A Framework for Social Ontology

1. What Social Ontology Studies

The aim and problem of social ontology is to understand the nature of the social world. The term social ontology might not actually have been the best choice to name this project: the term social metaphysics is more apt, because to some people, ontology is a narrower pursuit. Quine, for instance, argued in the 1950s that ontology is basically a matter of taking inventory of the objects in the world. Philosophers working in social ontology, though, generally agree that the project is broader than cataloging what entities exist: we want an account of how the social world is built. What are its building blocks and how do they come together to build it? Social ontology should not be thought of as the study of “ontological claims” such as “social groups exist” or “there are no social spirits.” But instead, it is the study of ontological building relations between different kinds of entities. Consider, for instance, the intentions or attitudes of groups. Suppose that the Boston Red Sox has the intention to win the World Series. One question is whether there are group intentions such as these at all, but beyond that, we want to know what it takes for there to be such an intention. An answer will involve three parts:

Social ontology tries to make progress on clarifying all of these in the context of specific topics: group intentions, laws, corporations, property, institutions, social groups, and so on.

To begin an inquiry in social ontology, we need to choose which entities to work out the ontology of, that is, where to focus our attention in analyzing the social world. There are many ontological categories that could be reasonable candidates: social objects, such as universities, corporations, crowds, and dollars; social properties, such as being president, costing 10 dollars, and being illegal; social events, such as World War II, the 1972 Olympics, and the coronation of Queen Victoria; social kinds, such as races, genders, and classes; and so on. Choosing a category is something of an art, because many of the categories are adaptable enough to effectively subsume other categories. So the choice is best made on the basis of economy, simplicity, and generality. Among the various candidates, social facts turn out to be a practical category for inquiry, because they are fine-grained enough to make the distinctions we want and general purpose enough to accommodate the other categories as special cases. Examples of facts include the following: Google is a profitable corporation, Assad is a war criminal, and the bourgeoisie is wealthier than the proletariat.

A fact, as typically understood in contemporary metaphysics, is a part of the world. We talk about facts: we say things such as “Google is profitable in the first quarter of 2015” to assert that a particular fact obtains. However, the facts are the worldly things, not linguistic ones. Facts are typically understood to be reasonably fine-grained: the fact Google is profitable in the first quarter of 2015 is different from the fact Google earned $6.5 billion in the first quarter of 2015. However, facts are not excessively fine-grained: the fact Cicero was a Roman orator is the same fact as Tully was a Roman orator, because Cicero and Tully are names for the same person.

Objects might seem better candidates for ontological investigation, because what we really seem interested in are questions of what the pieces of the social world are—how are social objects built? A problem with objects, however, is that we are not just interested in one or two facts about objects. If we are investigating the ontology of corporations, for instance, it is not very helpful to ask the question: what is a corporation? It is not too clear what this question is even asking. Rather, we need to investigate lots of facts about them: What are the parts of a corporation? Does it have any essential properties? What does it take for a corporation to survive over time? What is it for a corporation to take an action or to have an attitude or intention? And so on. Social ontology is not just concerned with what social objects exist. It needs to address a variety of properties of social objects, and it also needs to address non-social objects when they have social properties. ² Altogether, we need to break things down to the granularity that facts have anyway. So it just makes things less complicated to work with a fact-based ontology. To talk about properties or objects, we can also just talk about general facts: that is, facts having a form such as x is a war criminal or corporation y has intention j. This keeps things nicely explicit: instead of just talking about an object such as a corporation, we force ourselves to be clearer about exactly what aspect of an object or property we are giving an account of.

Which facts are the social ones? It is hard to give a satisfactory answer to this question. Does the category include facts about tables, chairs, and cars? About genetically modified organisms? About groups of animals? It is not clear what is at stake in circumscribing the social facts and may be pointless to engage in a lengthy exercise to pin it down. There may not even be a distinction between social and non-social facts: to be sure, some people see the project of social ontology as accounting for the social in naturalistic or individualistic terms. However, other projects are committed merely to clarifying how the social world is built of social building blocks. In any case, the project of social ontology does need to take a somewhat expansive view of the social world. Some theories of social ontology have restricted themselves to only a few examples—for instance, language or law or property or money or driving rules—and left it opaque whether a given theory was meant to be extended to others. My inclination is to keep an open mind about the domain of social ontology and understand the category of social facts broadly.

