Reviewed by
Armando Geller
Centre for Policy Modelling, Manchester, UK
Give me two reasons why a book criticising the increasing stance between social science research designs and the actual objects under investigation, i.e. society and related phenomena, should be of any interest to the social simulation community? First, social simulation is innate to the social sciences and not immune against some of the calamities conducted there. Secondly, social simulation is a technical discipline - comparable to orthodox quantitative social science methodologies - and therefore prone to lose contact with its object of research. Hence, Ian Shapiro's core stipulation in "The Flight from Reality in the Human Sciences", namely to conduct research not in a method- or theory-driven manner, but to adopt a problem- and question-driven style in order to describe and understand reality, also applies to the field of social simulation.[1] Reality, he declares in the introduction of his book (pp. 8-9), consists of causal mechanisms; science is potentially the best tool to investigate them and find out about their true nature. Within this critical line of thought Shapiro touches on a number of issues that are of immediate interest to computational modellers of social reality, such as epistemology, the role of evidence and theory and the relationship between them, good model design and empirical testing.[2] Shapiro - together with Alexander Wendt, with whom he wrote the first chapter - adopts the epistemological view of a realist.[3] He claims that realism makes a difference in the conduct of socio-scientific inquiry (for the following see pp. 19-50). Commonsense realists, he notices, accept reality as independent from the mind; scientific realism additionally encloses the idea that unobservable entities and causal mechanisms exist independently of our research endeavours. However, realists acknowledge any observation's theory-dependence, thus rendering any access to reality biased, although (well-established) theories refer to and are constrained by reality. Iterated contribution to scientific knowledge in the realist sense builds upon reasoning that is based on mature theories from observed effects to unobservable causes. Hence, research's starting point is reality, or more likely an aspect of it, with the aim to find and describe social mechanisms (I would add social processes and structures) and to describe and explain reality. Under the epistemological paradigm of realism the process of research is confronted with persistent epistemic insecurity, which is considered by Shapiro as, however, rather a strength than a weakness. Because the objects of investigation are open systems and realism is driven by evidence, the latter consequentially confines research the least possible.
Social simulation and its prevailing cognate methodology's, i.e. agent-based modelling, virtue are not only to analyse complex social phenomena, but also to afford pursuing this in an evidence-driven fashion (Geller and Moss 2008a). Social simulation therefore makes a difference by helping realism, as understood by Shapiro, to make a difference (cf. also Moss 2008). Social simulation enables the researcher to materialise this difference in that real social phenomena can be analysed and eventually understood by social simulation approaches in terms of social mechanisms, processes and (dynamic) structures. However, the more social simulation and its inherent computational methodologies as such begin to be a research design's rational, instead of letting the object of investigation determine the research process, the more social simulation also becomes pathological, analogous to, for example, rational choice. It is misleading to understand every terrorist organisation as a complex dynamic network and to assume that an underlying statistical characteristic of most social processes is a power-law. It is, by contrast, likely that multiple true descriptions of reality exist. Social simulation, understood as a complementary methodological approach to the more orthodox quantitative and qualitative methodologies, can constructively contribute to the development of complementary and alternative descriptions of reality.
The general consequence for the social sciences accruing from inhabiting a realist standpoint, Shapiro argues (particularly in chapter 2 and 5), is to develop and apply close-to-the-ground research designs that do not lose touch with reality. This, of course, does not only affect in-the-air epistemological questions, but also pragmatic methodological issues. Or to paraphrase in Shapiro's words: Being closer to the data and making realistic assumptions is likely to produce better models, i.e. better descriptions and explanations of reality and, consequentially, social mechanisms (pp. 86, 181). Hence, there must be some empirical tests available (or conducted) to scrutinise the data and the assumptions (pp. 83-86). This, of course, narrows a model with regard to its applicability to a particular case, but breaks up its constraining arbitrary domain restrictions and renders a research design susceptible for the requirements to understand reality (pp. 77-78). A further consequence of Shapiro's reasoning, and subject to the idea of multiple realities, is that a model's assumptions might not hold in so many cases, but one can justify why they hold.
I have argued elsewhere (Geller 2006, chap. 3 and 4) that it is important to apply the argument of close-to-the-ground research designs to the domain of social simulation. A construct-valid conceptual model based on evidence is likely to produce better results.[4] However, in social simulation it is not only a model's concept that needs to be validated against reality, but also its output, i.e. qualitative or quantitative data that subsequently can be successfully cross-validated. Cross-validation, as for example described by Moss and Edmonds (2005), is therefore an integral part of social simulation and should not be neglected. Actually it is tantamount to the so important empirical test demanded for by Shapiro that enables inference between the model and the target system (cf. also Edmonds 2000) and that supports avoidance of the flight from reality.
