Are Social Kinds Vague and Does It Matter?

Introduction

Vagueness is a familiar and pervasive phenomenon in the social world. Many of the categories through which social life is organized, both in ordinary scenarios and in institutional settings, present some level of indeterminacy. Since social categories structure our expectations, coordinate behavior, and ground rights and obligations, their vagueness can have tangible consequences on many levels. In everyday contexts, for instance, it may be unclear whether a person (including oneself) belongs to a particular social group (Richardson 2024). In legal contexts, vagueness can have even more direct and consequential effects, since the application of a norm may hinge on indeterminate concepts (Endicott 2000). In extreme cases, vagueness can generate institutional disruption: electoral outcomes have been contested due to uncertainty about what counted as a valid vote, as in the case of the infamous “hanging chads” in the U.S. presidential election of 2000. ¹ Vagueness is also the source of challenges in social sciences: monetary economics, for instance, has long struggled to provide a stable and principled characterization of what counts as money. This bears direct consequence on classificatory practices in economics and has had concrete institutional consequences, as central banks have adopted different operational definitions of money, thereby shaping how they measure monetary aggregates and conduct monetary policy.

Yet social ontology has paid remarkably little attention to vagueness and the literature on this topic has remained scarce. Given the prevalence and significance of such examples, this neglect is rather surprising. This raises a natural question: why has vagueness remained largely peripheral within this field?

A possible explanation may lie in the central role that language and intentionality have traditionally played in theories of social ontology. Social reality has often been regarded as belonging to an ontological domain distinct from that of natural phenomena and natural kinds. While most accounts in social ontology (Ásta 2018; Epstein 2015, 2025; Guala 2016; Hacking 1999; Khalidi 2010; Mason 2021; Searle 1995, 2010; Tuomela 2002a) hold that social reality is in some sense ontologically dependent on humans, a prominent and influential tradition has offered a more specific characterization: the existence of social kinds must be traced back to the human capacity to form judgments and entertain beliefs. This commitment has been pushed further in influential work (Ásta 2018; Hindriks 2006; Richardson 2024; Searle 1995; Thomasson 2003a, 2003b) which holds that what depends on human minds is not only the existence of social kinds, but also the conditions that determine what counts as an instance of a particular kind—that is, their membership conditions. For these reasons, social kinds have traditionally been separated from natural kinds, as the former have generally been considered conventional and their terms’ extensions stipulated.

This way of framing social reality has made available both an intuitive explanation of vagueness in the social world and an apparently straightforward strategy for managing it. The first maintains that social kinds inherit their indeterminacy from the vagueness of the mental attitudes that bring them into existence; the second holds that if those constitutive attitudes can be sharpened, then the resulting kinds will likewise cease to be vague. Despite their initial appeal, these proposals end up trivializing the phenomenon: they make vagueness appear as a merely superficial obstacle by assuming that one can remedy the fuzziness of social kinds simply by manipulating their conceptual boundaries.

The aim of this paper is to assess whether this traditional and largely popular framework—and its strategies for handling vagueness—are theoretically and empirically adequate to address this phenomenon. The answer will be negative insofar as the traditional framework commits to the mind-dependence of social kinds’ membership conditions. This commitment inevitably renders social kinds epistemically transparent—the idea being that one has infallible epistemic access to kinds’ nature and their terms’ semantic extensions, without any need for empirical discovery. Specifically, I will argue that epistemic transparency, in this context, becomes a theoretically burdensome commitment that is incompatible with vagueness under any of its interpretations. Indeed, theories that rely on mind-dependent membership conditions cannot accommodate the existence of vague social kinds and thus risk eliminating vagueness from the social world altogether. This is deeply problematic, as social reality is not devoid of vagueness and indeterminacy: vague social kinds are exceedingly common and they must be confronted in the social sciences, legal contexts, and ordinary scenarios. The inability to accommodate such a pervasive aspect of social reality threatens to render these theories empirically and theoretically inadequate to the very phenomena they aim to explain.

Accounting for vagueness should be a desideratum for any theory that aims to offer an accurate and comprehensive characterization of the social world. Addressing vagueness is therefore not simply a general philosophical challenge, but a domain-specific issue that should concern social ontologists qua social ontologists.

This result is particularly striking, given that vagueness is often assumed to pose a greater challenge for realist accounts of social kinds than for anti-realist ones. Contrary to this assumption, I will argue that vagueness is more threatening for anti-realist frameworks: realists can easily accommodate vagueness as they do not need to commit to an overly restrictive conceptual machinery that effectively rules it out. Furthermore, the Homeostatic Property Cluster theory (Boyd 1989, 2000) provides a valuable conceptual resource to handle vagueness from a realist perspective.

The structure of the paper is as follows: in the second section, I make the case for vagueness in the social world, specifically by focusing on two examples: money and war. These two cases show how vagueness is routinely encountered and attested by social scientists, generating a relevant amount of disagreement among researchers, while also bearing important practical consequences. I conclude that theories in social ontology ought to account for vagueness (under at least one of its main interpretations) if they aim to be descriptively adequate with respect to the social world. In the third section, I will present a prominent tradition within social ontology, characterized broadly by the relevance given to language and intentionality in explaining social reality and, more specifically, by its commitment to mind-dependent membership conditions. This characterization is not intended to be exhaustive of all positions in social ontology; however, the commitment to mind-dependent membership conditions will be shown to be widely shared and embedded in influential theories in the field. In a subsection of the third section, drawing on Amie Thomasson’s work, I will show how this commitment turns the traditional framework anti-realist about social kinds. After having outlined the framework, the fourth section will examine two attempts to address vagueness using the conceptual resources provided by the traditional framework. The fifth and sixth sections present my argument concerning the epistemic transparency engendered by mind-dependent membership conditions and its incompatibility with vagueness. According to this argument, the commitment to mind-dependent membership conditions makes it impossible to accommodate vague social kinds. This is problematic, since it yields a picture of the social world as devoid of vagueness and indeterminacy. Yet vague social kinds appear to be pervasive, as attested by both the social sciences and the philosophical literature. The seventh section explores possible strategies for preserving mind-dependent membership conditions by examining whether certain subsets of social reality might be free of vagueness, drawing in particular on debates about vagueness in the philosophy of law. I argue, however, that the only available option is to give up the mind-dependence of social kinds’ membership conditions in order to avoid the challenges and paradoxes it generates. Finally, in the eighth section, I argue that realist accounts of social kinds are better equipped to accommodate vagueness, in particular by appealing to the Homeostatic Property Cluster (HPC) theory. I also show how this framework applies to the case of money and discuss the advantages it offers in handling vagueness.

Vague Social Kinds

Vagueness has been a central topic in philosophical inquiry since the last century, generating substantial amount of debate across a wide range of subfields, but for the present purposes it will suffice to recall the most salient aspects of this phenomenon. Vagueness concerns indeterminacy, typically manifesting in borderline cases. Consider a person gradually losing their hair: “John is bald” is clearly false initially and clearly true at the end, yet intermediate stages resist precise truth-value assignment. This indeterminacy can be understood in three ways. First, as a semantic phenomenon, vagueness reflects the limits of natural language. Second, an epistemic view locates vagueness in our limited knowledge: an omniscient observer could, in principle, identify the precise transition from non-bald to bald. Third, an ontic perspective treats vagueness as inherent to the world, emerging in objects, properties, and facts themselves. ²

As noted above, vague social kinds arise both in everyday practical contexts and in more theoretical settings. Importantly, vagueness can generate substantive disagreement and divergence among researchers. ³ Social scientists routinely grapple with vagueness and sometimes treat it as an ineliminable phenomenon. One might therefore expect social ontologists to engage more directly with the conceptual and methodological implications of vagueness, both for their own theorizing and in order to make sense of social scientific practice. In what follows, I consider two cases—money and war—whose vagueness produces both significant practical consequences and persistent theoretical uncertainty in their classification.

The concept of money is notoriously riddled with difficulties for monetary economics as there is hardly any consensus on what qualifies as money. This issue has been brought back into the spotlight by cryptocurrencies (Passinsky 2020), however it is part of a long-standing debate about what counts as a monetary asset. This is a problem which bears many practical consequences, particularly for the way institutions measure the amount of money circulating in a given economy, which, in turn, also influences their monetary policies. Despite efforts to standardize measurement practices globally, differences can still be found among national banks: for instance the Bank of England (BE) excludes units or shares by money market funds, while the European Central Bank (ECB) includes them in its statistics. This discrepancy can be attributed to the different definitions of money adopted by the two banks: BE uses an institutional definition, whereas the ECB uses a functional one.

