Agent-based Simulation for Complex Social Systems: Support for the Developer

The successful implementation of policies in complex social environments, require a deep understanding of interdependencies between many actors with different perspectives. In order to understand, analyse and design such complex systems, advanced modelling tools are required. Models of many types have been extensively used for researching complex so-cial systems. Modelling is especially suitable when prototyping or experimenting with the real system is expensive or impossible. In one extreme of the scale, text-based models are well suitable to support argumentation and decision-making by groups of people, but are not computable. On the other extreme, mathematical models, including dynamical systems, statistical models, differential equations, or game theoretic models, enable a precise and computable representation of the system, but their complexity rises exponentially as the complexity of be-haviours grow, so that describing complex individual behaviour with equations often becomes an intractable task. In (Ostrom, 1988), simulation is described as a third way to represent social models, being a powerful alternative to these types. Simulation has a high descriptive power and can easily be run on a com-puter. Moreover, simulation can represent non-linear relationships, which are often tough problems for the mathematical approach. In particular, Agent-Based Modelling and Simulation (ABMS) is a power-ful approach for the analysis of complex social systems. This comes from the fact that agent related concepts allow the representation of organizational, so-cial and behavioural aspects of individuals in a society and their interactions. In ABMS, system behaviour emerges as the result of the combined activity of many (tens, hundreds, thousands, millions) individuals, each following its own behaviour rules, living together in some environment and communicating with each other and with the environment. ABMS can support both qualitative and quantitative evaluation methodologies. With the advent of software platforms in recent years, agent-based modelling (ABM) has increased in popularity among social scientists (e.g. Repast (North et al., 2006), Netlogo (Tisue, 2004), Swar