Parsing the Crowd: Politics, Social TV, and the Power of Analytics
19 October 2012

by tmartin on October 19, 2012

by Teresa A Martin

HOW DO YOU FEEL ABOUT CROWDS? Me, I can get overwhelmed.

Like the other week at OysterFest in Wellfleet … when this wildly popular October event turned the street into pedestrian gridlock  … and suddenly my imagination clicked into high gear and I felt like I was fleeing from the jaws of a giant mobile bivalve in some late night horror flick rendition of Revenge of the Raw Bar.

So I can completely understand why any of us an feel a bit overwhelmed by the virtual crowds of the Internet. During this week’s second presidential debate some 109,560 comments per minute popped into the Twitter-sphere alone.

Like millions of others, I was watching the TV and my connected device at the same time, being both observer and participant, in the grandest tradition of live debates. I figure side commentary in the agora entertained more Ancient Greeks than the actual words of the Athenian debaters. Some things never change!

The millisecond that the phase “binders of women” spilled onto stage, the tweets went wild and it didn’t take a genius to know we’d just heard the “story” of the evening.  To get more insight into this moment – and indeed, into the dynamics of the crowds following all the debates – I turned to a research project called The Crowdwire (http://thecrowdwire.org/).

The Crowdwire lives within BlueFin Labs (http://bluefinlabs.com/), a Cambridge-based spinoff of the MIT Media Lab. Fun fact for Cape-types: the director of The Crowdwire project, William Powers, lives on the Cape, creating another one of those links that fill our region.

Bluefin Labs specializes in something called “social TV analytics,” which involves a great deal of algorithm development and data crunching on the words, text, and use patterns found in the Twitter stream. Basically, it applies mathematics to the sound of the digital crowd to make sense of the crowd’s murmurs – a sort of mathematical translator of digital babel, if you will.

In its core business the company applies these analyses to the hot area of “social TV,” a phenomena in which people no longer sit passively and absorb television and its “messages from the sponsor” but instead gather around that giant water cooler in the cloud, Twitter, to talk with each other about the shows, the personalities, and the products.

For Big Brands, this virtual water cooler provides a virtual cornucopia of feedback and data – and the consumer goods giants have been queuing up to companies like Bluefin Labs for a handle on what millions of tweets really mean.

And what bigger brand can we find than the Office of the President? Especially now, when the debates have become popular TV?

Media reports say that about 67 million people watched the first debate and about 65 million watched the second one. In addition, millions – some estimates put it at 3-4 million more – watched the debates via some sort of non-television device. The presidential debates beat out the Summer Olympics in the ratings game and only the Super Bowl drew more viewers than the debates this year.

That’s one big crowd.

Some portion of the crowd brings its social TV habits along to this debate show. Slightly more than 10 million tweets pinged out during the first debate and more than 7 million comments filled the Twitter stream during the second one.

And so, enter The Crowdwire. Powers and BlueFin took the company’s analytics tools and applied them with a research lens to the presidential and vice-presidential debates and indeed to the entire ongoing campaign.

For example, one question the group had was whether the pundits’ areas of focus mapped at all to what people were actually talking about. In other words, the team wondered if all those talking heads synch with what everyone else cares about?

During the vice-presidential debates, the traditional experts yammered on about the role of the moderator. The Crowdwire project used  a set of analysis algorithms that parse both literal and related remarks and found that in the Twitter and Facebook worlds, people talked about the moderator, Martha Raddatz, 10% more than the candidate, Paul Ryan.

The Crowdwire project also uses language-based analytic tools to break social media commentary into three groupings: Politics – the horse race aspect of the campaign; Policy – the actual issues of the campaign; and Personality – the human aspect of the candidates.

“We determine what statistical proportion of the conversation – a metric called “share of voice” – is devoted to each of the P’s, and then track how the balance fluctuates over time,” the project’s blog explains.

In other words, by following the social media stream and using the Bluefin tools to decode the sounds of the crowd, the project can watch how events shift what people talk about and respond to.

Pollsters, of course, have been trying to figure this out for years. The voice of the crowd, while not necessarily representative of all people and all groups, adds an interesting layer of texture to the discussion. Polls report what people say when asked a question — but the crowd records the trends in what people say to each other when no one is looking.

Much like the water cooler discussions of yore, that reflection can be both more honest, more brutal, and more insightful than we might like – and might even bring us all a little self-awareness of how and why we respond to events around us.


Find The Crowdwire project at: http://thecrowdwire.org/

Leave a Comment

Previous post:

Next post: