Reviewed by
Petra Ahrweiler
University of Hamburg
This promising initiative contains a collection of disciplinary applications for agent-based modelling. It is divided into four sections. The first section following the introductory chapter of the editors is dedicated to socio-economics. The second and largest section deals with population and demography. The third section introduces applications from ecology and environment while the last section which is titled 'general aspects' tackles trans- or interdisciplinary issues. Presenting and combining different approaches, the aim of the book is to describe the motivation for setting-up agent-based models and for studying emerging patterns in complex adaptive systems. The editors´ introduction (Billari, Fent, Prskawetz and Scheffran) is well written and contains much useful information for scholars not familiar with agent-based modelling. However, there is not much reflection of current discussions within the agent-based community, but instead some quite clear-cut positioning concerning fundamental topics. This e.g. refers to the maintained easy blend between deductive and inductive principles in agent-based modelling, i.e. the epistemological claims of social simulation, which is much discussed in the community.
The editors identify agent-based modelling as an emerging field in studying complex adaptive systems. They rightly assume that the micro level interactions between heterogeneous agents which lead to macro-level system behaviour form the primary modelling challenge in the social sciences today. This emphasis is particularly illuminated by the book and stresses the most important advantage of agent-based modelling.
The social sciences are concerned with models containing agency and structure. Most targets are complex, involving many parameters, many relations, and much feedback. Individual action formation, individual decision behaviour and the strategies of intelligent social actors are generated and permanently changed in context-specific interactions and, via feedback, with other actors each taking into account experience-based expectations and the basic contingency of potential actions and actors. Such behaviour of individuals on the micro level leads to emergent social phenomena on the macro level, including again all kinds of feedbacks between these two levels. Models often describe evolutionary processes containing high dynamics and uncertainty. Agent-based modelling is centred around emerging structures and the interactions of agents specifying the respective micro-macro-link (see especially the contribution of Conte, Paolucci and Di Tosto). The whole book can be read as a clarification and illustration of this point following the editors' introductory integration of the book's contributions into their sketched framework.
Contrary to insights from complexity science and promoting linearity and mono-causality, there is still some belief that the social sciences are in an immature state of pre-science, and that we simply have not yet discovered the simple laws helping us to predict social reality. All contributions of this book present evidence for an opposed view. Using agent-based models allows us to find "the trade-off between simplicity and abstracting in modelling, and taking into account the complexity of [...] reality" (Pyka and Grebel). Pyka and Grebel deal with the fact that socio-economic phenomena are located in contexts and are subject to qualitative change: an agent who is exposed to certain strategic behaviours of other agents is located in a technological, regulatory, political, economic context, has a certain experience and history with suffering strategic behaviour, is situated in a certain competition/cooperation structure with other agents, his/her knowledge stock is changing over time etc. These characteristics influence and shape entrepreneurial behaviour and, therefore, the evolution of industries. Pyka and Grebel successfully show that the socio-economic realm bears a clear elective affinity to agent-based modelling as the most adapted modelling strategy.
The agent-based approach allows the representation of heterogeneous agents that have individual and varying stocks of knowledge moving and interacting within and with a permanently changing environment where changes are linked to agent action. Using this approach, we are able to model uncertainty, historicity, environmental change, effect of failure on the agent population, agent learning and agent co-operation. Ecological and environmental research questions seem to be a natural area of application for this kind of modelling as is clearly shown by the contribution of Gebetsroither, Kaufmann, Gigler and Resetarits.
The demographic chapters study in detail what happens when, favouring representations close to reality, more parameters are considered in modelling (e.g. Aparicio Diaz and Fent): it rises the explanatory and predictory power of the model (the same observation applies for the contribution of Edmonds later in the volume). This insight generates many possibilities for future research, e.g. to use empirically-grounded agent-based models for generational accounting where general equilibrium theory have not provided consistent results.
The book manages to show the need of various social sciences to describe and understand complex phenomena, where agency and structure is involved and intertwined. For some of the most important processes empirical research is not enough or not even possible. Agent-based simulation offers to produce continuous and dynamical data flows based, where necessary, on empirically-grounded and history-friendly models. This allows for a procedural representation of social science theories with high explanatory power and for scenario modelling which is of some use for strategic decision making and policy issues. This book shows that it is possible to promote this motivation for agent-based modelling through a wide range of disciplinary applications.
Return to Contents
of this issue
© Copyright Journal of Artificial Societies and Social Simulation, 2007