We have completed maintenance on DiscoverMagazine.com and action may be required on your account. Learn More

Emerging Technology

Pass your e-mail through some new software and the answer will become obvious

By Steven Johnson
Apr 1, 2003 6:00 AMJul 18, 2023 7:57 PM


Sign up for our email newsletter for the latest science news

In his classic novel Cat's Cradle, Kurt Vonnegut explains how the world is divided into two types of social organizations: the karass and the granfalloon. A karass is a spontaneously forming group, joined by unpredictable links, that actually gets stuff done— as Vonnegut describes it, "a team that do[es] God's Will without ever discovering what they are doing." A granfalloon, on the other hand, is a "false karass," a bureaucratic structure that looks like a team but is "meaningless in terms of the ways God gets things done."

Illustration by Leo Espinoza

No doubt you've experienced these two types of networks in your own life, many times over. The karass is that group of friends from college who have helped one another's careers in a hundred subtle ways over the years; the granfalloon is the marketing department at your firm, where everyone has a meticulously defined place on the org chart but nothing ever gets done. When you find yourself in a karass, it's an intuitive, unplanned experience. Getting into a granfalloon, on the other hand, usually involves showing two forms of ID.

For most of the past 50 years, computers have been on the side of the granfalloons, good at maintaining bureaucratic structures and blind to more nuanced social interactions. But a new kind of software called social-network mapping promises to change all that. Instead of polishing up the org chart, the new social maps are designed to locate karasses wherever they emerge. Mapping social networks turns out to be one of those computational problems— like factoring pi out to a hundred decimal points or rendering complex light patterns on a 3-D shape— that computers can do effortlessly if you give them the right data.

Until software designer Valdis Krebs came along, however, there wasn't an easy way to translate social interactions into a machine-readable language— short of following people around, anthropologist-style, noting whom they called or whom they chatted with at the watercooler, and then typing it all into a PC. "In the late '80s," Krebs says, "I was taking two graduate classes at UCLA— a class in organization design and a class in artificial intelligence. I was real busy at my day job, and I had a lot going on in my personal life, and I started thinking, 'Boy, it would be great if I could figure out a way to do one project to hand in for both classes.' " It seemed like an unlikely combination, until a friend showed Krebs an article about an early rendition of social-network-mapping software. "I looked at the article and had that 'aha!' moment: 'Here's the project for both my classes.' "

Krebs has spent most of the last 15 years honing his mapping software, which he called InFlow. He quit his day job in 1995, after IBM agreed to license the technology, and now he makes social maps full-time. Krebs is half sociologist and half digital cartographer: Many of his organizational maps are based on surveys taken of employees answering questions about whom they collaborate with, what their work patterns are. That data is then fed into InFlow, which paints striking visual portraits of social structures in organizations. They look almost like images from a chemistry textbook— dozens of molecules strung together in an intricate shape, each one representing an employee. The links between each person are a way of visualizing the flow of information through a company. "The maps show how ideas happen, how decision making happens, who the real experts are that everybody goes to," Krebs says. They show the karass buried inside the granfalloon.

Of course, modern corporations no longer need surveys to make sense of their employees' social interactions. With the rise of e-mail, chat rooms, bulletin boards, and Web personals— the watering holes of the digital realm— our social interactions now leave behind an increasingly long trail of data. And that makes them easy to map.

"If we're going to spend more of our social life online," Judith Donath says, sitting in her office at the MIT Media Lab, "how can we improve what that experience feels like? How can you convey the sense of being in a crowd or the movements of a crowd?" Stylish, and aided by a subdued, affable vocal style, Donath runs the Media Lab's Sociable Media Group, exploring what we can do with all the digital data we're implicitly collecting about ourselves.

"You have this enormous archive of your social interactions, but you need tools for visualizing that history, for feeling like you're actually inhabiting it," Donath says. Turning her sleek, black flat-panel display toward me, she loads up Social Network Fragments, created by Danah Boyd, a grad student, and Jeff Potter, a programmer. The program is visually stunning, if somewhat overwhelming: a floating mass of colored proper names projected over a black background and clustered into five or six loosely defined groups. It looks more like a work of information sculpture than a supplement to e-mail software.

