Facebook. Nearly a billion of us use it, but does it affect how we make important decisions? James Fowler at the University of California San Diego has been looking at the social networking site Facebook and how it can influence voting behaviour...
Chris - You famously showed that if people have overweight friends on Facebook then they're more likely to put on weight themselves, and now you're turning to politics.
James - Yes we are and this week, we reported the results of an experiment that we conducted on election day in the United States in 2010, where we randomly assigned the ‘Get Out the Vote’ messages to 61 million people on Facebook. There were 2 different kinds of messages. One with faces of friends and one without, and there was also a control group that received no messages. And the results of the experiment showed that the message directly mobilised 60,000 people who saw the message, but it also mobilised an additional 280,000 people who were friends of the people who saw the message. And we were able to do this because in the United States voting records are public and we were able to match about 6 million of the Facebook users to publically-available records to see whether or not they actually got to the polls.
Chris - So you could check up on them as well. So just to recap, you had 3 groups of people. A control group who you just looked at and just monitored their behaviour, you’ve got a group who get a little button which says, “go and vote” and another group that say, “go and vote and by the way, here are some friends of yours who voted.”
James - That’s correct and the interesting thing about the 2 different kinds of messages we showed people, there was one that had pictures of their friends who had clicked on the button during the day and another one didn’t have these same pictures. What we found was that, the one that had the pictures, actually was the one that did all the work. The one without the pictures, the rate of voting for the people who got that message was exactly the same as for the people who saw no message whatsoever. And so, these pictures of your friends are what seemed to be doing most of the heavy lifting in terms of the total effect of the experiment on real world behaviour.
Chris - But do you know James, whether it is just seeing pictures of your friends that makes you attend to that bit of the screen more and for longer so you'll notice the, “go and vote now” message more or is it that the friends have voted and this makes you think, “Wow! I’d better keep up with what my friends are doing”?
James - That’s a good question and we would like to figure this out I think in a future experiment. Right now, we just know that the message works, but it’s possible that it’s just something that draws your eyes to the message because, you know, just like when you see a list of random words, if your name happens to be in that list of random words, you're likely to be able to pick it out very quickly. Seeing faces of friends, that might cause you to attend the message, but one of the things that we were able to show was that, not only was it the case that people who saw faces of their friends voted more, but the friends of the people who received the message voted more. And so, the consequence, there was something that happened on the way from you getting the message to your friend voting that probably had something more to do than just attending to the message, had to do with the process of social influence.
Chris - So, what are the implications then? You put this message up and people see that their friends have reacted to it and you end up with a bigger result through the social networking impact then the primary advertising in the first place. Facebook must be delighted.
James - That’s right. Yeah, it was really interesting. To make it just perfectly clear, for every one person whose behaviour was changed directly, there were 4 friends whose behaviour was changed. And the interesting thing is I think that campaigns, they’ve gotten very smart in the last few years in countries like the United States and in Europe, and they do experiments. They do this message testing on a large scale, but I think that they don’t necessarily utilise all these new social network information that we have and what this really shows is that if you were just looking at the people who receive the message, you'd be missing the whole story. The network is really key here.
Chris - Is the fact that voting is something that people should universally do also important because if you had conversely put up an advert saying, “Mrs. Whoever bought a Ferrari” now that’s only going to be relevant to a subset of people and a subset of friends, and therefore, it’s unlikely to have the same impact as if it said, “go and vote” because that’s something that doesn’t attract a physical cost and also, is something everyone should do?
James - I think that’s exactly right and in particular, I would expect these social effects to be stronger for things that we think of as social behaviours or socially appropriate behaviour. Another thing that we found in the study, just like we have found in other studies, is that there are some people who clicked on the “vote” button in this case, but then we went to check up on them, they didn’t actually vote. And I think one of the reasons why is because it’s highly socially desirable to be seen as participating in politics. When it comes to buying a Ferrari, you might get some social benefits out of owning a Ferrari, but it’s not about people saying, “This is your obligation to own a Ferrari.” So, I do think that it may be the case that we wouldn’t see the strength of effects for other kinds of things online.
Chris - What's worrying me though is the sorts of numbers that you're producing from this study of people whose voting behaviour was influenced. If you look in the state of Florida, in the Bush/Gore election, that was one with nearly 600 votes only – a tiny number – you're influencing orders of magnitude more people with this study. So, does this mean that potentially, a savvy politician using Facebook could actually totally skew an election result then?
James - I do think that these are going to be effective tools, but everyone is going to know about them. And so, I think that what you'll see is, both sides are trying to use the tools and both sides trying to mobilise their voters. And it’s important to remember that although it was a very large number of people – 340,000 people – this is out of 61 million and so, it’s a small effect, but because it’s spread out over a very, very large number of people in a very, very large network, I do think it’s possible that as we learn more about how these processes work that you could do targeting, and you could get those 537 voters in Florida that you would’ve needed to change the outcome of the election.
Chris - James, we’ll have to leave it there, but thank you very much. That’s James Fowler from the University of California San Diego, and he published that report this week in the journal Nature.
Part of the show Silicon Sailors - Robots take to the waves from the 16th Sep 2012