Encoding Protest Duration In An Agent-Based Model As Characteristic Phase Transitions

Protests and civil unrest events carry high societal impact and are examples of complex interactions and collective behavior. Agent-based modeling (ABM) is one approach to simulating emergent phenomena seen in protests by leveraging individual behaviors derived from theory or observation. The utility of these models is immense; however, techniques for understanding the theoretical consonance of these complex aggregate behaviors is missing. For example, protest dynamics can range from small-scale, more frequent protests to nation-wide events that can last several days. This work focuses on characterizing the duration between population level shifts from protest and non-protest states using a characteristic, stylized, network model of individual interactions. The model encodes the population (macro-)level protest states as a well-known phase transition (double-well potential function) dependent on individual (micro-)level interaction characteristics. The model is fit to the ABM in distribution and the process of rioting is captured by a reduced set of parameters.