Abstract for: Building More Robust System Dynamics Models Through Validation

Today human and nonhuman societies face many complex problems that require understanding, not merely through individual relationships, but entire systems so as to recognize unintended positive or negative consequences. This paper presents an overview of system dynamics, from modeling to critiques, as well as an explanation of where the structural origins of behaviors arise. The discussion includes white box vs. black box models, what makes models useful, and how to improve the quality of models via structural validation, as well as how to make use of the power and adaptability of machine learning. To improve the quality of models, we argue that feedback loop dominance profiles assist to illuminate the underlying causal structure, thus clarifying the feedback-based explanation of dynamics. Methods of feedback system neural networks and feedforward artificial neural networks are offered as ways to do exploratory data analyses which can help to ease model conceptualization processes. From industrial, municipal, global dynamics, and their limits, the fields of system dynamics and machine learning offer new and emerging insights moving pragmatist inquiry forward in the 21st century.