See my earlier posts here and here.
Explanation in social science has to be a) reductionist in principle and b) holist in practice. This is because social facts and other social phenomena reduce to human actions and mental states, but reducing them to this for explanatory purposes would require knowing each and every action and plausibly-ascribed thought that had a causal role in producing the social phenomenon, and this is probably impossible. Likewise, the properties of a dodo entirely reduce to the atoms that compose it, but knowing precisely how they all inter-relate in producing the actions of a single dodo, let alone the species, is impossible. (Dodos are extinct, after all, so we can't reconstruct very much beyond generalities!) In social science as in biology, we can and should accept the view that all sociological phenomena reduce to human thoughts and actions while using entities for explanatory purposes that we do not necessarily have to explain in terms of those individual thoughts and actions.
There's an enormous literature on this subject, but this position sounds fairly self-evident to me. I suppose the problem for many social scientists is that the reductionist position is sometimes motivated by (right-wing/neo-liberal) political considerations - a confusion of methodological and ideological individualism - but in my case, at least, it is motivated by metaphysics. It is not logical to believe in a universe where everything is composed of elementary particles, where everything is material and natural, but where 'social' phenomena have an independent existence - as if the evolution of human beings created a metaphysical rift, generating forces incapable of being explained naturalistically. I find that view as troubling and incorrect as belief in the soul (a belief deliberately designed to create a metaphysical rift between humans and other natural phenomena).
I'm interested, therefore, in trying to elucidate the properties that allow social facts to exist - that is to say, the properties of mind that allow for this. I'm not all that bothered by the case for in-principle reductionism, because the case is already settled: humans are part of the universe, and all of their features have to reduce to the things that compose them. I'm more interested in understanding how these features reduce, not in whether they do (because they do). There have been plenty of attempts to do this, but one of the most comprehensive is John Searle's. It is also the only comprehensive theory of the reduction of social facts that you can buy in Waterstone's. Searle seems to have carved out a niche for himself as a philosopher of the people, because Making the Social World, his most recent book to tackle the subject, is not a popular work like his books on mind and consciousness, and it is hard to believe that the ordinary reading public would find it enjoyable (although you never know, and I'm gladdened by the knowledge that John Searle has decided to use his popularity to make higher philosophy more widely available).
Anyway, Searle's theory is certainly worth reading, and it really is a brilliant attempt to make social facts naturalistic. But it also has quite a few problems, not least of which is the decision to see what he calls 'collective intentionality' as different from beliefs about beliefs. Searle claims that social facts depend on collective intentionality, which he calls 'we-belief'. This is an irreducible belief of we-ness: we are doing something, rather than simply he and he and I and she are doing something. He claims that humans have the capacity for collective action as a result of a form of collective belief that doesn't reduce to a set of beliefs about the beliefs of others. In his earlier work - and this is the argument he uses to establish the truth of the principle he believes to be at the heart of social facts - he said that if collective intentionality reduced to beliefs about beliefs, then it would be potentially infinite, and nobody's head can hold that many beliefs. If I believe that we are working, and this belief reduces to beliefs about beliefs, then I have to believe that 1) I am working, 2) she is working, and 3) that she believes that I believe that she believes that I believe that she believes that I believe that she believes [ad infinitum] that we are working - or else we cannot work towards a common goal, except haphazardly. This was Searle's original argument; in his most recent work, he has allowed for some features of social action to reduce to beliefs about beliefs, but doesn't seem to want to get rid of the idea of pure collective intentionality.
This argument is not sound. Yes, it looks like a vicious infinite regress, but is it? No. In fact, the ability to have nested sets of beliefs of this form is key to most human thinking - see, for instance, Michael Corballis's The Recursive Mind (2011, Princeton University Press), a witty and enjoyable book outlining the role of recursion (which is what this set of beliefs is) in almost every sphere of human life. Far from producing a vicious infinite regress, reducing collective intentionality to recursive sets of beliefs about beliefs would make it easier to integrate the theory of social facts with the rest of cognitive science, and it would also entail the evolution not of a social module of human thought in a single bound, but rather the more gradual evolution of the capacity to think recursively combined with the ability to detect intention in others.
That isn't to say that there isn't a kind of thought called collective intentionality. Clearly, processing a long chain of recursive sets of beliefs about the beliefs of others would make social interactions an enormous cognitive burden (which may strengthen the case for the rise of human sociality generating higher intellectual capabilities in humans). With the paradox of the heap, humans think in terms of heaps while being able to understand the heap in terms of its grains, and we're not nonplussed when it turns out that removing the grains = removing the heap. In a homologous fashion, humans think in terms 'we' but are able, when necessary, to understand social interactions in terms of beliefs about others (and beliefs about others' beliefs, including beliefs about others' beliefs about our own beliefs, &c.). This gives people the flexibility to work with one another smoothly and efficiently without being entirely surprised when other people don't always work the way you think they will.
I'm sure that this is opaque to many people, even people who are interested in the topic. But it's important to find out just what the properties of mind are that allow for social processes. I think Searle is on the right sort of track, but he's also dead wrong about a number of things - collective intentionality's irreducibility is only one of them, and his view that rules and social facts are 'desire-independent' is also, to my mind, deeply flawed (this is a view based on his earlier work in the philosophy of action, where he takes issue with what he calls the 'Classical' model of action, derived from Donald Davidson; in fact, I see the idea of 'desire-independence' as the biggest hole in Searle's theory, but I won't go into it now). I recommend giving Searle's books a go - The Construction of Social Reality is a good place to begin, especially as he outlines the rest of his philosophy in it in a concise form, and then Mind, Language, and Society, followed by Making the Social World, his opus on the subject. Joshua Rust's books on Searle, including his popular account of Searle's philosophy (John Searle, 2009, Continuum Books), are also a good source for Searle's theory, especially as Rust is far from un-critical. What Searle has provided is a good but flawed model that we can and should spend time hammering out into a more plausible, realistic model, ideally one with practical uses in the social sciences.
I should be clear: it isn't necessary to have this theory in place in order to undertake research. That's why I say we should all take sociological reductionism as true in principle, and use a theory along Searle's lines to show how this is possible, while continuing to use theories that solve problems in preference to uniting completely with higher-level social theory. I don't think you'd require a theory of human social ability in interpreting archaeological finds or ethnographic data, necessarily. The empirical work should take pride of place there, and the theory should be enough to explain it and to link it up with other, reductive, views, but need go no further than this (just as we don't need to refer to the properties of fermions when discussing the evolution of the dodo). And Kent Flannery and Joyce Marcus have managed to write a beautiful, brilliant book on the origins of inequality, including a discussion of the principles of social logic in different sorts of societies, without involving a theory comparable to Searle's, just as we can investigate human language and society without necessarily unifying it with the study of the brain. It would just be nice to have a way of making humans natural - of letting us see just how we work in a way that is consistent with what we know about the universe.