David Lazer’s interdisciplinary team includes social scientists, graphic designers and data miners. Together they’re using computational modeling to gain insights on society. Photo by Brooks Canaday.
For the last several weeks, Northeastern University researchers have been using computational models to distill massive amounts of presidential campaign data into nuggets of information that the human brain can comprehend.
From a “Debate Tweet Meter” to an analysis of super PAC funding, the team has tried to “illuminate processes by which money is raised and language is produced,” explained David Lazer, a professor of political science and computer and information science whose lab is leading the effort. “The machinery around both deeply affects our democracy.”
While Twitter is an obvious go-to source for lots of data on voter sentiment, other sources — such as the RSS feeds of mainstream media sources, the political “blogosphere” and campaign ads — leave traces of the linguistic strategies intended to sway that sentiment.
To untangle the sources of those strategies, Lazer’s interdisciplinary team of social scientists, data miners and graphic designers is developing visualization tools that tell the story behind the language. Assistant research professor Yu-Ru Lin, who leads the Debate Tweet Meter project, sifts through and analyzes large data sets including Tweets or financial contributions. Assistant research professor Mauro Martino turns those data into dynamic visual representations, while postdoctoral researchers Drew Margolin and Sasha Goodman use the information to make inferences about social processes.
“The beauty of my lab is that we have these different types of people with different skills and perspectives,” Lazer said. “And then we shake them up and cool stuff comes out.”
The group is also probing the financial structures behind language. “A lot of the money supports expenditures on language,” Lazer said, referring to the spending of political campaigns and political action committees.
He noted that focus groups and surveys, for example, could be used to help campaigns tailor their message to elicit a desired response. From there, the message percolates through society, leading to “linguistic homogeneity.”
Using content from television commercials, various types of websites and language used by the candidates themselves, the researchers are developing what they call the Invisible Networks Project. “We’re looking at the shared chunks of words that are articulated by politicians and the media,” Lazer said. “They are readily identifiable if you look at the data, because it’s exactly the same quotation.”
By identifying these texts, the team is constructing a visual model of the network of language that pervades our world and influences our everyday experience.
“A critical element of a democracy is for people to be exposed to different points of view,” Lazer said. “Ultimately we’re all subject to the same laws and the same policies.” Lazer’s team is working to reveal those views by laying bare the machinery of money and memes in politics.
