Navigation – Plan du site

AccueilNuméros9-2Revue des livresComptes rendusSteven F. Railsback and Volker Gr...

Revue des livres
Comptes rendus

Steven F. Railsback and Volker Grimm, Agent-Based and Individual-Based Modeling. A Practical Introduction

Simon Scheller
p. 407-413
Référence(s) :

Steven F. Railsback and Volker Grimm, Agent-Based and Individual-Based Modeling. A Practical Introduction, Second Edition, Princeton: Princeton University Press, 2019 (Second edition), 340 pages, ISBN 978-069119083-9

Texte intégral

Afficher l’image
Crédits : Princeton University Press

1You have probably encountered the acronym ABM before, but in the unlikely event that you have not: Agent-based modeling (which is what ABM stands for) is a formal modeling technique in which complex (social) systems are represented on the level of individual agents and their mutual interactions, seeking to identify by means of simulation analysis how macro-level patterns emerge from agent interactions.

2In recent years, agent-based modelers (such as myself) have become eager and proud to proclaim that their once nerdy niche methodology has moved closer to the mainstream of scientific research in numerous fields. One might argue about the degree to which this is true for different fields. Yet, papers using ABM are increasingly published also in mainstream science journals, and publications range across a large variety of disciplines (Macal, 2016). Economics is no exception in this regard (see e.g. Tesfatsion and Judd, 2006; Janssen and Ostrom, 2006; or more recently Geanakoplos et al., 2012; Klimek et al., 2015; Tesfatsion, 2018).

3Not just advances in computational infrastructure account for this growing success. While ABM can still be seen as an evolving research program, important attempts towards unifications and standardisations in model analysis and presentation have been made. These attempts center around issues such as robustness and sensitivity analysis, conceptual tools like the so-called ODD (for: “overview, design and details”) protocol or the modeling cycle, or the use of NetLogo as a widely shared software platform for implementing ABMs. If you don’t know any of these: Do not worry, we will get to them.

4While such standards usually emerge at the frontier of current research enterprises, teaching ABM to students and junior researchers constitutes an important aspect for the adaptation and the widespreadedness of such standards in a larger community. In this article, I discuss Railsback and Grimm’s Agent-Based and Individual-Based Modeling (henceforth abbreviated R&G’s ABIBM), which has become a standard introductory textbook for aspiring agent-based modellers since the publication of its first edition in 2011. A quick and certainly highly unreliable Google search for “Agent-based Modeling syllabus” evidences this, as six out of the first ten syllabi (which include literature at all) use R&G’s book as a primary resource. With now its second edition out in 2019, a closer look at ABIBM is warranted.

5Before discussing ABIBM, let us ask: Why should economists and other social scientists even care about ABM? First as already argued, there is an increasing chance for having to engage sooner or later with research that involves ABM. Second and more importantly, ABM research promises to illuminate micro-macro relationships in complex social systems by modeling possible causal mechanisms explicitly. This helps in uncovering, understanding and explaining mechanisms of social dynamics, even in cases where data are hard or impossible to obtain. In this respect, the ABM approach is diametrically opposed to other data-driven methods, and thus promises a highly useful complementary tool for social science research in many different areas.

6R&G intend ABIBM as a guide to ABM at advanced undergraduate or graduate-student level, both for university class teaching and self-study. Most importantly, the authors promise what one could call an “overall package” for using ABM to address scientific questions, including technical aspects of implementing, analysing and presenting ABMs as well an insight into ABM’s methodological, theoretical and conceptual aspects (xii). In short: Work through ABIBM, and get started as an ABM researcher. Let us see how the book lives up to this goal.

7As its subtitle promises, ABIBM really is a practical introduction to ABM. Reading the chapters is only one part of working with this book. Already on page 16, readers are invited to get started with NetLogo, and first code examples follow merely a few pages later. Each chapter is accompanied by a substantial set of practical (programming) exercises. While these start easy, they become challenging rather quickly and require continued effort to get through. To assist learners, documentation is very detailed for modeling and coding examples and exercises throughout the book. Additionally, plenty of supplementary material is provided online, and there exists an online platform for interaction with one’s fellow ABM apprentices. The exercises also frequently feature examples from the NetLogo model library, which somewhat alleviates the burden of active programming before being able to look at and work with ABMs scientifically. This underlines one crucial characteristic of ABIBM ’s approach: Theory and practice consistently go together, so you get started thinking right away about models, how they are used and what matters for good modeling.