Key to this broad treatment of facts is to notice that not every social fact is known about, thought about, or conceptualized. A well-known role for social science is to identify categories and kinds in a society, which the members of that society are too entrenched in to notice. Some social facts and categories are partly or completely formalized, with explicit rules and statutes defining and governing them. Others are entirely unnoticed and unremarked on. And most are in the middle with partial recognition and partial formalization. In social ontology, we need to make sense of all of these.

2. Ontological versus Causal Explanation

A crucial point of entry into social ontology is to distinguish it from a different inquiry: the inquiry into the causes and interactions among social phenomena. It is one thing to investigate how the social world causally works—what sequences of events lead to one another or what mechanisms are operative. It is another to investigate what the social world is—its building blocks or what it consists of. Although this distinction is reasonably commonplace nowadays, historically, it caused a lot of trouble. Particularly in the long-standing debates over individualism versus holism in the 1950s, there was a tight association between the rejection of holism (an ontological point) and the advocacy of an individualistic methodology regarding social explanation. ³ In the following years, it became clear that these perspectives could be separated. One could reject holism about the social world without having to endorse individualism about social explanation. ⁴ A standard distinction came to be made between ontological individualism—a thesis about social ontology—and explanatory individualism—a thesis about social explanation. ⁵

Although this now-standard distinction roughly makes the point, it is not exactly right. First, there is a problem with contrasting explanatory claims and ontological claims: ontology too involves explanation. To be sure, scientific explanations involve much more than explanations of ontology. Typical scientific explanations involve causes, mechanisms, and more. However, purely ontological explanations can be made, and these are part of the practice of science. Second, there can be obstacles to both ontological explanations and to causal explanations. The opponent of individualism in methodology, for instance, might oppose one or both kinds of explanation: the ontological explanation of social facts in terms of individuals and the causal explanation of social facts in terms of individualistic mechanisms. We need, in other words, to distinguish four sorts of questions, listed in Table 1.

OPEN IN VIEWER

Table 1.

Four inquiries: determination vs. explanation, ontological vs. causal.

	Ontology	Causes
Determination and dependence	(a) What are the ontological relations between social facts and other facts that build them? (e.g., are social facts exhaustively determined by individualistic facts?)	(c) What are the causal relations between social and other facts? (e.g., are social facts caused by individualistic mechanisms?)
Explanations and methodology	(b) Can (and should) we explain the ontological determination of social facts in terms of other facts? (e.g., in terms of individualistic facts)	(d) Can (and should) we explain the causal mechanisms involving social facts in terms of other facts? (e.g., in terms of individualistic mechanisms)

Category (a) consists of purely ontological questions about the way the social world is built. We might ask, for instance, about whether social facts are exhaustively determined by facts about individual people, or whether social facts are emergent, and so on.

Included in category (b) are problems about reduction. Supposing that there is a given ontological relation between social facts and some other facts, is it possible to describe or explain this connection? In saying that there can be obstacles to ontological explanation, that is to say that a particular answer to questions in category (a) do not entail the answers to those in category (b). Even if we suppose that institutions are built out of their members, ⁶ it is a further question whether institutions can be broken down into their members for explanatory purposes.

Category (c) includes questions about the causal relations among social facts and between social and other facts. We might ask, for instance, whether changes in social facts always proceed through a particular non-social mechanism. Moreover, category (d) includes questions about describing or explaining the causes of social facts. One answer to this category of questions is “explanatory individualism” or “methodological individualism.”

The standard distinction between ontological individualism and explanatory individualism separates questions in category (a) from those in category (d). However, if we want to isolate the inquiry into social ontology alone, we need to leave out (c), as well as do our best to make it explicit when we are making claims about (b), that is, ontological explanations, as opposed to making claims about ontological facts alone.

