So, a couple weeks back I did a panel for a seminar run by the Institute for Humane Studies, and ended up chatting afterward with a couple of the students and Dan Drezner. I mentioned in passing that I’d done a previous panel a year back on the application of social network theory—about which I’d forced myself to learn something for that purpose, having had my interest piqued by interviewing Duncan Watts—but never written up any of the stuff I came up with. I was afraid I’d forgotten all of what I came up with (though I probably still have the notes somewhere) but finally managed to dredge up the following, whcih might be interesting enough to be worth posting:
You can imagine a network, in the abstract, as consisting as a bunch of nodes and the connections between them. Now, what researchers have found is that human social networks (among others) display an interesting kind of small-world connectedness. That means that, for the most part, you can get from any one node to any other node via a relatively small number of connections, as in the game six-degrees of Kevin Bacon. Now, here’s why this is interesting: Imagine all your nodes (say people) as a huge circle, where each node is connected to the N adjacent nodes on either side. This is a highly clustered network, where you can get to nearby people in relatively few steps, but some nodes are going to be as far as (Nodes/2N) steps away. On the other hand, you can imagine a “randomly wired” network, where each node is hooked to N others at random. Here, too, you’ll get some very short chains between nodes, but unless N is huge, mostly there’s going to be a lot of steps between any given pair of nodes. The “small world” effect only shows up in the middle, with networks that exhibit a mix of clustering and “random” links. (The extreme example, of course, would just have every node connected to the same supernode, leaving everyone at most two steps away, but we’re assuming there are practical limits to the number of connections any one node can bear, even if some are far more connected than others.) And that looks a lot like our actual social networks (and, for that matter, blog networks). Most of your interactions are within a social circle, where a lot of the people you know also know each other—but everyone has “random” links outside, to people in other social circles. And, of course, we sometimes have distinct or only partially-overlapping social circles based on different affilitaions: Lots of blogrolls have a section for personal friends, a section for bloggers in the same city, another for political comrades, perhaps more for edubloggers or econbloggers or envirobloggers or feministbloggers.
OK, bracket that for a second. Imagine you’re trying to introduce an innovation—say the fax machine—into a network of people, and you can start by airlifting in a dozen or so. How do you target them? If you’ve got a highly clustered (part of a) network, and drop them all there, the innovation spreads rapidly to the rest of the adjoining nodes, because each person knows lots of other people with a fax machine, increasing the value of getting one. But it spreads slowly from that first cluster, since people outside it aren’t likely to know anyone they could use one of their own with. On the other hand, you could disperse them wildly across a random network. Well, the situation’s even worse there. What you want to do is disperse them in a way such that there’s an incubation “cluster,” but also fax-links to other clusters so you more quickly hit critical mass in far flung parts of the network.
Now, finally, I promised this would be relevant to blogging somehow. Imagine you want to spread a meme throughout the blogosphere. What’s the best insertion point? Well, from the above, one idea might be to try to take advantage of this structure of clusters overlapping among multiple dimensions. Say you’ve got a post idea that applies some interesting economic idea to education. You kickstart the meme in a cluster of education blogs, who (if you’re lucky) begin discussing the idea back and forth among themselves. Then, though, the key is that the econ (or politics, or philosophy, or whatever) dimension gets activated as econbloggers or polibloggers come in contact with one of the nodes in the edublog cluster, allowing it to make the rounds in each of those clusters. (Imagine the starting cluster as a circle of nodes with lots of lines connecting them and then a few lines on each node shooting off the circle to make contact with other, similar clusters.)
Obviously, I know little enough about network theory that this may be either too obvious to deserve mention or, on the other hand, horribly wrong in some way. But there you have it.