Posted by Rob Knies

Generic image 

How can influential people be identified on social networks? Can Twitter data identify the political ideology of legislators? What, exactly, does it mean that something on the web has “gone viral”?

Mushrooming interest in social networking raises many intriguing questions, and in an age of big data, the explosion of content in those networks offers a new means of exploring the answers to those questions. Such opportunities provided the basis for CAOSS 2012, the Workshop on Computational and Online Social Science, held Oct. 12 in the Altschul Auditorium at New York’s Columbia University.

The event was organized by Sharad Goel and Jake Hofman of Microsoft Research New York City, researchers who are shaping a new field at the intersection of computer science and social science just as it begins to coalesce.

Their timing couldn’t have been better. About 250 students, postdoctoral researchers, faculty members, and industrial researchers traveled from around the United States to attend the event.

“One of our main objectives,” Goel explains, “was to showcase the breadth of approaches to computational and online social science, and I believe we were quite successful in that regard.

“The plenary speakers—which included computer scientists, an economist, and a sociologist—each offered novel perspectives on this emerging research area. Simply hearing their presentations back to back drove home the point that the field is inherently interdisciplinary.”

It’s useful to identify precisely what they mean by “computational and online social science.” The first part calls upon more than 20 years of Microsoft Research machine-learning expertise to transform huge amounts of social-media data into useful insights that can uncover patterns and relationships previously hidden. The “online” part addresses the practice of using the web as a platform for running experiments, enhancing both the scale and efficiency of traditional methods. Together, such opportunities enable researchers to explore new and longstanding questions in social science largely thought impossible just a decade ago.

Take the idea of defining what it means to “go viral.”

“It’s not clear what people have in mind,” Hofman says. “It’s used in so many different ways. Certainly, in the scientific community, it’s used in a more precise way, and it’s not clear if that more precise way actually maps onto what we observe.”

The fact that a concept such as “going viral” can be observed these days is incredible in itself. The terminology, of course, is designed to mirror the progress of an epidemic, but in this case, real-time data can be examined to provide observations and to draw conclusions.

“We can actually look at how things spread,” Hofman says, “and now that we can do that, what does that notion of ‘viral’ mean, given that we can look at the actual infection pattern?”

Those are the kinds of phenomena discussed during the workshop. The titles of the talks presented themselves lend an idea of the kind of freewheeling discussion they engendered:

  • Measuring and Propagating Influence in Networks, by Sinan Aral of New York University’s Stern School of Business.
  • Inferring Causality in Observational Data About Social Networks, by David Jensen of the University of Massachusetts Amherst.
  • How Users Evaluate Each Other in Social Media, by Jure Leskovec of Stanford University.
  • “Which Half Is Wasted?”: Controlled Experiments to Measure Online-Advertising Effectiveness, by David Reiley of Google.
  • The Virtual Lab, by Duncan Watts of Microsoft Research New York City.

The workshop also featured six graduate-student-led short talks that, in aggregate, proved a particular success.

“Another of our goals was to encourage students to participate in the workshop,” Goel says, “and to that end, we had some great short talks and posters by graduate students.
“The talks were scheduled to be exactly nine minutes, and after the allotted time was up, the audience was instructed to start clapping, which worked amazingly well.”

Chalk that up as yet another promising early step for a nascent research discipline.

“The field is in a funny place,” Hofman says, “because it’s not computational enough for the computer scientists and a little too computational for the traditional social scientists.

“But the field is beginning to define itself more sharply. Having the opportunity to be part of that shaping is why we were keen to do the workshop.”