Systematic analysis of micro dynamics in agent based simulation

Micro dynamics analysis in agent-based simulation plays an important role when the simulation is used for decision making in social systems. Since the analysis has been carried out through trial and error of analysts, its quality highly depends their skills and it is impossible to find out all significant relationships between macro phenomena and agent behavior. This paper proposes a novel analysis methodology not dependent on individual skills in order to investigate agent behavior systematically. By the methodology, analysts can generate meaningful hypotheses to explain a target macro phenomenon based on the agent behavior. We verified the effectiveness of our proposed methodology with the simulation of checkout area in a supermarket and showed that the method can generate hypotheses including the findings that were obtained by a specialist in agent-based simulation.