Lecture on Nov 4, 2009. (Slides)


  • Required

    • Voyagers and Voyeurs: Supporting Asynchronous Collaborative Information Visualization. Heer, Viégas, & Wattenberg. (html)

    • Designing for Social Data Analysis. Wattenberg & Kriss. (pdf)

  • Optional

    • Design Considerations for Collaborative Visual Analytics. Heer and Agrawala. (html)

    • Many Eyes: A Site for Visualization at Internet Scale. Viégas et al. (pdf)


nmarrocc wrote:

Sense.us was pretty cool. Collaborative use of graphs is pretty interesting. I also like the idea of collaborative visualisation building as inmanyeyes . Its interesting to think of graphics as a dialog. If you think about it the purpose of a graphic is to convey information. A static image can say several things but you cant have a dialog with it. Its like when Socrates was commenting on books and how terrible they were because they kill the dialog. Now with computers we can get that dialog back.

have a visualization on my Its also interesting to see all the different places dialoging takes place. In some instance users took their observations or new graphics back to their blogs so they could dialog with their established user base. Ultimately you can have no dialog if none is looking at your visualization. Here we have situations where the whole story could be spread over many different sites or services. For example Icouldfacebook page, but I could have pulled that from manyeyes , and users their could also have their observations on their own blog. This could be a problem if one were, for example, interested in seeing all theinteresting observations people have made about the Baby Names Voyger application. They would have to scour all kinds of different sites to see the whole picture.

vagrant wrote:

Beyond the demos that we have seen in lecture sessions, what struck me about the first reading was the concept of roles in social data. What I like about this concept is that data visualization can be more than just for a narrow target audience--e.g. researchers, economists, etc. That data can be an interactive experience has been a theme both subtle and overt in this course, but the way it is described here--beyond actions but as roles--expands upon the boundaries of consideration I weigh when observing visualizations, and will likely stick with me going forward.

Data--particularly statistics that describe past-to-current events--seem to lend themselves naturally to various forms of discussion, from interpretations of causality to predictions of future trends. Having emphasized these ideas in my past two projects, I wonder if there is a space for this sort of debate in the general sense. Metacritic comes to mind, but what might a visual Metacritic offer?

jqle09 wrote:

I wonder how the dialogue around a graphic differs from place to place. I would guess closed communities online offer much more compelling conversations than blogs and sites open to the web at large, but I think it would be interesting to see if and/or how within particular groups ideas might converges towards a certain consensus varying across the different groups. This might also be interesting data to visualize.

@vagrant - Maybe a visualization for metacritic would involve viewing the type of expectations people had for a movie (maybe through words in conversations about it online), aggregating that into a rating score for how they think the movie might do and then seeing how the actual rating fairs as the movie is released and more people review it. I think it might actually look a lot like what you did for your assignment 3.

I was really intrigued by the idea that experiencing socially interactive visualizations lends itself to two modes where we are switching between being a voyager and a voyeur. As I think someone noted during lecture, these exact two modes seem to apply to the way we interact online. I constantly switch between these two modes when browsing social sites like facebook, link aggregation sites (digg, reddit) and the like, by viewing the data that is presented on the sites and then being a voyeur by reading the data (comments, likes, 'is a fan of' posts) created by others. then finding new information to voyage. Though in this context the two modes seem more intertwined.

Being voyager and voyeur seems to hinge the ability to save the state of a visualization and I am trying to think if there is general a way to save a visualization's state if many parameters can give the same objects for visualization but at the same time single parameters can allow for different ways to view the visualization objects.

jieun5 wrote:

I recently learned about an awesome site visual complexity, a "unified resource space for anyone interested in the visualization of complex networks" with a goal "to leverage a critical understanding of different visualization methods, across a series of disciplines." It is the most comprehensive data visualization database page I've personally encountered. (There's another project based on this that allows for an easy tag-based navigation, reMap, which is also fun to explore.)

What's cool about this resource is that, among other things, it lists in the bottom "most visited projects", "most commented projects", and "popular searches." Thus, it promotes navigation through its content based on what other users have done. The site offers a blog-like feel to each examples, with consistent organizations showing the author, organization, year, URL, and description, plus comments field. Though most visualization examples currently lack comments, I see a good potential for this website given its clean interface and comprehensive collection of high-quality visualization examples over broad disciplines.

malee wrote:

To take a step back, a current trend (not specific to datavis, but for tech in general) is minimizing the effect that distance has on human-human interaction. Examples include collaborative tools (eg google docs), various forms of communication (eg videochat), and the notion of keeping in touch through online presences (eg facebook). Another trend is one of shorter attention spans. Blog posts are shorter than books, youtube videos are shorter than films, and tweets are just too short for their own good (: )).

