I've been playing around with some visualization tools and looking at visualizations this week and I'm always disturbed by the seductive nature of visualizations..
What I mean: there are some that result in some very, colorful pretty output but it's very difficult to hone in one what individual data nodes represent and even more importantly, it's really tough to discern where the data are coming from. Personally, I think that a visualization is useless if you don't have that latter piece of information. Many Eyes is a lot better than SciMinder when it comes to this, yet SciMinder is often cited for the "oooh, pretty" factor. (And don't get me wrong, I say, "oooh, pretty" as well :)) I think a very useful purpose of visualizations is to be able to provide an anchor so that we can discuss our assumptions and the questions that drive the data collection (and then come to some conclusions about what we can fairly do with the data).
Yet, I wonder how many people looking at visualizations actually bother to trace the source of the data or ask what's been included/excluded when creating a visualization?