Facebook's political team predicted the site would be the most accurate predictor of Congressional races, bar none. Were they right? Not exactly.
This was the first election year in which Facebook and Twitter had achieved ubiquity. Even during Barack Obama's Facebook-fueled campaign of 2008, the network was a long way from the half-billion user base it attained this year.
Social media is like the telephone in that it changed the world once it had connected everyone; unlike the telephone, it connects people in clusters. The user, in social media, is overtly connected only to people they choose to associate with.
Because Facebook's user base is a pretty accurate cross section of American voters as a whole, Facebook's political team thought its vast expanse of clustered associations should translate nicely into the most accurate polling predictor of the midterm elections.
Interestingly, they were wrong in the exact areas you'd think they would be right: the most high-profile races.
Facebook predicted that candidates with more Facebook fans than their opponents would win their races. While they were right in that 74 percent of candidates with more fans did win, the most talked-about races, which should ostensibly get the most attention and therefore have the most fans online, all contradicted the Facebook-fan precept.
Christine O'Donnell, Meg Whitman, and Sharron Angle all had more Facebook fans than their counterparts, yet all lost. Meg Whitman, who spent three times as much on her campaign as opponent Jerry Brown, had fully twice as many Facebook fans as Brown.
Over the summer Malcolm Gladwell bemoaned the "weak ties" problem of social media. He asserted that Facebook and Twitter merely constitute a facsimile of connectedness by lowering the bar for what qualifies as "connection." He even claimed that Twitter's value in Iran's post-election riots of 2009 was overestimated.
Sharron Angle had a significant advantage in Twitter followers over opponent Harry Reid (despite Reid's Twitter rapport with Lady Gaga) and yet Angle lost.
A media-savvy friend dropped a new adage on me recently: "In old media, content was king, but in new media, conversation is king." Along these same lines of logic, the Harvard-developed Crimson Hexagon program analyzed Twitter in the month leading up to the midterm elections and attempted to generate predictive polling data not based on the number of followers, but on the content of conversations.
Although Sharron Angle had more followers than Reid, Crimson Hexagon's analysis predicted that Reid would win 55 to 45. Reid ended up winning with 50.2 percent.
One way to interpret social media's failure in predicting some of this year's critical elections would be that Malcolm Gladwell was right -- our digital connections are just not valuable enough to serve as real-life predictors. Another way to interpret it would be to say that, much like the telephone, it's not just the people you're connected to but the content of the conversations you have once you're connected that indicate the real human intentions flowing through the digital medium.
Perhaps by 2012 we'll have not only a pervasive digital social graph in the form expansive social media, but we'll actually have figured out what it all means.