(Newsroom America) -- Army and university scientists are turning to problems with social media to create social sensing as a scientific discipline. For the Army in particular, this emerging science space, they say, will better help commanders assess and comprehend the accuracy and true meaning of information on the battlefield.
"Humans are prolific generators and communicators of information. In Army operations, commanders rely on information provided by Soldiers to make decisions. They also require information from automated sensor systems to understand, for example, where troops are located in time and space," said Dr. Lance Kaplan, U.S. Army Research Laboratory researcher in the Networked Sensing and Fusion branch.
"This traditional pipeline has worked well when the flow of information was relatively small. Now, however, with proliferation of computers and wireless communications, the volume of information produced and shared in the network of Soldiers and sensors can be extremely large. It becomes difficult to assess and comprehend the accuracy and true meaning of such information on the battlefield," he said.
A small team of ARL scientists, along with scientists from Rensselaer Polytechnic Institute, Notre Dame and the University of Illinois-Urbana Champaign found a somewhat similar problem occurs in social media.
"Although the Army battlefield information problem is not identical to the social media problem, they have some important similarities from the scientific point of view," Kaplan said.
In social media, the volume of the data is no longer manageable for manual processing. It's also very difficult to ascertain the credibility of social media reports, especially given that the reliability of the people posting reports is uncertain. A well-documented social media problem is spreading rumors. Automated systems are needed to extract relevant and reliable information from the vast amount of data generated by social media.
Challenges and opportunities associated with this research are described in a perspectives paper recently accepted for publication in IEEE Computer entitled, "The Age of Social Sensing," by Kaplan and Dong Wang of Notre Dame University, Boleslaw Szymanski and Heng Ji of Rensselaer Polytechnic Institute, Tarek Abdelzaher of University of Illinois-Urbana Champaign.
The IEEE Computer article is the first-ever decisive manifesto introducing the burgeoning science of social sensing to the entire computer science research community, Kaplan said.
"In light of the growth of social media, it demonstrates the need for a science of social sensing to extract reliable and relevant information from noisy, conflicting and equivocal human generated reports," he said. "It lays out for the very first time the challenges and research opportunities by building upon the successful history of signal processing and sensor fusion."
He also credited the emergence of the field of social sensing within collaborative research as originating from the ARL Network Science Collaborative Technology Alliance.
Their paper addresses how traditional sensor processing techniques convert sensor stream in target identification labels and tracks through classification and multi-target tracking algorithms, for example, Kalman filters. The key idea advanced in the IEEE Computer article is that one can design a new "macroscope" to similarly label and track objects by collecting massive human generated data from social networks; however, to make this feasible challenges in cyberphysical and in social and linguistic spaces must be addressed.
"The cyberphysical space considers challenges of how human may distort a ground truth in light of their own biases and their social contacts and algorithmic methods to estimate and compensate for source reliability and polarization. The social/linguistic space consider challenges for natural language processing methods to interpret text data and understand the context of how and why the report was generated," he said.
The article surveys initial research to address the social sensing vision, but demonstrates the need for a more holistic multi-disciplinary approach that combines social and cognitive models, linguistics, estimation theory, information theory, and reliability analysis, with the goal of putting social media exploitation on well-understood analytic foundations, not unlike fusion of hard data from physical sensors and signals.