“the main point of information visualization is not to decorate, nor to show-off but, to help in making the right conclusions on unexplored data”

I recently googled for information on visualization, or casually, “infovis”. I ran into a weblog on Information Aesthtics (”infosthetics”). Visualizing social networks seems to be one of the leading topics in this blog, too. Network visualization is, as Janne recently pointed out, technically difficult as such, but also conceptually problematic. And network visualization is only one visualization task in our work.

The term infosthetics reminded me of the seminal books by Edward Tufte, that crystallize making good visual representations in conjectures of graphical excellence and a couple of quantitative measures. Tufte’s books are a delight to read – not only because of their highly professional graphical standard – but because of the interesting stories. The most startling of them is perhaps the possible role of a disinformative data visualization in the tragedy of th e explosion of space shuttle Challenger (STS-51-L). Its, of course, often too easy to second-guess; to get another perspective, see also a thoroughful criticism on Tufte’s claims on the Challenger issue by Wade Robinson and others.

Another seminal classic is “Semiologie Graphique” by the French semiologist-cartograph Jacques Bertin. He is attributed the first detailed theory and taxonomy of data graphics. Bertin is a semiologist and he treats maps and data graphics as a language where the elements of visualizations are signs. The theory involves detailed taxonomy graphical attributes, such as, luminance, color, texture, position, and size of objects, classifying them according to their most suitable purpose of use. The theory preludes the modern psycho-visual, scientific approach to the theory behind data graphics and information visualization. The thesis by Jarkko Venna, recently defended in my “academic home”, is worth reading as an example of psycho-visually based visual data mining on high-dimensional data and similarity structures.

Visualizing information is difficult in general, and it is not a question of computer graphics alone. The visualizations have a multitude of roles. They can be a form of aesthetics; they can be promotional, informative, pedagogical and most importantly in data analysis, be tools of analysis that “amplify cognition” – just as Card, Mackinlay, and Shneiderman put it (see here).

Aesthetics, or elegancy of the visualization helps in achieving in this task at least when understood as a form follows function principle. However, the main point of information visualization is not to decorate, nor to show-off but, to help in making the right conclusions on unexplored data and in communicating them to others. That’s why information visualization is difficult.