Animation and The Value of Visualization
From CS294-10 Visualization Fa08
Lecture on Dec 8, 2008
Readings
- Animation: Can it facilitate? Tversky, Morrison and Betrancourt. (pdf)
- Visualization Research Challenges. NIH/NSF Research Report. Johnson et. al. (pdf).
- The Value of Visualization. Van Wijk. (pdf)
Kuang - Dec 08, 2008 02:45:12 pm
The value of visualization paper is fantastic! It may be controversial, but the visualization-as-technology "calculus" is really good for me in terms of placing everything we've discussed into a connected (hate using this word loosely:) paradigm for the practice.
The author is fairly up-front about how it's hard to quantify the various elements of the model, and that its utility is limited as such. But, for students of visualization, I think it is an informative model for "placing" various efforts in visualization.
The discussion at the end about visualization-as-art and -as-science: I found these useful counter-opinions, but less compelling.
My first thought was that I wish this would have come earlier in the semester... but then, maybe we didn't have enough background to fully appreciate it. Look forward to reading others' comments.
Witton Chou - Dec 08, 2008 03:09:20 pm
just b/c we love them so much... thx Ketrina for all the great drawings to walk into class and enjoy:
Seth Horrigan - Dec 08, 2008 06:45:48 pm
I supposed that Tversky would answer "no", to his question "Animation: Can it facilitate?" As he said of those systems where animated versions provide better performance, "the animated graphics present information not available in the static versions." While I do not doubt that the studies he referenced suggest that animated versions of the information presented do not achieve significant improvement over static versions that have all the same information, his insistence that the static versions compared to must present all the same information misses the point. When trying to describe an inherently mobile system, or a system that varies over time, describing all the information in a static medium may be hard or impossible. If the system described is animated, it probably makes sense to show it as animated.
There are cases where this may not be the case. As Tversky mentions, systems composed of discrete units may lend themselves to static representations of the discrete steps where each is described, but if that system is animated, this merely moves the work of the animation to the imagination of the reader. Like he mentions, a progress bar should be animated. The information varies over time, and the information varies in real time. Perhaps you can figure out how to represent this statically, but why bother.
I suppose what I am saying is: animated visualizations may not perform better than static visualizations when both contain the same information, but there are some forms of data that cannot be visualized without animation, and for everything else, if it is more complex to create a static version and this does not outperform an animated systems, go with animation. It is probably more natural, and contortion is not an end to itself.
Chris - Dec 09, 2008 02:45:28 pm
I too had mixed feelings about the remarks in the Tversky paper.
With respect to the argument that "animations may be hard to perceive," I found the remark that "sketches of the trajectories of pendula, propelled objects and dropped objects by novices and experts alike are often incorrect" to be revealing of the fact that we may be looking at two different things. What I believe animations in this case are very effective at conveying are a semantic description of motion (e.g, "it's falling/walking/running"). The author is concerned more with a quantitative analysis of the motion (e, g" "what is the position at time T").
A relevant question which is not discussed with respect to this distinction is "how do we plan to use this data?" If we only need a semantic understanding, then the animation will do fine. If we need to read off a position at a given instant of time, then clearly we will need to analyze a static snapshot.
Ketrina Yim - Dec 09, 2008 11:42:35 pm
@Witton: Thanks for putting the last 294-10 'toon on the wiki! I hope everyone enjoyed viewing it as much as I enjoyed making it, and all of the other ones, too.
Though Van Wijk's use of calculus seems a tad excessive, it does reveal the importance of considering the "cost" versus knowledge gain of visualizations. We've learned about some quite impressive visualization techniques in this class, but they all require significant investments, usually in the time the designer needs to learn to use the tool. Pie charts and bar graphs may not be the best way to present data, but they are very easy to make, thus giving them a low development "cost". This highlights the need for user-friendly interfaces in visualization creation tools. Only then will current visualization research be more widely adopted.
Speaking of animation in visualization, Le Grand Content depicts numerous aspects of life as graphs. Van Wijk may find it neither effective nor efficient, but it has a high artistic value at least.