Although it is crucial to distinguish the ontology of social phenomena from causal connections, it is sometimes confusing to do so. One pitfall is trying to distinguish the two in terms of time, or the past. Ontological relations are sometimes thought to be synchronic, or simultaneous, and causal relations diachronic, or over time. However, this is a mistake. Many (if not most) social phenomena are built of diachronic parts.

To see how this works, consider a footprint. For something to be a footprint involves its having been marked by a foot, at some point in the past. ⁷ An instantaneous snapshot of the world—right this second—is not enough to determine if a particular mark is a footprint or not. A mark that looks like a footprint might be a faux-footprint. Even a perfect foot-shaped mark struck by hand is not a footprint. The following fact ontologically depends on facts about the past:

In particular, by the fact,

This might be confusing, because fact (5) itself involves a causal relation. It is the fact that there was a particular causal sequence of events—a foot strike causing a mark. However, fact (5) is just a complex fact, and notice that the relation between (5) and (4) is not a causal one. Fact (5) does not cause fact (4) to be the case. Fact (5) is part of what it is for (4) to be the case. In other words, fact (4)—a fact that obtains now—ontologically depends on a pair of causally linked historical events having taken place.

The same goes for many, if not most, social facts. The fact Obama is president ontologically depends on his having been elected in the appropriate manner in the past. The fact John Roberts is chief justice of the Supreme Court depends on his having been nominated by the president and approved by the Senate in the past. And the fact Assad is a war criminal depends on his having performed certain acts in the past. These facts all have diachronic building blocks.

This, incidentally, provides a quick test to apply to any theory of social ontology. If a theory only gives us synchronic building blocks for the social facts, then that theory must be mistaken. For instance, a theory cannot be right if it takes the social world to ontologically depend only on the current beliefs, attitudes, and even current practices of community members. ⁸

An even more basic test of adequacy for a theory of social ontology, though, is whether it makes a clear distinction between claims of the form,

and claims of the form,

If this distinction is obscure, then it is not clear the theory is talking about social ontology at all.

Some people voice skepticism about the distinction between ontological and causal determination. They worry that it is impossible or incoherent to cleave the facts that are causally related to some fact from those that are ontologically related to it. This is an interesting concern, which warrants investigation. However, to date, I have not encountered compelling reasons for doubting a sharp distinction. For example, a prominent reason people give for erasing the causal/ontological distinction is to observe that many social kinds have “looping effects.” ⁹ That is, characteristic of certain kinds is that the categorization itself affects our practices. Moreover, those affected practices in turn affect the category, and so on, looping from category to practices to category to practices. However, that is not reason for skepticism about distinguishing ontological from causal determination. The reason can be seen in the footprint case. It may be that facts about a social kind ontologically depend on causally connected diachronic facts, including practices of applying categories. ¹⁰ So some facts may be both causally related and ontologically related to facts about that kind. However, we can still sharply distinguish those facts that play an ontological role (what we might think of as those that are “constitutive” of the category) and those that do not.

3. Grounding and Anchoring

Now, I will turn to two fundamental notions at the heart of social ontology: grounding and anchoring. To begin with a fairly straightforward example, consider the following facts:

The question for social ontology is not just to notice that these facts obtain but to investigate why they do. In virtue of what are these facts the case? Again, this is not a question about causes. There are, of course, causes for each of these: Perhaps Assad became a war criminal because of pressure from his advisors. Perhaps because of the way his mother treated him when he was a child. Perhaps it was his training as an ophthalmologist. Those are questions for social science but not for the social ontology of (6). Instead, what ontologically explains (6) are facts such as

The relation between facts (9) and (10) and fact (6) is not causal; instead, Assad having performed those acts is what it is to be a war criminal. To apply a notion gaining prominence in contemporary metaphysics, the acts “ground” the fact that Assad is a war criminal. They are the metaphysical reason that (6) obtains. ¹¹

One of the key tasks for social ontology is to work out how facts such as (6), (7), and (8) are grounded in general. That is, to fill out a formula such as

The task is to relate social facts to the other facts that ground them. We can call this “the grounding project.”

