If you are an epidemiologist, that is, a scientist who studies diseases and their spread throughout a population, early detection of health trends is crucial to staying ahead of an outbreak and potentially saving many lives.
In a new TED talk (and video), Harvard Professor of Medicine Nicholas Christakis, reveals how data from social network sites can be used to predict a disease outbreak–before it grows into a major health crisis.
Professors Christakis and Fowler explore the nature of social Networks (SN) and reveal the rules that govern how SNs form. Knowing these rules allows us to predict, and possibly prevent, new disease epidemics.
Professor Christakis and his colleague James Fowler spent a lot of time mapping the intricate interconnections that comprise our various social networks. Their work shows that the dynamics of SN (such as which persons in a network constitute its major nodes or “hubs”, members that are the most connected to) can be used to predict the spread of a disease within an interconnected population.
On a lighter note, the same techniques can be used to predict how “good” an idea (or even a new product) is by tracking how it spreads through a network. Of course, whether and how an idea or virus spreads through a network depends on the nature of the relationships between members of the network; members can be friends, family, co-workers, colleagues, teammates, neighbors and/or or sexual partners. Not all connections in a network are equal.
To learn more about this network mapping, watch the video below of Nicholas Christakis speaking on how social networks predict epidemics (article continues after the video).
Of course, network mapping is not perfect. Not every idea that initially spreads through a social network becomes dominant. In reality, ideas suddenly die out, just like germs. Chaos and stochasticity (non-determinism) play a role even in more reliable models. Nodal analysis has its limits.
Years ago, Stuart Kaufman put forth his theory of “adjacent possibilities” which holds that an idea will catch on, or be readily adopted, if it is “sufficiently adjacent” (conceptually or technically closely related) to existing ideas that currently dominate a given market, or “field of possibilities”. There may be over-lap here with social network theory: what is “adjacent” to me, is another member of my social network–someone who is socially related to me–and thus potentially influenced by me (or I by she/he). How close and connected I am to others (and thus, how alike, or “adjacent” in our thinking) will in part determine the possibility of an idea catching on.
This work by Christakis and Fowler has uncovered a remarkably accurate means of forecasting all manner of social trends. In the case of a possible influenza outbreak, the information can be used to ramp up vaccine manufacturing and distribution in a targeted fashion to the maximum number of people, well ahead of the contagion wave. This will aid what is known as herd immunity which has been demonstrated to be the best way of preventing an epidemic in a given sub-population, and so preventing its spreading to a larger population (a pandemic).
This use of social network mapping by public health officials to predict epidemiological trends seems related to ecologists using the Google Page Rank algorithm to analyze real-world food webs (another form of network) — an unexpectedly useful discovery which I write about here.
Christakis has also analyzed the spread of other behaviors through social networks, such as smoking and obesity (videos of these talks are also available on the Youtube channel).
Based upon the comments on the TED website, however, it seems there’s one thing that was not predicted: the strong concern expressed by viewers of this video for data privacy (which is a somewhat mute concern if one is posting personal data on the Web), but also, how unknown numbers of others will use it to their advantage (usually meaning for profit), or, for government monitoring of its citizens.
Data mining and usage in the Internet age is a double-edged sword. But at least, in the case of public health, it’s use could very well save countless lives.