Part of the difficulty of converging on a single definition can be attributed to the fact that many objects can perform monetary functions while displaying highly heterogeneous sets of properties. Such heterogeneity has generated challenges for sharp classifications in economics. Historically, a wide range of objects have served as means of payment, including banknotes, financial instruments, stones, cattle, cigarettes, and shells. These objects are not only physically different, but they also perform monetary functions to varying degrees and for different reasons. For instance, commodity money such as cigarettes can be accepted in trade both because it has a valuable use outside its monetary function and because, in the right context, it can be treated as a means of payment. By contrast, fiat money such as national currencies holds little physical value, at least insofar as one considers the physical qualities of banknotes and coins; nevertheless, it enjoys a high degree of acceptability in trade, as they are backed by central banks. Furthermore, in actual economies multiple objects can work as means of payment simultaneously, while differing greatly in their degree of acceptability in trade.

For these reasons, Frasser and Guzmán (2020) propose to renounce finding a sharp distinction, in favor of a model which integrates vagueness into the way we understand money. Indeed, by surveying contemporary monetary economics the authors found that the modern understanding of money does not allow to regard anything in an object’s fundamentals—namely its physical qualities—or in the beliefs attached to it, as intrinsic essential qualities enabling a categorical divide between money and non-money. Instead, the authors maintain that the passage from what counts as money and what does not is a gradual and vague one that does not allow for a precise cutting point. Such a view is expressed by the Liquidity degree view, endorsed by Frasser and Guzmàn, which does not posit an absolute standard for moneyhood and holds that in an economy multiple objects circulate as means of payment that differ gradually in their liquidity value. The value of liquidity can be described as the degree of acceptability in trade that a particular object enjoys, which determines how useful it will be for exchange. Such liquidity can be measured on a scale that lacks a critical point above which something can be considered money, there is only a continuum going from “not accepted in trade” to “universally accepted.”

Similar issues have been raised by the concept of “hybrid wars” in the field of political science and legal studies. Hybrid wars are conflicts where different types of warfare are mixed, while never reaching open armed conflict. It can be defined as a situation where “the line between war and peace time is rendered obscure. This means that it is hard to identify or discern the war threshold. War becomes elusive as it becomes difficult to operationalize it” (Almäng 2019, 191). Understanding whether a conflict qualifies as a war is relevant from a legal standpoint, since a state at war possesses rights and duties that states at peace do not have. It is also important from a scientific point of view, since it allows researchers to categorize the event correctly and to use it reliably for inductive inferences. However, the concept of “hybrid wars” by standing as a borderline case between peace and war, disrupts our ability of both adjudicating legally the conflict, and categorize it reliably.

Whether the vagueness surrounding such kinds should be understood as a consequence of linguistic indeterminacy, epistemic limitations, or the underlying structure of the kinds themselves depends partly on one’s preferred theory of vagueness, and partly on one’s broader commitments in social ontology. What matters for present purposes is that an adequate account of social reality must be able to accommodate vagueness under at least one of its standard interpretations. This is not a marginal requirement, as failing to satisfy it would result in depicting the social world as devoid of any vagueness or indeterminacy, a picture that is incoherent with the vagueness encountered in both ordinary and social scientific contexts.

In the next paragraphs, I will sketch how a prominent tradition in social ontology—one that assigns explanatory primacy to language and intentionality and treats social kinds as having mind-dependent membership conditions—has tended to encourage a deflationary attitude toward vagueness. Most importantly, I argue that this traditional framework by committing to the mind-dependence of membership conditions and thus to epistemic transparency about kinds, becomes structurally unable to account for vagueness and vague social kinds.

The Traditional Framework

There is an unquestionable sense in which things like marriages, money, governments, and cocktail parties can exist only insofar as humans exist alongside them. There cannot be a possible world with a functioning economy, an ongoing war, or a poorly governed state in the absence of any humans. ⁴ This intuition is broadly shared across social ontology, from social metaphysicians to game theorists. Yet this is also where theoretical paths diverge, as different approaches interpret this form of dependence in markedly different ways.

An influential tradition has taken human minds to be responsible not only for the existence of social reality and its kinds, but also to be the determinants of their nature. This idea has a strong initial appeal. Consider money: its existence presupposes humans, and its value appears grounded entirely in shared beliefs that is an exchangeable good, rather than in any physical property of the banknote itself, which is just a piece of paper. On this view, people’s shared beliefs and judgments ⁵ are able to endow physical objects with properties that they could not possess in virtue of their physical make-up alone or to bring into existence non-physical entities that can interact with us causally. Crucially, however, individual beliefs and judgments are insufficient; only shared, collective mental attitudes can perform this task. What emerges is a domain of kinds that are ontologically subjective—as they depend on mental representations—yet epistemically objective, in virtue of their intersubjective stability.

This perspective has fostered the idea that social reality is a projection of human minds onto the physical world, making social ontology largely an a priori enterprise concerned with understanding how the social world and its kinds come about on the basis of mental attitudes. Whenever we consider social phenomena, it seems that mental attitudes do not just play a causal role in their existence, but rather that the contents of our beliefs and judgments can (collectively) determine the conceptual boundaries of a specific phenomenon and also what counts as instance of it. This, in turn, has supported a sharp contrast between natural kinds, taken to be independent of our mental attitudes, and social kinds, whose nature appears to be shaped (at least partially) by human beliefs and judgments. ⁶

Part of this distinction hinges on the type of dependence attributed to social kinds with respect to mental attitudes: (1) causal mind-dependence or (2) constitutive mind-dependence. The former is generally considered unproblematic for realism about kinds. There is, indeed, a meaningful sense in which many natural kinds, like synthetically fabricated chemical elements or artificially selected animal breeds, can be described as mind-dependent. Human minds intervened in their creation, and may even have guided and planned it, thus occupying a clear position in the causal chain that eventually led to their fabrication. Constitutive mind-dependence, on the other hand, is a more problematic notion, that if attributed to social kinds would hinder realism about them. According to constitutive mind-dependence, minds and their attitudes do not merely participate in the causal history of a social entity but literally constitute it. Consider the case of marriages: there seems to be nothing over and above the belief that two people are married that ultimately constitutes the binding contract between them. ⁷

The notion of constitutive mind-dependence has come under controversy, as it is difficult to cash out exactly what it supposed to entail. ⁸ However, a way of capturing desiderata similar to this idea is to refer to the more specific claim that social kinds’ membership conditions are mind-dependent. This latter claim is largely shared in social ontology (Ásta 2018; Hindriks 2006; Richardson 2024; Searle 1995; Thomasson 2003a, 2003b) and is held to capture the idea that beliefs and judgments not only participate causally in bringing social kinds into existence, but also fix the kinds’ boundaries. Consequently, what it means to be a social kind, or an instance of it, is dependent on the mental attitudes that we entertain about it.

As mentioned, this claim is largely shared in social ontology, and it is embedded in theories that have shaped the debate in social ontology, such that of John Searle (1995) and in later influential theories like that of Ásta (2018) which was developed almost two decades after Searle’s first formulation.

Searle’s account of institutional facts is a paradigmatic case of this sort of mind-dependence. According to his theory, institutional facts arise through collective intentionality and are governed by constitutive rules of the form “X counts as Y in context C.” What distinguishes institutional from merely social facts is that they require the assignment of a status function—one that only holds in virtue of collective acceptance. Social kinds in Searle’s theory are dependent on beliefs or, more precisely, the existence of things of kind K is dependent on the acceptance of certain conditions concerning the nature of Ks. In other words, whenever we have a collective acceptance of constitutive rules, we are laying out the sufficient conditions for something to count as an institutional fact Y and to be an instance of a certain institutional kind K.

A similar commitment is found in Ásta’s conferralist framework. Social properties such as being black, gay, married, etc. are context-dependent properties conferred by other agents. Crucially, this conferral can succeed even when it fails to track the relevant base property: what matters is how the object or person is taken to be, not what it actually is. The umpire may misjudge the ball’s trajectory (base property), but cannot be wrong about calling a strike, since it is their judgment that constitutes it as such. What determines the conferral is exclusively the mental attitudes of those conferring the status. Thus, mental attitudes set up the membership conditions of kinds and fix the extension of their terms.