These are not bad trends, but it would indeed be a shame if these trends totally replaced the 'old' way of doing things. Books still have their place, snail mail still has its uses, and distance will always matter (a la http://www.crew.umich.edu/publications/00-04.pdf).

It's interesting to examine this lecture's topic of social analysis within this context, and it'll be interesting to follow these trends in the future.

cabryant wrote:

Collaborative data visualizations provide opportunities, not only for the generation of new understandings and ideas, but social data aggregation as well. Currently, a great deal of social data is generated by surveys, which are limited by the imagination of the surveyor. Visualization sense-making allows users to define personally relevant information, and potentially serve as the basis for new data sets and visualizations.

One example from lecture is the depiction of book distributions across libraries and the personally encoded reading lists that resulted. A next step might be an aggregation of this data, followed by subsequent exploration (e.g. breaking down distributions by demographics crossed with library use, etc.) In essence, this process would allow "survey domains" to develop naturally from social activity.

Similarly, behaviors related to social data analysis roles (e.g. achievers, socializers, explorers, and killers) could be quantified with respect to an application like NameVoyager. Information concerning quantities, demographics, and interactions between these groups would likely yield insights into (online) social interactions.

bowenli wrote:

I think one of the interesting facets of collaboration and remote interaction is how much of it is based off of recreating reality. Collaborative visualizations is one area that really enhances the experience beyond what you can do in reality.

On a related note, in determining telepresence, it may be useful to look at Steuer's paper on defining "virtual reality" - in which he talks about two key dimension: interactivity and vividness. Both these affect the way we perceive how real something is to us.

vad wrote:

I wonder what kinds of cool collaborative visualization gadgets would be possible with Google Wave (if for some ungodly reason you do not know what Google wave is check it out NOW http://wave.google.com/). As far as my understanding of developing for wave goes, I think that it will be remarkably easy to develop visualizations with cutting edge interactivity for wave. Wave has the potential to make Protovis hugely popular as I think that it will be very easy to integrate them together.

nornaun wrote:

What I think would be nice to have for a collaborative visualization system is a feature that enables the user to link between related datasets. For example, when I make an annotation for a certain visualization about car usage, I can posted a link to oil prize visualization. The link will appear in both for others to comment or look at. I believe that such feature will lift integration between datasets and utilization of information to another level. This will certainly requires a basic central protocol for datasets to be able to link to each other. Still, it can be achieved with the online database like Many Eyes which has variety of datasets and search function already implemented.

fxchen wrote:

vad, I was actually thinking the same thing. What is the best way to integrate collaborative tech like Wave (I think it supports integrated spreadsheets) with visualization.

I remember seeing SenseUs from before but I was blown away by the search capability. The bookmarking that allowed for search presents a pretty power way to explore data.

zdevito wrote:

One thing that was brought up in class was the concern that providing cues as to the popularity of an item on sense.us was that it would homogenize the interaction with data and prevent active exploration. Another concern is that making data visualization a social interaction can introduce bias into the data that was not intended. While the author of a visualization may take care to make a particular view of the data as unbiased as possible commenters have less incentive to interpret data accurately. A few carefully placed biased explanations can skew the impression of a dataset. While this may not be a large concern for fairly uncontroversial census numbers in topics like healthcare or politics people with an agenda may seek to mislead using the collaborative tools.

rnarayan wrote:

sense.us and the idea of collaborative visualization is really a pioneering effort in visual sensemaking. Few thoughts that come to mind are:

a. i agree with @zdevito about introduced bias - in general filtering noise from signal is a difficult task without some form of NLP and semantic understanding of comments, etc.

b. one way to further develop this project could be to cull the heuristics gathered from the community to build a expert system of sorts - this can be useful tool to confirm or disprove various hypothesis by repeated runs on new data. It can be an inductive approach or a Bayesian rule-based approach.

wchoi25 wrote:

One of the most fascinating points made in lecture for me was the harry potter book visualization where users interacted with the visualization to create their own "fingerprint" of the collection of books they have read. It is a really interesting problem of how best visualizations can support simple annotations as these such that the emerging pattern across tens or hundreds of social users is evident.

Also worth thinking about is how sense.us and these viz sites like many eyes differ from wikipedia. In wikipedia, there's an emphasis on editing one reference document collaboratively, refining over it to produce something with higher quality. In sense.us, there doesn't seem to be a "master" graphic for each query. Instead, there are all kinds of views generated by different users and while they are easily reachable and connected to each other, the navigation through them becomes a much more formidable task than say, doing a wikipedia search on a topic and finding the page for it. I wonder 1) if it's valuable to be able to create these master visualizations and 2) how the tools should support these activities.

Leave a comment