CHARLOTTESVILLE — On the Monday after last month’s violent rallies, Phil Bourne learned during a meeting that fellow University of Virginia officials had shared feelings of regret, dissatisfaction and frustration about the disturbing turn of events.
“It was clear we hadn’t made as much out of social media as we could have,” Bourne said. “So we started to wonder if there were ways to use social media and big data to handle these things in the future.”
That discussion engaged several faculty members’ interest and since has rapidly expanded into a research project involving students and state agencies.
Bourne, director of the Data Science Institute, which compiles large-scale data analysis, realized the project also would be a good fit for his top students.
Though just in its beginning stages, Bourne hopes the project will help the university and law enforcement respond more effectively and efficiently to potential crises like the August protests.
Released a month after the turbulent weekend, a UVa adiminstrative report confirmed that incomplete intelligence analysis about the Aug. 11 white nationalist torchlit march to the Jefferson statue at the Rotunda allowed the situation to turn violent.
“The school relied on official intelligence and did not effectively use information offered from social media and the website It’s Going Down,” the report said. “Police also relied too much on inaccurate information deliberately planted by organizers of the rally.”
Don Brown, a leader of the project and the founding director of the institute, said his team hopes to address those two problems: slow gathering of data and an inability to distinguish between false from true information.
The first step, he said, will be to hear from students who learned of the march to the Rotunda and rapidly organized a counter-protest while police still were sorting through decoys planted by march organizers.
The “smokescreens,” Brown said, included misleading information spread to the police that the marchers would meet at Darden Towe Park, that the gathering would be brief and that there would be a motorcycle rally.
“It was like a light bulb went off,” he said, about the moment when he realized that counter protesters had gleaned information from social media faster and more accurately that university and law enforcement officials had.
“I didn’t know anything, I was sitting at home. But we have people right here at UVa who knew what was going on. What were they looking at so that they were prepared?”
Brown wants to see the public social media posts the students saw, but he also wants to see a much wider range of messages from public and private accounts, which may help determine patterns of speech and language of people planning violence.
But it’s not enough to have the messages themselves: in order to predict and prevent future incidents, Brown also needs the data behind the messages, such as the timestamps and location.
By mapping the data, the language used in the messages and the connections users have to each other, he hopes his team can determine tip-offs for future escalation and violence.
“It was the largest thing law enforcement has had in 20 years,” Brown said. “We don’t have much to build on from a data perspective.”
By combing through the locations of posts, coded language and usernames, he hopes, patterns will emerge that will help police identify and weed out similar smokescreens in the future.
The project has generated a lot of interest from other UVa faculty members, as well as law enforcement and the Governor’s Data Internship Program.
The institute is interdisciplinary, and by bringing on advisers with backgrounds in computer engineering, religious expression and political extremism, Brown believes the teams can find creative solutions.
Arlyn Burgess, associate director for operations and strategic initiatives at DSI, said some of those partnerships still are taking shape, but the group is determined to connect with governmental, academic and private organizations.
The project is one of 13 graduate ‘capstone’ programs offered by the institute this year.
Currently, the capstone students are writing out their research questions and developing a research plan, Brown said.
The capstone program runs for two semesters, and Brown said he hopes students will be able to publish their findings in April.