For some years, the fields of software engineering and artificial intelligence have been making use of the concept of interacting autonomous agents. Although the term agent has been defined and used in several ways by other scientific groups, for the purposes of software engineering it has been sufficient to start with a minimal definition of an agent, such as the following (Ferber 1999): an agent is a physical or virtual entity that:

The definition corresponds closely to a kind of living organism whose behaviour is aimed at satisfying its own needs and fulfilling its own objectives. Adaptation among such an assembly of agents may occur in two ways: by altering individual characteristics (learning), or as a collective process that brings reproductive processes into play (evolution). We can think of this combination as adaptation that is simultaneously individual and collective. A broad range of agents that are simulated in several different contexts, as described in the second part of this book, are summarised in the Table 1.1.

In his quest to find a general biology, Stuart Kauffman suggests that the biosphere got itself constructed ‘by the emergence and persistent co-evolution of autonomous agents’ (Kauffman 2000: 3). He goes further by suggesting that there may be a fourth law of thermodynamics that roughly states that biospheres maximise the average secular construction of the diversity of autonomous agents and the ways those agents make a living to propagate further. Although this raises a labyrinth of issues about the core features of autonomous agents and their abilities to manipulate the world on their own behalf, it also has important ramifications for the study of human ecosystems. Complex webs of interacting life within ecosystems need to be recognised and addressed in any simulation that attempts to explore alternative co-evolutionary pathways for the future.

Kauffman defines an autonomous agent as a self-reproducing system able to perform at least one thermodynamic work cycle. If true, this means that all free-living cells and organisms are autonomous agents. It also suggests that we lack a concept of propagating organisation. Complexity in organisation rises from the subatomic through the atomic, molecular, cellular, organismal, societal, and ecological system levels. Coevolving autonomous agents may be co-constructing and propagating organisations of work simultaneously at all of these levels. This poses a major challenge for our new kind of science. Nevertheless, more ambitious simulation models such as the Management Strategy Evaluation (MSE) approach for the North West Shelf region (see chapter 14) are beginning to grapple with several levels of the agent hierarchy.

The time when systemic Dynamical System Modelling (DSM) and atomistic (Agent-Based Modelling (ABM)) approaches were described as inherently conflicting is gone. Pioneering work has proven that these approaches are largely complementary (Carpenter et al. 1999; Janssen et al. 2000). DSM provides an elegant analytical framework to study dynamic equilibrium of a system, to the extent that the global parameters can be made explicit. On the other hand, ABM demonstrates an ability to demonstrate emergent phenomena, but is not a predictive tool. Hence, ABM could help discover some emerging values of system-wide parameters and DSM might provide bounding behavioural domains for agent-based simulations.

In chapter 5, David Newth describes how, from social systems to computer networks, graphs can be used to describe the way in which components in large systems interact. In mathematical terms, a network is represented as a graph, in which nodes represent network elements and edges define relationships between elements. According to the author, this powerful paradigm now needs to be coupled to agent-based approaches in order to refine our understanding of human ecosystems.

A comprehensive series of research tends to illustrate how social network structure can be used to determine the characteristics of individual actors, and how social network structure and associated dynamics can modify the behaviour of these actors (Borgatti and Foster 2003). But much remains to be said about the influence of autonomous and partly rational actors on the structure and evolution of these social networks. Here lies the crucial question of how social coalitions and factions form and dissolve over time. Coming back to our previous analogy between human body integrity and environmental sustainability, it is a matter of knowing how system structure and fuctions co-evolve in order to maintain the resilience of a given human ecosystem.

In chapter 9, Lisa Elliston and Steve Beare illustrate the impact of such an organic link with their simulation of a pest invasion in Northern Queensland. Ways in which wheat farmers and contractors interact shape the network over which Karnal bunt, a disease of wheat, diffuses and contaminates various parts of the countryside. Conversely, the presence of contaminated areas influences the behaviour of local farmers and the strategies deployed by the Quarantine services. In a different context, Pascal Perez and colleagues (chapter 10) describe an equivalent co-evolutionary process of interactions between illicit drug markets and law enforcement strategies. So called hot spots move around the urban environment due to endless adaptive strategies deployed by police forces and crime syndicates.