Numbers Game: IBM’s “Many Eyes” Portal Turns Data Visualization into Community Art

11/5/07Follow @wroush

With the wrong visualization tools, data can be deathly boring—just think of all the dry, meaningless PowerPoint presentations and Excel spreadsheets you’ve endured in darkened lecture halls and conference rooms. But with the right tools and context, data can come alive, as Yale information designer Edward Tufte has famously argued and you’ll understand yourself if you’ve seen the inspiring 2006 and 2007 TED videos of Swedish researcher Hans Rosling using his Trendalyzer software to illustrate global demographic trends.

The folks at IBM’s Lotus division also seem to understand the power of good data graphics, and last week I had the opportunity to walk across Rogers Street from the Xconomy world headquarters to the division’s Cambridge lab to talk with the brains behind Many Eyes, the company’s grand experiment in collaborative data visualization.

At the Many Eyes portal, anyone can register for a free account, upload a data set, and select one of about 16 ways to display it, from a traditional bar chart or fever chart to a sophisticated scatter plot, tag cloud, or treemap. For example, here’s a tag cloud I created in about 15 minutes, by pasting the text from last week’s Xconomy blog posts (excluding news briefs) into the Many Eyes upload page.

Xconomy Word Frequency Tag CloudClick on this graphic to go to a live version of the tag cloud, where you can mouse over individual words to see how many times they turned up in our stories. Be sure to try visualizing using both the one-word and two-word modes by clicking the radio buttons. In one-word mode, you’ll see that the most common word in Xconomy stories last week—and probably every week—was “company,” followed by “Taylor” (Eons CEO Bill Taylor was the subject of a long profile by Bob on October 30) and “EMC” (which has captured a lot of mindshare recently due to a string of acquisitions and the mind-boggling rise of share prices in its subsidiary VMware). The most common two-word phrases were “Virtual Iron” (a VMware competitor), “gene therapy,” and “operating systems.”

None of that is particularly earthshaking, of course. But it’s fun. It’s the kind of information that would have been hard to obtain before Many Eyes came along without resorting to specialized software. And most importantly, it’s shared. If you clicked on the tag cloud graphic above, it took you to a public Many Eyes page, where anybody can view it and opine upon it.

“The traditional view of data visualization is that it’s solitary, like looking through a microscope,” says Martin Wattenberg, leader of the Visual Communications Lab at IBM Lotus’s Collaborative User Experience (CUE) Research Group. “But really, it’s the stories people tell around visualizations that make them interesting. We wondered if that is something you can design around. And we decided the best way to test that was to build a site and deliberately design it around collaborative visualization.”

Wattenberg’s group launched Many Eyes in January as part of IBM’s Alphaworks, a Web environment where early adopters can test specific IBM software innovations before they get incorporated into products (and 40 percent of them do). Since then, users have uploaded more than 8,700 data sets and saved more than 6,000 visualizations based on them.

Often, the conversations about the visualizations are as interesting as the visualizations themselves. That’s partly because of the community customization features built into the Many Eyes interface. Even if someone else uploaded a data set, for example, a visitor can select a customized view of that data, then save and share a snapshot of that view for discussion. For one dataset, giving a breakdown of federal spending from 1962 to 2004, a visitor isolated a view showing a huge spike in spending on “deposit insurance” from 1987 to 1992 and asked what had caused it. Within days, other visitors had provided an answer: “Appears to be the savings and loan bailout cost,” one wrote. Wattenberg calls this exchange an example of “social data analysis in motion.”

But as any veteran Washington watcher knows, quantitative data can be fuel for political as well as social discussion, and Many Eyes has seen its share of polemics. The Sunlight Foundation, a nonprofit organization in the nation’s capital devoted to using technology to ensure greater transparency in government, has made extensive use of visualizations from Many Eyes to illustrate the epidemic of congressional earmarks. For fun, in September Wattenberg posted Alberto Gonzales’s testimony to Congress regarding the firings of U.S. attorneys. Users immediately used the data to create word trees, which showed that the word “don’t” in Gonzales’s testimony was most frequently followed by the word “recall,” giving a rather blunt illustration of the former attorney general’s self-serving memory lapses. But within 90 minutes, Wattenberg says, someone else (presumably a Republican) published an equally damning word tree of Bill Clinton’s testimony in the Monica Lewinsky scandal.

Of course, new tools are always vulnerable to abuse. Many of the thousands of visualizations published by Many Eyes users are mystifying, nonsensical, or just plain painful. “We definitely see people make visualizations that just aren’t the right type for their data,” says Wattenberg. “But I make an analogy to the early days of the Macintosh and desktop publishing, when people would use twelve different fonts in the same document, just because they could. Nowadays most people are very typographically literate. I think we’re likely to follow the same course with visualization tools.”

I asked Wattenberg what unexpected findings were emerging from the Many Eyes experiment so far. Most important, he said, was the way Many Eyes visualizations have become part of a larger conversation going on across the Web. “Most of the conversation is happening on blogs around the site, rather than on the Many Eyes site itself,” he says. That’s still important data for IBM, Wattenberg says, because the company is interested in helping people use data in the ways that feel most natural to them. “These are all cases of people talking about numbers—which people do in business all the time.”

“We had expected to create a conversation” within the Many Eyes site, Wattenberg says. “But at this point we feel like it’s enough to be a component of the larger Web community.”

Wade Roush is a contributing editor at Xconomy. Follow @wroush

By posting a comment, you agree to our terms and conditions.