8Part I of ABIBM mainly caters to the primary needs just described: It gives readers an idea of what modeling is about and how it is done. Most importantly, R&G’s well-known modeling cycle (8) describes the practice of ABM as an iterative process: Formulate a research question, derive hypotheses, implement a model, analyse it. By going through this cycle over and over again, each of these aspects is continuously refined. Importantly, this problem-oriented spirit of using ABM carries through the entire book, and advises readers to develop habits of mind that are essential to ABM and science more generally.

9Part I also invites the reader to implement her very first model in NetLogo: The butterfly-hilltop-model is a simulation of how distinct movement corridors emerge when butterflies search for hilltops. When I started working with NetLogo as a young PhD student (using ABIBM ’s first edition), I still recall my own surprise after seeing butterfly highways emerging after having written only a few lines of very simple code describing very simple movement procedures. This not only vividly illustrates ABM’s potential to explain “much from little”, but also embodies many people’s continued fascination with ABM.

10R&G’s maybe most well-known methodological contribution is their ODD-protocol as a framework for describing ABMs based on characteristic model components (35). ODD stands for “overview, design and details”, which constitute their three proposed steps of model description. The protocol starts by requiring a general overflow and purpose, followed by increasingly detailing distinct model components. Certainly, not all ABM publications explicitly follow the ODD-protocol and its many subcategories. Yet, the fundamental principle of spelling out a model’s components in a structured and standardised way casts into a concrete form how formal modeling induces transparency. This is equally important for research communication and replication. Such a protocol also forces the modeler herself to be clear, consistent and thorough in her thinking and writing.

11Part II introduces and discusses several common components of ABMs in detail, leading the reader through the “engine room” (233) of ABM. Each chapter provides a closer look at a standard ABM building block, what it can be used for, and how it is implemented. Apart from some components that are essential to most ABMs (like sensing, scheduling or adaptive behaviour), here you can pick and choose what you want to focus on depending on your interests and needs. ABIBM is modular in this regard: You need not read all chapters and do all the exercises to learn how to use ABM for your purposes. Further, different components are mostly presented on the basis of different models. Hence, wrapping one’s head around each model demands substantial resources in terms of focus, time and effort, especially when one wants to do all the exercises thoroughly. While doing so certainly pays off in terms of acquired skill and experience, those readers with a more restrictive schedule will face trade-offs at this stage.

12Part III turns the reader’s attention towards applying ABMs to the study of real systems. It does so mainly by elaborating on the core challenge which every ABM researcher faces: How do we know that our models are appropriately “realistic”, and how can we learn from them? In response to this challenge, R&G propose and elaborate their principle of pattern oriented modeling (234): As an agent-based modeler, we should design our models so that they reproduce not just one pattern we are interested in. Instead, we should evaluate to what extent they fare well on a multiplicity of such criteria, thereby comparing larger ensembles of models. On that basis, ABM can also be used for theory development: If we analyse our models in this pattern-oriented fashion, we can try to make deductions about, say, how individuals behave in certain situations, or what cognitive processes we can reasonably ascribe to them.

13There is an ongoing debate within the ABM community about the issue of learning from very simple models, which is particularly prominent in the social sciences. Notably, while most of R&G’s examples are taken from ecology, the two simple models the authors engage with are Duffy (2006)’s minimal intelligence trading model (R&G, 253f), and Schelling’s (1971) segregation model (R&G, 286ff). The former model supposedly shows that an equilibrium price emerges quickly even on a market with minimally intelligent traders. Schelling’s model, simulating the residency decisions of agents of two distinct groups, supposedly shows how spatial segregation can emerge already from rather moderate preferences for neighbor-similarity. I deliberately use the word “supposedly” here, as there is no general consensus on what these models really show. Arguably due to their very abstract nature, both models stand somewhat apart from the examples taken from ecology.