Usually, there is not just one set of grounds for a given social fact. Notice that fact (9) is sufficient on its own to ground fact (6). That is all it takes for Assad to be a war criminal. Equally, fact (10) is also sufficient on its own to ground (6). Typically, the fact that grounds a social fact is not essential; there are many possible or actual metaphysical reasons for the social fact to obtain.

Also, just because we have given one set of grounds for a social fact does not mean that the grounding project is finished. Facts (9) and (10), for instance, are themselves complex social facts, including things such as the Syrian civil war. Social facts are grounded by facts at many levels. For certain purposes, a theorist might be satisfied with (9) or (10) as an ontological explanation of (6). However, for other purposes, one might insist on going further and work through the grounds for (9) and (10).

If we can work through the grounds of a social fact—such as coming up with (9) and (10) as grounds for (6)—then, we have given an ontological explanation for the obtaining of the fact. The grounding project is fundamentally a project in explanation. As I pointed out earlier, however, it may be that certain ontological explanations are impossible. We might never, for instance, be able to work out the grounds for certain facts about institutions. They may be too complicated, or there may be more basic reasons preventing that from being done. Still, even if we cannot work out specific grounds for specific social facts, we might still be able to say something about social ontology in general. It might still be possible, for instance, to demarcate the sorts of grounds that social facts in general have. The ontological individualist, for instance, might argue that all social facts are grounded by facts about individuals. However, she might deny that, for a given particular social fact, we can work out what those grounds are.

Below, I will return to applications of the grounding project. The grounding project, however, is only one of two tasks for social ontology. One sort of question about facts such as x is a war criminal is about its grounds: in virtue of what such a fact obtains. However, there is also another set of facts that has a different metaphysical role in x is a war criminal. What, we might ask, make these the conditions for being a war criminal? Why do war crimes include such-and-such atrocities against civilians? Why is it that certain acts against soldiers count as war crimes and certain ones do not? What about the world and about us sets up the conditions for x is a war criminal?

An obvious, if oversimplified and not-quite-accurate, answer is that these conditions are a product of agreement or convention. The agreement or convention is not itself part of the grounds of facts such as (6), (7), or (8). Genghis Khan was a war criminal because he committed various atrocities in wartime. That is what it takes to be a war criminal: these atrocities are the grounds. Separate from the grounds are the facts that set up the conditions for war-criminality. That is the facts of agreeing, or convening, or whatever else.

This other set of facts I will call anchors, and the putting-in-place relation I will call the anchoring relation. It is a formula such as (11) that is anchored: a formula that articulates what grounds what. The anchors of (11) give a metaphysical explanation as to why x is a war criminal has the grounding conditions it does. This explanation relates a set of facts—facts about people agreeing, convening, or whatever else—to the conditions for being a war criminal. Like the grounding relation, the anchoring relation is not causal. There are many causal factors, for instance, that led to the Geneva Convention of 1948. Among them are the sentiments of the population on hearing of the atrocities of the Second World War. These sentiments may have been important causal factors leading up to the Geneva Conventions. However, it is the conventioneering in 1948, not the sentiments causing it, that partly anchors the conditions for x is a war criminal.

One theory of how (11) is anchored is this: by explicit agreements. Theories of convention, such as Lewis (1969), provide a different kind of answer. Still other theories take additional facts into account. For instance, certain theories of jurisprudence hold that a variety of factors figure into determining the conditions for facts such as x is a war criminal. They note that part of what puts these conditions in place are the facts involved in enacting statutes: votes and other actions taken by national legislatures, as well as approvals by national ministers, presidents, and other executives. However, they add that it is not enacting statutes alone that determine what the conditions for war-criminality are. Judicial interpretations over the years, for instance, also play a role in determining what the grounding conditions are. Moreover, actual decisions by juries and courts play a role as well. Equally important are acts and practices themselves: part of what goes into determining the conditions for war-criminality are actual tokens of acts performed in armed conflicts. All of these may play a role in determining which acts count as war crimes.

