Tuesday, February 20, 2007

Social Networks fall apart

I posted a few days ago a model of a book-buying network over time. The researchers found that over time, a network which began as an undifferentiated and evenly connected network (every node attaching to other nodes with about an average number of links and without favoring any particular type of node) gradually changes. Certain links become favored and other links diminish in strength. Eventually where there was one network, three separate ones emerge.

This is an interesting phenomenon with implications for how a society governs itself. Where else does this segmentation occur? Why? And what social engines can affect it, making segmentation more pronounced or less pronounced? I thought I would take a look to see if I could find more information about it.

I found this paper which asks, "Why do populations often self-organize into antagonistic groups even in the absence of competition over scarce resources? " I can't promise that I understand the findings correctly (and I definitely don't understand them completely), but they do seem to find that especially in a large population, social segmentation is likely to occur.

Here is why: They assume, based on previous models, that actors change based on influences of their neighbors, either through attraction to similar actors or revulsion from dissimilar actors. In previous models in which actors tried to be more like their similar neighbors and tried to associate with similar actors, researchers had found a powerful force that led to total homogeneity of a network. But if actors instead emulate similar neighbors not out of homophilia but out of xenophobia, distinct communities are likely to emerge and can even become persistent indefinitely.

The authors of The Federalist Papers decided that a large republic would be more stable than a small one because the number of factions would be so large that none could overpower the entire state. These findings seem to support the Framers' conclusion from 200 years ago. In a large network such as the United States, individuals are driven, not into homogeneity but a number of disparate groups.

On the other hand John Dewey still pointed out the homogenizing effect of modern technology. He was speaking of the railroads, but the internet is but a railroad for ones and zeros. And homogenization still seems to happen. The article's findings don't say that a population splinters itself into as many camps as their are actors. But the model only accounts for two camps. If people are driven only away from those dissimilar and driven to become more like those they are driven too, will each actor eventually create its own unique camp?

Also, what role does linking play? I didn't understand, from the article, how one actor is linked to another. How does an actor acquire neighbors? What happens to a network when new links become possible? Is a network's size determined by its number of actors, or by the distance from one actor to another? I could imagine that the population trends toward heterogeneity when there can be greater distance between actors, but that when the distance shrinks (because new technologies, for example) the effect toward homogeneity becomes stronger. What role does distance between actors play in the balance between segmenting and unifying forces?

No comments: