What are mechanisms in social science?
- 165 Downloads
Why should we introduce the notion of ‘analytical sociology’ into the field of sociology, and why should it be linked to the concept of ‘mechanism’? With these two principal questions, Pierre Demeulenaere, Professor of Sociological Theory and Philosophy of the Social Sciences at the University of Paris-Sorbonne, opens his Analytical Sociology and Social Mechanisms, a collection of thirteen papers written by social scientists and philosophers of the social sciences (1). Not every contributor should be considered an analytical sociologist. Rather than being a manifesto either pro or contra analytical sociology and the use of mechanisms, it is an attempt to reflect upon the key issues involved in sociological explanation (3). Even though several chapters raise very interesting points, the overall impression one gets from this book is that analytical sociology fails to redeem its main promise, viz. to add clarity, precision, and conceptual rigor to sociology, especially regarding one of its most central concepts: ‘mechanism’.
Analytical sociology is not a new sociological paradigm, but an effort to clarify analytically the basic epistemological, theoretical, and methodological principles underlying any satisfactory way of doing social science (1). It provides an updated version of methodological individualism, often called structural individualism, which combines its predecessor’s basic focus on the individual actors underlying all social phenomena with an updated view of these individuals, including a new theory of action and special attention for the actors’ (social) environment (10–12).
One of the central concepts figuring in analytical sociology is ‘social mechanism’: “Whenever we start explaining ‘why’ something happens, beyond mere description, we are necessarily led to introduce some type of causal linkage of elements that in turn raises the question of mechanism. Analytical sociology is impelled in this way toward the study of mechanisms and their functioning. Emphasis on the notion of mechanism corresponds to an evaluation of the proper role of causal linkages in the social sciences” (2).
Yet there is no consensus on the notion of ‘social mechanism’ in analytical sociology. This is not a new complaint. Ten years ago, James Mahoney (2001, pp. 577–581) has listed no less than 24 different definitions of ‘causal mechanism’ in his review of Peter Hedström and Richard Swedberg’s pioneering (1998). Has analytical sociology made any progress over the past 10 years? I’m afraid it hasn’t. In Analytical Sociology and Social Mechanisms, one hardly finds explicit definitions of ‘mechanism’, while the concept is used implicitly in many divergent ways. These can roughly be classified in three groups that should not (but often seem to) be used interchangeably.
A first group can be called complex systems mechanisms: systems of lower-level components the organized behaviors of which result in some higher-level phenomenon. This is the standard view of mechanisms in philosophy of (the life) science(s). According to Demeulenaere “a mechanism is [a] set of elements and their causal links that regularly lead from an initial social state to a subsequent one” (12; see also 188–189). Michael Schmid writes that “social scientific explanations are to be couched in the form of micro foundational multi-level explanations of macroscopic states of affair with reference to a substantive theory of individual action” (136). And Keith Sawyer claims that mechanistic explanations “commonly describe processes in very complex systems that underlie what, on the surface, appears to be a simple causal relationship” (78) and that “[a] mechanistic explanation of an event […] traces the causal processes leading up to that event by describing the components of a complex system and their interactions […]” (78). Other contributors, such as Dan Sperber (64, 75), Gianluca Manzo (268), and Michael W. Macy et al. (252), implicitly adopt a complex systems account of mechanisms as well.
A second group can be called mediating mechanisms: they connect causes and effects on the same level of organization. Jon Elster defines mechanisms as “frequently occurring and easily recognizable causal patterns that are triggered under generally unknown conditions or with indeterminate consequences” (50, emphasis omitted). And Robert J. Sampson “conceptualize[s] a social mechanism as a plausible contextual process that accounts for a given phenomenon, in the ideal case linking putative causes and effects” (230). (Later in his essay, Sampson (234–239) gravitates toward a complex systems concept, as he discusses the relation between macro-level processes and their micro-level composition.) Examples of Elster’s mechanisms include emotional mechanisms (fear, anger, …) and psychic needs (for autonomy, for cognitive consonance, …). These do not consist of a lower level underlying or constituting a higher-level causal relation, but only of causal patterns from, for example, a particular type of belief to a particular type of emotion. This is what distinguishes them from complex systems mechanisms. (Note here that in his earlier works, Elster (1983, p. 24; 1998, pp. 47–48) did adopt a complex systems concept: mechanisms enable us to go from the larger to the smaller, from societies to individuals: “To explain is to provide a mechanism, to open up the black box and show the nuts and bolts, the cogs and wheels of the internal machinery.”).
The third group consists of mechanisms as theoretical building blocks. Petri Ylikoski offers two concepts of ‘mechanism’ (160). A-mechanisms are particular causal chains, “whatever connects the cause and effect”, and thus can be regarded as mediating mechanisms. B-mechanisms, by contrast, are theoretical building blocks. A-mechanisms can consist of several B-mechanisms. Åberg and Hedström (201–226) seem to implicitly adopt both Ylikoski’s notions of mechanism.
These three groups of definitions are not incompatible. Mediating mechanisms can be seen as molecular complexes of single theoretical building blocks, as Ylikoski suggests. And the micro-level of complex systems mechanisms can be regarded as complexes of mediating mechanisms. (Elster’s indeterminate causal patterns that are triggered under generally unknown conditions closely resemble the referents of Sandra Mitchell’s pragmatic laws; for the relation between pragmatic laws and complex systems mechanisms, see Leuridan (2010).) So in the end, complex systems mechanisms are complexes of theoretical building blocks. Yet still they are importantly different. Hence, after reading Analytical Sociology and Social Mechanisms, one cannot avoid the impression that there still is not even a hint of consensus on the concept of ‘social mechanism’. Some authors even use different concepts of ‘mechanism’ interchangeably. For example, Elster describes his mediating mechanisms as “the basic building blocks […] of explanation in the social sciences” (61, my emphasis), while reusing phrases from his earlier, complex systems definition: “nuts and bolts, cogs and wheels” (1983, p. 24). Given analytical sociology’s defining characteristics (clarity, precision, and conceptual rigor), this is rather discomforting. This volume could have been an opportunity to meet this old complaint. It is a missed opportunity. (Analytical sociology is not the only approach to the social sciences which uses different concepts of ‘mechanism’ interchangeably though; see Weber and Leuridan (2008) for a brief discussion of Morgan and Winship’s (2007).)
In my opinion, analytical sociology should primarily adopt the complex systems definition, as do Hedström and Bearman (2009, p. 5) in their handbook of analytical sociology, since it best serves its needs. Two major issues which are repeatedly touched upon in Demeulenaere’s introduction are the nature of social causation and the relation between individual actions and social phenomena (the tenability of individual or structural individualism). Yet despite its aims, the book as a whole does not satisfactorily deal with any of them. First, causation is mentioned by most if not all of the contributors, but hardly anyone gives an explicit account of the concept or it is characterized negatively as being different from mere correlation. (A noteworthy exception is Petri Ylikoski, who adopts Woodward’s interventionism.) The topic of causality is extensively debated in the mechanisms literature. Broadly speaking, two big camps seem to be arising: on the one hand, those who adhere to an Anscombian theory of causation as productivity, detached from the notion of regularity, and on the other hand, those who prefer a more regularistic (often Woodwardian) account. Second, the contributors which most clearly express their views on the relation between individual actors and social phenomena are those who adopt a complex systems view of mechanisms: Demeulenaere (14) and Schmid (144–145) only allow for individual actors to figure in social mechanisms, whereas Sawyer would also include smaller-scale social units (79–80). Analytical sociology could benefit even more from the mechanistic literature in this respect. One of the most interesting accounts of the relation between a mechanism’s lower-level entities and activities and the resulting higher-level behavior is offered by Craver (2007). Even though his ‘mutual manipulability’ account of constitutive relevance faces some interesting problems (see Leuridan 2011), it may help to get a grip on discussions about methodological individualism.
Three chapters, I think, deserve special attention. First, perhaps the best contribution is Petri Ylikoski’s “Social mechanisms and explanatory relevance”. Ylikoski discusses mechanistic explanation within the framework of his contrastive counterfactual account of explanation and Woodward’s manipulationist account of causation. His main claim is that whereas the idea of mechanism has many uses in the philosophy of science (159–160), it is not very helpful in dealing with the problem of explanatory relevance (160–162). Second, Pierre Demeulenaere argues in “Causal regularities, action and explanation” that mechanism-based explanations should not be strongly contrasted to covering-law explanations. Causal regularities, he claims, are very important in human social actions (177–181) and hence in social science explanations (185–189). His paper is interesting for anyone who likes to stress the role of regularities in mechanisms, yet I have doubts about his devotion to old school covering laws—a devotion which he fortunately relaxes at the end. Finally, Michael Schmid offers an interesting four-step ‘logic’ of micro-foundational or mechanistic explanations in the social sciences: explaining the actions of individual actors in terms of their capacities, opportunities, …, determining how the various authors link their action with each other, identifying the resulting collective consequences and, finally, judging how these collective consequences affect the individuals’ subsequent decisions and actions (140–143).
The interested readers will find interesting contributions addressing interesting topics, but they should still not look for a final answer to the question “what are mechanisms in social science?” in analytical sociology.
- Craver, Carl F. 2007. Explaining the brain: Mechanisms and the mosaic unity of neuroscience. Oxford: Clarendon Press.Google Scholar
- Elster, Jon. 1983. Explaining technical change. Cambridge: Cambridge University Press.Google Scholar
- Elster, Jon. 1998. A plea for mechanisms. In Social mechanisms, ed. Hedström and Swedberg, 45–73.Google Scholar
- Hedström, Peter, and Peter Bearman. 2009. The Oxford handbook of analytical sociology. Oxford: Oxford University Press.Google Scholar
- Leuridan, Bert. 2011. Three problems for the mutual manipulability account of constitutive relevance in mechanisms. British Journal for the Philosophy of Science. doi:10.1093/bjps/axr036.
- Morgan, Stephen L., and Christopher Winship. 2007. Counterfactuals and causal inference: Methods and principles for social research. Cambridge: Cambridge University Press.Google Scholar