| Network theory |
[Nov. 18th, 2004|10:57 pm] |
There was an interesting article in the Communications of the ACM this month (I can't link directly because it's a member site), about network theory. It sounds pretty dry, but it's really fascinating stuff, not because of the algebra, but of because the possibilities it opens up. Network theory (here is a Wikipedia link for more info) is a branch of mathematics focusing on the correlation between network nodes. You can think about it in terms of computer networks, sure, but you can also think about it in terms of social networks. The article's basic gist was this:
Mathematician Paul Erdos claimed that when you tried to work out how many connections a node in a random network would have, there was a bell curve, centrered around a mean average number of links per node. If the average number of links was more than one, the whole network is connected. If it's lower than one, then the network is likely to be a largely disconnected set of islands. So, basically, the probability of a node being connected in a network is based on a mean average of links per node, and it's the same across all nodes. You're unlikely to see a node with lots and lots of links.
But, what people have found recently is that the probability in a random network isn't distributed that evenly. You *do* often find nodes with lots of links in a random network. This imbalance is interesting from a resilience point of view, says the article, because if a node goes down you either don't get much of an effect at all, or you get a catastrophic effect that bollocks up large parts of the network.
Well, all that's interesting enough - I've just written an article on self-healing networks and it would have been nice to have put that in, but among the discussion of real world stuff like Cisco routers and self-configuring switches there just wasn't room. And frankly applying this idea to the resilience of computer networks doesn't interest me anyway.
What does interest me is that firstly, how this applies to social networks, and secondly, how this applies to information. If key nodes in a random network are heavily connected, then they are what Malcolm Gladwell calls mayvens. Socially well-connected people whose opinion lots of people trust. People like Robert Scoble are mavens, (albeit it in a particular sphere), because he reads lots of people's stuff, and lots of people read his stuff. Viral marketing companies have been on to this for ages. They find 13 year old girls who are popular among their peers and give them free Britney Spears CDs and suchlike, so that they tell their friends, because the teenager's testimonial has more impact in that social group than a full page ad in Teen magazine does. It's sort of like the keyhole surgey approach to marketing, rather than the saw-off-the-leg-with-an-axe approach. Anyway, that's all accepted wisdom, and really, who cares? Marketing shmarketing. Blah.
But what I want to know is how are people taking these ideas and applying them to social activist networks? For example, political groups. Environmental campaigns. Civil rights groups. Marketing companies are well funded and can do this sort of thing relativley well, taking time and money to pinpoint the well-connected nodes in social networks to sell their wares. Can it be used to promote causes, and foster activities? Does the next Moveon.org - whatever it will be - know about this? One of the problems with real grassroots socially progressive groups is that they've always been so fragmented. Communication is often terrible. It's why so much potentialy good stuff never happens.
So, there's the application of network/graph theory to the social sphere. But then what about the application of this to information management? I've been thinking for a long time about how to use map metaphors to illulstrate the importance of relationships between items of information. Not just mind maps - I thnk they have lots of problems, such as an over-reliance on hierarchy and a lack of depth to their depiction of relationships between information items. But how about using that network theory and the idea of disproportionally connected nodes to make automatic decisions about what pieces of information in your visual map are more important? If you mixed with with a geographic metaphor to serve as a visual aid, I think this could be very powerful. There's a product idea in here, I think. I wish I had the time and the smarts to write it. |
|
|
![[info]](http://l-stat.livejournal.com/img/userinfo.gif){kind=small}
dannybradbury's journal