My aim is not to argue for a particular theory of anchoring here: these are just gestures at theories of anchoring. The key point is what these are gesturing at: they are attempts at giving a metaphysical explanation as to why x is a war criminal has the grounding conditions it does. These explanations relate a set of facts—facts about legislatures, executives, courts, juries, and tokens of acts in armed conflict—to the conditions for being a war criminal.

Anchoring and grounding: these are the two fundamental aspects to the building of the social world. Correspondingly, social ontology consists of two distinct projects. The grounding project is the inquiry into the conditions for social facts to obtain. What facts in the world are metaphysically sufficient reasons—that is, grounds—for social facts of some kind? The anchoring project is the inquiry into what puts those conditions in place. What sets up the grounding conditions for social facts, to be what they are?

4. Application and Significance of the Anchoring Project

There are many sorts of social facts. Some are quite formal, with explicitly set out grounding conditions. Others are just as strongly present in our social context, but without our ever noticing or conceptualizing them. The grounding–anchoring framework allows us to make sense of these differences.

The example I have been discussing—facts of the form x is a war criminal—has its grounding conditions laid out fairly explicitly in legal codes. As I mentioned, the statutes alone do not typically give the exact grounding conditions. Even so, x is a war criminal is a reasonably formalized category of facts, almost entirely laid out in written law.

Legal examples similar to this, however, can be a little misleading, if they are used as a model for social ontology: they can improperly suggest that all social facts are so explicit. To see that this is not so, contrast two different categories that have very similar grounding conditions, but that are anchored very differently. Contrast, for instance, facts about dietary prohibitions in different cultures. In Jewish scripture and rabbinical writings, the conditions for being an unkosher animal are laid out. These conditions are highly explicit and formalized. Anchoring the grounding conditions for x is unkosher are detailed rabbinical deliberation and enactment procedures. However, this of course is not the only way to anchor dietary prohibitions. Contemporary Americans regard as taboo many animals, including dogs, cats, insects, most lizards, some rodents, and so on. Many of these are not codified in laws, but are arguably as complex and elaborate as the distinctions between kosher and unkosher foods.

Working out the grounding conditions for x is an American-forbidden food is not simply a matter of examining enacted statues and case law. To understand the difference between this sort of informal category and the formal kosher category, we need to look at the anchors. What make the difference is not the grounding conditions themselves. Both x is an American-forbidden food and x is unkosher have complex grounding conditions. What makes the latter formalized is that the conditions for being kosher are anchored in more explicit ways than are the conditions for American taboos.

Or to make the example starker, consider a society with no legal system, and perhaps even no writing system, but that has dietary categories whose application conditions exactly match the kosher/unkosher ones. That is, the foods forbidden in this community are precisely the same as those forbidden in rabbinical Judaism. The grounding conditions for a fact of the form x is a forbidden food in that culture are identical (no pigs, no shellfish, meat slaughtered in a certain way, and so on). What make the kosher/unkosher categories formal are not the grounding conditions but how those grounding conditions are anchored. In the rabbinical case, they are mostly anchored by legal enactments, and in the case of the non-literate and non-legal society, they are mostly anchored by practices in the society. In this example, we have two sets of social facts, with identical grounding conditions. However, the grounding conditions of these facts are generated by different ways of anchoring, or what we might call “anchoring schemas.”

A common temptation in developing theories in social ontology is to hope to find the single mechanism by which social categories are set up. It is more plausible, however, that anchoring takes place in many ways. The anchoring project, plausibly, needs to look for a variety of anchoring schemas and to develop an understanding of why each of the various schemas works to set up social facts to have the grounding conditions they do.

Social categories, such as kosher, are practical tools that societies generate and apply for a variety of purposes. The labor required to set up the tools can come from one of many sources. In a society without a legal system, certain formal enactments are not available as anchors. The virtue of formal enactments is that they can make it easy to set up social categories that work for practical purposes. Formal enactment alone can be enough to make it determinate what the boundaries of a new social category are. It can direct members of the society where to go to investigate whether a category applies to a particular instance. However, a formal legal system is not, of course, required to set up practical social tools. Instead, the tools may be set up by widespread and long-standing regularities in practice, or they may be set up by widespread beliefs and attitudes, or by both, or by other means entirely. With any of these “setting up mechanisms,” a social category can receive enough definition to be practically usable.

