Facebook Gets Results? 2010 Vote Experiment Worked, Scientists Say
If you logged onto your Facebook account on November 2, 2010, saw a cluster of familiar faces on your "wall" whose owners indicated that they voted, and it rousted you out of your chair for a trip to the polls, thank political scientist James Flower and six colleagues for the poke.
The nonpartisan go-vote messages, with links to sites showing local polling places and the appearance of voting "friends" on Facebook walls were part of an experiment involving 61 million "subjects" on Facebook. The goal of Dr. Fowler's team was to gauge the ability of on-line social networks to convey messages that affect real-world actions.
The answer: They certainly can, at least for voting. And the most effective means of steering the behavior of a large group appears to involve highlighting the actions of friends with whom an on-line network member is most likely to have a real-world, face-to-face relationship.
The results of the experiment are being published in Thursday's issue of the journal Nature.
"The network is the key," says Fowler, a researcher at the University of California at San Diego whose work focuses on social networks and their influence on fields ranging from public health and politics to personal tastes.
"If we want to make the world a better place on a massive scale, we should focus not just on changing a person's behavior, but also on utilizing the network to influence that person's friends," says Fowler, who led the study.
Nationally, the experiment's effect appears modest. The team, which includes Cameron Marlow, an in-house researcher at Facebook in Menlo Park, Calif., estimates that as a result of the get-out-and-vote message, some 340,000 people voted who otherwise were unlikely to vote in the 13 states that make voting records public. Those states account for 40 percent of the nation's registered voters.
The number of additional voters prodded into action might actually have been higher, Fowler adds. Roughly one-third of the voters participating in the 2010 mid-term elections had voted ahead of Election Day, either via absentee ballot or early-voting programs. But the team included these voters in its calculations, diluting the effect of the Election Day-only message in the team's calculations. Other potential voters – he estimates around 20 percent – probably saw the message too late in the day to reach polling place in time or saw it but simply couldn't get to the polls.
The party-neutral, go-vote cues appear to have appealed to self-identified liberals and conservatives in equal measure.
This is not the first instance in which Facebook has posted nonpartisan Vote messages on users' news feeds. But it is the first time the service's effort has been harnessed to see what sort of effect online networking services may have in turning those missives into action.
Of the 61,279,316 randomly-selected Facebook users included in the study, 98 percent received an information block on their news feed that had five features: a simple "Vote" logo, a link to a list of polling places, an "I voted" button, a counter tallying the total number of Facebook users who indicated they had voted, and a row of images of "close" friends who indicated that they had voted.
The remaining users were divided into two groups: one whose news feed displayed everything the large group saw – except the friends' photos; and a group that received no get-out-the-vote message at all.
Although tens of millions of users were participating in the experiment without their knowledge, the researchers say they went to great lengths to protect Facebook users' privacy– something on which Facebook insisted, Fowler said.
The team found that the news-feed message without the faces triggered millions of acts where people engaged in some form of online political communication or sought more information. But the people most likely to head for the polls were people who saw the faces of friends who had voted.
The team estimates that perhaps 60,000 voters were directly affected by the message, but "social contagion" via links with friends, and even friends of friends, led to another 280,000 voters casting ballots that day that otherwise wouldn't have voted.
The next step in testing the power of online social networks to trigger large-scale changes in social behavior is to see what types of messages are the most effective and what sorts of people see most influential, Fowler says.
After looking over the results and the approach the team took to test the notion that online social networks can affect off-line behavior, "I think they have shown that the message mattered," says Thomas Ferguson, a political scientist at the University of Massachusetts at Boston, who focuses some of his research on voter behavior and was not a member of the research team.
While the team foresees a beneficial use for such online nudges to encourage participation in what might generally be termed positive activities, Dr. Ferguson sees a potential downside for those who dislike unsolicited messages on their Facebook account.
"The world in which we used to get dozens of calls at night toward Election Dayfrom people calling us to vote" is long gone. "Spam has a whole new future," he says.