Based on chapter four, Shapiro can be characterised as a foundationalist. He believes in the accumulative nature of the production of knowledge. We can learn in chapter two that he is an advocate of inductive theory-building too. Every time a realistic model helps to better understand a particular real case it adds up to our knowledge about a set of comparable cases. From this set of cases, a theory can be eventually derived that describes a particular social phenomenon in general and that allows predictions with regard to causalities and mechanism in other cases. [5]
Shapiro's book can be read as an attempt to develop and justify an alternative research design in the social sciences, which is conceptualised around the idea of a "problematizing redescription as a problem-driven enterprise" (p. 202). The idea of re-description stems from Shapiro's self-understanding as a political theorist. Political theory, he argues, should acknowledge as its ultimate quest to scrutinise accepted accounts of reality (p. 181). For scientific reasons it is important to correct erroneous, but nevertheless influential descriptions of reality; it is important for political reasons as well, namely when these faulty descriptions shape political reality (pp. 202-203). Such a self-understanding of critically questioning should also be adopted by non-theorists in their conduct of scientific research. A consequence of this is that the synchronous and sequential existence of multiple realities is accepted. As stated afore, this not only has implications for scientific fairness and efficiency, but is also of ethical importance to the social sciences (see in particular chapter three).
This makes a crucial argument for social simulation as well. While social simulation is within its own boundaries a strong, well-defined and established domain, look for example at the number of highly sophisticated opinion dynamics models, it still has to prove its value to mainstream social science. [6] This can only be done by developing more, in number and content, realistic models describing and explaining real social phenomena. While it is obvious to "us" that (agent-based) social simulation models are predefined to depict, describe and explain social mechanisms, processes and (dynamic) structures, it appears to be less acknowledged in the social simulation community that social simulation can make an important contribution in providing complementary and alternative views of socio-scientifically established knowledge and that it is therefore not only a strong alternative, but also a strong complementary tool in the social sciences' aim of accumulating knowledge about our society. For all this - and a lot more - Shapiro's book provides a very well annotated and fascinating, although not always easy to read, argument framework with easy to express practical implications - how easy Shapiro's ideas are to implement and how willing social scientists are to implement them has first to stand its empirical test. But in any case, the following, according to Shapiro (p. 95, footnote 16), holds: "What differentiates the problem-driven researcher from the method-driven one is that the former will endeavour to give the most plausible possible account of the phenomenon that stands in need of explanation, whereas the method-driven researcher will study only those problems, or aspects of problems, to which his or her methods can be applied."
Notes 2Being a political theorist, Shapiro also thematises a number of other issues in his book. Although being more than just noteworthy, I leave them uncommented here, as they bear no direct relation to the point I think Shapiro implicitly makes for social simulation.
3Until now I have not really understood why Shapiro is reluctant to make use of the term "critical realism" as introduced by Bhaskar (1979).
4Geller and Moss (2008b) have defined evidence "as information that is derived from case-studies, empirically tested theories, the high quality media and engagement with stakeholders, domain experts and policy analysts and makers".
5It has puzzled me for a while how to understand Shapiro's notion of prediction, which he uses so often. I guess this is how it should be understood and not as a prophecy of somehow obscurely quantitatively proxied and measured social behaviour (cf. pp. 189-198).
6A reasonable indicator for this fact is the low number of social simulation articles in non-computational and -modelling, mainstream social science journals.
ReferencesEDMONDS, B (2000) The Use of Models - Making MABS More Informative. In Moss, S and Davidsson, P (Eds.) Multi-Agent-Based Simulation. Second International Workshop, MABS 2000. Berlin: Springer. pp. 15-32.
GELLER, A (2006) Macht, Ressourcen und Gewalt: Zur Komplexität zeitgenössischer Konflikte. Eine agenten-basierte Modellierung. [Power, Resources, and Violence: On the Complexity of Contemporary Conflicts. An Agent-based Model]. Zurich: vdf.
GELLER, A and MOSS, S (2008a) Growing Qawm: An Evidence-Driven Declarative Model of Afghan Power Structures. Advances in Complex Systems, 11(2), pp. 321-335.
GELLER, A and MOSS, S (2008b) Modelling Power and Authority: An Emergentist View From Afghanistan. In Edmonds, B and Moss, S (Eds.) Handbook on Simulating Social Complexity. New York: Springer. (forthcoming)
GREEN, D P and SHAPIRO, I (1994) Pathologies of Rational Choice Theory: A Critique of Applications in Political Science. New Haven: Yale University Press.
MOSS, S (2008) Alternative Approaches to the Empirical Validation of Agent-Based Models. Journal of Artificial Societies and Social Simulation, 11(1): https://www.jasss.org/11/1/5.html
MOSS, S and EDMONDS, B (2005) Sociology and Simulation: Statistical and Qualitative Cross-Validation. American Journal of Sociology, 110(4), pp. 1095-1131.
PAGE, S E (2008) Book review. Social Simulation: Technologies, Advances and New Discoveries. Journal of Artificial Societies and Social Simulation, 11(2): https://www.jasss.org/11/2/reviews/page.html
Return to Contents
of this issue
© Copyright Journal of Artificial Societies and Social Simulation, 2008