Even in a society with a well-developed legal system, most social categories are not formally anchored. Moreover, the social categories that are formally anchored are frequently approximations of extant informal categories. When we do move to formalize a social category, the formal or legal anchors often remain just some anchors among many. In particular, just because we have introduced a legal dimension to a social category does not mean that the law has superintendence over the category. For instance, the category of marriage has an enormous number of legal connections and rules affecting it. However, marriage is larger than just the legal regulations. Marriage is a long-standing and widespread institution across the world and remains anchored, in part, by historical and ongoing practices. These practices can lag changes in legal regulation, or the law can lag changes in practice. This means that definitions in the law can get it wrong about marriage. The law has enormous bearing on marriage, but the category is a hybrid of law and sociology. It is easy to make the mistake of assuming that legal specifications are definitive. For some categories, they may be, but for ones with widespread and long-standing practices, such as marriage, we can be misled about grounding conditions if we assume that the law exhausts the anchors.

Understanding the anchors of a category of social facts can be enormously useful for working out its grounding conditions. If we wish to know the grounding conditions for x is a war criminal, we could learn a certain amount by probing our own intuitions or by asking the wider population about their beliefs, but it is certainly useful to know to look at the Geneva Convention, subsequent legislation, the cases of the International Criminal Court, and so on. To figure out the grounding conditions for x is unkosher, it is useful to know to look at rabbinical writings. Moreover, to figure out the grounding conditions for x is a forbidden food in the society without a legal system, it is useful to know to look at their practices.

The anchoring project is not only an instrument for helping us figure out the grounding conditions for social facts. It is largely a project in understanding how the social world is built, providing answers to certain broad and philosophical questions about the nature of the social world. In social ontology, we are not just concerned with what social facts obtain and why, but with why we have the social kinds we do, and what puts them in place.

The anchoring project also has potential applications in social critique. We can understand the grounding conditions for a social fact, but we may only be able to critique it when we see how and why it is anchored. For instance, if the grounding conditions for a particular social fact are partly anchored by a widely held false belief, then a way to “debunk” the concept may be by criticizing the anchors. ¹²

Finally, anchoring has potential to play a role in certain modeling projects in the social sciences. As I will discuss below, inquiry into grounding is usually more pertinent for building models. However, in certain domains, we are frequently anchoring and reanchoring new concepts. Investigating the anchors for a certain sort of fact can also be pertinent to modeling it, particularly in domains such as financial innovation, where new kinds are being anchored all the time. Still, what tends to matter for modeling tends to be understanding the grounds for social facts, not why those social facts have the grounding conditions they do.

5. Application and Significance of the Grounding Project

The anchoring project is of great philosophical interest, but it should not overwhelm the significance of the grounding project. Many of the philosophical projects in social ontology, and the bulk of projects in social science, depend more on getting the grounds for social facts correct than on coming up with an account of anchoring.

One thing that leads to confusion of grounding and anchoring is that some projects in working out grounds do not seem to be properly “philosophical.” It is not a philosophical project, for instance, to work out the grounding conditions for x is unkosher. This is an interpretive and descriptive project, a matter of investigating texts, practices, and characteristics of the natural world. However, that does not mean that social ontology is the study of anchoring rather than grounding. More general questions about grounding are, in fact, the bread and butter of social ontology. Even asking about food prohibitions in a more general way—for example, what are the grounds in general for facts of the form x is a forbidden food in culture y?—is a similar sort of question as more typical topics in social ontology, such as questions about the nature of corporations or money.

