current page

history

user

Lecture on Oct 26, 2009. (Slides)

Readings

  • Required

    • Animation: Does It Facilitate? Barbara Tversky, Julie Morrison, Mireille Betrancourt, International Journal of Human Computer Studies, v57, p247-262. 2002. (pdf)

    • Animated Transitions in Statistical Data Graphics. Heer & Robertson. IEEE InfoVis 2007. (pdf)

  • Optional

    • Smooth and Efficient Zooming and Panning. Jack J. van Wijk and Wim A.A. Nuij, IEEE InfoVis 2003, p. 15-22. (pdf)

    • Effectiveness of Animation in Trend Visualization. Robertson, Fernandez, Fisher, Lee, Stasko. IEEE InfoVis 2008. (pdf)

    • Animated Exploration of Graphs with Radial Layout, Ping Yee, Danyel Fisher, Rachna Dhamija, and Marti Hearst, IEEE InfoVis 2001 (pdf)

    • Principles of Traditional Animation Applied to Computer Animation John Lasseter, Proceedings of SIGGRAPH 87, Computer Graphics, 21(4), pp. 35-44, July 1987. (acm)

    • Animation: From Cartoons to the User Interface. Bay-Wei Chang, David Ungar, UIST 1993. (pdf)

Comments

vad wrote:

I was very excited when the animation of the sterling engine was shown in lectures. I love engines, and I have learned everything I know about engines from static and animated visualizations. I want to share some of my favorite visualizations:

AnimatedEngines a mechanical visualization gold mine.

Ford Engine 1 really gives you a good idea of where all the moving parts are inside the engine (as well as oil and coolant). Notice the effective use of transparent mesh in 3D and cheesy music.

Ford Engine2 a very cool video. (Shorter than the one above)

How Rotary Engines Work a really cool design that takes some time to understand. A great example of a cycle by cycle design.

malee wrote:

I appreciated Tversky et al's critical analysis of animation. While studying different visualizations, it is easy to be swept up by the coolness of a visualization and forget the whole point of creating the viz in the first place. As a researcher, it is also an easy trap to interpret experimental results that exaggerate the effectiveness of one's own system. The congruence and apprehension principles provided a nice framework within which to examine animation, and I liked how the Animated Transitions paper listed design considerations specifically within this context.

On another note, I have been noticing just how much more aesthetically pleasing the diagrams are in the newer papers (such as Animated Transitions and Pivot Graph). It would be interesting/fun to consider aesthetics as another metric of visualizations. It seems that people would pay more attention when interacting with a pretty viz than one from 1995, and so perhaps this increased engagement leads to greater effectiveness? : )

vagrant wrote:

I found the reading "Animation: Does It Facilitate?" to be comprehensive. In particular, I found the distinction between animation and interactivity to be poignant. That animation without interaction may produce results little better, if not worse, than those of static graphical visualization makes sense to me given the examples described in the article.

Interactivity as the key advantage that animation offers excites me from a primal passion. As an avid video gamer, I felt that the discussion in the InfoVis 2007 paper would be of great interest to video game developers. Before the crux of the actions any video game is grasped comes the presentation of the game--from the menu to the in-game user interface. In both categories the importance of easing the player--be he a veteran or an inductee into the pastime--into the experience is critical to whether the learning curve is a ten-minute affair or one that takes hours. Beyond video games, I believe that the user interface of any application that requires frequent data/status lookups should meet a higher minimum standard of clarity. Good animations could almost guarantee such a goal.

Finally, I found the examples shown this week in lecture to be more stimulating than those that came before it.

jieun5 wrote:

I really liked the systematic approach used to evaluate animation in the Heer & Robertson paper. Granted that efficacy in choices of graph types and transitions between them depends much on the nature of data, I still really admire the way experiments were tightly controlled to compare static vs. animation vs. staged animation.

Here are a few questions that came to me as I was reading through the article:

- The two experiments are described as 3x2 within-subjects design for each transition type. Are the results shown in Figure 6 and 7 "average" error across all subjects? Or is it average across repeated trials for a given subject? If it's the former (which I think it is), I'm pleasantly surprised by the lack of variability found across subjects. My intuition tells me that there should be great across-subject variability on tasks measuring cognitive performances in general; for instance, we could easily have a scenario in which the best performance for person A is using "animation" may still be worse than the worst performance for person B using "static"... But the black bar denoting variance(?) on the Figures 6, 7, and 8 seem quite small.

