Soon after Twitter launched in 2007, it dawned on social scientists, epidemiologists and policy wonks that the social media platform could be used to gather and analyze big datasets quickly and at low cost. Mental health and crime were two obvious topics. Law enforcement agencies also crept into the game, using Twitter and other social media services to document gang affiliations and investigate crimes.
But there has also been considerable skepticism about whether the messy datasets that social media produce meet the standards needed for reliable science and law enforcement.
A new paper by second-year Epidemiology and Biostatistics postdoctoral fellow Alicia Riley and her former University of Chicago mentor, Forrest Stuart, PhD – now an associate professor of sociology at Stanford University – seeks to establish best practices for further social media-based research and to help focus community violence prevention efforts.
The paper, published July 30 in PLOS ONE, uses a mix of qualitative fieldwork and computational methods to identify moments of heightened collective traumatic grief in Chicago. Such moments often lead to escalations in gang-related violence. They also signal community-wide traumatic stress that may result in a range of poor health outcomes.
“Expressing threats online is art, it’s culture – it’s how people make a brand for themselves,” explained Riley. “It doesn’t necessarily translate into anything. But in much of the way that criminal justice agencies have been trying to use big data, they seem totally unconcerned if those threats are real,” said Stuart.
In their new paper, Stuart and Riley, along with co-author Hossein Pourreza, used hand-coding of Twitter posts and interviews with youth to develop key words that would identify relevant Twitter content. For example, they linked a handful of phrases, such as “RIP” and “rest up,” to Tweets expressing grief for lost young people.
The researchers also catalogued the names of gang factions, which in Chicago are synonymous with mini-neighborhoods. They included tweets that used both a neighborhood/gang-related word and a grief phrase. The neighborhood/gang keywords also replaced the much-touted use of geolocation tags to link content with a specific place because, in reality, nearly all Twitter users turn off geolocation tagging.
With grounding in the local context, Riley and Stuart were able to identify days over a five-year period in which particular neighborhoods posted many tweets expressing collective grief. They verified that approximately 90 percent of these spikes corresponded to a shooting or similar event.
The study suggests that social media data analyses can help direct public health interventions in troubled neighborhoods if combined with fieldwork and community partnership. The researchers report that community groups that intervene to stop cycles of violence say they need a head start, even if just a day or a few hours, to be able to stop more violence before it happens, instead of scrambling to mitigate the fallout after it does.
“We want to push these methods in ways that are more equitable, less about putting people in prison and more about wrapping our arms around them,” Stuart said.
Riley said she expected many of her colleagues to be surprised that she would engage with computational work, of which she and Stuart have been critical.
“We wanted to do this in a way that foregrounds the many ethical and methodological concerns and that focuses on the community-level, not on individuals,” she said. “There’s still a reason to not let these tools be totally in the hands of economists, data scientists and law enforcement. We need to be in that sandbox challenging dangerous uses of social media data and modeling a better way.”
The researchers hope that further studies of violence and trauma in Chicago could piggyback on this community-level approach. They also want to establish the need for qualitative fieldwork in computational studies of community drivers of health.
Postscript: Riley recently accepted a position as assistant professor in the sociology department at UC Santa Cruz beginning in Fall 2021, after the conclusion of her postdoc next summer. Congratulations!