Collaborative Visualization

From CS294-10 Visualization Fa07

Jump to: navigation, search

Lecture on Oct 8, 2007

Slides

Contents

[edit] Readings

  • Voyagers and Voyeurs: Supporting Asynchronous Collaborative Information Visualization, Heer, Viegas & Wattenberg (html)
  • Design Considerations for Collaborative Visual Analytics, Heer & Agrawala (html)
  • Many Eyes: A Site for Visualization at Internet Scale, Viegas et al. (pdf)

Online Collaborative Data Sharing and Visualization Sites

[edit] Mark Howison - Oct 08, 2007 01:19:08 pm

I think it is exciting that visualization is going through a similar social awakening to the one that happened about 20 years ago in education research, when researchers started gravitating toward social theories of learning and away from purely cognitivist and information-processing perspectives.

I was also glad to see the methodologically-sound use of kappa analysis for inter-rater reliability in the sense.us user study. The only thing missing would have been a full coding scheme displayed as an appendix, but I can understand that space constraints wouldn't allow this.

[edit] Robin Held - Oct 11, 2007 05:27:52 pm

I found the reading entitled "Design Considerations for Collaborative Visual Analytics" to be very informative. As the title suggests, it provides a good overview of considerations that should be kept in mind when one designs a visualization system for use by multiple users. I imagine it has the potential to help guide the development of the field over the coming years. In particular, it exmplifies a way to systematically approach optimal designs, without having to rely on multiple, obscure sources in the literature. As a result, it hopefully makes life easier for newcomers to collaborative visualization, who want to design efficient new systems without having to read multiple volumes of work on visualization in general.

[edit] David Jacobs - Oct 12, 2007 01:44:36 pm

I know it's not entirely relevant to this lecture, but I found another one of those "Spot the Differences" games. This one has an interesting twist in that both scenes are presented side-by-side, and are actually a stereogram (cross your eyes to fuse the images). I find the task much easier than in previous examples, since there's a more natural "wrongness" about the conflicting areas. In any case, it's worth checking out just for the pretty vector art and 3D animation. 5-Differences Game

[edit] James Andrews - Oct 17, 2007 03:54:21 am

Although visualizations like the one Jeff showed, of the census data, can certainly be interesting on their own, I'd worry that digging through the raw data would sound too much like work to attract many users; many who would find it a compelling experience would not even try it out, because it 'sounds' fundamentally boring. I think the way to bring someone in to a data set is not to just give them the data set, but to start them out with additionally an article or presentation that uses the data set and gets them interested and thinking of their own questions. The article or presentation should ideally, in a clear fashion, integrate the interactive visualization system; visualizations shown for the purpose of the presentation should be easy to immediately interact with. It would be very cool to see a project that took something like Hans Rosling's "casts" and integrated the visualization in to that experience, so at any time the presentation could be paused and the visualization under discussion could be explored.

[edit] Amanda Alvarez - Oct 17, 2007 07:06:03 pm

I was glad that Jeff brought up cues, in this case the scented widgets. These give you a trail to how other people have already interacted with the visualization, and allows a kind of latent, anonymized collaboration. This got me thinking about the use of cues in situations that are the complete opposite of the ones Jeff talked about (asynchronous/remote), ie. a single user at a single place and time. The possibility of the user being able to manipulate the visual cues provided by the display itself is a cool idea (this is a very different use of the word 'cue'). This encompasses not just manipulation of visual variables, but also adjusting the information content of the display based on the eye's innate cue constraints (accommodation, focus, disparity, etc.). This could be especially interesting when tied to animated visualizations, given how people use (or don't use) motion cues. The basic idea is one of weighted widgets which would control display properties and allow the user to optimize the view and thus the process of sensemaking.

[edit] Ken-ichi - Oct 18, 2007 09:21:11 am

In response to James' comment, I'm not sure collaborative visualizations really need to be broadly appealing. I agree that "digging through data" probably seems too much like work for most people, and as a result I doubt sites like Many Eyes or Swivel will ever reach YouTube levels of popularity. However, I bet these kinds of systems would be great in small, focused groups, like conferences, or within research labs, where everyone really is interested in data and their implications. I'm currently wondering whether such a system might work for a project I work on, in which UC scientists are researching the effects of large scale forest management practices in the Sierra. Many public groups are very interested in the data that come out of this research, and though not all of them are as interested in the Web as we are, they might be drawn to these sorts of collaborative visualizations.

[edit] Mcd - Oct 21, 2007 10:34:26 pm

I think it's exciting that these collaborative tools are available for discussing visualizations. When I think of, for example, Google Trends, some of the most interesting bits are the annotations of major news events related to the terms visualized. Visualizations are very good at the 'what' of patterns in data, but rarely communicate the 'why.' Because the factors behind patterns and relationships in data are often external to the data themselves, discussion and annotation presents I think the best way to greatly improve visualizations' power. Good work, again, Jeff!

[edit] N8agrin - Oct 22, 2007 02:12:53 pm

Collaborative visualization tools seem to have incredible potential. It's amazing that we've had the notion of the visual interface for over 20 years and, yet only now have we come to understand how powerful it is to be able to share information with one another, provide the opportunity for others to comment or modify that information and provide some method of accountability for the annotations they make. The development and implementation of sense.us seems to capture a lot of the dynamics of collaborative data interpretation extremely well, not a surprise, considering the authors of the paper are well know in the visualization and CMC fields.

The thing that really struck me from the sense.us paper was the observation that each comment in the system had to be tied to the state of the visualization at the time the comment was made. Preserving this state is crucial to keep the comment in its original context, but also provides other collaborators a jumping off point from which they can see what others had successfully discovered in a visualization and what manipulations of the visualization they might try to further extract information from the underlying dataset.

The one aspect of sense.us that confused me slightly were the "doubly-linked conversations". Based on my reading, it seemed to me that the authors were suggesting the links going from the vis to the discussion and vice versa constituted a double-link. I think the utility of the feature is further enhanced by displaying the state of the visualization at the time the comment was left (as I had pointed out in my previous paragraph). I had a difficult time understanding, however if the double link implied a literal link inside the visualization and the conversation, or if it implied the link existed between the conversation and the visualization's prior state. I would assume the latter.

[edit] Hazel Onsrud - Nov 04, 2007 01:18:55 pm

Especially for the purpose of this class, I found the topic hub about many eyes itself pretty interesting. http://services.alphaworks.ibm.com/manyeyes/browse/visualizations?topichub=MhVnJsOtha6ryUNzlNwJ2- Take a look at the data sets they include. It would be interesting to map this temporally.



[add comment]
Personal tools