Dan Sperber over at his website, also forthcoming in Pierre Demeulenaere, ed., Analytical sociology and social mechanisms:
There are several approaches in the social sciences that seek to provide causal explanations of social phenomena neither in terms of general causal laws nor in terms of case-specific narratives, but, at a middle level of generality, in terms of recurrent causal patterns or “mechanisms” (Hedström & Swedberg 1988). Typically, these approaches invoke micro-mechanisms to explain macro social phenomena. Most of them, ‘analytical sociology’ in particular (Hedström 2005), are versions or offshoots of methodological individualism. These individualistic approaches either stick to the “methodological” in “methodological individualism” and leave aside ontological issues, or else they are also individualistic in the metaphysical sense and deny the existence of supra-individual social phenomena that cannot be analysed in terms of the aggregation of individual actions (see Ruben 1985).
The ontological challenge to which individualism responds is that presented by holistic approaches that place the social on a supra-individual level of reality. Another possible challenge, coming not from above but from below, that is, from the natural sciences, is generally not considered. The individuals invoked in individualism are not so much the individual organisms recognised in biology as the individual agents recognised in commonsense ontology. Individual agency is taken as a primitive in this approach, rather than as a tentative construct that should be unpacked and possibly questioned by psychology and biology.
Most mechanistic approaches, whether their individualism is just methodological or also metaphysical, show little interest in providing the social sciences with a naturalistic ontology, that is, one continuous with that of the natural science. The main goal of this chapter is to outline such a naturalistic ontology. But why should we want such an ontology in the first place? I don’t, by the way, believe that the social sciences in general should systematically work within naturalistic ontology: many of their goals, concern and programs are better pursued with the usual commonsense ontology. But when it comes to providing a scientific causal explanation of social phenomena, there are at least two reasons to prefer a naturalistic approach. The first reason is trivial: To the extent that it is possible, we would prefer our understanding of the world to be integrated, both for the sake of generality and for that of coherence.
The second, more interesting reason to want a naturalistic ontology has to do with the quality of our causal explanations. Either the laws of physics admit of exception and social events provides such exceptions (and there is a Nobel Prize in physics to be won by doing sociology!), or else whatever has causal powers in the universe at large and among humans on earth in particular has them in virtue of its physical properties. Of course, this does not mean that social scientists should get involved in the physics of social causality. What it does mean though is that, when we attribute causal powers to some social phenomena, we should be able to describe it in such terms that its physical character is not a total mystery but raises a set of sensible questions that can be passed on to neighbouring natural sciences, psychology, biology, and ecology in particular, that directly or indirectly do ground their understanding of causal powers in physics.