How Network Structure Can Provide Advanced Warning About Epidemics
Hi Everyone!
In this post I’m going to write about a TED talk I recently watched that talks about how knowledge of network structure can be used to give advanced warning about epidemics. The talk was given by Nicholas Christakis and can be seen here.
In this talk Christakis talks about how people occupy different places in social networks. Some people are more central in their network while other people hover more towards the edges. This is important because your location in your network can affect how susceptible you are to contagion. If you occupy a central position in your network—say you’re the party planner—then you’ll be more likely to pick up some juicy gossip then if you’re located on the edge of your network and hardly ever leave your room. At the same time, however, if you occupy a central location you’ll also be more susceptible to catching contagious diseases, such as colds.
This is important because this information can help give us advanced warning about when a disease will strike a population. You see if we know that the people at the center of a network have picked up a disease then we can predict that this disease will spread to the peripheries of the network in a certain amount of time. In the TED talk Christakis found that people in the center of the network caught a virus on average 16 days before people on the edges. That’s over two weeks advanced warning!
Now this raises the question: how do you tell what location someone occupies in a network? One way is simply to interview everyone in a particular network and painstakingly record each one of their ties to other people in the network. Although this can work for small data sets it can be a huge burden when your network has too many people. Fortunately, Christakis provides an alternative method to find central people— pick a random sample of people then ask them to name a friend. The idea here is that your friends on average have more friends than you do. Don’t take this as an insult, it isn’t your fault, it’s just a property of networks!
One way to think about this is to image social networks as concentric circles. The most central rings indicate people with the most friends (think party host) and the periphery rings indicate people with the least friends (think recluse). Now imagine we pick a random person in this circle. The chances are that if we ask this person to name a friend he will be more likely to pick the party host than the recluse—in other words he’s more likely to name someone in a central ring as a friend then someone in a periphery ring.
If any of this seems interesting I’d recommend you check out Christakis’ book, Connected. This book goes into detail about how ideas spread through networks. It even shows how obesity can spread through networks. If your friend’s friend’s friend gains weight then are more likely to gain weight too. Crazy stuff!
very interesting!
[…] us to provide better warning on epidemics (one of the students wrote a nice blog on this video here). On this video, Prof. Christakis briefly mentions Google Flu Trends, a small Google project […]