The program was featured as a work of art in a gallery show in New York City in the summer of 2002. But the data it represents are culled from mundane sources: the addresses of e-mail messages sent or received. By looking at the names of people whom you send messages to or receive them from, and who gets cc'd or bcc'd on those messages, the software builds a portrait of your social networks. If you often send messages to your entire family, the software will draw links between the names of all the people you've included in those messages; if you cc a few colleagues on a message to an important client, it will connect those names as well.

Assuming you have a significant amount of e-mail traffic, the software will create a remarkably sophisticated assessment of your various social groups, showing you not only their relative size but also the interactions between different groups. If your college buddies have grown close to members of your family, you'll see those two groups overlap on the screen, like two crowds huddled next to each other.

If these visualizations are interesting for individuals, they're even more interesting for large organizations, where social networks can play a key role in the success or failure of the operation without any individual really knowing where all the networks are. Every large organization has its granfalloons and its karasses. You have your executive vice president for sales, and the 10 deputies who report to her— that's a granfalloon. The karass is the group of 10 people from 10 different divisions who come together to make sure the new product ships on time. Granfalloons are what you see in the annual report and the business plan; the karass is what actually happens on the ground, when things are going well. It's that implicit social structure that both Donath and Krebs are after, in their different ways.

Intelligence analysts once assumed that terrorists organize in isolated cells. But social-network maps revealed that the 9/11 hijackers' cells morphed into a hub-and-spoke pattern with an obvious leader: Mohammed Atta. The active structure resembled that of an IBM project team.

Social mapping is not just for corporate sociologists. Krebs has used his software to analyze the social networks visible in book-buying patterns on Amazon.com, by tracking the "people who bought this book bought these other books" feature. The software starts with one book and follows the links out to five books connected by an Amazon customer's purchasing habits; then the software moves on to 25 books connected to the five. (If he's attempting a particularly broad study, he'll do another sweep.) Then the InFlow software creates a map showing clusters of books that are often purchased together— and by association, clusters of book buyers with shared interests. These are implicit social networks, not explicit ones; you don't necessarily know the people in your cluster, but you have a lot in common nonetheless.

Not surprisingly, social-network software is ripe for political analysis. "A few weeks ago," Krebs says, "I got into a discussion online about the state of the country politically, and some people were arguing that the country was really divided, that we were back to where we were after the 2000 election. One side can't stand the other side. And I started thinking, I wonder if you could see evidence for this in the book-reading networks." Krebs used InFlow to analyze the network of book purchases surrounding two best-selling titles, one from the left (Michael Moore's Stupid White Men) and one from the right (Ann Coulter's Slander).

"What I got were two cliques that were about as distinct as they could be. I kept looking for paths that crossed between them. Every time I tried to follow one of these paths, I'd go out three or four steps, and then boom, I'm right back in the clique." Most strikingly, the two networks intersected only on a single title: Bernard Lewis's What Went Wrong. Otherwise, the two groups were engrossed in entirely different reading lists, with no common ground.

Those two cliques make it clear that tools designed to detect social networks are just as good at detecting antisocial behavior as well— for sniffing out the karasses that don't ever speak to each other or those linked by one solitary thread. For corporate managers and sociologists alike, this may prove to be the software's most useful function. It shows us the gaps in the network, the borders that no one dares cross.

Learn more about InFlow and the work of Valdis Krebs: www.orgnet.com.

Read about Danah Boyd and Jeff Potter's Social Network Fragments project at smg.media.mit.edu/projects/SocialNetworkFragments.

1 free article left
Want More? Get unlimited access for as low as $1.99/month

Already a subscriber?

Register or Log In

1 free articleSubscribe
Discover Magazine Logo
Want more?

Keep reading for as low as $1.99!


Already a subscriber?

Register or Log In

More From Discover
Recommendations From Our Store
Shop Now
Stay Curious
Our List

Sign up for our weekly science updates.

To The Magazine

Save up to 40% off the cover price when you subscribe to Discover magazine.

Copyright © 2024 Kalmbach Media Co.