GMU pioneers a new approach to harvesting GEOINT
The Center for Geospatial Intelligence at George Mason University is conducting extensive hands-on studies to understand how “people act as a new type of sensor,” according to Anthony Stefanidis, director of the center. “The question then is how to extract actionable knowledge from this new type of sensor,” Stefanidis said.
Stefanidis, also the director of GMU’s USGIF-accredited graduate programs in geospatial intelligence, collaborates with professor Arie Croitoru and senior researcher Jacek Radzikowski, both of the GEOINT center, as well as professor Andrew Crooks of GMU’s computational social science department, to harvest data from various social media sites using keywords or locations. They then parse the data using geolocation information and store it into a “social media ingestor” for analysis to contribute to many different projects.
Analysis of social media data is in the process of being integrated into the program’s overall curriculum, Stefanidis said. The department anticipates introducing a graduate course in geosocial analysis by spring 2013.
Analyzing the words and photos in social media provides first-hand understanding and knowledge directly from the field that you can’t get from the news. For example, during the Occupy Wall Street protests, the team was able to analyze space and time patterns to pinpoint clusters of activity and determine which clusters were doing what. Or, as was the case with Occupy Oakland in California, they were able to identify key players in the movement and track their influence over a particular period of time. This level of analysis, the GMU team said, wasn’t available five years ago.
This new data is significant, because it yields understanding of the human landscape and moves beyond standard terrain features. Satellite images of buildings don’t say much when examining a particular event, Stefanidis said. But a keyword search for places, people, or news can reveal an understanding of connections and how spaces are shaped. “For all of the multibillion dollar satellites in the U.S., nothing will give you all of this information,” he said. But, what is the relevance of this form of data collection for the intelligence community? No building came over from Afghanistan to attack us,” Stefanidis said. “People came over, and you can track them, their movements and connections.”
In addition to monitoring individuals, this type of analysis can help draw conclusions from current events. An example of this is the conflict in Syria. The team pulled a map of the country from Al Jazeera and started mining social media feeds for the keyword “Syria.” What they found were tweets coming out of certain urban spaces in the country, and then references to Syria coming out of Western Europe. “We try to reveal and map all of these invisible connections,” Stefanidis said.
For example, the high concentration of tweets in Western Europe can demonstrate which European communities have an interest in Syria, or help answer questions such as whether the British election has an impact on Syria or vice versa.
Croitoru said the lab is constantly changing and evolving its methods, and has its system set up and ready to provide real-time, rapid assessment as soon as an event occurs. He used as an example the August 2011 earthquake in Mineral, Va. Monitoring tweets allowed the team to generate a much quicker assessment of what was happening versus waiting for the announcement from the U.S. Geological Survey, Croitoru said. More than 15,000 earthquake-related tweets flooded Twitter within about an hour. The challenge then was to differentiate between the actual event—the people who felt the earthquake and were tweeting from the East Coast—and the cyber event generated by the response, such as that from earthquake veterans in California.
Stefanidis said analyzing tweets generated from the Washington, D.C., area is a good example of how studying social media can provide intelligence about where people work, live, and have the most access to technology. “What D.C. looks like at 2 p.m., and what it looks like at night—it is two different cities,” Stefanidis said.
Croitoru pointed to the March 2011 tsunami in Japan as an example of how this method can help analyze the way information promulgates throughout a society. Immediately following the disaster, Japanese broadcasting station Nippon Hoso Kyokai was highly successful at communicating information to the public through the use of Twitter and retweets. “This is a new type of geography that we are pursuing,” Stefanidis said. “We are looking at the non-obvious, and detecting and visualizing the invisible.”
Stefanidis and Croitoru said mining social media is becoming more prevalent across many disciplines, including computer science, GIS, and social sciences. While this new approach offers many opportunities, it also holds potential future challenges such as validation and biases that will need to be addressed. But, the potential benefits from such an approach far outweigh the relevant scientific and research challenges, Stefanidis said.