Learning What Makes a Society Tick

Abstract

We present a machine learning methodology (models, algorithms, and experimental data) to discovering the agent dynamics that drive the evolution of the social groups in a community. We use a parameterized probabilistic agentbased model integrating with micro-laws to present the agent dynamics. The micro-laws with different parameters present different actors' behaviors. Our approach is to identify the appropriate parameters in the model including discrete parameters together with continues parameters. To solve this mixed optimization problem, we develop heuristic expectation-maximization style algorithms for determining the appropriate micro-laws of a community based on either the observed social group evolution, or observed set of communications between actors without considering the semantics. Also, in order to avoid the resulting combinatorial explosion, we appropriately approximate and optimize the objective within a coordinate-wise gradient ascent (search) setting for continuous (discrete) variables. Finally, we present the learning performance from extensive experiments.