Discovering Agent Models using Process Mining: Initial Approach and a Case Study

Agent-based modeling is widely used for modeling and simulation of self-organizing sociotechnical systems that are composed of distributed autonomous agents. In these systems, macro level behaviors emerge from local micro level behaviors of agents that follow rules and interact with each other and the environment. Although the individual agents' behaviors are typically described by sets of simple rules, the many interactions, heterogeneous populations, and complex topologies can make it challenging, or even impossible, to predict or steer the emergent behaviors beyond micro levels. Hence, the actual behaviors of such systems are generally hard to know beforehand, and they need to be observed to extract realistic models. In this paper, we propose a proof-of-concept approach to discover agents' underlying models from log data generated from their behaviors, utilizing process mining. To conceptualize and demonstrate our initial approach, we use an illustrative example of the popular Schelling's model of segregation. Our findings provide encouraging initial evidence on how agent models can be extracted utilizing process mining techniques.