Visualizing Group Intelligence

Creating a common mental model

Steven Chien interviews Martin Wattenberg via a series of emails from May 15 – May 20

MARTIN WATTENBERG runs IBM's Visual Communication Lab where he explores information visualizations that help people make sense of data. One such project is Many Eyes, where the goal is to “harness the collective intelligence of the net” for discovery, insight and analysis. Martin was recently named “one of the world’s 100 top young innovators” by Technology Review. He is also well known for artistic data visualization, in which information sources as varied as music, museum collections and Web searches are rendered visually.

Assignment Zero interviewed Martin about how visualization helps foster the exchange of ideas and insights, and about "sensemaking."

Q: What are some of the pros and cons of extracting the wisdom in crowds via visualization?

A: Visualization serves several purposes: defining common ground, attracting a crowd, making complex information accessible to a larger audience. These are the pros. There are definitely some negative aspects. People can fixate on superficial aspects of a chart or graph. And understanding and using complex visualizations sometimes requires a high degree of visual literacy, or at least a willingness to explore.

Q: How did you get involved with "collaborative sensemaking"?

A: I've been interested in data visualization and individual sensemaking since my days in financial journalism at SmartMoney.com. Part of my interest in the collaborative angle comes by osmosis. I work with world experts in collaborative computing and their perspectives are contagious! I also was inspired by a few particular examples: the intense blog discussions around the "NameVoyager" baby name visualization, and the storytelling behavior Fernanda Viegas witnessed around what were initially private email visualizations.

Q: Your Many Eyes project mentions that visualizations offer a social side, where "discussion and storytelling are just as important as data analysis." What is it about visualization that sparks discussion and collective insight? What impacts do you foresee from "democratizing" visualizations?

A: I don't really know. Those are exactly the questions we're trying to answer with Many Eyes. I can hazard a few guesses, though. On the serious side, visualizations can give a group of people a common mental model so they can have a productive discussion. Let's face it, it's boring and trite to complain about government spending in the abstract... but add a historical chart of the entire US budget and you get an interesting conversation. Visualizations are also just plain fun, and serve as an attraction and conversation starter.

Personally, I'd love to see a bigger emphasis on facts and evidence in public debates. Maybe Many Eyes can help that happen. I'm also curious to see if people use visualization in ways that I wouldn't expect, pointing the way to new research directions.

Q: What surprised you the most with your project?

A: I've been incredibly impressed by our users and their creativity. We've seen truly clever applications of standard visualization techniques, as well as some very cool artistic visualizations. For instance, one person used our tag cloud to create "litmashes" -- literary mashups -- that visualized the results of mixing various literary works.

Q: Do you really think there's wisdom in crowds? If so, what's the clearest example you know of?

A: There's always wisdom in crowds, just as there is always gold dissolved in seawater. The question is how to extract it! One of the clearest examples is Wikipedia, which illustrates three features of crowd intelligence. First, it can work: I find Wikipedia so useful that I probably turn to it at least once a day. Second, it works inconsistently: I still frequently find incomplete Wikipedia articles. Third--and this may be the most controversial point--crowds can be creative. If you look at the set of guidelines, processes and policies that Wikipedia has invented, it's an amazing structure that's not quite like anything else.

Q: You mentioned in a recent paper that conventional forms of organization seemed to have spontaneously emerged from Wikipedia. What do you make of the dynamic between this phenomenon and inconsistency?

A: That's a good point. Processes tend to evolve for cases where you need to reduce risk. A good example is the Featured Article on Wikipedia: there is really only space for one on the home page, and a huge number of people will see it. So a bad article would be pretty harmful.

Q: How would you address inconsistency as a feature of crowd intelligence? Do you try to combat it or do you accept it?

A: It took me a while to realize this, but inconsistency can be a feature, not a bug. If you really need absolute consistency in a system, you can get it, but usually at the price of a lot more effort and a lot less flexibility. To paraphrase Emerson, a foolish consistency is the hobgoblin of top-down systems.

Q: What was it that made you realize inconsistency can be a feature, and not a bug?

A: Consider the del.icio.us tagging model. Unlike librarians, del.icio.us users are completely inconsistent in how they categorize Web sites. But what this means is that you don't really have to learn a specific set of keywords--when I do a search I can look for the terms that I would use, and chances are a few other people have used them too.

Q: Taking a step back, what's really new about the whole concept of crowdsourcing? Where do you think it's going next?

A: I suppose crowdsourcing has been around as long as people have been telling the story of stone soup. What's new, in my view, is that the Web has made it orders of magnitude easier to gather a crowd. I do think there are some limitations and conflicts we'll see in the next decade. There's still only a finite amount of attention, so the number of crowdsourcing projects can't expand forever. And as people try to make money off of crowdsourcing, that may change the dynamics considerably.

Q: Is there really any money to be made with crowdsourcing? If so, why will some people work for free so that others can profit?

A: We'll find out about the money, but my guess is that it's there. As for why people would contribute, one answer is that people sometimes work for free just because the work is exciting and fun, even as others make money from their efforts. Linux is an obvious example. Another answer is that people may be paid for their work. One way this could happen is through a de facto micropayment scheme--Google AdSense shows anything is possible! Another possibility is that crowdsourcing will occur among paid workers in a large enterprise. A proof of concept is the "Intellipedia" project, where the Wikipedia model was applied to gathering knowledge in the closed environment of US intelligence agencies.

Q: What are your next steps for 'Many Eyes' and other visualization projects?

A: We have a bunch of plans, not all of which I'm ready to describe in detail. But I can tell you one high-level goal: we want someone to make an important discovery using Many Eyes. Our current research efforts are all aligned with this goal.

Q: How would you like to see information visualization evolve?

A: I'd like to see information visualization turn into a full-fledged medium of communication. That will require new tools to make it easy to create and remix visualizations, as well as universal access. I'd also like to see information visualization used for more types of data than pre-digested tables of numbers. We're starting to see more visualizations of "unstructured" data, but there's a long way to go.

Q: Finally, is there anything I've forgotten or missed? Any last thoughts you'd like to throw out?

A: If you want to see the future, look at what artists are doing today.

Click here for more information on Martin's research and artwork.