Both accounts converge on the idea that mental attitudes fix the membership conditions and extensions of social kind terms, illustrating how a significant tradition in social ontology conceives social kinds as robustly mind-dependent: not just as causally dependent for their existence, but as entities whose nature and instances are determined by collective beliefs and judgments.

Two Approaches to Vagueness

Johan Brännmark (2024) and Kevin Richardson (2024) have recently published work that strongly advocates for the recognition of vagueness within social reality. Brännmark offers an historical diagnosis of the reasons that precluded the discussion of vagueness within social ontology, while also providing an analysis of the sources of indeterminacy in social reality. Richardson, on the other hand, proposes a metaphysical argument that aims to explain how vagueness is engendered in social reality and how such vagueness can be ontic while not being problematic from a philosophical point of view.

Although not explicitly connected, both works build on the same core intuition. Drawing on the widespread view that social kinds are ontologically dependent on human mental attitudes, both authors explain vagueness as something inherited by social kinds from the beliefs and judgments that ground them. If social kinds are brought into existence by collective beliefs or judgments, and if these beliefs or judgments are themselves vague or indeterminate, then, so the argument goes, such indeterminacy is naturally, and unsurprisingly, inherited by the social kinds they constitute.

Here is the argument:

(1) If mental attitudes bring social entities into existence and,

Then

(3) the resulting entities will be vague.

Richardson (2024, 42) further points out that vagueness arises in cases where the membership conditions of a social kind depend on our mental attitudes, since any indeterminacy in our judgments is reflected in the indeterminacy of the kind’s own membership conditions. Thus, Richardson derives metaphysical vagueness from the semantic vagueness of judgments about kinds’ membership conditions (2024, 43).

Interestingly, Brännmark also suggests that given the argument outlined above, a second, and indeed opposite, line of reasoning can be plausibly developed. If we can exert control over the constitutive mental attitudes that bring social entities into existence, and if these mental attitudes can be rendered less vague or not vague at all, then the resulting entities, too, will exhibit a corresponding reduction—or elimination—of vagueness (Brännmark 2024, 73).

This latter point is particularly relevant, as it highlights a distinctive strategy for managing vagueness that would be available only within social ontology. In general, vagueness gives rise to a range of challenges, from issues of correct categorization to problems concerning inductive inference, that are not easily resolved. Yet, in the case of social ontology, there is a risk of trivializing this issue, insofar as it is assumed that social kinds possess conceptual boundaries that can be collectively fixed and subsequently adjusted or sharpened. Rather than circumventing the problem, this strategy seems to allow for the direct elimination of vagueness by refining the very elements that ground social kinds. This would be a rather unique result, as such a strategy is typically unavailable in other theoretical contexts. Importantly, even this strategy presupposes that there has to be an initial stage where vagueness has to be genuinely present.

Whether these strategies can genuinely succeed remains an open question. In the next sections, I present a novel argument showing why these accounts, and more generally all accounts that commit to the mind-dependence of social kinds’ membership conditions, cannot accommodate the phenomenon of vagueness.

Can the Traditional Framework Accommodate Vaguness?

Thus far, we have seen how the claim about the mind-dependence of social kinds’ membership conditions entails epistemic transparency for kinds. As a matter of fact, the traditional framework by committing to this claim inevitably turns anti-realist about social kinds, as it defies the principles of extensionality, error, and ignorance which, according to Thomasson, make up realism.

Naturalistically inclined philosophers might be disappointed with this outcome, as it leaves little room for empirical research about social reality. However, even those content with this view have reasons for concern.

In the next section, I will present a novel argument contending that the epistemic transparency engendered by the commitment to mind-dependent membership conditions is incompatible with vagueness under any of its standard interpretation: semantic, epistemic, or ontic. Specifically, epistemic transparency rules out the possibility that there could be any genuine vague social kind. I will also defend my argument from possible objections and argue that it holds even on weaker readings of epistemic transparency.

