Classically, science has relied on two main approaches: experimental (focusing on the observation and analysis of data) and theoretical (emphasizing mathematical analysis of the logical consequences of basic axioms). Over the past 50 years, the proliferation of inexpensive computers has led to increased use of a third approach, simulation, that combines mathematical precision (via a computer program, which is a formal mathematical object known as a partial recursive function) with principles of experimental design and data analysis from classical experimental science. This new way of doing science is particularly important in domains such as social science, where the complexity of realistic mathematical models frustrates analytical solution, and where practical and ethical barriers preclude thorough experimentation.

Watts and Gilbert are major contributors to the field of computational social science. Gilbert, in particular, is the long-standing editor of the standard journal in the field, the Journal of Artificial Societies and Social Simulation. What sets this contribution apart from general works on social simulation (such as Gilbert’s standard text [1]) is its focus on a single social phenomenon, in this case the process of innovation. This focus has at least two consequences.

First, it allows them to demonstrate the importance of a fundamental principle of simulation science, the comparison of different modeling approaches to the same problem (a process that Robert Axtell has called “docking”). A critical question in simulation is determining which aspects of a model’s behavior are reasonable reflections of the reality being modeled, and which are simply artifacts of the modeling method. Approaching the same problem with different tools is an essential technique for teasing apart these two influences.

Second, the focus on a single phenomenon makes the book a contribution to the study of innovation, not just a treatise on simulation. While students of any aspect of social science will benefit by the insights into social science simulation that the authors present, the book will be especially attractive to researchers in innovation, even those who might not have considered simulation as an important tool up to this point. In this spirit, the authors offer simulations for a wide range of innovation theories, including both simulations for theories that previously were not computational, and attempt to replicate previously published simulations of innovation. (In the latter case, their effort highlights the inadequate documentation of some published models, a fault that the authors avoid by making executable versions of all of their models available on the web.)

Setting aside the introductory and concluding chapters, the book’s six main chapters introduce successively more complex models, exploring three themes: how innovations diffuse, how they are generated in the first place, and how multiple innovations interact as they diffuse through a shared space. These three themes are interwoven throughout most of the chapters.

Chapter 2 exhibits a range of diffusion models, including three different computational approaches to an epidemic model: system dynamics, discrete-event simulation, and agent-based simulation. It also discusses the probit model and models based on competing technologies, such as order models, stock models, and network externalities.

In chapter 3, the authors introduce a structured environment (a social network) through which innovations diffuse and interact, and show the impact of path dependencies. Chapter 4 uses insights from the study of complex adaptive systems, such as Kauffman’s NK model, to study the tradeoff between exploration and exploitation in innovation networks, and gain insight into collective learning among innovators.

Models of innovation in science, where bibliometric data offers a much richer source of real-world calibrating data than is available in some other areas of innovation, are the focus of chapter 5. Chapter 6, studying different constraints on diffusion, is based on a paper published by the authors in 2011. It brings together a wide range of constraints, including some that have already been discussed in earlier chapters.

Chapter 7, on technology evolution and innovation networks, begins with the work of Silverberg and Verspagen, and then weaves in other work on this theme, showing how simulation can support the scholarly conversation on a focused topic.

This volume is an important step in the development of simulation as a third mode of science, alongside theory and experiment. It constitutes an impressive argument for the contribution of simulation to a specific area of social science, while highlighting a number of challenges (such as model docking, model documentation, and model calibration) that need to be taken more seriously in the community. It will be of interest both to students of innovation, and to users of simulation in other areas of social science and beyond.