- The remark in Section 5.2, that the concept of % increase and decrease can be confusing, caught my attention. I remember having struggled with this concept growing up-- that increasing a price by 10% then decreasing it by 10% of the resulting price does not get me back the original price, that it matters with respect to *which* value (original vs. new) when working with percents. Though the paragraph says that "participants were informed of the difference between negative and positive changes, and practice trials revealed correct answers so subjects could calibrate their estimates", I wonder whether these calibrations were "enough" to last through the test session.

- I wonder exactly what made the subjects to laugh upon first viewing the multi-staged stacked bars transition. :) Perhaps it anthropomorphized the data too much?

- I wonder about the feasibility of having (semi-)automatic determination of automated transitions in presentation tools, as mentioned in the Conclusion. I feel that the results of the experiment are specific to a given set of data, and that the optimal animation parameters will depend highly on the nature and complexity of data-points, and possibly even on the experience and cognitive abilities of the audience...

nmarrocc wrote:

Vagrant makes a good point about animation in video game menus. It seem to me that almost all of the newer video games i've played animate their menus. I'm not really sure if the animation is superfluous or not. it doesnt really seem to add any additional functionality. Sometimes it can help you navigate, by maintaining orientation or providing intuition, but i think the real reason they add it is that its cool, as an art form its just more fun and interesting to interact with stuff that moves.

I never really thought about the distinction between animation and interactivity. One could argue that animated menus are a form of interactivity and not pure animation. What are some other interesting ways to animate things to increase intuition? The jiggles in the force feedback graph give the user clues as to how the graph will react to specific actions. What are some other animations that can be used to give users clues as to how to interact? The best intuitions are those that reflect real-world things. There is a really cool demo I saw last year of an interactive desktop model that was 3d. http://www.youtube.com/watch?v=M0ODskdEPnQ Instead of using 2d windows to represent files and applications it used 3d graphics of paper. When you move the paper around it animates to look like moving around real pieces of paper. One could argue that this form of animation seems superfluous but I think that it gives the user a more intuitive feel for how the model is supposed to work

bowenli wrote:

Tversky: I wish this paper was a bit shorter. The paper cites many other works so you would expect it to be some kind of summary. However, it's pretty long and it's really hard to judge the claims without going into some kind of detail and reading the papers. Overall it seemed a bit unbalanced, with the "what" section being too long, and then the "why" section leaving more explanation to be desired.

The congruence and apprehension principles sound like re-applications of considering people's mental models. Also, I would agree that the results should be considered for complex systems only. For example I think @vad's engine videos are pretty good. Recently, in me116 my group presented a pair of crutches we designed. The physical model was junk, but having a video in CAD of the mechanics really helped get the message across.

@malee I think this is one of the fundamental questions in design. Does aesthetics come at the price of functionality or can they coexist? Here's Norman on design and emotion: http://www.ted.com/talks/don_norman_on_design_and_emotion.html

Heer, Robertson: I think I understand the motivation for the research as described in the introduction, however I'm not sure that the way the experiment was set up really tests comprehension of translating one graph to another. If you ask participants to focus on specific values, then I think that optimizes for certain types of transitions. But will those transitions lead to comprehension of the graph as a whole - and the remembrance of what the previous graph was? I'm not sure.

alai24 wrote:

I think Animation: Does It Facilitate? brings up some very interesting points when it starts discussing the equivalence between static and animated graphics. One can take the fact that some of the animated visualizations present information not available in the static graphics as evidence of animation's ability to condense information in space.

Additionally, the difficulty people have with accurately predicting trajectories underscores was a factor I had never considered when dealing with animations. There is a pretty big limitation on the amount of information animations can convey due to our funky acceleration perceptions.

cabryant wrote:

I have been contemplating animation (and non-interactive video, in general) a great deal lately, as part of a self-directed assessment of the effectiveness of various educational technologies within the context of the Learning, Design, and Technology program. In fact, just prior to this lecture on animation, I had formulated the following statement regarding video as a pedagogical tool:

“When well constructed, the simultaneous presentation of information across multiple modalities is emotionally engaging and evocative. However, such presentations run the risk of being confounding and distracting. Above all, the question of passivity is critical. The viewer has little choice but to process the linear story being presented, as opportunities to explore tangential questions or even interact with the artifact in a constructive manner are limited or nonexistent.”

