InfoVis impressions, part 3: "The Impact of Social Data Visualization"

This is part 3 of my impressions of InfoVis 2007. Click here for part 2.

This panel essentially continued the discussion from the “Infovis for the Masses Panel” the previous day, and, to me, was the best of the conference. The presenters were Brent Fitzgerald from Swivel, Ola Rosling (son of Hans Rosling) from Google/Gapminder, Fernanda Viegas from Many Eyes, and Martin Wattenberg (subbing for Warren Sack) talking about examples of social visualization on the web. The panel was chaired by Robert Kosara from the University of North Carolina at Charlotte.

Kosara kicked off the discussion by motivating the idea of focusing on the visualization needs of “everybody.” He pointed out several important functions for popular visualization, including using it to address data that is public but not readily available to most people (census data, federal spending data, the type of information visualized by They Rule, etc.). He also emphasized the value of visualizations as political statements and as storytelling (referencing Hans Rosling and Al Gore’s recent uses of infovis). Finally, and appropriately, he asked the audience whether we can use visualization to change the world, and whether anyone wants to use visualization for anything beyond presenting at InfoVis every year. Slightly rhetorical, but poignant!

Next, Fernanda Viegas spoke about the various different ways people are using visualizations on and around Many Eyes, including as political commentary, art (“lit-mash” with tag clouds, etc.), as focal points for conversation, and as educational tools. She pointed out that these are all “new” ways of using visualization that see it as a medium for communication rather than scientific tool. This, I think, is one of the most significant points you can make about what visualization “for the masses” is about.

Following Fernanda, Brent Fitzgerald talked about Swivel, contrasting its more private data model to Many Eyes’ open one. He described their goal as creating the “big database in the sky,” where anyone can find the information they’re looking for, emphasizing Swivel’s support for collaborative data editing and analysis. Swivel aims to produce a “data marketplace” based on a social network of contributors.

Ola Rosling presented next, starting with a characteristically “Rosling-esque” demo of child mortality data from across the world using the Trendalyzer/Google visualization tools. He stressed the importance of animation in information visualization (particularly with regard to time-based data), arguing that “static representations are lying” by only showing a single snapshot of a changing data set. He also suggested that visualization literacy should be motivated by democratizing access to data, as well as simply creating visualizations of compelling data. Finally, he ended with a fantastic analogy that compared the information visualization community to the photography community; he asked us to imagine a photography conference where the photographers were all horrified by how their field was being devalued by the ubiquity of photographs taken by average folks with their cellphone cameras. It’s not a perfect analogy, but certainly an apt one, and one that I think touched on a real “fear” within the community. As Stephen Few pointed out in his capstone the next day, there was a lot of laughter after Ola’s analogy, but it was nervous laughter.

Finally, Martin Wattenberg presented on the popular use of visualization on the web by introducing the concept of “vernacular visualization,” the use of “unsophisticated” visualizations by non-(visualization)-experts to “get things done.” I really liked this framework for understanding non-expert use, as it relates closely to Thomas McLaughlin’s concept of “vernacular theory” from cultural/media studies, and the idea that non-experts (or non-academics) engaged in the production and consumption of (in this case) information visualization make use of a rich set of “theories” whose basis is not necessarily in traditional ivory-tower academic theory (and which may in fact break from it), but which are nonetheless very effective and meaningful for “getting things done.” Martin argued that in order to conceptualize “infovis for the masses,” we need to understand this vernacular. As an example, he pointed to the impact of xkcd’s “Map of the Internet,” a “vernacular” (literally a cartoon) visualization, and its influence on more traditional IP-space visualization within the infovis community (hint: evidently it quickly became one of the most cited examples of IP-space visualization in academic infovis papers!). He closed by identifying three directives essential to promoting social data analysis: 1) Love your data! 2) Enable conversation, 3) Visualizations can be very simple or very complex, and emphasized that “Anything can be successful [in popular visualization], so let’s try a lot of things!”

The discussion that followed in the Q&A session was varied and at times tense. The initial questions again reflected a lot of concern about the misappropriation of data on sites like Many Eyes, as well as a concern that visualization could be used in “unexpected (read: “negative”) ways. This culminated in what I perceived as bizarrely derogatory question from Ben Schneiderman, who asked the panel how we can move beyond the “modest” successes of their work, and increase the credibility and reliability of their data by “removing the garbage from view.” Aside from ignoring what these sites are actually about, this, again, struck me as characteristic of the dominant conception of information visualization at the conference as a truth-telling tool. It reflects a lack of awareness of visualization as a medium, with all the subtleties such awareness entails. As Martin pointed out, people misrepresent information in any medium, and several members of the panel reiterated that part of understanding visualization as a medium is gain a literacy that allows you detect the “lies.”

There was an interesting question about how to motivate an analytic approach (in visualizations) to understanding for non-experts, and how you might use elements of “artistic” infovis to bring people in to analytic infovis. This is something I’m interested in myself. Fernanda Viegas responded by suggesting that non-experts are not necessarily trying to do “analysis,” but rather trying to tell stories, or trying to see themselves in the data, and this is a type of use that should be taken in to consideration (and supported). Robert Kosara suggested that designers should look at “playfulness” as a means to capture attention, and that giving users the “means of production” with regard to visualization will help engage them beyond what happens when they are merely presented with a finished product. Ola Rosling emphasized the need to keep things simple, as non-experts will quickly turn away from a tool that seems too complex.

In a related question, Zach Pousman asked how we might evaluate the effectiveness of social visualization tools, or even identify the measure of effectiveness. Martin favored the “ethnographic” approach, arguing that since visualization is a medium, we need to look at how it’s being used and written about, rather than relying on traditional HCI metrics. The line of questioning ended there, but I think this an area in need of much more exploration; it is fairly obvious that the “small-N” evaluations attached to most infovis research cannot accurately capture real-world usage patterns, particularly when talking about social or mass-audience systems. Coming from media studies, I always think of this as analogous to the situation the television industry is facing, where the networks are only just now realizing that Nielson ratings don’t account for the myriad of new ways (YouTube, etc.) that viewers encounter and interact with television content. The same could be said about visualization.

The rest of the questions continued to revolve around issues of data provenance, and whether or not it was “wise” to empower non-experts with the “means of production” of information visualization. On the whole, I couldn’t help but find the conversation discouraging. I felt that, at best, most of the questions ignored what was significant about the popularization of infovis, and at worst, their rhetoric was actually quite demeaning to the concept. Even the perspectives that favored empowering non-experts were based on an undercurrent of intellectual superiority – there were many repeated metaphors of non-experts being “illiterate” or having a “child-like” ignorance of how to do proper visual analysis. The idea that they could be using visualization in meaningful ways that did not necessarily adhere to the traditional “values” of infovis remained largely unaddressed.