THE BLOG
11/02/2012 05:50 pm ET Updated Jan 02, 2013

Beyond Polls

With Election Day almost here and the presidential race tight, everyone's following the polls. Polls are appealing because they're so simple. They remind us that in the end we face a binary choice. Who will be the next president, Candidate A or Candidate B?

For all that polls tell us, however, they're mostly about numbers and don't really let us hear the voices of the people. But today it's possible to hear those voices - and thereby get a better sense of what's really driving the election - through social media. In just two election cycles, Twitter and Facebook have emerged as lively complements to polls, offering a constantly changing read-out on how every twist and turn in the election is being received out in the world.

The narrative of presidential politics still unfolds on TV, through debates, conventions and news coverage. But now the public is using social tools to talk back to the their TVs, participating in politics in a whole new way.

The social platforms are not a perfect mirror of the general public. But since the last election, their popularity has grown tremendously, and the election conversation they're hosting has exploded. The three presidential debates alone drew more than 30 million real-time tweets and public Facebook posts, more than twice as many comments as the last Super Bowl.

But ultimately the point is not sheer volume. Beyond the numbers, social media's strength is the messages that this huge trove of data is carrying about the state of our society and the issues that matter right now. Rather than answering a set series of questions like the ones asked by pollsters, citizens commenting in social media share whatever's on their minds with their own networks of friends and followers. This makes for an organic conversation that's rich with meaning.

To get at that meaning, you have to figure out who is talking and what they're saying. A few months ago, we launched a project called The Crowdwire that aims to do just that, using natural language processing algorithms and machine learning techniques to make sense of the election commentary in social media. Among our findings so far:

• Women, whose votes in certain states will reportedly decide this election, spoke up about the debates more heavily than men. And women were engaged more by economic questions than by issues foreign and military affairs.

• Large numbers of African-Americans commented about the conventions and debates, a significant fact given that a strong African-American turnout helped Obama's 2008 victory. A repeat of this phenomenon has been in doubt this time, which makes this demographic's social engagement with the campaign significant.

• During the debates, policy issues such as health-care, education and controversial topics such as the Benghazi attack often had a greater share of the conversation than pure political chatter - the "horse race" fixation that often dominates traditional political news coverage. In this election, social media's reputation as a center of frivolous discourse has often seemed undeserved.

• Though Twitter's audience skews young, issues of importance to seniors - Medicare in particular - figured prominently in the policy conversation, sparked by events such as a Paul Ryan's speech to an AARP convention.

In addition to reflecting public concerns and reactions, social media are redefining how elections unfold. With the candidates themselves, the journalists who cover them, and tens of millions of citizens all tweeting, this medium is changing the political process itself. For instance, old-school campaign slogans such as "I like Ike" have been replaced by viral phrases born in social media, typically in the form of Twitter hashtags such as #emptychair and #bindersfullofwomen. These quickly cross over to the mainstream news outlets and enter the public consciousness.

There are challenges in social-media analysis, including how to accurately detect sentiment. When a young person remarks that a candidate is "sick," they mean something very different from an older person using that word, and technology can't always make such semantic distinctions. To get the analysis right, careful human oversight is necessary.

As the medium matures, so inevitably will the tools we use to understand it. If all goes well, our machines will get better at understanding us, and in turn help us transcend number-driven politics, humanizing how we choose our leaders. Which will be a very good thing for democracy.

The Crowdwire is a non-commercial project of Bluefin Labs. It's on Twitter @TheCrowdwire.

Subscribe to the Politics email.
How will Trump’s administration impact you?