Archive for January, 2009

I’m always a little skeptical when lots of people become irrationally excited about something. From time to time, I get to cash in on this excitement. When that happens I also get to be on the receiving end of jealous rhetoric. On my first day of grad school, I went out to the local pub with some of my new classmates. On hearing that I would be working on social networks, I got a passel of questions asking why chosen subfield of Sociology was getting so much attention. I also got to hear why it wasn’t knew, but we all know that there’s nothing new under the sun anyways. Daniel Lemire has some interesting thoughts on this.

 

So what are social networks good for? Are they only good for throwing sheep? Are they just vast databases of dubious friend lists? My answers are: 1. I suspect a lot, 2. no, and 3. no. Before I start spewing rhetoric about online social network services, let me outline what I understand social networks to be, and what kind of tools they give us for thinking about the world. Coming from a social sciences background, social networks are simply collections of people and the relations between them. Now, relations can be anything. This means that for any group of people, we can induce any number of networks. For example, we could think about the acquaintanceship network (who knows whom), the communication network (who talks to whom), or any number of other networks.

 

What’s interesting about real social networks is what is often referred to as the small-world property (see the work by Watts and Strogatz). This doesn’t refer to any empirical measure. Instead, the small-world property refers to the fact that real world social networks seem to have higher clustering but lower path length when compared to random networks (where people friend each other uniformly at random). Perhaps even more interesting, it seems like people can actually navigate these networks without resorting to centralized directories. In other words, hey are “searchable.”

 

In my mind, online social network services will really come into their own when they provide mechanisms for people to take advantage of this property to solve individual problems. The challenge will be to find a way of incentivizing people to participate in search chains. For example, suppose two people separated by a chain of other people want to meet each other. How do we incentivize or compensate intermediaries to participate in such chains when their payoffs are minimal. Kleinberg and Raghavan have done some work on this in what they call query incentive networks, but I have yet to see any commercial applications.

 

Services such as StackOverflow (or Slashdot’s comment moderation system) use reputation systems to elicit query responses, but such services don’t take advantage of human ability to route queries. The reputation mechanisms serve more to regulate crowd behavior rather than as distributed “computing” networks. I think part of what could make social networks effective computing networks is the fact that individuals seem to be able to do this sort of social search.

Advertisements

Squirt!

Hello! Depending on who you talk to, I’m either a distracted sociologist or a lost computer scientist. Up until this past Spring, I was a Ph.D. candidate studying collective problem solving and social networks. My work till now has been mostly sociological in nature, but there has always been a computational component. I am now on leave, working on methods of audience segmentation for Internet ad targeting.

I am starting this blog in order to have a place to participate in online discussions concerning social networks (both real and virtual) and the various applications, algorithms, and conceptualizations surrounding them. It’s my way of keeping my hand in while I serve more commercial interests.




  • Recent Comments

    Daniel Tunkelang on Squirt!
    antisociology on Squirt!
    Daniel Tunkelang on Squirt!
  • Ten to read