Edward Tufte’s capstone presentation was the highlight of the conference. I have seen him talk before and he is a powerful speaker. His books on information display are well know and highly regarded. (See Amazon.com author pages: http://www.amazon.com/exec/obidos/subst/features/t/tufte/tufte-edward.html) His prior talk I attended included disparaging remarks on computer displays, mostly because of the available resolution. This turned out to be one of the sub-themes of his discussion. He started out with a graphic illustration of a display resolution scale by talking from different places on the room. On the high end is a quality print which gets approximately 4000 dots per inch. Mid range are common books such as romance novels. Near the low end are normal computer displays. Below this are the presentations generally displayed in “misshapen trapezoids” (overhead and video projectors). Far to many presentations are done this way and exhibit the worst of information visualization.
The primary theme of Tufte’s talk was “The Five Grand Principles of Information Display”. These are:
These derive from a parallel 5 principles of analytical thinking, not the current fashion of information design. When creating visualizations, the designer should ask: what is the analytic task? and How can I best support this task (invigorate it)?. A viewer should ask: what other displays of this data are possible? And what would I really like to see?
He stressed the need to provide quantitative documentation within the displays. Axis should be labeled and scales provided directly on the display. Avoid the ‘key’ box showing line/symbol meaning – label the line. He used past conference proceedings as an example of how well the visualization community was doing. Fully 80% of the images in the 1998 proceedings did not show scale, variables or other labeling. Only 2 % were adequately labeled. The other 18% had partial labeling.
Displays should be content driven. They are designed to show something and that should be the primary design effort. Avoid using ‘optically vibrating chart junk’ to add flash to a graphic (extra boarders, etc.). It is irrelevant and distracting from the content (unless the purpose is to distract from the lack of content!) The same can be said about computer displays. Modern windowing systems have an incredible amount of graphic junk that takes up valuable screen real estate. Tufte noted that at the lowest end of the information content per inch scale are web pages. First the windowing system grabs its area (including lots of advertising for the vendor), then the browser takes away a fair chunk to display its menus and advertisements, then there are the banner ads, frames, spinning logos, flashing text and other distracting web junk. When you analyze the amount of space for the actual desired content, it is often less than 10% of the available screen. Romantic novels have a hundred times the information density!
Tufte showed three first edition books from his private library as examples of good graphic design. One of these was Newton’s Principia Mathematica, the other two were by Galelo. All of these changed how people would reason about the world. They included graphics and tables but did so completely in context and without a “see figure X” reference. The graphics were done in-line. These were impressive and inspiring, although I do wonder if Tufte isn’t making too much money off his books. He paid almost $500,000 for the 1610 edition of Galelo’s “Starry Messenger”.
One of the hand outs from this talk was Tufte’s 30+ page booklet “Visual and Statistical Thinking: Displays of Evidence for Making Decisions”. This uses two case studies where statistical visual displays were crucial to life and death decision making. The first case study is John Snow’s famous discovery of the source of the 1854 London Cholera Epidemic. The second, in contrast, is the disastrous results of poor displays used by engineers to argue against the launching of the Space Shuttle Challenger in 1986. Tufte examines the graphics and statistical evidence presented in both cases and provides several alternatives to show better or worse approaches.
Good design is clear thinking made visible
Bad design is stupidity made visible. It reflects a deeper corruption and lack of quantitative knowledge.
Don’t demand understanding in 5 seconds. A good visual should encourage the user to contemplate it. The first response should be content related, not layout.
post speech discussion – Tufte has a PBS/BBC video in development – he hopes to do for video documentaries what his books have done for other media. Not your usual talking head!