A team of hundreds will scour the Web for signs of voter intimidation and poorly functioning polling places by Nanette Byrnes
While most news outlets on Tuesday will focus on which candidates and ballot measures are getting the most votes, a group of journalists and students will be mining social media to detect problems with the process of voting itself. Are voting machines malfunctioning? Provisional ballots being handed out? Voters being inappropriately asked for ID or being harassed? In short: are citizens able to exercise their right to vote?
Already stories have surfaced of long voting lines in Nevada and Georgia, where an unusually high number of voters cast their vote ahead of Election Day.
More than 150 people are expected to crowd into the student newsroom at the CUNY Graduate School of Journalism in New York’s Times Square. Professional journalists and students will join data experts from Google and other companies for pizza, coffee, and snacks, and to sift through a deluge of data on that day’s voting. It will be the hub of a group called Electionland, led by the nonprofit news site ProPublica, Google News Lab, CUNY, and others. They hope Electionland will leverage social media and other data sources to surface and flag suspicious voting situations.
Systematically monitoring voting is hard because it is supervised by local officials, making it tough for groups to highlight concerns quickly, says Derek Willis, a data journalist and developer at ProPublica and part of the Electionland team. “A national election is really 50 state elections,” says Willis.
Electionland is a bid to take the wide reach and accessibility of platforms such as Facebook and Twitter and use that to monitor and improve the voting process.
Willis and others involved readily admit that information coming from social media and Google search trends are at best an imperfect measure and at worst subject to manipulation. “It will be just as much about debunking false claims as identifying the problems,” says Carrie Brown, director of the social journalism program at CUNY.
Charles Stewart, a political science professor at MIT and an expert on voting, says he is “very, very worried” about the potential that Electionland’s focus on social media could be susceptible to manipulation and that news organizations, drawn in by the immediacy of the information coming through on social media, might rush out information before it is properly vetted for accuracy. Disseminating information on issues such as the length of voting lines, even if accurate, could have the negative effect of discouraging voter turnout, he says.
Just as problematic, much of social media is not representative of the broader public. “It is skewed urban, toward tech-savvy white men. It is skewed in a lot of ways,” says Stewart.
To minimize those issues, Electionland’s creators have tried to pull data from a variety of sources and create a system to critically vet the information that surfaces.
Though much of the data will come into Electionland from automated systems, its review will require a significant amount of human judgment. Six hundred student volunteers will use tools like Facebook Signal and Dataminr to pick up on hot issues on social media platforms. In addition, the public will be able to send concerns directly to Electionland via text message. The group will pull in leads on election irregularities phoned in to a hotline run by the Lawyers’ Committee for Civil Rights.
Students will try to verify the authenticity and significance of various signals with tools like reverse image searches (to determine if a photo has been in circulation before) and geolocation (to determine whether a text is coming from near a polling place in question).
Professional journalists will then review issues that pass the students’ evaluation, and if the pros agree, the leads will be sent to local reporters around the country who have signed up to check out questions about polling places in their area. “There is no magic box here,” Willis says. “As with most civic-oriented projects there is no algorithm that does the job without human input.”