If the argument holds, it would imply that the traditional framework is bound to depicting social reality as inherently devoid of any vagueness or indeterminacy. This is problematic, because vague social kinds are a relevant part of social reality and their vagueness is what explains the divergence and challenges in classificatory practices. Not being able to account for vagueness then results in empirical and theoretical inadequacy in respect to the social world and the challenges faced by social scientists in their inquiry about it.

For this reason, Richardson’s and Brännmark’s attempts to address vagueness from within the traditional framework ultimately fail. The strategy of introducing conventional cut-offs for kinds’ boundaries is appealing, but it can succeed only if the framework can already make sense of genuine vagueness at an earlier stage. In discussing these attempts, I will refer mainly to Richardson. Although both accounts build on a similar intuition, Richardson’s provides a more complete argument of the inheritance of vagueness, while Brännmark’s work is more focused on a historical reconstruction.

Why Social Kinds Cannot Be Vague

Recalling the three standard interpretations of vagueness mentioned in the previous section—(1) semantic, (2) epistemic, and (3) ontic—let us consider whether, under the traditional framework, social kinds can be vague in any of these senses.

Let us suppose that vagueness is a semantic issue. If we assume that social kinds have membership conditions that are mind-dependent, then the terms’ extensions associated with them will have artificial boundaries determined by the conditions set by people’s collective mental attitudes. Given that extensionality does not hold for social kinds under this view, whatever falls into such extension should be infallibly knowable a priori through conceptual analysis. Thus, there is a tension between the idea that we should be granted complete and infallible access to a term extension and the extension being vague. Epistemic transparency seems to be a burdensome requirement that cannot be accorded with the inevitable vagueness that many times can be found in terms’ extensions. This suggests that within the traditional framework, social kinds cannot be vague in the semantic sense.

Similarly social kinds cannot be epistemically vague. If the error and ignorance principles do not hold for social kinds, then we can know a social kind term’s extension a priori and we cannot be wrong about the concept’s boundaries. What counts as a member of a kind is established by us and we cannot be massively wrong about a fact we have infallible epistemic access to. If we cannot be ignorant about the conceptual boundaries of our social kind concepts, then there cannot be a form of vagueness engendered by our ignorance.

So far, it seems that social kinds cannot be vague if we interpret vagueness as a semantic phenomenon or as an epistemic one. What about ontic vagueness? Under the traditional framework, it is not clear what it would mean for a social kind to be metaphysically vague, as according to this view social kinds are not real. Richardson proposed a viable solution: explaining ontic vagueness as the consequence of a prior semantic or epistemic vagueness supposedly inherited by social entities from the mental attitudes that ground them. To get his argument started, Richardson requires that there can be semantic vagueness concerning the judgments that fix membership conditions. However, by accepting the mind-dependence of membership conditions, Richardson also accepts that the resulting kind will be epistemically transparent. As I have already argued, however, such epistemic transparency rules out both semantic and epistemic vagueness in this context. Thus, “the chain of transmission” that transfers vagueness from mental attitudes to social entities cannot get off the ground, as there is no semantic vagueness that can be “turned” into ontic vagueness. Consequently, Richardson’s argument fails, and the ontic option remains therefore unavailable.

To sum up, the supposed epistemic transparency of social kinds, which is direct consequence of adopting the claim about the mind-dependence of social kinds’ membership conditions, is incompatible with social kinds being vague.

Which Strategies for the Traditional Framework?

The preceding discussion shows that the traditional framework’s commitment to mind-dependent membership conditions carries a significant theoretical cost. Since this commitment entails epistemic transparency, it prevents the framework from accommodating the many vague social kinds that populate social reality. This shortcoming is not minor: it threatens the framework’s ability to offer an accurate and comprehensive account of the very phenomena it aims to explain.

Given this difficulty, the question is no longer whether vagueness poses a challenge, but how proponents of the traditional framework might respond. Two strategies seem available. The first is to restrict the framework’s scope by focusing only on social kinds that are not susceptible to vagueness. The second is to relinquish the claim that membership conditions are mind-dependent.

The first strategy would consist in identifying a sizable subset of social kinds that are not susceptible to vagueness. By narrowing the scope of the traditional framework to this smaller class, one could preserve mind-dependent membership conditions, at the cost of rendering the traditional framework a less explanatorily powerful theory of social reality.

