Concrete Ontology: Comments on Lauer,Little,and Lohse

The papers in this discussion are a very welcome contribution to current discussion of social ontology. They turn the discussion from broad conceptual analysis based on ordinary language toward approaches that see social ontology as built on and perhaps contributing to empirical social research. They differ on how they think this should be done and exactly what the ontological implications would be. I am not going to work through and comment on all the intricacies of the papers. Instead, I am going to focus on the general arguments and questions discussed and try to show how they might be further refined by focusing on a more applied or empirical level than is done by the articles here.

Much of the current social ontology literature is on my view largely an unfortunate and sad extension of unfruitful analytic metaphysics. Analytic metaphysics has somehow missed the major results of 20th century philosophy. There we learned from Quine and many others that there is no science-independent knowledge producing enterprise based on philosophical conceptual, linguistic a priori analysis. Analytic metaphysics will sometimes try to deflect these results by talking about ontology “being informed by” science, “going back and forth” with science, “woven into” science, or some such hedge, but in practice conceptual analysis generally marches along happily unrelated to anything of scientific concern. “Social ontology” unfortunately often reproduces this unfruitful trend and has very little contact with or use for real social science research. Of course, conceptual analysis can be a valuable source of possible distinctions, conceptual relations, clarification of claims, and so on. Yet, it cannot substitute for real empirical knowledge of the social world.

The papers in these exchanges elaborate this perspective in one form or another. The three contributors all agree that a priori, armchair conceptual analysis of social ontology by itself is not going to provide substantive empirical understanding of the social world. In that respect, all are critical of current mainstream social ontology. The OM! argument as discussed here asserts that social ontology is prior to empirical social science and all contributors reject this assertion. I of course concur.

However, the OM! thesis also claims that doing social ontology in its current mainstream form can contribute to the advancement of social science, and that is a separate thesis that could be true even if the contribution is not a priori. All of the contributors, I think, basically agree in some form that social ontology can matter so long as the contribution is seen as promoting social science research rather than providing independent knowledge of social reality. However, they disagree on how that will work and what the contribution is exactly.

I first take up the shared claim that abstract conceptual social ontological analysis can contribute to empirical research. An ideal assessment of this claim would specify various different aspects or types of social ontological analysis, measures of social scientific success, and provide evidence that the former promoted the latter. Of course, this seems like an impossible ideal, but it does not hurt to state what we would have in the best of all possible social research worlds. One way to start on such a project is to look at some things we may know about conceptual analysis and scientific progress in general.

One crucial thing we know from the history of science is that scientific progress requires both clearly formulated concepts and claims as well as appropriate workable concepts and relations that are sometimes really best thought of family resemblance concepts along with piecemeal accounts of their relations. The concept of “gene” embodies the latter. Historical and contextual investigations of research practice suggest that much science works without formalizable theories. Recognizing that would deepen the three contributions here which sometimes talk of “theory” without further qualification; Lohse is most explicit that the issues are best not put in terms of theories or solely in terms of them.

However, even if the place of theories in science is often overplayed, a real hindrance to social science research is often the lack of clearly specified claims. Key concepts are sometimes hardly explicated and often done so in different ways in the scope of single publications. Relations between key concepts are not specified, for example, the specific causal relations are left unclear. Science across the board has progressed by eliminating such ambiguities so long as doing so is not a straightjacket on dealing with empirical reality. Social ontology in traditional forms might contribute to these goals. Work on norms and conventions by philosophers comes to mind as a possible success story. Similarly, Mackie’s INUS approach to causation certainly has helped in political science and elsewhere.

There may be other useful routes for abstract social ontology to provide “production line” materials to use Lauer’s terminology. Admittedly, so far, there are few signs that contemporary analytic metaphysics applied to the social realm has produced much useful conceptual fodder for social research. But that does not seem inevitable.

The debate over ontological commitment between the contributors seems both too deep and hard to take on in this venue and at the same time not closely enough tied to useful social science applications and examples. The issue is too deep in that there is a large general debate on realism, antirealism, instrumentalism, and deflationism that has occurred over decades. However, whatever arguments pro and con have arisen in the general debates over realism, and so on in philosophy of science would apply to the arguments and positions developed in the three contributions. None of the contributors claim to make any new arguments on the general issues. Neither will I. Lohse and Lauer’s paper have sympathies, as do I, for deflationism, though what that comes to is not always clear.

Lauer’s original paper invokes quantifier variance arguments to show that conceptual social ontology can contribute to social scientific progress but can do so without ontological commitment to social entities. The replies do not really pursue the argument invoking quantifier variance. However, as I detail below, it can raise really interesting empirical issues in social research.

On my view, debates about “the meaning” of quantifiers or about realism versus instrumentalism (or is that antirealism or . . .) are too abstract and global to get very far. I take the ideas of Quinean holism and antifoundationalism along with some elaboration and clarification that you might call contextualism (Kincaid 2005; Williams 1999) to suggest that the realism-type questions raised by the contributors often have to be tied to specific empirical issues to be meaningfully and fruitfully addressed. The claim by Lohse and perhaps accepted by the other contributors that “realism is notoriously hard to defend in the social sciences” has dubious content on my view. ¹ Realism here means what? True, getting closer to the truth, well confirmed, getting more confirmed, and so on? Can we judge all of social science either pro or con here? I despair of progress on such global, empirically untethered questions (Kincaid 2000) and defer to more concrete, related questions that we may make some headway on, though we certainly can see traces of the bigger issues in the background. However, if we put global realism debates aside, the three articles focus on the need to tie social ontology to empirical issues in the social sciences which seems to me much more fruitful than the global realism-antirealism debates.

One more local question the contributions raise is whether taking social scientific theories (or scientific accounts to be less theory centric) as tools, without making a commitment to the entities they presuppose might be useful for scientific progress? Despite Little’s (2020) apparent claim about instrumentalism to the contrary, I think evidence from science in general shows that instrumentalist attitudes can and have contributed to scientific progress in certain times and contexts. Special and general relativity and quantum mechanics all benefited from making ontological interpretations secondary and getting on with empirical investigation. Current physics (and some considerable amount of chemistry) is awash in technical, formal techniques and models that are not given an ontological interpretation, without much concern so long as empirical progress happens, which it does.

Can we say anything similar for the social sciences? That seems to me a terrific and understudied question implied by debate over Lauer’s quantifier-variance motivated antirealism about the social sciences. For example, there is a story of the development of neoclassical utility theory that sees progress coming from giving up on psychological interpretations of choice theory and simply sticking to observable behavior (Ross 2005). Whether this form of instrumentalism has led to progress is controversial (but see below); whether and where instrumentalism has been helpful elsewhere in the social sciences is an open interesting question that these contributions raise.

These questions above are about usefulness of social theory without ontological commitment. What about ontological commitment itself? My natural inclination is to say that if we refer to entities that are essential to theories with real empirical success, then we are as committed to them as we are in anything else we believe which has empirical success. I take it that this is Little’s position. But as the contributors show, an alternative, roughly instrumentalist approach is a competing alternative. Lauer says something quite important in this connection: that a certain humility about social science results is often called for, because of all the uncertainties surrounding data and its interpretation.

I think we can go beyond blanket realism or instrumentalism if we look for more local issues and do some clarification. A first step, I would argue, is to see just when, where, and how specific social research with specific ontologies has empirical success. I have tried to do that for class, race, ethnicity, and caste in some previous work (Kincaid 2016, 2017). My arguments there are about particular social entities in specific societies at specific times, not about the existence of class, and so on sui generis. I would hope that these arguments show that to the extent we are realists about much of anything beyond the directly observable, then we ought to be realists about such social entities. These kinds of arguments are within science, not some kind of general philosophical argument such as inference to the best explanation. I take it that this is Fine’s position which Lauer mentions. It is philosophically deflationist but commonsense realist.

However, there are, nonetheless, many interesting complications for such an approach raised by these contributions. One is pluralism, nicely stated by Little’s contribution. Pluralism is attractive if an essentialist, natural kinds view of the world seems simplistic to you. Again, we could treat these as grand global claims about pluralism a la Quine and Goodman. More tractable, I believe, are pluralist claims restricted to specific areas that do not commit us to any grand metaphysical theses about universal inscrutability, no facts of the matter, and the like.

So, in the cases of social entities such as class, race, and so on that I have argued, we should sometimes take realistically, a certain kind of pluralism may well be right. I do not know. The use of class and so on in specific circumstances may help us to identify causes of social phenomena—there is considerable evidence that they do. However, that does not preclude using other ways to divide up the relevant social phenomena that are also explanatorily—causally—informative. The notion of a “ruling class” at certain times and places explains much. Does dividing those at the top differently into ruling elites also explain? That could well be the case and it could be that we can do so without contradicting the ruling class hypothesis. Marx himself moved between talking about the ruling class and ruling classes; maybe that shift is more than just an equivocation but a recognition that multiple complementing accounts might be given. These are the kind of empirical issues which gives “realism” and “pluralism” concrete implications.

Aside from the pluralism issue—but surely related—are definable instrumentalist versus realist issues in specific areas of social science research. A good example of something worth studying here are economic accounts of individual choice.

As I noted above, the history of choice theory in neoclassical economics can be read as a history of “purifying” the theory—of removing reference to the psychological entity “utility.” We now know that a set of choices that meet a set of consistency conditions can be represented by a utility function, where that just means it can be represented as maximizing a function of a certain sort, the moniker “utility” just being verbal. This account of choice is called revealed preference theory (RPT, though that is an unfortunate title; see below).

The often-raised question is whether this development away from psychological utility supports realism or instrumentalism. The question is often imbued with unhelpful philosophical terminology and baggage. More nuanced issues are actually involved. A helpful parallel comes from Dennett’s (1991) ideas about the reality of belief. Dennett famously initially tried to think of beliefs as like “centers of gravity” in physics, seemingly trying to skirt questions of reality in the case of belief as no more useful than questions whether centers of gravity are real. Eventually he produced a more nuanced view as outlined in “Real Patterns,” rejecting the instrumentalist interpretation of his work. Beliefs are real to the extent they allow us to find causal patterns in behavior that we cannot find otherwise. If they do help us find such patterns, they count as real or at least as real as other scientific postulates.

Applied to RPT, this Dennettian story suggests the following interpretation: utility functions should be taken realistically insofar as they pick out real patterns. (Here, I am following Ross 2005.) They pick out real patterns most convincingly when we can use the utility functions assigned to individuals based on specific sets of choices to predict and causally explain other choices in other circumstances, for example, when prices and budget constraints are different. When we can make these predictions, then we should take the utility functions as picking out real things even though they make no psychological commitment about mental states.

Should we be realist about RPT? Put that baldly, the question is malformed. We should ask instead whether we can use only observed choices to describe real patterns as elaborated above—to explain without making ontological commitments to underlying psychological states. That is an empirical question we know how to investigate in some cases. All the evidence points to “sometimes and sometimes not” answers with lots of work to be done on the details.

Questions like those around RPT illustrate the kind of social ontology and realism questions philosophers can contribute to and have. I am sure there are numerous other local ontological questions in the social sciences that philosophers of science can help clarify. So, to sum up, these papers are refreshing and point in the right direction on my view. But, we need to get more concrete.