Copyright Sociological Research Online, 1997


Halfpenny, P. (1997) 'Situating Simulation in Sociology'
Sociological Research Online, vol. 2, no. 3, <>

To cite articles published in Sociological Research Online, please reference the above information and include paragraph numbers if necessary

Received: 7/7/97      Accepted: 30/7/97      Published: 30/9/97


Suppose you want to explain the pattern of scientific publishing. The way you proceed is as much dependent upon your interpretation of 'explain' as it is upon the topic to be researched. This is because sociology is a multi-perspectival discipline, in which several different interpretations of the nature of explanation continue to compete. These provide different understanding of the goal of research, that is, different understandings of what constitutes an adequate or satisfactory explanation. They also influence the conceptualisation of the explanandum, that is, the identification of what it is that is in need of explanation.

That sociology is multi-perspectival is a commonplace nowadays. However, discussion and dispute continues both within perspectives on their internal specification and between perspectives on their relative merits. Accordingly, the number of perspectives and the location and nature of the boundaries between them remain active questions. And exponents of one perspective argue trenchantly that others are shadows of their own or have no place in sociology's attempts to make the social world intelligible. This is not the place to engage with such arguments, nor to attempt an enumeration of all possible perspectives. Instead, some rather simple characterisations of some major perspectives will be sketched out, with the aim of showing how simulation might be situated in relation to them.


Positivism is the place to start, for it has exercised enormous influence over sociology's understanding of its explanatory task (Halfpenny, 1982). Positivism is often taken as the embodiment of modernity, the vehicle that promised to deliver the fruits of the enlightenment to all disciplined human thought. Carrying such a baggage, positivism is notoriously difficult to define. Nevertheless, a quick characterisation for current purposes is provided by the deductive-nomological (d-n) schema: a satisfactory explanation is achieved when the explanandum is deduced from the combination of one or more laws and a set of initial conditions (Hempel, 1942). Even this simple schema can be extensively elaborated by varying the characterisation of laws that are central to its operation. A key issue in debates about the nature of positivism's laws has been how to retain the explanatory force of the d-n schema, which seems to require of laws that they be more than mere empirical regularities, without making them true by definition, mere conventional calculative devices for deducing conclusions from premises. On the empiricist side, heroic Humeans maintain that laws are induced from instances of empirical co-occurrence, but this seems to give no grounds for predicting future co-occurrences, for what guarantee is there that the future will follow the past? On the conventionalist side, laws are accepted when they enable verified predictions, but this seems to leave hollow their explanatory claims, for the question 'why do the predictions work?' remains unanswered.

This bifurcation of positivism into empiricism (or inductivism) and conventionalism is captured by the distinction that Crutchfield (1992: p. 68) makes (in the quotation that appears at the head of Byrne's (1997) article) between science and engineering: '... the scientist presumes ... to be focused on what the model means vis-à-vis natural laws. The engineering view of science is that it is merely data compression ...'. In other words, the conventionalist engineer is happy to condense experience into whatever formulae enable predictive calculations when they are slotted into the d-n schema, whereas the empiricist scientist wants to know whether the formulae are laws of nature.

Some simulations seem to situate themselves on one side of this divide, others on the other. The 'bottom up' empiricists seek to represent in their models the observed regularities that they believe are natural laws and their simulations seek to explore what outcomes result when multiple laws interact and initial conditions are varied; their goal is to examine what would naturally happen were circumstances different or allowed to play themselves out over time. The 'top down' conventionalists, on the other hand, construct and refine models to generate outcomes that are validated by subsequent experience; their goal is efficient prediction, regardless of whether the tools to achieve it mirror nature.


Realism has gained popularity as a sociological perspective because it seems to offer a solution to positivism's internal dissolution in to an unresolved debate between empiricism and conventionalism (Harré, 1970). Realism argues that explanation is a matter of identifying the real mechanisms, usually beyond direct observation, which are natural kinds and whose causal powers or powers of agency operate to produce the flux of observable events, including law-like regularities among those events. What for positivism are the source of explanations - laws - are for realism an object of explanation; empirical regularities are to be explained by demonstrating that they are an observable manifestation of the interplay of underlying generative mechanisms. The common example of realism from the natural sciences is the behaviour of observable physical entities being explained by reference to their atomic structures. These underlying structures are real existents, not merely conventional calculating devices, and a realist explanation is accepted when the existence claims of its proposed explanatory mechanisms (its models) are verified, for example, by developing instruments such as electron microscopes which enable them to be observed, at least indirectly.

In the social sciences, realism takes two forms: the micro version which maintains that the explanatory real mechanisms are to be found in the structures of the mind and the macro version which maintains that the explanatory real mechanisms are to be found in supra-individual social structures. Both versions face problems in establishing that their proposed explanatory mechanisms are real existents, particularly as it is difficult to see how instrumentation will assist in making them (indirectly) observable as it did in the atomic structure exemplar from the natural sciences. Moreover, the basis for developing models (descriptions of proposed real mechanisms) in the social sciences is less secure than in the natural sciences, where they are commonly analogues of already widely accepted and well-understood mechanisms. The paucity of established social science mechanisms means that disputes over the nature of minds (including their relation to brains) and of social structures (including their relation to human agency) continue unabated; for neither is there a widely accepted model.

Some simulations seem to aspire to realism. This is particularly so when it is argued that their models represent the essence of underlying mechanisms. The enticing possibility then conjured up is that computer simulations might give us access to the real underlying processes through which mechanisms generate observable social facts, just as new tools in the natural sciences extended our access to atomic structures, for example. However, the unresolved issue of how to establish the existence claims of proposed real generative mechanisms in the social sciences hits realist simulations hard. Appealing to accurate predictions is no answer, because this does not distinguish between a conventionalist calculative device and a real mechanism. The former produces accurate predictions but does not pretend to capture reality, whereas the explanatory force of the latter derives entirely from its claim to represent the real mechanisms operating 'beneath the surface' of the directly observable social world.


Throughout the history of the social sciences, interpretivism has been positivism's constant rival. It too is loaded with a heavy baggage of multiple and sometimes conflicting definitions, each emphasising different aspects of its rich tradition. Setting such complexity aside, for current purposes this perspective can be quickly identified in terms of its appeal to interpretive understanding as the essence of its explanatory claims (Turner, 1980). Groups of people construct and maintain shared interpretive frameworks which give meaning to their actions and interactions. Explaining the flux of social life involves learning and deploying these frameworks to make sense of group members' actions and interactions, that is, see them as reasonable in the circumstances. Of course, people belong to several or many social groups, and have available to them several or many interpretive frameworks, which are fuzzy-edged and partly overlapping, partly not. Some such frameworks are flexible and constantly transmuting, like those shared by a group of friends, others are more rigid and constraining, like those embodied in a legal system. Given people's multiple group memberships, identifying the (sub- )culturally appropriate meaningfulness of particular actions and interactions, for natives and sociological investigators alike, is frequently a sensitive and delicate matter of negotiation, of trial and error through a process of interaction. Nevertheless, the explanatory goal is clear: to achieve interpretive understandings that are congruent with those shared by participants in the group; in simple terms, to see the world their way.

It is important to recognise that these interpretive understandings are not the mental states of the actors, thought of as causal antecedents of their actions (Skjervheim, 1974). That approach to explaining action leads back to a version of positivism, seeking laws that capture hypothesised systematic relations between motives or intentions or other mental events and the subsequent actions they bring about. In interpretivism as described above, the interpretive frameworks are necessarily shared resources for making sense of actions and interactions, not private mental states that cause them. The interpretive frameworks are of course consciously entertained by group members to a greater of lesser extent, but the meanings that are interpretivism's explanatory goal are constitutive of the actions, not independent antecedents of them.

Rule-based, artificial intelligence influenced simulations might initially appear to be situated within the interpretivist perspective, especially as that perspective is sometimes described in terms of actors following rules which make sense of their actions (Winch, 1958), as if interpretive frameworks were etiquette guides, careful conformity to which would give immediate cultural competence. However, the defeasible, open- textured and flexibly applied rules that make up interpretive frameworks have yet to be adequately articulated by expert systems. Indeed, to some interpretivists, the rules that make up interpretive frameworks are essentially ineffable; they cannot be fully articulated because their application always depends in a way that is inexpressible in any formal calculus on their embeddeness in the whole interpretive framework. This is more than merely a matter of a high level of complexity that could be solved by a sufficiently diligent programme of analysis; instead, for some interpretivists attempting to analyse a framework necessarily alters it.

An ironic twist might be noted, which derives from recent sociology of scientific knowledge, inhabited by interpretivists of a highly constructivist bent (Woolgar, 1988). They might argue that the very practice of simulation could become constitutive of what is accepted as social scientific activity. What triumphs as science, they maintain, is not governed in any empiricist or foundationalist sense by success in mapping out the truths of nature. Instead, science, like all human productions, is what we convince ourselves it should be. Although this constructivist approach panders to the powerful urge to make social reality (in this case, what is understood as science) bend to our individual or collective will, it seems in danger of bumping its head on intransigent facts. Moreover, constructivism's explanatory force can seem merely whimsical, for arguing that simulation will become inscribed as science when its knowledge claims are socially legitimated does not seem to satisfy the question: 'why simulation and not, for example, cold fusion?'.

Situating Simulation

Occasional passing remarks above have hinted where some of the varied forms of simulation might be situated among sociological perspectives. This question can be pursued more closely by examining the particular example of simulation by Gilbert (1997) that appears in this issue of Sociological Research Online. On turning to his paper, it must be noted that he states that 'Simulation itself imposes no particular theoretical approach on the researcher: simulations may illuminate phenomenological as well as systems thinking; realist and relativist epistemologies ...'. I concur: research procedures can be adapted to any perspective. Although perspectives differ fundamentally insofar as they embody irreconcilable philosophical presuppositions about knowledge (including explanation) of the social world, they do not determine what procedures must be adopted to generate their particular notion of knowledge, and vice versa nor do procedures entail perspectives. This is why it has been possible in the preceding discussion to suggest that simulations of certain sorts might be situated in several different perspectives.

Nevertheless, one would expect Gilbert to place the particular simulations described in his article within one perspective (at least implicitly), thereby establishing the nature of explanatory goals against which the adequacy of his explanations can be assessed. This is true of anyone offering an explanation; the explainer and her/his audience need to know which perspective has been adopted so that the appropriate standards of adequacy can be applied to assess the explanation. An initial (explicit) statement of Gilbert's explanatory goal is found in paragraph 1.2: '...this paper ... will use simulation to posit a generating structure for science', which gives a firm indication that realism is the perspective within which his study falls. Further evidence to situate Gilbert's study within realism comes from the end of paragraph 2.2 where mechanisms are again mentioned. The realist interpretation is also supported in paragraph 1.5, where he writes that in his work 'Simulation is ... used to investigate a mechanism which might underlie "Lotka's Law"'. Such a statement is typical of the manner in which realism is invoked to provide an explanation of a law by identifying an underlying generative mechanism, thereby, it is argued, answering the question left open by a positivist explanation of why the law holds. This is exemplified in Gilbert's phrasing in paragraph 4.3: 'Simon's algebraic derivation [his mathematically expresses laws] still leaves open the question of the mechanism by which the observed distribution is generated.'

The observation that it is within realism that Gilbert situates his study is also supported in paragraph 1.3, where he draws a distinction between his own use of simulation and most other uses in sociological research, in which it has been 'a way of making predictions, for example about fiscal transfers in ten to twenty years time'. He draws this distinction between his own use of simulation and the use of simulation for prediction again later, in paragraphs 2.1, 2.2 and 3.1. These uses for prediction would seem to fall within the empiricist version of positivism, in which the future outcome of the continued application of currently observed regularities or laws are calculated. Gilbert's argument here is a common one in sociology, with realism adopted as an antidote to perceived limitations of the empiricist version of positivism, in which laws are used to predict but why the laws, thought of as empirical regularities, continue to hold in the future is not explained. But confidence in the view that Gilbert's use of simulation is situated within realism ebbs away when he goes on in paragraph 1.3 to state that 'In this paper ... simulation is used in a quite different way: to explore theoretical possibilities, to undertake the equivalent of "though experiments" and to understand the limits of and constraints on social life by constructing "artificial" societies'. This could perhaps be read as favouring realism, particularly if the models he posits are introduced as hypothesised real mechanisms whose existence is to be tested. However, the phrasing used seems to be more attune with conventionalism, where the explainer is seeking calculative devices that reproduce observed patterns but for which no claim is made that they represent real generative mechanisms.