14ABIBM ’s moderate focus on ecology is not necessarily a hindrance for scholars from other disciplines such as the social sciences. These models are sometimes more easily accessible since it is clearer what their components represent in the real world. In other fields, however, agent-based models can be employed somewhat differently at times. In the social sciences and philosophy for instance, models are often taken to provide “how-possibly” explanations, claiming not to provide an actual explanation for a certain phenomenon, but only that some micro-mechanism would be sufficient to generate it. The methodological discussions surrounding simplicity, descriptive accuracy and the fundamental mode of learning from ABMs in different fields are lively and multifaceted, see e.g. Grüne-Yanoff (2009); Lustick and Miodownik (2009); Reiss (2011); Fumagalli (2016); Reutlinger et al. (2017) as a starting point. Tracing these discussions obviously lies beyond what can be sufficiently conveyed by an introductory textbook such as ABIBM. Thus, R&G’s pattern-oriented modeling account should be considered as one voice in a broader debate. For ABM beginners, ABIBM nonetheless provides a good starting point for a worthwile broader exploration into these matters and offers a great introduction to ABM also for scholars from those other fields.

15Part IV introduces the most important concepts and strategies for model analysis. R&G’s basic idea consists in treating models as virtual laboratories for simulation experiments, where a range of scenarios can be explored systematically. Again, many practical examples are used to illustrate core tools, such as performing sensitivity and robustness checks with behaviour space, or producing contour plots to present one’s results in an easily accessible way. Standardised tools in this category are of central importance to the credibility of ABMs in mainstream research domains. Well-performed sensitivity, robustness and uncertainty analyses are the agent-based modeler’s answer to the question one frequently gets asked at non-ABM conferences: “How do we know your little toy model really holds?” Oftentimes, this criticism is not phrased as such, but rather comes in the form of “Have you tried modeling X?”, or “What happens if you change Y?”. Even though it is clear that one cannot pre-emptively answer all such questions, following the principles described in part IV provides a solid guideline for researchers as to what can reasonably be done to establish a model’s credibility.

16In summary, if properly worked through, ABIBM offers everything one needs to get started working with ABM scientifically, from basic techniques and strategies of model building, to the general ideas and core tools of how to do science with ABMs. It does so in an intuitive and accessible way, using numerous well-documented examples and exercises. All things considered, ABIBM is a great didactic tool, and I can recommend it wholeheartedly as a basis for both university courses and mavericks who want to get started all by themselves. For using ABIBM in university courses, it should be noted that one should take some time to carefully pick and choose the topics for a one-semester class so as not to overburden students with too much input, or failing to cover all areas sufficiently. This is especially true for picking exercises and examples, as they sometimes require quite some time to familiarise with. For maverick learners, it should be noted that patience and a certain amount of tolerance for frustration is demanded, as even a very good textbook can only provide guidance up until some point. In either case, you only really learn to use ABM if you get started with your own project. That is the real challenge, but can be very rewarding eventually.

17In comparison to the book’s first edition, not just the cover illustration has been updated from orange to blue. Most importantly, the second edition is based on NetLogo 6.0.4. This matters, as especially the transition from NetLogo 5 to 6 has brought some significant changes with it. Some functions and commands were added, some work differently than before, which often results in errors when models programmed in earlier versions are run in NetLogo 6. If you start with NetLogo as a novice, using version 6 is a no-brainer. Still, the transition is worthwhile also for experienced users, and certainly in order to avoid confusion in teaching. Furthermore, as is necessary for a rapidly evolving field like ABM, the second edition incorporates new sources, and keeps track of the field’s most important developments. The essence of ABIBM, however, has remained the same.

18Steven Railsback and Volker Grimm’s personality as enthusiastic ABM researchers perspires from every pore of ABIBM. They deserve much credit for the didactical approach to learning ABM which this book embodies. It is certainly no overstatement to say that the ABM research community would be much smaller without ABIBM as a starter-pack for young researchers. At the same time, the authors deserve credit for the conceptual advances they have brought to ABM as a research practice, which are all contained in this book. It remains to be seen how the field of ABM keeps on developing in the future, and whether or not it can truly consolidate as a mainstream research practice. Coaching young scholars towards following a structured and problem-oriented approach to ABM research is crucial in that regard. The other task that remains to be taken up over and over again is to reach out to non-ABM scholars and join forces on the actual research frontier, in order to make agent-based modeling an embedded endeavour altogether.

