A discussion at a seminar in Dagstuhl (Germany) on Information Visualization led to an interesting insight: what if we completely misunderstood who the users are for visualization? Especially in light of the current developments for broadly usable visualization, we need to rethink the types and levels of expertise that we can expect.
Two presentations in a session about Visual Analytics (by Stephan Diehl and Daniel Keim) presented work that had been based on visualization, but was not in itself visual. Diehl talked about a system for software development that was informed by insights from an exploratory visualization of version control data, and Keim gave a great overview over the problems when trying to “sell” visualization to potential users and funding agencies.
The current view of course is that visualization systems should be designed in a way that is useful to domain experts with little to no knowledge of visualization. In reality, of course these are not the actual users of visualizations: rather, the tools are run by their developers, in communication with the domain experts. Also, the idea of broadly available visualization tools á la Swivel and Many-Eyes presents us with a completely new type of user: the casual (but interested) non- or semi-expert.
So let’s face it: we’re deluding ourselves with our current user model. A much more realistic taxonomy of users (IMHO) is the following:
- Visualization Experts. We develop the tools, we use them. Simple as that. We use external data, and we communicate with domain experts, but we do not hand the tools over. We know how to read our displays, and so we can make things that are far more advanced and complex than we would expect somebody without experience in visualization to understand.
- Casual Users. The people who actually use visualizations they did not develop themselves are casual users, who are curious about something, or who just like playing with something visual. These users need general (i.e., not data-specific) tools that will be much simpler, and that will need to follow known interaction paradigms as much as possible.
Understanding this will make our visualizations much more useful in practice, and we will have more time doing productive work instead of chasing after users that simply don’t exist.