Virginia Tech team leads federal effort to forecast the flow of information online
According to a 2016 study by the Pew Research Center, a majority of U.S. adults now rely on social media as their primary news source. This means that when natural or man-made disasters strike, the quality of information shared online and how it spreads is critical; the ability of official evacuation orders to break through the noise of inaccurate or intentionally misleading posts could literally be a matter of life-and-death.
Collaborators with the Virginia Tech team will include experts in computer science, cognitive science, economics, and sociology from Stanford, Carnegie Mellon, Claremont, Duke, Wisconsin, USC, and the Institute for Human and Machine Cognition.
The Defense Advanced Research Projects Agency, or DARPA, has approved the allocation of more than $6.7 million in funding to the Virginia Tech-led team over the next four years. The project, titled Homo SocioNeticus, is a key component of DARPA’s new SocialSim Program, which will support fundamental research to develop technologies that afford high-fidelity simulation of online behavior at-scale and will guide the establishment of a new research community centered on pushing the limits of rigorous evaluation of human social simulations.
“For decades, cognitive science, psychology, and neuroscience have been developing models to understand decision-making on an individual level or, rarely, in small groups,” said Mark Orr, a research associate professor at the Biocomplexity Institute of Virginia Tech’s Network Dynamics and Simulation Science Laboratory and principal investigator of Homo SocioNeticus. “This team will be the first to test how today’s cutting-edge cognitive science, informed by social science, translates to networks the size of Facebook or Twitter, orders of magnitude larger than anything attempted before.”
In addition to boosting critical communications during disaster relief operations, one major potential benefit of the initiative is its capacity to help de-escalate conflicts without the use of armed force. Where adversaries seek to increase military tensions through online misinformation campaigns, this research may present a means of swinging public opinion back in favor of a peaceful resolution.
“Any time you can save lives without having to be aggressive, that’s important,” said DARPA’s SocialSim program director, Jonathan Pfautz. “At present, the government employs small teams of experts to speculate how information may spread online, but a system that can automate that process and provide greater accuracy could focus our efforts even further.”
The program also aims to lead the way in establishing guidelines for ethically sourcing data in large-scale studies of social media usage. With government agencies and private-sector marketers scaling up their efforts to forecast how large groups of people spread information online, clear, consistent guidelines about how to obtain consent and ensure anonymity are more critical than ever.
“DARPA wants to set the example and build a vibrant research community around this big, audacious problem,” said Orr. “The ability to accurately diagnose how large populations share data may still be several years off, but in the process we’ll be establishing a firm precedent for how this type of work should be done.”
More broadly, researchers say this project could provide a new framework for understanding the online “echo chambers” that have come to define today’s media landscape.
“The standard site of information consumption has shifted from centralized, largely uniform media news sources to decentralized, self-selected ‘information pockets,’ which has dramatic implications for how people understand events in the world,” said James Moody, professor of sociology at Duke University and member of the project team. “If we can effectively model how information moves across such landscapes, we may be able to help people see the blind spots in their own information sources as well as distinguish systematic distortions.”
The Homo SocioNeticus initiative will apply new methods developed at the Biocomplexity Institute of Virginia Tech to computationally model the behavior of large populations and simulate the interactions that drive information diffusion and evolution.
“Our researchers have developed ‘synthetic populations’ capable of simulating the way real-world communities interact on a massive scale,” said Chris Barrett, director of the Biocomplexity Institute and professor of computer science in Virginia Tech’s College of Engineering. “Granular representations of interacting individuals at the scale of regional and national populations, in combination with artificial intelligence approaches to cognitive processes associated with those individuals at that sort of scale, is a really big deal. When we achieve that kind capability, it will allow study of social aspects of cognitive processes at population scale and in unprecedented detail.”
The Biocomplexity Institute regularly conducts research through federal, state, and industry grants and contracts. Notably, this award is part of the institute’s portfolio of research programs that has received more than $103 million in new awards in the first half of FY 2018. For more information on the SocialSim initiative, visit the official DARPA project page.