Activity Matters: Modeling Complex Systems Beyond Complex Networks

Complex systems consist of a large number of strongly interacting elements, often denoted as agents. Complex networks are a specific representation of such systems, where agents are
represented by nodes and their interactions by links. This implies that all interactions are decomposed into binary interactions between any two agents. Hence, complex network approaches have proven useful if the focus is on the topology of sparse interaction networks [1]. Their results can be even misleading if a system instead of a static topology is characterized by bursts of activities, in which the sequence of interactions matters most [2,3]. A second limitation arises if the interactions of agents are not just constrained by timing, but also by energetic conditions to allow such interactions. To cope with these challenges, we need to extend the link-centric network perspective toward a node-centric agent-based modeling perspective. The activity of agents is expressed by a set of driving and driven variables [4]. Driving variables describe the input and transformation of energy, whereas driven variables describe the usage of energy, i.e. energy consuming activities such as link formation and communication. During the talk, I discuss the above limitations of network approaches using different examples. I further demonstrate the applicability of the agent-based framework to various processes of structure formation, from self-wiring of networks to online communication.