Infrastructure systems are large scale socio-technical systems, embedded within the society, supporting and enabling a myriad of societal processes. Understanding and shaping their evolution is essential if we are to rise to the challenges of climate change, resource scarcity, population growth and changing geo-political regimes.

We conceptualize infrastructure systems as complex adaptive socio-technical systems, evolving through constant parallel and distributed interaction of heterogeneous social and technical elements. This interaction leads to emergent patterns, manifested in the economic, environmental and technical performance of these systems. Technical capability, institutional and organizational structures, and economic and social reality determine which and how infrastructure systems are deployed. Once a physical structure is present, it in turn shapes how we operate and evaluate current and design future infrastructures. As these systems deliver new services, they shape the social systems that use them, giving rise to new institutions. This process of reflexive downward causation (systems acting as their own causal agents) leads to emergent structures that limit and constrain the evolution of the system itself systems.

Understanding these systems requires multiple formalisms and disciplines and an explicit and rational account of their emotion- and value-laden nature. Moreover, the exact prediction of the future states of the system is strictly impossible, due to the intractable nature of evolutionary processes.

Human intuition is a poor guide when dealing with such complex evolving systems spanning large physical, temporal and social scales. However, we can use models and simulations to enhance our thinking, turning computers into what Steve Jobs has called “bicycles for the mind”. One of the roles of models is to provide insight into the real world by creating simplifications. However, when dealing with complex adaptive systems, there is a lower limit to how simple these models can realistically be. Following Ashby’s “Law of requisite variety” – which requires that a model of a system must have at least as much internal variety as the system that it is trying to represent – we claim that we need modeling methods and tools that themselves are complex evolving entities.

In this paper we describe (1) an empirically tested evolutionary model development method, (2) an ecosystem of tools and (3) sets of best practices that meet this requirement. The method involves evolving a series of locally optimal agent based models based on socially constructed formal ontologies, and features stakeholder-driven mechanism generation and distributed collection and curation of data. Use of internet tools ensures maximum transparency of assumptions, mechanisms and data used in modeling, enabling a stakeholder-oriented, iterative and collective sense-making process.