Reviewed by
Giangiacomo Bravo
Department of Social Studies, Linnaeus University
Agent-based modelling is becoming a mature social science instrument. This includes, even if it is not limited to, a move from the early highly abstract “thought experiments” to models having the goal to reproduce specific features or geographical areas of the real world with a certain degree of detail. It also means that these models have not only academic implications but are increasingly used by policy-makers. While this should be viewed as a success for the ABM community, it also means that more responsibility is placed on the shoulders of modellers. The challenge indeed is how to develop large scale, complex models that are informed by empirical data both in the definition and the calibration phases. This book addresses this challenge by providing a comprehensive framework that helps modellers in the characterisation and parametrisation of agents (two future volumes in the same series will focus on the modelling of the social network and the biophysical environment).The CAP (Characterisation and Parametrisation) framework, although deriving from a previous work that dealt with social-ecological systems (Smajgl et al. 2011), is a context-independent description of the key steps and decisions that modellers are confronted with while developing empirically-based ABMs. This implies that here readers can find guidelines and lessons that can be applied to any human-related situation, independently of the research question and the method used to gather the relevant empirical data.
The book structure includes a first chapter that introduces the CAP framework and presents a useful decision tree helping researchers to select the right methodological option at each stage of the model development. All other chapters but the last one, which discusses the framework robustness, show examples of researches applying empirically-based ABMs to situations ranging from socio-ecological systems to market and work dynamics, farmer’s choices and power structures. Although not always perfectly, the framework is quite robust and well performing in most if not all these situations. Especially interesting is the range of different research methods that have been successfully employed in the different chapters of the book. These include qualitative methods - such as expert knowledge, participant observation, focus groups and interviews - and quantitative ones - from census data and surveys to behavioural experiments. It is worth noting the effort made by the authors of the empirical chapters in explaining their methodological choices, which will greatly help future developers in their selection of the most appropriate methods to design and parametrize their own models.
To sum up although Smajgl and Barreteau’s volume should not be viewed as a "cookbook" (as the authors highlighted several times), it is at least a nice handbook both for newcomers and expert modellers. Although the CAP framework cannot cover all the needs of a growing community of scholars, it is a progress towards a greater standardisation of the procedures adopted in the model design and parametrisation. More importantly, this can be seen as a further step in the process leading ABMs to become a mainstream instrument both within and outside the academia. This is a desirable outcome not only from a scientific point of view but also to promote governance-styles based on sound and verifiable research rather than on unverifiable administrators’ rules-of-thumb.
References
Return to Contents
of this issue