Two powerful but not entirely related trends have, over the last decade, revolutionized the way national-level political campaigns are run. There’s 1) the explosive growth in social media chatter, and 2) the increasing data-savvy among better-run campaigns. Blending the two together gives you a third micro-trend: So-called "social media intelligence."
If you’re Mitt Romney, being able to parse huge quantities of data generated by Twitter, Facebook, Tumbler and online comments can help you not only understand whether your take on immigration is capturing the public’s imagination, but who exactly is talking about which aspect of the issue.
For a candidate, such insight can function both as an early-warning system and the conveyor of a hyper-charged social savvy — kind of like knowing the topics that will come up in conversation before an important cocktail party.
MindMixer, an online platform for civic engagement, bought a fellow Omaha company called VoterTide back in March. On paper, the acquisition made a great deal of sense. Still a young start-up, VoterTide had gotten traction selling social media insights to candidates on both their own campaigns and those of their opponents. Candidates who win elections the data-driven way are reluctant to give up the advantage when they come into office.
But in practice, it’s proving more difficult than expected to port social media intelligence from campaigns to cities. A great deal of that, explains Nick Bowden, MindMixer’s cofounder and CEO, has to do with the simple structure of a political campaign. You have a candidate. He or she has a name. Bowden explains how his social intelligence tools, rebranded Community Insights, would have parsed the New York City comptroller’s race.
For example: "They can say, ‘Okay, here are the stories trending around [the keywords] Eliot Spitzer, and they’re coming from BusinessWeek, the Huffington Post and the AP. Here are the top 10 words that are used to describe that article and the top 10 Twitter hashtags being used."
"Cities," though, "are complex and dynamic, and what people are talking about is across the board." No doubt, the data is there. MindMixer looks at millions of pieces of social media information sourced from the vendor DataSift. DataSift gathers both real-time and historical data from scores of online sources, standardizing and packaging it for easy analysis. And people are posting about the city services they come into contact with on a near-daily basis.
"But the words that people use to talk about cities [are] indirect," Bowden says. "People don’t always write, ‘I wish that the city of Los Angeles would fix the 405.’" Instead, they might write about ‘L.A.,’ or ‘the highway,’ or just simply whine about the traffic being bad without specifying the location.”
When people write about their experience with a restaurant on Twitter or Facebook, the data shows, they tend to include the name of the restaurant. "But if you go to a park and the playground’s broken," Bowden says, "you don’t necessarily mention the name of the park. That makes it a lot more intangible, so you have to build in predictive elements."
To some extent, the solution is baked into the technology. Only between 5 and 10 percent of tweets are geo-tagged with their location, Bowden says, but a good 60-70 percent of the social media data they’re analyzing comes from Twitter. That creates a small but useful pool of content that contains geographic information. Location metadata can help tie a comment made in the digital ether to an actual spot on the map.
MindMixer is also working with researchers in linguistics and sentiment analysis at the University of Nebraska to better understand how people communicate about city life. Backstopping that approach is the collection of several months’ worth online conversations before looking for insights in the day, as a way of teaching the tool how locals talk. (MindMixer has trial Community Insight projects up and running in Los Angeles, on transportation, and in Chicago, on education.)
The trick that Community Insights will have to pull off to succeed is capturing the full panoply of digital civic conversation without including so much noise that the data becomes worthless. But Bowden and company are up against powerful forces.
"Celebrities always break it," Bowden says. "Justin Bieber will tweet, ‘Hey, I’m driving the 405,’ and that gets 10,000 retweets and Facebook ‘likes,’ but it doesn’t really mean anything."
Bowden is reviewing Los Angeles’ collected transportation chatter as we talk by phone. The city’s dashboard is swamped, he says, by generally useless conversation about Elon Musk’s proposed Hyperloop, which has a suggested terminus in L.A. "We thumbs that down," Bowden says.
For cities to become social media intelligent, though, the process of parsing tweets, posts and comments needs to be automatic and nearly seamless, Bieber and Musk be damned. City officials across the country talk of knowing that they’re required today to participate on Twitter and Facebook, but having a sense of flailing around in the dark on there. Ultimately, Bowden says, the goal is to equip a city with "a portrait of who in their communities really drive the conversation." That would mark a return to a sort of intimate town-hall setting, without the self-selection that can rob in-person politics of their power.
"Cities at some point have to become aware of the implicit conversations that are happening," Bowden says. He’s hopeful that his tools will get there, but he admits that those working inside City Hall today lack the social insights of those still trying to get there. "Somebody," he says, "has to crack that code."