Marketers aren’t the only people who see social networks as a trove of information. Before Facebook and Twitter, and the like, academic researchers studying human social behavior had to conduct and collate surveys laboriously, asking subjects about the number of their friendships, their closeness and strength and duration. Now much of that information can be teased out of the digital traces people leave online as they connect, share, and forward stupid joke e-mails.
Getting access to all the data from the companies can be tricky, of course, even in anonymized form. That’s part of the reason that Google (GOOG), Facebook (FB) and Yahoo (YHOO) have been able to attract top social scientists with the promise of working with their data troves—some of that work is for the benefit of science, some of it is just finding better ways to get people to click through ads.
But one doesn’t have to be a professional social scientist to be curious about what Facebook can tell us about social life—especially one’s own. One of the more fascinating figures in the world of data science is Stephen Wolfram, a British scientist, originally trained as a physicist, who among other things created the “answer engine” called Wolfram|Alpha that both Microsoft’s (MSFT) Bing search engine and Apple’s (AAPL) Siri rely on.
Over the past couple years, Wolfram has been occasionally posting on his blog about how he brings data science to bear on his own online habits. Last spring he wrote about what he had learned by tracking and plotting every e-mail he had ever sent, along with every meeting he had ever scheduled and even every keystroke he had ever made (7 percent of his keystrokes, he discovered to his surprise, were backspaces).
Wolfram recently announced the launch of a new program called Personal Analytics for Facebook. Already a million people have downloaded it and plugged in their own data—those who do so can opt to be a “Data Donor,” allowing Wolfram to include it in his research dataset.
Last week Wolfram wrote a blog post detailing some of what that donor data has yielded so far. He was able to chart, for example, that the number of Facebook friends a person has peaks in his late teenage years and then steadily declines. He also found that most of a person’s Facebook friends tend to be his own age until he hits his mid-50s, when they start to skew younger, with a particular concentration in the age cohort about 30 years younger. (Wolfram doesn’t suggest this, but one explanation for this phenomenon is that parents become Facebook friends with their kids’ friends).
One of Wolfram’s other findings from his first pass through the data is that the median number of Facebook friends in his donor data is 342, or 229 if he corrects for sampling errors. That brought to mind the work of Robin Dunbar, the evolutionary psychologist who has famously argued that the maximum number of meaningful social relationships a person can have (as he puts it, “the number of people you would not feel embarrassed about joining uninvited for a drink if you happened to bump into them in a bar”) is 150.
I forwarded the Wolfram post to Dunbar and asked for his thoughts. He had some reservations about the methodology but was particularly intrigued by Wolfram’s finding that social networks tend to fragment into distinct clusters—corresponding, Wolfram suggests, with such things as school, family, work, or one’s neighborhood—and that the most common number of clusters is three. “It has been known for a long time that triadic sets like this are particularly common in natural relationships/networks (the pattern known as triadic closure), but no one really knows why,” Dunbar wrote in an e-mail, “(there are some theories in social psychology, but I am not overwhelmed by them).”
When I plugged my own Facebook data into Wolfram|Alpha (it’s easy, just click here to get started), I was struck by a few things. For one thing, everyone I know really is married, as I am. I have very, very few single Facebook friends: 85.9 percent are married, and another 7.4 percent are in a relationship.
And using the tool that maps out my friend network visually, I discovered that while my friend graph has three centers of gravity, two of them are very close together and bleed into each other. There’s a tight cluster of my high school friends, the majority of whom I am not actually in touch with. Then there’s a broad scatter of dots where my college friends, work friends, and the friends of college friends and work friends all blur together. If you were trying to illustrate the incestuous, small-world nature of the social life of a Brooklyn-based, expensively educated magazine journalist, my graph would be a pretty good exhibit.