Statistics, networks, and obesity: not what we thought

Sway of Social Networks on Behavior May Be Overestimated –

One of the most fascinating pieces of social science analysis that I’ve read in recent years is Nicholas Christakis and James Fowler’s book, Connected: The Surprising Power of our Social Networks. No, it is not a book primarily about facebook, but about personal ties and connections and how they can have profound influences, even through indirect ties of “three degrees of separation.” It’s also a rare example of a fun read from social scientists.

Among their most interesting and well cited findings is that you can “catch” obesity by being in a network of more obese people and vice versa. The argument is plausible as norms of body image and acceptability and information travel between friends and friends of friends. I have no expertise in social network analysis, and as the findings from the book had largely appeared previously in top peer-reviewed journals, I took these to be incredibly powerful results.

Well, the NY Times today brings to light a series of harsh reviews of the work, which eviscerate the authors’ claims. It’s pretty depressing. Not because I have any particular attraction to the obesity is contagious theory, but because it brings to light just how difficult it is to establish scientific claims — and not just in social science, but in the bio-medical sciences as well.

In a response to the critiques, Fowler and Christakis say that they are aware of the limitations of their methods and this is how science proceeds, but that until there’s a better model and evidence, they are sticking to their guns. I suppose that is indeed all we can do. But it still leaves me a bit unsettled when I ask myself which social scientific claims do we really believe with a very high degree of certainty… and the answer is, I can’t think of any.