The Handbook of Agent Based Modelling

This book presents a detailed technical exploration of complex adaptive systems and agent based modelling. Complex systems are found ubiquitously in both naturally occurring and man-made contexts. The flocking and swarming of animals, spontaneous emergence of traffic jams, financial markets and modern economies are all examples of complex systems composed of large numbers of heterogeneous, interacting agents. The multiplicity of microscopic interactions between these agents gives rise to the endogenous emergence of macroscopically observed phenomena (such as market crashes and economic recessions) that is at times counter-intuitive and surprising. The most practical method for the modelling, analysis and characterisation of complex systems is agent based modelling. An agent based model builds a computational (algorithmic) representation of a system from its microscopic, heterogeneous constituents (the agents) and their interactions with one another. From these simple building blocks emergent, non-equilibrium behaviour can endogenously arise that is both unexpected and non-intuitive. The emphasis throughout the book is on introducing practical examples of complex systems and showing how they can be modelled and analysed using agent based techniques. The goal of the book is to highlight the power (and weaknesses) of agent based techniques for analysing complex systems and to equip readers with the tools necessary to undertake their own explorations in this endlessly fascinating subject. The first part of the book describes what complex systems are and how agent based modelling is an appropriate framework for analysing them. A detailed discussion of the factors one must consider when designing, calibrating and validating agent based models is also presented. Validation and model risk management in particular is a key concern when designing and using agent based models due to the freedom one has in specifying a model's characteristics. The second part of the book introduces a number of practical examples of agent based models across a range of disciplines (chosen to highlight the most important characteristics of complex systems). Classic models including the Ising and Schelling models, Kuramoto’s model of collective synchronisation, ant dynamics, litter collection by termites, the Game of Life, forest fire models, the Nagel-Schreckenberg model of vehicular traffic and the Bak-Tang-Wiesenfeld sandpile model of self-organised criticality are introduced and analysed. Further examples of more complex agent based models characterising swarms of agents, evolutionary dynamics (prey-predator ecology, the Daisyworld model, the Turing model of morphogenesis and the CCC model) and models of agents with bounded rationality competing for limited resources (the Minority Game) are also discussed. The final part of the book considers advanced, contemporary applications of agent based models. Agent based computational economics is discussed and macroeconomic agent based models analysed in detail. These models display a very rich phase behaviour providing significant insight into the microscopic origins of emergent macroscopic behaviour. We also briefly consider how one can integrate macroeconomic agent based models with models of climate dynamics to capture the joint co-dependent evolution of economic-climatic systems. Several chapters are then devoted to considering how one can build agent based models of financial systems. Starting from a detailed discussion of the structure of the financial system as a complex network, we consider models for financial stability and the impact of high-frequency trading on price formation, liquidity and the prevalence of flash crashes. Looking to the future there exists the possibility of integrating agent based models with techniques from machine learning utilising big data. This opens up the possibility of being able to build ever more realistic representations of agent behaviours.