I attended Pat Hanrahan's talk the other day at Adobe. It was on visualization techniques. It was interesting although I found it a little basic. The techniques he described have a long history, dating back to Minard's map of Napoleon's Russian campaign. Phase Space representations and simplified distorted topologies were both shown and discussed. These are certainly valuable visualization techniques, but both have been around for a while.
Phase space representations came of age in 1963 when when Edward Lorenz was plotting weather data and derived the Lorentz Attractor which is a phase space visualization of the data he gathered.
Hanrahan's primary point, however, that the visualization should be chosen to work best with the context in which the data needs to be delivered, is spot on. In many respects, a contextually approrpriate visualization can present data in a manenr that highlights the information content specifically.
This is especially important for self-adapting content. You can extrapolate most of the same principles when the issue of transforming or adapting content to different devices and contexts.