Wednesday, August 31, 2011

Reaction: The Value of Information Visualization

Right off the bat, I appreciate that the paper refers to Information Visualization as a proper "field" of study, and gives it a short name, InfoVis. It's a small thing, and it's probably obvious to anyone taking the class, but perhaps those outside Comp Sci. don't understand that there's tremendous depth to visualizations. The authors make the point later that the depth isn't obvious, but I feel like they overplay it.

I'm not sure that I agree with their conclusion that InfoVis is hard to quantify. Their reasoning, that visualizations are used to amplify cognition, and that cognition amplification has no "ground truth," seems spotty. Why can't we look at a hard to read, flat chart full of data, and a good visualization of the same data, and not easily say that the visualization enhances our understanding of the data? I think the improvement is obvious if the visualization is good enough, if not completely measurable. It appears to me that they're calling for some kind of measurable cognition amplification, which doesn't seem necessary to me if the improvement is obvious. Also, this statement to me seems blatantly false: "Unfortunately, however, activities like exploration, browsing, gaining insight and asking better questions are not ones that are easily amenable to establishing and measuring a ground truth." On the contrary, it seems like asking better questions itself helps to establish a ground truth, even in the context of InfoVis.

I'm trying to come up with a few analogies here, though, but I'm struggling.

I do agree with the authors that InfoVis might not be useful if a person has a specific question about some data, but that it does help ask the right questions potentially, and they're useful when a person doesn't know exactly what they're looking for in some data. However, I don't agree that this in any makes InfoVis hard to justify.

I also like this statement: "visualizations can support more efficient task performance by allowing substitution of rapid perceptual inferences for difficult logical inferences and by reducing the search for information required for task completion." If visualizations need extra justification, I think that's a good line.

I wish I knew what the authors were trying to say about data mining. I feel like that simple dataset could be build the graph that human users drew, but I don't know for sure.