WORKSHOP DISCUSSION DOCUMENTATION

Network of Experts (N/X) Workshop, "Multimedia Visualisation of Massive Military Datasets",

Aalborg, Denmark, October 3-4 2001

A brief presentation/discussion of 13 guidelines for effective visualization (PowerPoint presentation)

Tore Smestad

FORSVARETS FORSKNINGSINSTITUTT
Norwegian Defence Research Establishment
P.O. Box 25 No-2027 Kjeller Norway

This presentation/discussion took part in the discussion period just before lunch on October 4th (from 1155 to 1207 according to my diary). It was initiated by Martin Taylor during a discussion earlier that day to present old ideas of mine, see "Working paper" of 1993 titled "Outline for concepts and principles of effective visualization" on www.vistg.net. This material was first presented publicly at the TTCP Visualization workshop in San Diego March 8-10 2001.

To make the presentation/discussion as visual and to the point as possible, I got arranged for three overhead pictures simultaneously, side by side, at the front. In the middle was a picture presented by Bill Wright the previous day that seemed to match a number of the 13 guidelines.

A set of 13 guidelines for effective visualisation

  1. Place a lot of messages into single images.
  2. Make illustrations with high image-information and a low ratio between the irrelevant image information and that of the coding-information.
  3. Partition an image into gradually smaller parts.
  4. Arrange the visual structure so that 3-7 visual units may be naturally focused at a time.
  5. Select diagram forms with a high signal sensitivity of the message(s) to transmit over the actual range of messages(s).
  6. Use obvious coding schemes; don't change them, clarify them explicitly if not obvious.
  7. Select diagram forms and coding that are likely to resemblance humans mental image.
  8. Make illustrations with a visual structure and content according to prior knowledge.
  9. Use several diagram forms in the same image.
  10. Select diagram forms representing different levels of abstraction in the same image.
  11. Pair many instances of n-tupple elements of the selected diagram instances by the connection metods.
  12. Use transparent and adjacent connection to minimize the position-uncertainty of visual units.
  13. If the illustration does not contain all messages, select diagram forms with connection keys to (potentially) supplementary illustrations (that is: furnish valence to the illustrations)

The guidelines are motivated by information theory, and might be seen as a compilation of rules and principles behind a vast number of illustrations that intuitively are good, a lot of them presented in the books of Edward Tufte.

My key point was to show how the display of Bill Wright can be claimed to agree with a number of these guidelines - a display ("Risk Movies") that I believe is a very efficient display for interactively presenting a lot of complex numerical and qualitative relations aggregated from stock marketdata around the world.

The 13 guidelines use a number of both familiar terms and perhaps unfamiliar ones in this context like "diagram form", "signal sensitivity", "connection method", "adjacent connection", "transparent connection", "valence", etc. (All terms in italic in the 13 guidelines are defined in the working paper.) A few of these concepts were described using the shown illustration of the conceptual frame of reference. Especially the three "connection methods" (adjacent, transparent, and expanded) were described being quite central. The display of Bill Wright uses the two connection methods adjacent and expanded.

Guideline 2 was omitted in the discussion since it took too much time to explain. The other guidelines from 1 to 4 seemed to agree well with the display of Bill Wright. The 5th seemed also well met. It states that when changing somewhat of the essential messages, the appearance of the display should also change noticeably. The guidelines 6 and 7 also seem well met, probably also the 8th. Guideline 9 is well met since the display uses several diagram forms (at least six, the number depends on a definition of "diagram form"). Guideline 10 is to some extend met, at least if one define "abstraction" in terms of aggregation. Number 11 is well met since all instances of diagrams and data are somehow connected to others, mainly by adjacent connection. Guideline 12 is not very relevant here, but the bars in the table at the "floor" can be seen as an example of this. Guideline 13 is not relevant here since only one illustration (image) is used. One can instead say that Guideline 1 is met to the extend that all the separate illustrations normally necessary for conveying all this information are effectively condensed in one by use of connections by the "valence" supplied by each diagram form.

The brief presentation period ended with a lot of comments and viewpoints from the audience. Most of them addressed the meaning of the concepts and guidelines related to the sample display used here. There were comments stating agreements to guidelines with the same interpretation as I also would have formulated (which I found very pleasing to hear, I admit!). The discussion had to be aborted due to overtime for lunch.