The same sort of ambiguity between conventionalism and realism occurs in paragraph 3.1 where there is first another formulation that aligns with realism: Gilbert says he will be drawing on approaches 'simulating what are judged to be the most significant underlying processes at work'. But this is soon followed the statement that simulation is 'a type of theory development, in which the simulation is treated as a method for the formalization of theory. Once the theory has been refined, it can then be tested deductively in the usual way'. 'The criteria which should be used to judge the work', he continues, 'focus ... on ... "what are the necessary and sufficient conditions for a given result to be obtained?"'. Generative mechanisms here give way to theoretical formalizations, and these in turn become computational routines, the necessary and sufficient conditions to deduce the observed pattern; realism retreats to conventionalism. As the paper proceeds, his task becomes one of specifying a formula that will reproduce aspects of the observed pattern of publications.

This retreat from realism is not unusual in sociology. It arises because, although elaborate philosophical arguments have been constructed to recommend realism to sociologists (Bhaskar, 1979), there are no uncontroversial answers to the two crucial questions of what the real generative mechanisms might be and how their existence might be demonstrated. To maintain a realist stance, Gilbert would have to maintain that the computational routines of his model are the real mechanisms and that their existence is demonstrated because the results they produce look sufficiently similar to some of the observed feature of scientific publication. But this is a highly attenuated realism that does not seem to offer any added value over conventionalism, which makes no claim as to real existents. It merely maintains that its formulae provide a convenient way of calculating one set of observations from another, without any claim that they represent the way the world works. Accordingly, if Gilbert's study is situated within conventionalism, it does not illuminate anything real about the operation of scientific publication, in the sense that it really works in the that way his computer is programmed to when it simulates some of the observed patterns of publication. In the light of this, a statement in his concluding paragraph (8.6) is unfortunate: 'The model succeeds to the extent that the processes underlying the institution of science are illuminated, not because it manages to mimic the precise quantitative patterns of science.'. This reasserts realism's goal and reminds the reader how far short of it his conventionalist study falls.

The conclusion is that simulation of the sort exemplified by Gilbert's article is an elegant technical exercise situated within the conventionalist corner of the sociological perspective of positivism. Although his presentation appears to promise a project situated within realism, in the face of the key problem of that perspective, of how to convince that the models are real generative mechanisms, that is, how to establish their existential claims, it retreats to conventionalism and provides no more than a complex calculation that reproduces some patterned features of its object of study. This is not, of course, to devalue its achievements as a conventionalist study; as indicated at the beginning, sociology is multi-perspectival and explanations should be assessed according to the criteria of adequacy associated with their own perspective. No attempt is made here to elevate the achievements of one perspective above those of another. Instead, my goal has been to see how to situate simulations within perspectives so that their achievements might be appropriately assessed.

Finally, please note that I was invited by the editor to write a short 'thought-piece' and that is what I have done. This piece is clearly not a fully researched article and it undoubtedly fails to do justice to the complexity and subtlety of simulations in sociology, just as its portrayal of perspectives is heavily simplified. Nevertheless, I trust that it serves its purpose of provoking discussion and debate about simulation's explanatory force.


BHASKAR, R. (1979) The Possibility of Naturalism: A Philosophical Critique of the Contemporary Human Sciences, Brighton: Harvester.

BYRNE, D. (1997) 'Simulation - A Way Forward', Sociological Research Online, vol. 2, no. 2, <http://www.socresonl>.

CRUTCHFIELD, J. P. (1992) 'Knowledge and Meaning: Chaos and Complexity' in L. Lam and V. Naroditsky (editors) Modelling Complex Phenomena. New York: Springer-Verlag.

GILBERT, N. (1997) 'A Simulation of the Structure of Academic Science', Sociological Research Online, vol. 2, no. 2, <http://www.socresonl>.

HALFPENNY, P. (1982) Positivism and Sociology: Explaining Social Life. London: George Allen and Unwin.

HARRE, R. (1970) The Principles of Scientific Thinking. London: Macmillan.

HEMPEL, C. G. (1942) 'The Function of General Laws in History', Journal of Philosophy, vol. 39, pp. 35 - 48.

SKJERVHEIM, H. (1974) 'Objectivism and the Study of Manchester (Part I)', Inquiry, vol. 17, pp. 213 - 239.

TURNER, S. P. (1980) Sociological Explanation as Translation. Cambridge: Cambridge University Press.

WINCH, P. (1958) The Idea of a Social Science and Its Relation to Philosophy. London: Routledge & Kegan Paul.

WOOLGAR, S. (1988) Science: The Very Idea. London: Tavistock.

Copyright Sociological Research Online, 1997