It was gratifying to see how closely this assessment meshed with the conclusions of Tversky, Morrison, Heer, and Robertson regarding the animation of data visualization. More importantly, their tools of analysis, such as the principles of Congruence and Apprehension, provide a more formal system of assessing educational videos, which, without much effort, could be considered animated data visualizations!

Despite the apparent downside, the use of animations to effect emotional engagement cannot be discounted. This is underscored by Robertson's finding that users prefer animation to static multiples even when they impede performance and diminish accuracy of assessment! This is an important datum when considering how to engage potentially uninspired students in the process of learning. Furthermore, several of the dangers of animation (such as user-appropriate timing) may be mitigated by subjecting it to interactivity. That is, providing animation as a secondary interactive tool in a data visualization provides the advantages of both static and dynamic visualization while diminishing the inhibiting factors.

aallison wrote:

"animations must be slow and clear enough for observers to perceive movements, changes, and their timing, and to understand the changes in relations between the parts and the sequence of events. This means that animations should lean toward the schematic and away from the realistic"

I think this is a great point brought up in the Tversky paper. Animation and movement can be considered one of many sources of graphical noise added to your graphic. If you choose to use animation, you must reduce the overall amount of cognitive noise by reducing the complexity of your displayed object down to its essential elements . The end goal is to tell your viewer what to focus on solely through the structure of your graphic.

"Interactivity may be the key to overcoming the drawbacks of animation as well as enhancing its advantages."

I think this is extremely important. Most video tutorials, for example are worthless to me because they are too slow. I feel enslaved by their speed, and so seldom use them. But allowing the animation to be tailored to an individuals abilities or preferences, the animation can more effectively convey its information without distracting the viewer. I think SCPD recorded lectures took a good step in this direction by allowing the user to speed up/slow down the playback rate of the video. Very few people have time to sit through a lecture more than once, but the ability to jump to the important parts and speed through the boring ones allow me to consume the information much faster than before.

fxchen wrote:

I asked a question during lecture on strobe lights, Mike sent me some links I thought I would share: sensitivity - refresh rate http://en.wikipedia.org/wiki/Flicker_fusion_threshold

These two articles relate to how we perceive persistent light and effectively, how we perceive animation. The two articles were complimentary in presenting information on the psychophysics of light and human vision.

Anyways, my curiosity was piqued and I looked up some news I remember from a couple of years ago. Pokemon aired an episode where they set the flicker at the exact frequency to trigger seizures (http://www.cnn.com/WORLD/9712/17/japan.cartoon/). This danger matters to viz designers (obviously, using flashing lights to trigger seizures is bad!). more importantly, I think accessibility is sometimes forgotten in design; for example, I remember recently attending a talk where the visualization was way too small to read legibly.

nornaun wrote:

While reading Heer&Robertson, I was comparing animation of transition with beginner mode in software. For example, an expert user would prefer using keyboard shortcut rather than using the mouse to interact with GUI since the former is much faster. Likewise while transition helps viewers understand the relation between data in different presentation, users who are already used to it might find the animation slow annoying. If you know where everything is, there is no need for 3-second delay of showing the transition. Still, transition is useful for the users who just start exploring with data set. Considering that the data sets are more for new users, unless you are a researcher who look at the same kind of data everyday, it is more reasonable to make animation as a default and enable the option to disable animation.

zdevito wrote:

Tversky, Morrison, and Betrancourt discount many studies that show a positive result for animation because "in the animated graphics, more information was presented or the same information was presented differently, better, in the animated than in the static graphics." Here the authors are trying to study the benefits of animation alone rather than confound it with extra information. Consider the study of the study by Thompson and Riding on the effectiveness of animated diagrams of the Pythagorean theorem. The authors note that Thompson and Riding exclude fine-grained transformations in the static diagrams, confounding the results. However, what if they included a static visualization that showed _every_ frame of the animation as a (very long) static visualization. It is pretty clear that this visualization would have so many frames that it would be less effective than the tuned static visualization or the animated version of the same frames. This visualization contains the same amount of information as the animated one but it probably would be non-optimal. So equivalent information is not really a fair comparison as perfectly equivalent information is non-optimal.

tessaro wrote:

