The MIT Technology review has an amazing piece on Facebook’s Data Science Team: America’s top sociologists, big data software designers and analysts dealing with what is probably the largest database ever built on and by people. I really, really invite you to read the whole article. Otherwise, here’s a (too large) selection of extracts:
For one example of how Facebook can serve as a proxy for examining society at large, consider a recent study of the notion that any person on the globe is just six degrees of separation from any other. The best-known real-world study, in 1967, involved a few hundred people trying to send postcards to a particular Boston stockholder. Facebook’s version, conducted in collaboration with researchers from the University of Milan, involved the entire social network as of May 2011, which amounted to more than 10 percent of the world’s population. Analyzing the 69 billion friend connections among those 721 million people showed that the world is smaller than we thought: four intermediary friends are usually enough to introduce anyone to a random stranger. « When considering another person in the world, a friend of your friend knows a friend of their friend, on average, » the technical paper pithily concluded. That result may not extend to everyone on the planet, but there’s good reason to believe that it and other findings from the Data Science Team are true to life outside Facebook. Last year the Pew Research Center’s Internet & American Life Project found that 93 percent of Facebook friends had met in person. One of Marlow’s researchers has developed a way to calculate a country’s « gross national happiness » from its Facebook activity by logging the occurrence of words and phrases that signal positive or negative emotion. Gross national happiness fluctuates in a way that suggests the measure is accurate: it jumps during holidays and dips when popular public figures die. After a major earthquake in Chile in February 2010, the country’s score plummeted and took many months to return to normal. That event seemed to make the country as a whole more sympathetic when Japan suffered its own big earthquake and subsequent tsunami in March 2011; while Chile’s gross national happiness dipped, the figure didn’t waver in any other countries tracked (Japan wasn’t among them). Adam Kramer, who created the index, says he intended it to show that Facebook’s data could provide cheap and accurate ways to track social trends—methods that could be useful to economists and other researchers.
Other work published by the group has more obvious utility for Facebook’s basic strategy, which involves encouraging us to make the site central to our lives and then using what it learns to sell ads. An early study looked at what types of updates from friends encourage newcomers to the network to add their own contributions. Right before Valentine’s Day this year a blog post from the Data Science Team listed the songs most popular with people who had recently signaled on Facebook that they had entered or left a relationship. It was a hint of the type of correlation that could help Facebook make useful predictions about users’ behavior—knowledge that could help it make better guesses about which ads you might be more or less open to at any given time. Perhaps people who have just left a relationship might be interested in an album of ballads, or perhaps no company should associate its brand with the flood of emotion attending the death of a friend. The most valuable online ads today are those displayed alongside certain Web searches, because the searchers are expressing precisely what they want. This is one reason why Google’s revenue is 10 times Facebook’s. But Facebook might eventually be able to guess what people want or don’t want even before they realize it.
Recently the Data Science Team has begun to use its unique position to experiment with the way Facebook works, tweaking the site—the way scientists might prod an ant’s nest—to see how users react. Eytan Bakshy, who joined Facebook last year after collaborating with Marlow as a PhD student at the University of Michigan, wanted to learn whether our actions on Facebook are mainly influenced by those of our close friends, who are likely to have similar tastes. That would shed light on the theory that our Facebook friends create an « echo chamber » that amplifies news and opinions we have already heard about. So he messed with how Facebook operated for a quarter of a billion users. Over a seven-week period, the 76 million links that those users shared with each other were logged. Then, on 219 million randomly chosen occasions, Facebook prevented someone from seeing a link shared by a friend. Hiding links this way created a control group so that Bakshy could assess how often people end up promoting the same links because they have similar information sources and interests.
He found that our close friends strongly sway which information we share, but overall their impact is dwarfed by the collective influence of numerous more distant contacts—what sociologists call « weak ties. » It is our diverse collection of weak ties that most powerfully determines what information we’re exposed to.
Marlow says his team wants to divine the rules of online social life to understand what’s going on inside Facebook, not to develop ways to manipulate it. « Our goal is not to change the pattern of communication in society, » he says. « Our goal is to understand it so we can adapt our platform to give people the experience that they want. » But some of his team’s work and the attitudes of Facebook’s leaders show that the company is not above using its platform to tweak users’ behavior. Unlike academic social scientists, Facebook’s employees have a short path from an idea to an experiment on hundreds of millions of people.
In April, influenced in part by conversations over dinner with his med-student girlfriend (now his wife), Zuckerberg decided that he should use social influence within Facebook to increase organ donor registrations. Users were given an opportunity to click a box on their Timeline pages to signal that they were registered donors, which triggered a notification to their friends. The new feature started a cascade of social pressure, and organ donor enrollment increased by a factor of 23 across 44 states.
