Agent-based Modeling Automated: Data-driven Generation of Innovation Diffusion Models

Abstract

Simulation modeling is useful to gain insights into driving mechanisms of diffusion of innovations. This study aims to introduce automation to make identification of such mechanisms with agent-based simulation modeling less costly in time and labor. We present a novel automation procedure in which the generation of diffusion models is automated. It comprises three phases: (1) preprocessing of empirical data on the diffusion of a specific innovation, taken out be the user; (2) automated inverse modeling of decision models from a decision model library for their capability of explaining these data; (3) policy simulation automatically assesses user-chosen policy interventions in their potential of influencing the spreading of the innovation. We present a working software implementation of this procedure. We applied this tool to data-analysis on the diffusion of a sustainable innovation, water-saving showerheads. The proposed procedure successfully generated simulation models that explained available diffusion data. This provided a proof of concept. Further, it progresses agent-based modeling by providing model validation by design and by enabling detailed bottom-down modeling at the lower complexity of top-down modeling. We believe the proposed approach can widen the circle of persons that can use simulation modeling and better understand and shape innovation.