In John Guare’s Six Degrees of Separation, the character of Ouisa Kittredge finds the eponymous six degrees both a comfort and torture—the latter because “you have to find the right six people to make the connection.” At Social Networking Week, a conference hosted at U of T by Bell University Laboratories in conjunction with U of T’s Department of Sociology that ran from Oct. 30 to Nov. 2, researchers met to discuss the new vistas of angst—and insights—opened by the rise of the Internet and its information technologies. One of the greatest research tools is the ability to visually represent social networks. In large part, it’s the same technology at work on your Facebook page, though researchers are able to put it to some surprising uses.
One example is U of T mechanical engineering professor Mark Chignell, who, with a team of grad students, is visualizing social networks in the public health sector as a means to evaluate the processes currently being used. Social network diagrams are often called “egocentric” because they are organized around one individual. One project Chignell’s team is working on is a “patient-centric” model, mapping the process a patient moves through at Toronto General Hospital—a complex one encompassing discussions between doctors, nurses specialists, and outside caregivers who look after a patient once discharged.
For Chignell, visualizing these relations means pinpointing where errors happen. Chignell cited a recent study by the Institute for Medicine estimating that anywhere between 44,000 and 98,000 people die in U.S. hospitals each year due to medical error. Chignell’s study of TGH found several points where such errors could be introduced, from communication to diagnostics, treatment to preventative care.
“It’s not surprising that errors are made, because there are so many points of contact for an error-prone process to express itself,” Chignell said. “One of the problems is that doctors and nurses tend to speak different languages.” In some cases, Chignell suggested the solution could be as simple as offering better technology for conferencing.
Social network diagrams are nothing new. Sociograms, introduced into psychiatry in the 1930s by psychiatrist and professor Jacob L. Moreno, depicted interpersonal connections through lines and dots, largely the same way as the social network diagram widget available on Facebook today. But whereas on Facebook you can upload your friends list to create a diagram very easily, a sociogram in Moreno’s day took much longer, representing hours being interviewed by a therapist.
But even as the Internet has made the basic network diagram easier to produce, it has also created new communities to challenge researchers. Presenting at the conference was Caroline Haythornthwaite, professor of library and information science and the University of Illinois.
Together with PhD student Anatoliy Gruzd, Haythornthwaite uses Natural Language Processing techniques to mine for social relations amongst the bountiful data on online bulletin boards such as Usenet. It’s easy to map a chain of posts—not so to visualize how posters relate to one another.
“We don’t want to know just how many people answer the questions,” Haythornthwaite said. “We want to be able to see the network structures: who is doing the answering, when, under what circumstances, for what kinds of relations.”
The human brain easily recognizes that “Frederick” may become “Fred” over a chain of posts. But watching programming try to mimic the same process, it becomes clear just how complex that thought process actually is. How do you “teach” a program to recognize Fred responding to Dan’s question about something Mary said, but to not recognize “Shrek” as a community member?
How do you recognize names that aren’t capitalized? NLP is designed to read documents such as newspaper text—proofread sentences with proper grammar—not computer-mediated text with all its different spellings, its typos, its half-finished sentences. Having come up with the right iterations to solve these issues, Gruzd can save Haythornthwaite the hours she would have spent manually mapping out an online social network. Solving each of these issues is slow-going.
There’s no question though that visualizations of personal information hold a strong fascination. “People want to look at these pictures,” said Fernanda Viegas. She is part of IBM’s Visual Communications Lab, a small team that develops visualization tools for researchers.
The traditional node-link network diagram used by Facebook is intuitive, Viegas said, because we seem to naturally understand a node to be an identity, though the multi-variant pivot graphs developed by her lab are actually better for representing large-scale networks. That people like to look at the pictures is an important aspect for Viegas. “What are people doing with these visualizations? We tend to think that these are really cool for data analysis. They are. They are valid tools. But they only go so far,” she told the audience at the conference.
“One of the things that people hardly think about, in terms of visualization, is that they are really good conversation catalysts and that they actually function as social artifacts. People love to look at themselves. They love to look at their social network, and they love to tell stories about it.”
The lab’s latest project is a public website, many-eyes.com, which allows a user to upload data into one of various ingenious visualizations. All visualizations and data uploaded to the site are made public. The point is to foster discussion. The lab was initially surprised that members of the general public are uploading data—everything from “Co-occurances of Names in the New Testament” (guess who’s number one) to “John’s Freezer Contents (only the meat).”
Visual Complexity, a blog that showcases complex networks, has celebrated some Many-Eyes productions, as well as those of other sites, such as Vizster and Social Action. The same narcissistic fascination with ourselves, which drives sociology, could also explain the popularity of sociologists’ work.