I found the research on the efficacy of animation in visualizations fascinating. I also found the methodological challenges of conducting such research mentioned in previous posts of interest. How does one make best practice generalizations given the complexity and, in some cases, blurring of the distinction between static but interactive and passive yet animated? It can seem that in attempting to isolate a performance hit or boost given a singular approach, judgements about the more common condition of combinations of interactivity and animation remain unclear. For a graphic that includes both interactivity and animation in parallel, what elements constitutes their respective instances when they can coexist in some cases so inextricably? Is speed a good proxy for comprehension? Is preference for motion an indicator of something useful or anecdotal albeit worth noting. I know that when amazon.com makes changes to its website, it tracks the performance by running different layouts in differing areas to compare performance. I wonder if it would be possible for data graphics, like the online diagrams by the New York Times, to produce slightly differing versions for a given story. They may be able to poll user preferences or see if people go all the way through an interactive form as opposed to watching an animation. The specificity of the diagram content and the wide distribution may be an interesting test bed for indicating user preferences or for just asking better questions.

The link below is to a recent NYTimes animation of a journalist escape from captivity in Afghanistan.

http://projects.nytimes.com/held-by-the-taliban/#part-5

A companion to this could be a generated in the form of an interactive timeline. This may be an example where such a comparisons of interactive versus passive animation (with narration) may be possible. How to enlist users to rate the graphic or track its performance would be a large part of the design challenge.

joeld wrote:

I'm not a big fan of animation unless the animation actually shows some process that evolves over time or models some aspect of the physical world - like rotating a 3-d object or changing the view in some other fairly natural way. Having nice transitions between data views is snappy and looks cool, but reminds me a lot of the page-transition animation that was popular when powerpoint first was introduced.

That said I was quite interested in visualizations such as cone trees when they were first introduced in the 90s. In the heyday of Silicon Graphics there were those that thought that 3D languages like VRML would eventually become popular ways to view information. The reality - at least so far - is that 2D static data is much easier to create and to consume. Recent developments in web technology may change this. For example, the canvas html element has been used to create some interesting animations.

http://dev.opera.com/articles/view/creating-pseudo-3d-games-with-html-5-can-1/

Animation actually becomes essential when we need the change of view to unpack the 3rd dimension of data.

rnarayan wrote:

@joeld comments on 3D - in the field of modelling, 3D becomes almost indispensible to be able to simulate complex phenomena. For instance, animating an iso-surface (as opposed to mere points) reduces the visual processing overhead required for sense-making. A while back I had taken this COMSOL Multiphysics modelling course which started out as a finite-element analysis tool for simulation of fluid dynamics, but has since been applied for many other domains. Plz see: http://www.comsol.com/showroom/animations/

@nornaun comments on (Heer, Robertson) transition animation - i am afraid i disagree here - firstly the transition animations are not just for recasting the same data with different plots - with substrate transformation, the gridlines move to show scale changes and with data schema transitions, it can morph and/or interpolate orthogonal changes. In these cases, it seems indispensible to have these transitions in place - without it would be like having no perspective guidelines when changing the viewpoint in a video game or Google earth. Also, @joeld - IMHO, it is not at all appropriate to compare these animated transitions with Powerpoint slide transitions, which has no role in enhancing apprehension, but only there for the visual pizzaz - the controlled study also confirms this with positive responses across subjects with little variability - the only complaint i can at all register is of course against Microsoft - since we dont see the product of this and other research in actual implementations of real products (unlike Apple)...

Tversky, Morrison, and Betrancourt + @zdevito comments - i agree that it is not only not optimal, but also posit that there is really no perfect information (or procedural) equivalence between a sequence of static small multiples and the same in animated form. What seems to be also discounted (in the paper) is the "connection of the dots" and animation by way of explicit sequencing of these underlying statics in a sense generates the extrapolation which in most non-trivial cases is anything but a linear mapping. In the case of COMSOL above, the path trace is by way of a solution to a set of partial differential equations.

wchoi25 wrote:

The concluding insight from Tversky et al's "Animation: can it facilitate" that the limitations of animation as a visualization medium could be overcome with interactivity was the most striking to me. When thinking about how animation compares against several well-chosen snapshots of the same data, there seems to be something about images that sticks to our memory better. If animations could be interacted with in a much more fluid manner than just basic stop/start functions, such as saving snapshots to separate panels for comparison, they would be able to minimize the sacrifices that come with them while at the same time allowing for a much more detailed representation of data.

I think in many cases it comes down to the ease of comparisons. Static images as in small multiples are easily compared, yet comparing animations that show the same data with certain tweaked parameters is a much more difficult task.

Leave a comment