2019 IEEE Symposium Series on Computational Intelligence (SSCI) | 2019
Who Did This: Identifying Causes of Emergent Behavior in Complex Systems
Abstract
Understanding the causes of emergent behavior in terms of system components, their attributes, and their interactions is critical in order to influence the appearance of desired emergent properties and the reduction of any undesired effects. Several methods have been proposed for the identification of the causes of emergence, but they rely either on significant a-priori knowledge or have only been applied to small scale models. In this paper, we apply two existing methods to two large scale models and discuss their effectiveness in identifying system components of interest and their overall applicability. We employ the Dynamic Cluster Index method to identify candidate agent subsets containing agents most likely to be relevant in obtaining an emergent behavior, and the Interaction Count method to identify agents who interact the most within the model and thus are likely to lead to the appearance of emergent behavior. We discuss the benefits and disadvantages of both methods and identify future areas of research.