Agent-Based Models. Nigel Gilbert
EC1Y 1SP
United Kingdom
SAGE Publications India Pvt. Ltd.
B 1/I 1 Mohan Cooperative Industrial Area
Mathura Road, New Delhi 110 044
India
SAGE Publications Asia-Pacific Pte. Ltd.
18 Cross Street #10-10/11/12
China Square Central
Singapore 048423
ISBN: 9781506355603
Printed in the United States of America
This book is printed on acid-free paper.
Acquisitions Editor: Leah Fargotstein
Editorial Assistant: Claire Laminen
Production Editor: Gagan Mahindra
Copy Editor: Michelle Ponce
Typesetter: Hurix Digital
Proofreader: Barbara Coster
Indexer: Integra
Cover Designer: Candice Harman
Marketing Manager: Shari Countryman
Series Editor’s Introduction
Almost 50 years ago Thomas Schelling published the first agent-based model (ABM) in the social sciences. It showed how relatively modest residential preferences on the part of individual households could result in marked patterns of neighborhood residential segregation. Since then, and especially recently, applications have blossomed in many fields ranging from opinion dynamics to supply chain management, from language evolution to disease epidemiology, from consumer behavior to urban planning. The second edition of Introduction to Agent-Based Models targets this broad audience. The author, Nigel Gilbert, is one of the founders of computational social science and an authority on agent-based models.
As Professor Gilbert defines it, agent-based modeling is “a computational method that enables a researcher to create, analyze, and experiment with models composed of agents that interact within the environment.” ABMs range from highly abstract simplified models to facsimile models that attempt to replicate real observations. They explicitly link micro and macro levels of analysis, as illustrated by Schelling’s model of households and neighborhoods. Because agent-based models incorporate dynamic interdependencies among the individual agents, the consequences for macrolevel change in these models are emergent, frequently nonlinear, and sometimes surprising, as was the case with Schelling’s model.
Like the first edition, the second edition of Introduction to Agent-Based Models is for beginners. It is suitable as a supplement for undergraduate as well as graduate courses in formal models, simulation, and computational social science; it is also a quick first introduction for any interested social science practitioner. The author carefully defines concepts, outlines the steps involved in planning, building, and reporting ABMs, and includes a helpful glossary. Readers are shown how to use the NetLogo modeling environment, freely available to students, teachers, and researchers worldwide, to build and run a simple ABM. NetLogo helps readers get their feet wet, even those with little background in coding. The second edition of Introduction to Agent-Based Models retains the strengths of the first but updates the material, expands the coverage of verification, validation, and documentation, and addresses some new topics such as the use of ABMs to inform public policy. As was true for the first edition, the goal is to make readers better consumers of published ABMs and to provide the foundation for them ultimately to be creators of these models.
Agent-based modeling is a fast-moving area, especially in breadth of application. In addition, ABMs are increasingly a focus of interdisciplinary collaboration, between social/behavioral scientists from different disciplines (e.g., sociology and geography), between social/behavioral science and natural science (e.g., environmental science), and between social/behavioral science and computer science. Depending on purpose, the rules central to agent-based models can be derived from theory, past empirical research, and/or conversations with local experts. Indeed, ABMs are increasingly used in community-based participatory research. Given these trends, the need for a generally accessible primer is even greater now than when the first edition was published in 2007. This second edition fully satisfies that need.
Barbara Entwisle
Series Editor
Preface
Agent-based modeling is a form of computational simulation. Although simulation as a research technique has had a very important part to play in the natural sciences for decades in disciplines from astronomy to biochemistry, it was relatively neglected in the social sciences. This may have been because a computational approach that respected the particular needs of the social sciences was lacking. However, in the early 1990s the value of agent-based modeling began to be realized, and, since then, the number of studies that have used agent-based modeling has grown rapidly (Hauke, Lorscheid, & Meyer, 2017).
Agent-based modeling is particularly suited to topics where understanding processes and their consequences is important. In essence, one creates a computer program in which the actors are represented by segments of program code, and then runs the program, observing what it does over the course of simulated time. There is a direct correspondence between the actors being modeled and the agents in the program, which makes the method intuitively appealing, especially to those brought up in a generation used to computer games. Nevertheless, agent-based modeling stands beside mathematical and statistical modeling in terms of its rigor. Like equation-based modeling, but unlike prose, agent-based models must be complete, consistent, and unambiguous if they are to be capable of being executed on a computer. On the other hand, unlike most mathematical models, agent-based models can include agents that are heterogeneous in their features and abilities, can model situations that are far from equilibrium, and can deal directly with the consequences of interaction between agents.
Because it is a new approach, there are few courses yet available to teach the skills of agent-based modeling, although the number is increasing, and there are few texts directed specifically at the interested social scientist. This short book introduces the subject, emphasizing the decisions that a social scientist needs to make when selecting agent-based modeling as an appropriate method, and offering some tips on how to proceed. It is aimed at practicing social scientists and graduate students. It has been used as the recommended reading on agent-based modeling for a graduate-level module or doctoral program in computational social science, and it is also suitable as background reading in postgraduate courses on advanced social research methods. It would be a good preparation for any of the textbooks that provide a more in-depth guide to agent-based modeling (e.g., Hamill & Gilbert, 2015; Heppenstall, Crooks, See, & Batty, 2012; O’Sullivan & Perry, 2013; Railsback & Grimm, 2012; Squazzoni, 2012; Wilensky & Rand, 2015).
A knowledge of and experience with computer programming in any language would be helpful but is not essential to understand the book.
The book concludes with a list of printed and Web resources, a glossary, and a reference section. (The glossary terms will appear in bold at first use in the text.) Because the field is growing so rapidly, it has been possible to mention only a few examples of current research and some textbooks that provide more detail on some topics. There is much more that could have been cited if there had been space. In particular, the book mentions only briefly two closely linked areas: network models and game theory models, both of which are covered in much more detail in other SAGE volumes such as Knoke and Yang (2008) and Fink, Gates, and Humes (1998).
A website to accompany the book at study.sagepub.com/researchmethods/qass/gilbert-agent-based-models-2e includes an annotated exemplar model using NetLogo.
Acknowledgments
This book is born of some 25 years of building agent-based models, both large and small, and in domains ranging from science policy to anthropology. What I know about agent-based modeling has benefited immeasurably from the advice and companionship