allow (but not require) that the link to the original data stays explicit and available for the interested recipient. It should arrive in a standard format, accompanied by semantic markup, that is easily adapted to any other suitable use of the information. Most importantly, it shoud put tools in the *audience's* hands to navigate and investigate the data set -- in the two-plus-one dimensions of screen and the many additional dimensions afforded by "Escaping Flatland" techniques.
The educational import of a tool like this is also immediately apparent. An increasingly larger proportion of enjoyable jobs will require (or be well improved by) the ability to steer a course through a sea of numbers -- find correlations, recurring patterns and anomalies, analyze them visually and quantitatively, and (most importantly) communicate them lucidly. There are not enough (ubiquitous, free, powerful, straightforward) tools which give the tactile experience envisioned here.