Designers Jörgen Abrahamsson of Malmö City and Johan Nystrand of Google met up with Jonas Löwgren on Sept 28 to talk about information visualization. Much of the discussion centered on Manuel Lima’s Information Visualization Manifesto and general aspects of infoviz. There was also a fair bit of nerding around specific visualization challenges.
The first principle in the Manifesto proposed by Lima is that Form Follows Function. His point is that data as such are inherently formless and that the form of an information visualization must follow from the purpose of revealing (insights about the data).
Using this as a starting point, we concentrated on the notion of communicative intention: What is it that we want to say with an information visualization? This question may be a useful starting point for deciding on form and interactive behavior.
If the intention is to communicate at a glance how people voted on a four-choice question, for instance, then it would make good sense to stick to conventions and present the results as a pie diagram or a bar graph. After all, there is an established genre of simple information visualizations for simple datasets, that is reinforced many times every day through Excel diagrams, basic news graphics and so on. More complex intentions and more complex datasets would require other forms.
An obvious corollary is to state that people should be encouraged to use hand drawings and other simple techniques to visualize general ideas and relations, rather than using tools that force them to make something look much more precise and complex than it is.
Which leads on to aesthetics of information visualizations. Lima holds that “[a]esthetics are an important quality to many Information Visualization projects and a critical enticement at first sight, but it should always be seen as a consequence and never its ultimate goal.” This harmonizes quite well with the general view of infoviz as visual communication: The emotional, affective and otherwise aesthetic aspects of the chosen expression are parts of the planned communication and influence how it is received.
Simplicity and clarity are often assumed to be core aesthetic values in infoviz — the standard example here would be a diagram by Edward Tufte next to an autogenerated Excel diagram of the same dataset, plus a few comments about chart junk and data-ink ratios — but they may actually be bundled with the generally assumed goal to reveal. A simple and clear visualization may be seen as revealing since it “suppresses the noise” and “focuses on what is important.” However, if the intention is to communicate a sense of the complexity and hidden depths of a dataset, then it may be more appropriate to design for exploration by expressing aesthetic qualities like responsivity and pliability.
What we are talking about here is in fact examples of professional design skills in infoviz, including a repertoire of generative examples and a developed ability to judge the predicted goodness of a particular infoviz treatment. How can such skills be disseminated and put to valuable use? Writing a manifesto is obviously a step in that direction, as is collections of generalized examples called patterns (a good example is infodesignpatterns.com). A more developed arena for infoviz criticism would be another useful addition.
A prevalent strategy is to encapsulate visualization knowledge in generic tools, such as the diagramming functions in Excel. Even though such tools arguably do more visual harm than good today, we may hope for a development similarly to what happened in desktop publishing. Broadly speaking, the period between 1985 and 1995 was typographic hell in the field of business communication. When everybody got the tools on their own desks to do rudimentary graphic production, then everybody used the tools. The resulting mess of ugly documents and presentations motivated the gradual development of better templates and a better understanding of when to ask for professional help. It is interesting to note that Excel’s cousin Word today has traces of communicational intent embedded in its formatting palettes: If you use a standard template, you mark a line as a second-level header rather than as a bold-faced, 16-point piece of blue text. Introducing semantic markup may be a feasible way forward for generic visualization tools.
Moving on to the nerdy bits, we talked about the temporal dimension in infoviz. Good old Gapminder was an early and influential example of using animations to visualize time. We observed that it seems difficult to use animation for variables that are not purely temporal. But what about interactive controls?
Assume that we have a set of data listing home addresses and disposable incomes for people in a city. Assume that we plot each person as a dot on a city map. Now assume that there is a disposable income slider filtering the dots. It seems safe to expect that playing around with the income slider will reveal the socio-economic topography of the city with some accuracy. An animation, on the other hand, would probably not support such discoveries equally well.
It was also noted that motion under the user’s control makes it easier to discern and explore the qualities of the motion than a straight animation would, even if it uses traces. In general, Gestalt qualities of motion seems to be an underexplored field in infoviz. A survey of existing research (1) supports this observation, and also suggests similarly to our discussion above that interactivity may be a promising way to make motion more broadly applicable in infoviz.
1. Tversky, B., Morrrison, J., Betrancourt, M. (2002). Animation: Can it facilitate? Int. J. Human-Computer Studies 57:247-262.