GENNET: Advanced System for Synthesis of High Fidelity Social Networks to support SHIELD

Our objective is to provide a semi-automated system to generate and verify scalable, high fidelity, multi-dimensional, social networks with realistic distributional, temporal and spatial characteristics as networks and as event sequences. This enables the user to generate networks varying in size, node attributes, geo-temporal characteristics, consistency with other networks, and reflection of known subpopulations, in formats used by standard network analytic and temporal network packages and algorithms. Generative algorithms will be informed by theory and empirical data regarding social media choice, network generation, structure and message content. We build synthetic networks with limited message content using a multi-algorithm approach employing subpopulation identification, empirical seeding, constraint satisfaction, multi-network alignment and topic filtering. A secondary feature of the proposed system will be the ability to layer onto the synthetic network hidden-covert networks or purposive networks designed to accomplish some task. Veridicality will be assessed on multiple dimensions; network density, distribution of centrality measures, temporal characteristics, spatial characteristics, and evolutionary properties. Deltas between temporal slices will be validated for key actor and topology at the subpopulation and network level. Synthetic and real networks varying in size, media usage, and member activities will be compared at the node, group, motif, and topology levels.