Artificial Agent Society Simulations in an Encounter-Based Normative Action Environment

Abstract

The purpose of the study is to investigate potential relationship between agents' socialness and society's behavior predictability in an encounter-based normative action environment. For this purpose, we proposed a hypothesis and tested it against different simulation setups in the context of classical single source shortest path problem. By the end of simulations, it is observed that the hypothesis holds for both norm internalization and spreading measures when the agents in society have some degree of autonomy. That means for our setup, we conclude that lower degree of socialness results in lower behavioral predictability of the society when the agents have some degree of autonomy.