The most widely discussed topic in social ontology is also an inquiry into grounds: namely, the grounds of facts about group attitudes and group actions. Take a fact of the form group x intends J. Enormous attention has been paid in social ontology to explaining how facts such as these are grounded. Bratman (2014), for example, can be understood as providing an account of at least one way such a fact can be grounded for a particular sort of highly coordinated group. Or take a fact of the form group x judges P. List and Pettit (2011), for example, develop an account of the grounding of such facts. ¹³

No doubt, the work of these theorists would also profit from more attention to how facts such as these are anchored. Bratman, List, and Pettit, for instance, embed their theories of group attitudes and action into a loosely functionalist theory of group minds. Part of the explanation for these facts having the grounding conditions they do, in other words, is that they play their part in a certain kind of functional system. However, most of the topics surrounding this are hardly addressed in their work or the broader literature: exactly what the elements of this functional system are, how aspects of a group can realize such a system, and how playing roles in functional systems figures into anchoring facts about group attitudes. This is a topic that needs attention. Even so, the primary questions they are addressing are ones about grounding: what grounds various facts about group intentions, attitudes, and agency?

Social ontology is largely an inquiry into the grounds of social facts. It is striking, however, how limited a repertoire of social facts theorists tend to examine. It is not enough to ask—what is a corporation, or what is a social group? The social world includes a diverse range of facts, which may be grounding in surprisingly different ways. Consider, for instance, a group such as a basketball team. A grounding project needs to address the grounds of a wide variety of facts, such as the following:

This is only the beginning of a list. Getting clear on this diversity of facts is crucial in building models as well. When we model social facts, we are typically interested in a wide variety of them. To build good models, we need to understand the grounds of more than just a narrow subset.

For modeling the social world, inquiry into grounds is typically more important than inquiry into anchors. Sometimes, a model will just work out relationships between high-level social facts and other high-level social facts, such as the relation between inflation in an economy and the aggregate level of unemployment. However, much of the time, modelers attempt to break things down into their parts and then model the causal changes in social facts via causal influences on the broken-down parts. This “breaking down” is, at its heart, a matter of working out grounds. Rigorous philosophical inquiry into grounding has the potential to contribute to doing a good job with this crucial step in building models.

6. Structuring Social Ontology

Distinguishing grounds and anchors does not provide any particular guidelines on the correct theories of grounding or the correct theories of anchoring. It leaves open the possibility that there is a single way that social facts are grounded. Moreover, it leaves open the possibility that those grounding conditions are all anchored in one particular way. The distinction does not force a committed theorist to abandon a favored theory: the heart of most any current theory in social ontology is logically compatible with the grounding–anchoring framework.

It is, however, incumbent on a theory to be clear about the roles that a given set of building blocks plays in the social world. Consider, for instance, the roles that practices may play. Take a particular kind of dance: a tarantella, say. Various practices of dancing in particular ways, in particular contexts, are among the building blocks of a tarantella. However, it is ambiguous to put it this way. One role for dance practices is in grounding facts about tarantellas. A particular fact of the form x danced the tarantella, for instance, is plausibly grounded by dancing practices. However, in addition, a distinct role for dance practices is in anchoring facts about tarantellas. That is, dance practices may also play a role in explaining in virtue of what a fact of the form x danced the tarantella has the grounding conditions it does.

To put this differently, there are millions of tokens of tarantella dances, stretching back over time. Those tokens are, on one hand, tokens of tarantella dances, and on the other hand, they are also routines that anchor the category tarantella to be what it is. There is nothing wrong with practices playing this dual role: the very same dances are just doing two things. (In fact, particular tokens of a dance actually do lots of things: those very same tokens also have causal effects, for example.)

One of the insights of theories of practice is that tokens of our practices are involved in setting up our social categories. However, it is still easy to overlook the dual role for practices: A token of dance practices can play a role in anchoring categories, and the same token can play a distinct role in grounding some set of facts. A theory of practices should distinguish two metaphysical roles that the practices can have.

