Order this book
The book is organised in two parts: Part I is dedicated to introduce the main concepts and methods employed in the two domains to be bridged in the book, i.e. behavioural sciences and in CSS. Part II presents methods to be employed in the behavioural computational social science.
After the short introductory Chapter 1, Part I presents the integrative approach captured in the book: Chapter 2 introduces the explanation of social phenomena as a major objective of CSS. The key point stressed is “causality”, i.e., the aim “to identify the complex set of causes that generate social phenomena” (p. 12). Related methods of CSS (like agent-based models) are discussed.
Chapter 3 turns to the behavioural sciences: In particular, concepts and methods for empirically observing the behaviour of individuals are presented, which is regarded as a prerequisite to explain individuals’ behaviour and to integrate behaviour as set of causes in explaining social phenomena.
Chapter 4 comprehensively deduces reasons for an integration of behavioural sciences and CSS which, in a nutshell, means to join “a modelling paradigm with empirical analyses” (p. 43): For instance, for agent-based modellers an integrated approach promises to capture empirically founded behaviour. However, it is argued that the behavioural social scientist could benefit from empirically sound models too, for instance, by the capability to investigate dynamics and long-time horizons which are hard or too costly to capture in experiments with real individuals. The chapter ends with the conclusion that beyond the contributions of Behavioural CSS to its root domains an integrative approach per se helps to “overcome ‘false’ disciplinary boundaries” (p. 55).
Part II takes a more practical and technical perspective. Particular focus is put on methods for integrating tools originated in the two domains into a unified approach of behavioural sciences and CSS. Part II is organized in four chapters of which the last presents a comprehensive example application of the methods presented in the preceding chapters 5 to 7.
Chapter 5 introduces the so-called “behavioural agents”: this term captures models of individuals’ behaviour where the agents collect information from the environment (input) and transform this information into actions (output) – in a completely transparent and empirically salient way. Thus, model calibration according to empirical data is extensively discussed. However, the key feature of “behavioural agents” is that the explanation of their behaviour is left outside the model.
In contrast, Chapter 6 characterises key features of so-called “sophisticated agents”: employing sophisticated agents in a model allows for capturing the driving forces (causality) for the evolution of behaviour. Hence, modelling sophisticated agents allows for studying the causes of agents’ behaviour which are mapped by models of cognitive processes or structures. For this, the chapter further on discusses key tools and algorithms like reinforcement learning, genetic algorithms and learning classifier systems.
In Chapter 7 methods for mapping and analysing social networks are presented including remarks on social network analysis (SNA) and how to generate empirical data on social networks relying on the potential of “Big Data”. The chapter also comprises suggestions on how to generate models of social networks for simulation purposes and on spatial models of interactions.
Major tools presented in chapters 5 to 7 are integrated in an illustrative application study presented in Chapter 8: The example relates to the well-known social dilemma of voluntary public goods provision. After presenting key theoretical and experimental results of prior research the key features of the tools presented before are employed: behavioural agents (characterized by parameters calibrated according to experimental results), learning agents and different interaction structures (one of which according to “social network analysis”).
The Appendix contains a technical guide to the example model implemented in NetLogo combined with BehaviorSpace (https://ccl.northwestern.edu/netlogo/docs/behaviorspace.html). The code of the model is explained stepwise.
In sum, this book could be regarded as a plea and impressive demonstration for studying (the emergence of) social phenomena based on empirically founded models of individuals’ behaviour. With this, the book argues in favour of overcoming traditional borders between scientific disciplines and of integrating ideas of various disciplines into Behavioural CSS. The book does not only present this unifying approach but boils down a tremendous number of concepts, methods and tools to some relatively few pages (in sum 188 pages). Hence, this may explain why fairly often the descriptions are concerned with classifying the several tools rather than describing them in more detail. Nevertheless, while other sources may be required for descriptions of the tools in detail for purposes of implementation, the book provides a lot of valuable general orientation to the reader which ideas and tools to employ in a certain model.
Finally, it is worth mentioning that the book is part of the “Wiley Series in Computational and Quantitative Social Science” which is of certain interest for JASSS readers. The list of titles published in the series can be found here: http://eu.wiley.com/WileyCDA/Section/id-811877.html
Return to Contents of this issue
© Copyright JASSS, 2017