Agent-Based modeling of protests and violent confrontation: A micro-situational, multi-player, contextual rule-based approach

Abstract

We propose an innovative Agent-Based model of street protests with multiple actors: police agents, three types of protesters (“hardcore”, “hangers- on” and “passers-by”), and “media” agents that seek to witness and publish episodes and situations of violence. Agents have multiple goals and action selection is performed using a “personality” vector together with context rules that provide adaptation. Protesters turn active or violent according to the threshold rule proposed by Epstein, and police agents arrest violent protesters within their move range if they have sufficient backup. The model was applied to a scenario where policemen defend a government building from protesters and described several emergent crowd patterns in real protests, such as clustering of violent and active protesters and formation of a confrontation line moving back and forth with localized fights. Violent behavior was restricted to the initially more aggressive protesters and did not propagate to the bulk of the crowd.