Haut de page


Duffy, John. 2006. Agent-based Models and Human Subject Experiments. In Leigh Tesfatsion and Kenneth L. Judd (eds), Handbook of Computational Economics: Agent-Based Computational Economics, volume 2, Amsterdam: Elsevier, 949-1011.

Fumagalli, Roberto. 2016. Why We Cannot Learn from Minimal Models. Erkenntnis, 81(3): 433-455.

Geanakoplos, John, Robert Axtell, J. Doyne Farmer, Peter Howitt, Benjamin Conlee, Jonathan Goldstein, Matthew Hendrey, Nathan Palmer, and Chun-Yi Yang. 2012. Getting at Systemic Risk via an Agent-based Model of the Housing Market. American Economic Review, 102(3): 53-58.

Grüne-Yanoff, Till. 2009. Learning from Minimal Economic Models. Erkenntnis, 70(1): 81-99.

Janssen, Marco and Elinor Ostrom. 2006. Empirically Based, Agent-based Models. Ecology and Society, 11(2): 37-50.

Klimek, Peter, Sebastian Poledna, J. Doyne Farmer, and Stefan Thurner. 2015. To Bail-out or to Bail-in? Answers from an Agent-based Model. Journal of Economic Dynamics and Control, 50: 144-154.

Lustick, Ian S. and Dan Miodownik. 2009. Abstractions, Ensembles, and Virtualizations: Simplicity and Complexity in Agent-based Modeling. Comparative Politics, 41(2): 223-244.

Macal, Charles M. 2016. Everything You Need to Know about Agent-based Modelling and Simulation. Journal of Simulation, 10(2): 144-156.

Reiss, Julian. 2011. A Plea for (Good) Simulations: Nudging Economics toward an Experimental Science. Simulation & Gaming, 42(2): 243-264.

Reutlinger, Alexander, Dominik Hangleiter, and Stephan Hartmann. 2017. Understanding (with) Toy Models. The British Journal for the Philosophy of Science, 69(4): 1069-1099.

Schelling, Thomas C. 1971. Dynamic Models of Segregation. Journal of Mathematical Sociology, 1(2): 143-186.

Tesfatsion, Leigh. 2018. Electric Power Markets in Transition: Agent-based Modeling Tools for Transactive Energy Support. In Cars Hommes and Blake LeBaron (eds), Handbook of Computational Economics, volume 4, Amsterdam: Elsevier, 715-766.

Tesfatsion, Leigh and Kenneth L. Judd (eds). 2006. Handbook of Computational Economics: Agent-Based Computational Economics, volume 2. Amsterdam: Elsevier.

Haut de page

Pour citer cet article

Référence papier

Simon Scheller, « Steven F. Railsback and Volker Grimm, Agent-Based and Individual-Based Modeling. A Practical Introduction »Œconomia, 9-2 | 2019, 407-413.

Référence électronique

Simon Scheller, « Steven F. Railsback and Volker Grimm, Agent-Based and Individual-Based Modeling. A Practical Introduction »Œconomia [En ligne], 9-2 | 2019, mis en ligne le 01 juin 2019, consulté le 04 juillet 2022. URL : ; DOI :

Haut de page


Simon Scheller

Munich Center for Mathematical Philosophy, Ludwig-Maximilians Universität Munich

Haut de page

Droits d’auteur


Creative Commons - Attribution - Pas d'Utilisation Commerciale - Pas de Modification 4.0 International - CC BY-NC-ND 4.0

Haut de page
  • Logo Association Œconomia
  • DOAJ - Directory of Open Access Journals
  • Revue soutenue par l’Institut des sciences humaines et sociales du CNRS
    CNRS - Institut national des sciences humaines et sociales
  • OpenEdition Journals
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search