Social Data

           In a technologically dominated world where information regarding just about every transaction, mobile phone, and social network can be traced, tracked, and potentially sold, Facebook reigns supreme.
          
          It’s not that Facebook is covertly dealing people’s personal data or transgressing ethical boundaries, but the website has access to personal information on a commercial scale that has never before been possible. Users fear the impending doom of the their privacy while advertisers lick their chops at just a small taste of the internet giant’s trove of information that could be used to better influence people to purchase products.

          The ability of Facebook to enact a profitable model to suit advertisers based on this unprecedented user basis has been perhaps the largest business challenge the company has faced. Many of the question marks surrounding the business side of Facebook were well founded in the floundering ads on offer along the margins of pages that at best vaguely applied to the user’s actual interests, though never creating the connection that companies sought to establish with their customers. Though Facebook’s new advertising plan will be their most successful to date and puts the company on a track toward truly influential advertisements, there is still much that can be accomplished through a better understanding of how social networks impact our individual decisions.

          Facebook could be ‘the big one’ as it occupies the unique fault line intersections of big data and social networking. Up until recently these two key components were simply ignored or avoided due to the difficulty in grasping their applicability. The emergence of computational social science as a viable and robust discipline has helped to give rise to a new way of demonstrating what we do and why we do it. That Facebook has begun to cash in on using the social connections between users as a means of influencing behavior shows a positive shift in understanding complex human network interactions; however, while the company’s newest advertising attempt that uses friends’ existing ‘likes’ as a means of endorsing products on your own Newsfeed does tap into social networks, it is not adequate in addressing the ways social connections impact and inform what influences our behavior. Pairing the product with a friend is a good start, but there is a great deal of unlocked social potential left to be explored.

          A big part of what influences our social network comes from the types of bonds we have formed with those around us. Strong bonds are the kind generally formed through shared experiences, shared values, and face to face relationships. Weak bonds are formed with coworkers or people with whom you belong to a certain group, but lack anything else in commonality. In his October 2012 article in The New Yorker Malcolm Gladwell cites these bonds in his explanation of the ‘Arab Spring’ and the Egyptian Revolution. Gladwell felt that the Twitter revolutions that had been occurring in places like Moldova and Iran were not as strong as real revolutions because they relied on weak bonds in an effort to mobilize activists. Weak bonds are great online to the extent that they do not actually ask something of the individual, such as digitally signing a petition, but are less successful when it comes to rousing groups of people to meet, protest, and potentially lose their lives in an attempted government overthrow. Gladwell emphasizes that strong interpersonal bonds between friends and family are the real reasons that the revolution in Egypt was so successful. In much the same way we would expect Gladwell to forecast that the new ‘like’ endorsed Facebook ads will likely be very successful when they ask very little of the consumer, but they will fall flat when encouraging people to actually change their behavior by attending an event or to partake in a new fashion movement at JCrew.

          Facebook has an opportunity to utilize these weak bonds in effective ways that Gladwell overlooked in his article, namely homophily. The idea behind homophily (literally love of the same) is that we as humans model our behavior in large part on those around us, and the more similar other people are to us (or how we want to be) the more likely we are to see ourselves doing what they do. There is an incredible body of evidence that supports this with some startling results. Robert Cialdini notes in Influence that there are patterns among suicide victims that not only extend to race and gender, but also to the medium used to take their own lives. If an individual dies in a plane or car you are more likely to see subsequent waves of suicides by similar individuals in car or small plane crashes. Homophily helps to explain how suicides can become socially contagious. Our vulnerability to the halo effect and our bias towards experts are just a few other examples of how we can form strong associations through weak bonds. Researchers have found in order of importance that race, ethnicity, age, religion, education, occupation and gender are extremely powerful characteristics that we relate to in forming our in-groups. If Facebook can effectively define these similarities between their users through mining personal data they can empower influence through relatively weak bonds by matching endorsement with the correct type of reference person.

          While the psychology of weak bonds gives some wiggle room in application, strong bonds are still a much stronger (by definition) connection and a better way of influencing people’s decisions. There is no shortage of data when it comes to the power our social in groups have in forming our behavior. Depending on what the people around you do, your strong bonds can impact whether or not you are obese, happy, or going to vote. In the last case a study comparing messages sent through Facebook about ‘getting out to vote’ found that people were much more likely to vote if they saw the message accompanied by their strong bond association’s face reporting the fact that they voted (or at least were planning to). It makes sense though that we model our behavior along those who we have chosen to keep closest.

          Facebook’s new ads in which friends who have legitimately liked products serve as unwitting brand ambassadors is based on these concepts of weak and strong bonds. Beyond throwing an ad out and hoping you see it, and that it applies to you, the new ad format ensures that the message is associated with at least a weak bond or, hopefully, a strong bond. If it is a weak bond and the product is a big ask, for example switching your cell phone carrier, the response will likely be poor. While it is clear Facebook can discern who you have the most contact with the program is limited by a double coincidence where the condition of advertising to you personally requires that you have strong social connections who have already ‘liked’ their product.

Location of Ten DARPA Balloons

          Big data is revealing what is turning out to be the most effective and efficient means of engaging social networks to influence change. MIT’s Sandy Pentland is at the forefront of the aggregation and interpretation of big data for studies that run the gamut of disciplines. A program he helped run in Switzerland reduced power usage by 17% simply by having partners enroll in the program, and when one person used less power the other received a very small financial incentive. In this way Pentland was able to influence the individual’s behavior by impacting the social fabric that informs how they act as a member of group. And while a 17% reduction seems small, it is the same outcome as if the power company were to have doubled their rates, all for a couple dollars incentive a month . Pentland’s team took a similar approach in a DARPA competition to find ten red weather balloons set aloft at once in ten random places around the country. While other teams tried to encourage individuals through one-off incentives, Pentland and his team set up a scheme to tap into people’s social lives. If you found a balloon you would receive $2000 and the person who referred you would receive $1000, and the person who referred them $500 and so on. By finding all ten balloons released within six hours Pentland demonstrated the importance of empowering others to influence their own social networks.

Facebook Must Find A Way To Sift Through The Noise

          Facebook possesses incredibly important information about people’s social networks and has the means to disseminate messages that can have a profoundly influential impact on people. This is as alluring from a business side as it should be for those trying to make changes for the better. Pentland’s studies, along with those done by others, show that it can be relatively inexpensive to create the kind of changes in a person’s social fabric that can drive them to adopt or change a behavior. At the moment Facebook is still mired in impersonalized banner ads and product endorsements based on weak bonds which do not account for any of our homophilic biases. While using strong bonds will improve outcomes for those advertising on Facebook, true influence will be best achieved when advertisers, corporate or otherwise, can offer incentives that nudge social networks, rather than simply try to singularly address individuals. Pentland’s DARPA case shows that a Facebook campaign that could reward members of a group who referred participants to the campaign, whether through a purchase or otherwise, would spread the impact of the message throughout a group’s social fabric. Big data and computational social science are showing better ways to harness the social aspects of social media.

Deeper: One of the main spin-offs in a discussion about big data is the important issue of user privacy, as discussed in this New York Times article. There may be an unprecedented amount of information being generated everyday, but at some point there has to be a better understanding of what is and is not an acceptable amount of data mining. This is an issue especially now over the Snowden-NSA affair.  You can also hear Sandy Pentland talk at length about this issue from an interview done by Edge. A great introduction to some of Pentland's stuff can be found in this Google Talk he did last month (if the video comes in partway through just cancel the bracket to start from the beginning) to promote his new book. I found these two other papers helpful in understanding social bonds and the internet and social networks.