Michael So - Dec 12, 2008 01:30:31 am
I like the comment on the Tversky paper, particular about the reason animations fail due them being comprehended as discrete steps. Even if it is perceived as discrete steps, I believe there are instances where animation can still be of some use. If for example if what is being visualized are assembly instructions, a static graphic with arrows depicting the path of an object may be adequate. But in those instances, an animation can confirm the user's interpretation of those arrows, or even perhaps correct the user's interpretation.
And even if an animation is comprehended as discrete steps, it shouldn't imply that animation is less effective than a list of discrete static frames. I believe there are instances where an animation is more effective in those situations. An example would be learning the steps to accomplish a yoga pose. Viewing a series of static graphics takes time to analyze and interpret, whereas in an animation, the user can just mimic what he or she says and follow along with the animated tutorial. The animation can sort of simulate having an instructor there to teach you versus reading an illustrated textbook.
Simon Tan - Dec 12, 2008 06:44:02 am
I think what Tversky was trying to convey was a concept that we've seen repeatedly in this course; there are a variety of techniques you can use to better present the data in a visualization, but the trick is to use them well. Just as you can use color poorly and result in a worse visualization that you would have otherwise, I think it would be easy to mis-use animation and accidentally obscure the intent of a visualization. For example, if you don't have experience as an animator, you could easily make mistakes such as having two animations too intertwined with each other, resulting in a perceived causation.
I'm curious as to how Prefuse/Flare attempted to attack this problem with their animation libraries. I know I've seen Flare-based visualizations where one type of chart or graph turns into another (e.g. bar graph into pie chart); were principles from research used to create the animations or are they of a custom design?
Scott Murray - Dec 14, 2008 08:47:50 pm
I'm happy that we ended the semester with a lecture on animation and the use of motion in visualization—I only wish we had discussed it sooner. Motion is going to play a significant role in all the new possibilities for user interfaces and interactive, dynamic data visualizations in the future. It will help to get a head-start on understanding appropriate use of motion and developing a language to use with it.
For an innovative use of motion in data visualization, see Carolin Horn's Anymails project.
David Poll - Dec 15, 2008 12:00:59 pm
I don't have anything terribly insightful to say about Animation except that I wholeheartedly agree that it can make or break a visualization (bad animation can be REALLY bad). I did, however, want to thank everyone for their great insights. It's been a pleasure working with all of you, hearing what you have to say, and showing how incapable I am of seeing colors and space :). A big thank you to Maneesh as well. I really enjoyed the class, and got a lot out of it. Now I'm off to work on UI frameworks at Microsoft :) Wish me luck!
Witton Chou - Dec 15, 2008 06:03:18 pm
I wholeheartedly agree with Scott's comment - I too wish we had discussed the use of motion and animation sooner. It feels like adding a dimension of time opens many new doors. I think animations, if done correctly, can really facilitate one's understanding of a visualization. However, it is also possible that the animation can turn out to be distracting; so there is a delicate balance between what is helpful and what is extraneous.
I think Apple does a great job with using animations on the iPhone. Though at times the phone will lag, hang, and crash while processing data or animations, it has a very welcoming interface. Its response to motion on the touch screen and the corresponding visual animations in scrolling through a page are very intuitive and help the user track where they are and how they got to a particular segment of the page. User interfaces and their products' look and feel are a large part of Apple's success.
Calvin Ardi - Dec 15, 2008 08:36:12 pm
@Ketrina, I remember looking at the formulas on the slides and wondering what that was all about, but I think we're supposed to take it as some sort of generalization and go from there (e.g., if some variable is extremely high and another is extremely low, this would result).
It's interesting as to how animation can convey more information than static images, or even a series of static images. There's obviously a lot of different ways and techniques to animate and choosing the appropriate one can enable the viewer to understand more fully what is being visualized. Howstuffworks usually takes advantage of Flash and other multimedia to animate various things, ideas, and concepts (the four stroke combustion cycle was one we covered in class). A few weeks ago we read about a system that would allow users to peel or cut away layers of the human body; the authors found that without an animation, processing the before and after states took a much longer time as opposed to adding an intentional delay and the animation of actually peeling away the layer.
I also enjoyed the Digg stack; a lot of discussion ensued in class about it's effectiveness and initial confusion, but each element and encoding in the visualization had some sort of purpose to it. It's certainly neat (and, apparently, a screensaver) while it may not have a lot of usefulness to it, it's interesting to see all the "diggs" happening as opposed to refreshing the front page constantly.
