© Copyright JASSS

JASSS logo ------

Agent-Based Modeling and Network Dynamics

Namatame, Akira and Chen, Shu-Heng
Oxford University Press: Oxford, 2016
ISBN 978-0198708285 (hb)

Order this book

Reviewed by Andreas Koch
University of Salzburg

Cover of book The book Agent-Based Modeling and Network Dynamics by Namatame and Chen begins with a remarkable statement justifying its central concern which is the integration of network science and agent-based modelling (ABM): in order to examine dynamic processes of social networks it would be methodologically and theoretically beneficial to incorporate the spatial context, which has been largely neglected by equation-based approaches of network analysis: “what distinguishes agent-based modeling from conventional equation-based modeling lies in the spatial factor or the geographical specificity” (page 1). The idea of considering (geographical) space a necessary ingredient in social network analysis (SNA) has been convincingly outlined in this book from the methodological perspective; the theoretical argumentation, however, lacks epistemological rigor and plausibility.

Namatame and Chen claim to extend spatial approaches beyond Euclidean geometry by including topological concepts of space into the analysis of social interactions. To this first formal distinction of network approaches a second is added that takes the interdependence of individual acting and social structuring into account. While agent-based models of social networks emphasise the formation process of networks and the diverse qualities of topologies that emerge – from local neighbourhood to global social connections – the network-based agent-based models mainly focus on the impact of networks on the micro-macro relationship of how individual behaviour (and decision-making processes) is embedded into and dependent on social interaction, and how social structure emerges from individual actions.

With this two-dimensional framing of the interplay between ABM and SNA the book presents a convincing delineation of the growingly appreciated co-evolution of network science and spatial conceptualisations in chapter 2. Here Namatame and Chen provide the reader with a historical review from the early checkerboard model of the 1970s to the large(r) topological models representing macro structures in economy and society. The historical background is mixed with the spatial extension – from adjacent cell neighbours in Schelling’s segregation models to complex topologies in macro-economic models of different networks (random, scale-free, and small-world). This is done by expanding on the temporal dynamics topic, including phenomena such as avalanche effect and hysteresis effect. One guiding principle here is to start with game theories of small networks (in size and complexity), which are then “enriched” with social theories and large-scale networks.

Chapters 3 and 4 introduce the two types of models mentioned above, and the remaining six chap-ters are dedicated to the dynamics of networks; firstly from a theoretical point of view (“diffusion”, “cascade”, and “influence” dynamics) and secondly from an application domain perspective (among others, “network risks” and “economic crises”). Namatame and Chen put much effort in describing and explaining the methodological foundations, their strengths and weaknesses. Thus, they succeed in presenting a coherent idea, illuminating the need for modelling (structuring) and simulating (temporalising), and thereby contextualising action with interaction, the individual with the social. The search for adequate entities, relationships, and scales (O’Sullivan and Perry, 2013) has been done with great elaboration.

What fell short in the book is a contemporary, comprehensive, and critical discussion of the theoretical assumptions and consequences utilised throughout the chapters. Most assumptions are dealt with in a manner too simplistic. For example, on page 71 Namatame and Chen write: “How well agents achieve their goals depends on what other agents are doing”. Do agents – do we – really depend on what other agents are doing? And what about them? If, however, we would take this as truth for a moment, how could then “… their collective behavior [be] extremely difficult to predict” (page 68)?

This dependency structure has been maintained and, moreover, linked to another problematic relationship which is the determination by the spatially local: “Agents’ behavior and decisions are generally dependent on the information that they received from their neighbors” (page 114). If this was true, then the entire exploration into social network topologies would be useless. Though spatial context varies between strong determination and extensive fuzziness it remains largely unclear which spatial theory can be applied to which social context. In chapter 5, Namatame and Chen discuss models of diffusion dynamics, starting with a contingent differentiation of innovation models and information models of diffusion processes. They claim that the first can be related to epidemiological models, i.e. innovations spread like infections, but the latter is more complicated due to social influence (page 138). This is stated without further explanation. Indeed, their definition of social influence, presented in chapter 7, is quite poor: “Social influence is defined as a change in an individual’s thoughts, feelings, attitudes, or behaviors that results from the presence of others. Social influence is also defined as a change in an individual’s actions that results from interaction with another individual or a group”. Deeming social equivalent to a set of others and suggesting a perceivable presence appears to be a problematic theoretical approach. The conclusion of “dynamic” diffusion models in chapter 5 remains therefore less dynamic than it is stated by the authors (page 161).

At least two more issues are missing: (i) the geographical and topological mobility of agents is not considered explicitly; (ii) the technological impact on information dissemination as a space-substituting technology is not taken into account sufficiently.

Compared with other books on this topic, for example Understanding Large Temporal Networks and Spatial Networks by Batagelj et al. (2014), the book by Namatame and Chen comes with only a few graphical representations – and none of them in colour – which makes it quite hard to imagine the abstract formulation of methodological approaches.

In a nutshell, readers who are interested in the methodological strengths and challenges of ABM in examining social networks will find in-depth descriptions of techniques linking ABM and SNA, including mathematical explorations. Those who wish to know more about the theoretical foundations behind SNA will very likely be dissatisfied with the theories presented in the book.

* References

BATAGELJ, V., Doreian, P., Ferligoj, A., Kejžar, N. (2014), Understanding Large Temporal Networks and Spatial Networks. Exploration, Pattern Searching, Visualization and Network Evolution. Oxford and Chichester (UK), and Hoboken (USA): John Wiley & Sons .

O’SULLIVAN, D., Perry, G. L. W. (2013): Spatial Simulation. Exploring Pattern and Process. Oxford and Chichester (UK), and Hoboken (USA): John Wiley & Sons.


ButtonReturn to Contents of this issue