The best candidates for this subset seem to be social kinds of the third type, namely those that require propositional attitudes both at the type and token level. As Khalidi (2015) notes, these kinds tend to be present in heavily regimented environments, where they are defined by explicit rules and laws. For example, the conditions to become prime minister or a permanent resident are explicitly and arbitrarily chosen, and the functions, duties, and responsibilities associated with these roles are precisely specified through a network of legal norms. These kinds seem to have entirely conventional natures, which fits well with the idea that membership conditions are mind-dependent. Because their membership conditions are carefully stipulated, borderline cases seem less likely to arise, and where vagueness does emerge, it can in principle be corrected by refining the rules and laws that define the kind.

Legal and institutional kinds thus seem to fit the description of kinds whose nature is shaped by human beliefs and judgments and explicit stipulation. Yet vagueness also arises within legal kinds and is a well-known and contentious issue in both legal studies and the philosophy of law.

The law, by its very nature, requires stability, and judicial decision-making needs sharp lines in order to be applied clearly and consistently. However, the use of vague terms in the formulation of laws puts at great risk the fair and consistent application of legal principles. Timothy Endicott (2000) discusses an example from the Criminal Justice and Public Order Act, a law enacted in Britain in 1994, which employed the term “serious distress” to mark the threshold at which a certain level of loudness becomes a public offence. Clearly, both loudness and distress vary along a continuum, and identifying the exact point at which a sound causes “serious distress” invites a soritical series that cannot easily be resolved. The need in legal reasoning to impose crisp boundaries even across a gradual continuum, however, can have its downsides. Adam J. Kolber (2016) suggests that drawing clear lines where there would be ranges of intermediate cases in which the law does not clearly apply in either way, means that individuals can face harsh or no consequences depending on which side of the line they fall. ¹² Soritical series jeopardize the applicability of laws and open the door to judicial discretion without any clear guidance from the law.

Despite the fact that legal and institutional kinds may seem to possess membership conditions that can be fixed by our beliefs and judgments, they display fuzzy boundaries too as debates in both legal studies and in the philosophy of law illustrate. Given that there are no subclasses of social kinds that can be entirely not susceptible to vagueness, it seems that the only remaining option is to give up the claim that social kinds membership conditions are mind-dependent.

Can Realists Account for Vagueness?

Although the ability to fix kinds’ boundaries may initially appear to constitute a decisive advantage for anti-realist approaches, I have argued that this commitment ultimately excludes vagueness altogether. This is a significant limitation, since vague kinds are pervasive in social reality.

This raises the question of whether an alternative framework is better suited to accommodate vagueness. I suggest that realist approaches are, in fact, better positioned to account for the phenomenon. First, realists are not required to adopt the demanding conceptual apparatus that, within anti-realist frameworks, renders vagueness unaccountable for. Second, the broader philosophical literature on kinds provides well-developed resources for modeling vagueness from a realist perspective. In particular, Richard Boyd’s (1989) Homeostatic Property Cluster (HPC) theory, arguably the dominant account of kinds, was explicitly designed to accommodate the fuzziness often present in scientific classification. Although originally developed in connection with the biological sciences, it has since been extended to the social sciences (Ereshefsky and Reydon 2015; Mallon 2008 [2024], 2016).

On the HPC view, kinds are not defined by necessary and sufficient conditions. Rather, they are characterized by clusters of properties that tend to co-occur due to underlying homeostatic mechanisms. Such mechanisms are what lend kinds their objectivity and support inductive inferences about their members. ¹³ Importantly, there can be cases of imperfect homeostasis in which members of a kind possess varying sets of properties, sometimes only minimally overlapping with one another (Boyd 1991). This allows for substantial intra-kind variation and accommodates vague extensions.

This framework is well suited for capturing social kinds such as money. ¹⁴ The liquidity degree view maintains that there is no categorical threshold distinguishing money from non-money and historically a wide range of entities have played the monetary role: banknotes, financial instruments, cigarettes, and shells. This heterogeneity suggests that money is not governed by a fixed set of membership conditions, but is instead better modeled as a cluster kind. In this context, the relevant homeostatic mechanisms are plausibly provided by market conventions, institutional regulation, or monetary policy which can stabilize and reproduce correlations among the properties typically associated with monetary instruments (e.g., liquidity, acceptability, divisibility, and store-of-value capacity). Crucially, properties need not co-occur uniformly across instances of the kind, rather homeostatic mechanisms may generate monetary assets that possess heterogeneous and only minimally overlapping property sets. Accordingly, the existence of borderline cases is not anomalous but expected.

