Computing Social Organizational Structure from Externally Observable Communications Patterns

The coordinated behavior of social groups, such as military units, is mediated by communication between its members. Since the pattern of communication within a group is constrained by the roles and relations among the members, we can infer a great deal about the command structure and social attributes from intercepted radio communications. In operational settings where data are limited, content may be encrypted, and computational resources for data analysis are bounded, approaches are needed that can operate quickly using only "surface features", or externals, of the communications. In this paper, we present a descriptive framework in which the multiscale grammatical structure of social communications is linked to a layered structure of the interaction protocols governing human communications. Using this framework we create a taxonomic classification of communications protocols and their associated observable patterns. We describe a system architecture for the analysis of communications and the inference of social structure that is realtime and adaptive. We discuss the decomposition of transmitters into self-contained interacting groups, the decomposition of transmissions into coherent conversations, and the resolution of these two views of social communications into a structural representation of the organization.