Keeping these distinct is helpful both for assessing a theory of social ontology on its own terms and also for assessing competing theories. A given theorist of practices, for instance, might hold that practices exhaustively anchor the grounding conditions for social facts. That is to say, it is practices alone, and nothing more, that fully explain why social facts have the grounding conditions they do. A different claim by a theorist of practices might be that practices exhaustively ground social facts. That is, social facts obtain in virtue of a pattern of practices obtaining, nothing more. A theorist of practice might embrace both—it could be that social facts are exhaustively grounded by practices and that their grounding conditions are exhaustively anchored by practices. This is not a contradictory position. However, such a theory faces enormous pressure: if practices are setting up the grounding conditions for social facts, then why should those grounding conditions themselves consist only of practices? This is an explanatory burden for the theorist that makes both claims.

A similar point can be made about a different kind of theory in social ontology: individualistic theories of the social world. What role do individuals play in making social facts obtain? A natural understanding of ontological individualism is as a theory of grounding, not anchoring. The individualist about facts such as Google is a profitable corporation, Assad is a war criminal, and the bourgeoisie is wealthier than the proletariat, on this view, holds that these facts obtain in virtue of facts about individuals. Once we fix the facts about individual people and their relations, that exhaustively determines that these social facts obtain. However, that leaves unaddressed whether individuals play an exhaustive role in anchoring. The individualist might again embrace both: that social facts are exhaustively grounded by certain sorts of facts about individuals and their relations and that these grounding conditions are also exhaustively anchored by those very same sorts of facts about individuals and their relations. Again, this is not contradictory, but it faces pressure: if facts about individuals exclusively anchor the grounding conditions for social facts, then why should those grounding conditions also consist only of facts about individuals?

It is likely that good answers to the anchoring project will be very different from good answers to the grounding project. Consider, for instance, John Searle’s “collective acceptance” theory of social facts. ¹⁵ Searle argues that social facts are put in place by a community’s collective acceptance of a constitutive rule. (Constitutive rules can be understood as a very simple way of giving the grounding conditions for social facts. ¹⁶ ) This theory is essentially a theory of anchoring. He argues that certain facts about collective acceptance put in place the conditions for the creation of certain social objects. And in Searle’s view, collective acceptance is just a matter of all the members of the community being in a certain cognitive state. On Searle’s theory, the anchors for social facts are built out of psychological facts about members of the community.

Certain old-fashioned theories in social ontology also took social facts to be built out of the psychological states of members of the community. ¹⁷ These “psychologistic” theories seem to have understood the subject matter of social theory to be the study of the “group mind.” These theorists were concerned with phenomena such as the “criminal crowd,” a crowd manifesting irrational desires that were different from the desires of the group members. They attempted to analyze facts about the group mind in terms of facts about the individual psychologies of group members. Unlike Searle’s, these theories are not theories of anchoring; they are theories of grounding. They take social facts—that is, facts about group minds—to obtain when certain facts about the minds of group members obtain. On the surface, the fundamental building blocks in both Searle’s theory and in old-fashioned psychologism are similar. However, they are theories attempting to answer very different questions.

In this discussion, I have not advocated a particular answer either to how social facts are grounded in general or to how the grounding conditions for social facts are anchored in general. My own view is these are both done in many ways. The grounds for social facts are radically heterogeneous. Social facts are neither exclusively grounded by facts about individuals, nor by any well-circumscribed set of other facts. We set up the grounding conditions for social facts to be a grab bag. The anchors for the grounding conditions of social facts are also radically heterogeneous. Their grounding conditions are not exclusively anchored by collective acceptance or by any other well-circumscribed set of other facts. In many cases, they are anchored by a mix of historical tokens, miscellaneous features of the environment, legal enactments, community beliefs and practices, and more. It is not that either grounding or anchoring is incomprehensible, or chaotic, or immune to theoretical analysis. There are often well-ordered mechanisms we use for anchoring and grounding, in building a social world that is practically useful, cognitively undemanding, and yet comprehensible enough to community members. However, that does not imply that we have one single anchoring schema, or that the anchors are simple, or that the grounds for social facts need to be chosen from a limited set. Rather, the opposite.

However, my aim has not been to argue this. Instead, it has been to propose some structure for dividing up theories and parts of theories that are often confused. In pursuing the projects of social ontology, we need to cleave social ontology from questions that are about causation and not ontology. Moreover, we need to divide inquiries in social ontology into at least their two key projects.