The HPC framework is compatible with the standard interpretations of vagueness. While one may wish to provide a more specific account of whether monetary vagueness is ultimately semantic, epistemic, or ontic, the HPC theory does not require a commitment to any one of these options. Rather, it provides a general metaphysical model of kind structure that can accommodate fuzzy boundaries regardless of how vagueness is ultimately diagnosed. This constitutes a further advantage over the traditional framework, which cannot accommodate vagueness under any of its standard interpretations.

Indeed, the traditional framework requires that the membership conditions of money are collectively accepted and epistemically transparent. This entails that the extension of the term “money” is set by conditions that specify what it means to be that specific social entity. This framework not only makes for a less flexible account of money uncapable of accounting for the heterogeneity of monetary assets, but also implies that (at least) the relevant experts must have an infallible epistemic access to the kind’s semantic extension. This implication is difficult to reconcile with the persistent and substantive disagreement within monetary economics over the definition of money and over which entities fall under its extension. Such disagreement strongly suggests that even experts lack infallible epistemic access to the boundaries of the kind, and that this is at least partly explained by the fuzziness of those boundaries.

Concluding Remarks

A central aim of this paper has been to show that vagueness is a topic social ontologists should take seriously, insofar as it raises non-trivial and domain-specific challenges for accurately representing social reality. Vagueness is not a marginal phenomenon in the social world, but a pervasive one: it arises in ordinary contexts—for instance, when considering the indeterminate membership conditions of social groups—as well as in theoretical and scientific contexts, such as monetary economics, where persistent disagreement about what counts as money poses several challenges to researchers and central banks.

For this reason, any theory of social ontology that aims to provide an adequate account of social reality must be able to accommodate vagueness under at least one of its standard interpretations. Whether vagueness should ultimately be understood as semantic, epistemic, or ontic will depend partly on one’s general theory of vagueness and partly on one’s broader commitments in social ontology. What matters for present purposes is that social ontology must leave conceptual space for vagueness in some form, if it is to make sense of the indeterminacy that can be found in social categories and social scientific practice.

I have argued, however, that a prominent and influential tradition in social ontology, the traditional framework, cannot meet this requirement. By endorsing the mind-dependence of social kinds’ membership conditions, the traditional framework inevitably incurs a commitment to epistemic transparency. Yet epistemic transparency is incompatible with vagueness on any of its standard interpretations. The result is that the traditional framework is structurally forced to depict the social world as fundamentally devoid of vagueness and indeterminacy. This is an implausible consequence, given the role that vagueness plays not only in ordinary social life, but also in classificatory practices in fields such as economics and political science.

This problem cannot be resolved by attempts to explain vagueness from within the traditional framework. Recent proposals by Brännmark and Richardson appeal to the plausible idea that vague beliefs and judgments may generate vague social entities. However, insofar as the framework’s commitments entail epistemic transparency, this strategy cannot succeed: the very feature that supposedly explains the conventional character of social reality also prevents vagueness from emerging in the first place.

The traditional framework is therefore faced with a dilemma. If it aims to accommodate vagueness, it must abandon mind-dependent membership conditions and epistemic transparency. Giving up these commitments would amount to a significant departure from what has historically been presented as one of the main explanatory payoffs of constitutive-dependence views of social kinds—namely, the idea that the conventional character of social reality can be understood in constitutive terms, and that social ontology can proceed largely a priori by analyzing the logical structure through which beliefs combine in order to construct social reality.

The upshot is that, contrary to a widespread assumption, vagueness poses a deeper challenge for anti-realist approaches than for realist ones. Realist accounts can acknowledge the ubiquity of vague social kinds without being forced into a conceptual machinery that rules vagueness out altogether. If social ontology is to offer a descriptively adequate picture of the social world, it must give up the idea that kinds’ membership conditions are mind-dependent and that social reality is epistemically transparent.