A team of Carnegie Mellon University (CMU) researchers has developed a modeling and simulation package that allows users to create their own social-network models instead of having to call in professional modelers.
Using automated tools, open-source data and social-network analysis, the researchers created a model to address strategic questions, such as whether U.S. pressure can dissuade foreign leaders from going to war. This approach means that lower echelons, all the way down to the brigade level, can create their own social-network models to address local operations, said Lt. Col. Michael Lanham, an Army information systems management officer who is a Ph.D. candidate at CMU and a member of the research team.
“They can do this at the lower echelons without nearly as much dedicated modeling and simulation expertise and in less time,” Lanham said.
The project, funded by the Air Force Research Laboratory, is designed to show how multiple models could be used to mine mountains of open-source text, analyze it and then feed the results into a simulation. The CMU team used three tools: Automap for text mining, Organizational Risk Analyzer (ORA) for social-network analysis, and Construct, an agent-based computer simulation.
The researchers used a hypothetical India-Pakistan dispute set in 2002 as their test scenario. The question posed was whether U.S. diplomatic intervention could dissuade leaders who were inclined toward going to war from actually doing so. The answer was yes — but more interesting is how the conclusion was reached.
Automap trawled through 3,000 newspaper articles from Lexis-Nexis to identify words and key agents that fit nine categories in ORA’s social-network model, which uses hundreds of algorithms to identify groups, connection points and changes in networks. The researchers then manually deleted irrelevant data. The results were fed into Construct, which simulates the spread of information among the agents. Creating data lists took about 160 hours, plus another 40 for analysis and simulation.
“There is another way of building models when you don’t have dedicated modelers,” Lanham said. “You can have the system look at 10,000 or 25,000 text documents and say, ‘Build me a model of key concepts within that text.’ And then when you have that model, you can do reduction and see how much of that stuff is relevant.”
Lanham is now using these tools to assess command and control in cyberspace operations. He also sees this approach as particularly useful for counter-IED initiatives. By mining data from sources such as the Distributed Common Ground System, and overlaying it onto maps such as NASA’s wind topological software, over time, users could plot an entire IED production chain, Lanham said.
Because the tools are fairly simple to use, Lanham believes users could easily keep track of the social networks they most frequently follow.
“If you had this kind of ongoing social network that you maintain for your countries of interest, then you may get an insight into who you talk to on a formal and informal basis that you might not otherwise have,” he said.