January 09, 2012
In December 2010, a man in Tunisia set himself on fire to protest the local government's inaction against police misconduct. That act led to a wave of protests across North Africa and the Middle East, driven in part by the use of social media. Similar uprisings, guided and coordinated by social media users, have occurred in Bahrain, Syria, and Libya; and the U.S. military has been called upon to conduct stability, security, transition, or reconstruction operations in some of these countries.
There is widespread recognition that the Department of Defense and other government agencies need insight into human social dynamics—those aspects of collective human attitudes and interactions that shape perceptions, decisions, and actions, says John Gersh, of the Research and Exploratory Development Department (REDD), who recently edited a special issue of the Johns Hopkins APL Technical Digest describing Laboratory work in that area. "Each country is unique, but increasing our knowledge of common aspects of social identity, organization, and interaction will help us to understand, respond to, and maybe even forecast such events."
APL recently carried out a substantial independent research and development effort in this area. Several defense sponsors—including the Office of Naval Research and the Air Force Research Laboratory—have funded work derived from the program, and other potential sponsors have their eyes on APL's progress.
Human Language Technology
APL has long-standing expertise in technology that can automatically analyze images, as well as speech and text, in multiple languages. Building on these efforts, researchers are developing models and algorithms to extract data from text that may lead to a better understanding of social movements, political developments, and responses of local populations to counterinsurgency or stability operations.
The proliferation of Internet blogs and discussion forums has presented a gold mine for those seeking to measure the opinion of the masses, including the military. Using data gleaned from blogs geared toward politics, knitting, and tango dancing, APL researchers have extracted expressions of positive or negative attitudes from text—a process known as sentiment analysis. Such a capability would allow sponsors to detect shifts in both the volume and sentiment of mission-critical topics.
APL researchers have also been fine-tuning methods for detecting and classifying names in text, and applying that information to real-world situations—such as checking names against "no fly" lists.
Finding Online Artifacts
Every second, millions of people generate and share content online that leaves trails of so-called digital social artifacts—blog entries, Facebook updates, and e-mails. Recognizing an opportunity to develop insights from such data, APL researchers developed SocialRank, an algorithm that assigns a numerical weight to each node in a social network to measure the node's (or the person's) relative importance within the network.
"SocialRank demonstrates the utility of using social signals to highly accelerate the analytic workflow," says REDD's Jaime Montemayor. "Broadly, our ability to discover and track relevant social media can alert analysts to quickly locate and understand the onset of important events, such as natural disasters or terrorist activities."
Many of today's political conflicts are based on social identity differences—think of the Tutsi versus Hutu violence in Rwanda—and sides are drawn along ethnic, religious, and ideological lines. APL has developed the Social Identity Look-Ahead Simulation (SILAS), a game that models the effects of social identity on opinions and attitudes and provides a mechanism to make predictions about hypothetical conflicts.
National Security Analysis Department analysts used SILAS last year in a war game exercise called the Green Country Model. Since then, the Office of Naval Research has funded follow-on work, which is collecting open-text data from blogs, discussion forums, and Twitter, and then performing sentiment analysis on postings related to Nigerian politics. Researchers will then compare that data with polling data. If successful, this effort may help to establish a faster and more detailed source of data for verifying sociopolitical models.
"When I began working at APL more than 30 years ago, our challenges lay in developing technology and engineering systems to deal with threats to our national security," writes Gersh in the Technical Digest. "That has not changed, of course, but now we face a radically changed environment. Then threats changed slowly and were clear, localized, and well defined. Now threats change rapidly, are hidden by intent or by nature, and are worldwide and amorphous."
"In response, APL is applying longstanding capabilities and developing new ones to help our sponsors and the nation deal with this new environment. The modern imperative for U.S. counterinsurgency forces is ‘learn and adapt.‘ This imperative also applies to organizations like APL that support counterinsurgency forces and their mission."