Social kind realism as relative frame manipulability

Abstract

In this paper we introduce the view that realism about a social kind K entails that the grounding conditions of K are difficult (or impossible) to manipulate. In other words, we define social kind realism in terms of relative frame manipulability (RFM). In articulating our view, we utilize theoretical resources from Epstein’s (Epstein, The ant trap: Rebuilding the foundations of the Social Sciences. Oxford University Press, 2015) grounding/anchoring model and causal interventionism. After comparing our view with causal and principle-based (Tahko, Synthese 200(2):1–23, 2022) proposals, we motivate RFM by showing that it accommodates important desiderata about the social landscape (such as recognizing the context-relativity of social properties and the emancipatory dimension of social practice). Finally, we consider three objections. First, we tackle frame-necessitarianism (FN), the view that social kind frames are metaphysically necessary (and thus unmanipulable). Secondly, we engage with what Epstein (Epstein, Philosophy and Phenomenological Research, 99(3):768–781 2019a) calls UNIVERSALITY (the view that social kinds can hold in the absence of anchors) and we argue that it should also be resisted. Finally, we tackle a recent objection from Mason’s (Mason, Philosophical Studies, 178(12):3975–3994) essentialism about social kinds.

1 Introduction

In this paper we introduce a novel theory of social kind realism; what we will call the Relative Frame Manipulability account (RFM). According to RFM, realism about social kind K is true iff the frame under which K holds is relatively difficult or impossible to manipulate. Roughly, the frame of a social kind K specifies the properties under which instances of K are grounded. RFM claims that the relative difficulty of intervening upon and modifying the set of grounding conditions of K, determines whether realism or anti-realism about K is true.

In Sect. 2 we consider two rival accounts, and we compare them to RFM. Specifically, we take RFM to be more fine-grained while also providing an informative conception of mind-independence. In Sect. 3 we develop RFM by drawing on Epstein’s (2015) grounding/anchoring model and the causal interventionist framework. In Sect. 4 we present several illustrations of how RFM can be applied to concrete cases of social kinds. In Sect. 5 we motivate RFM by showing that it has important virtues: taking the realism/antirealism distinction to be a spectrum, accommodating the relational nature of realism and anti-realism claims (both in terms of the relevant agents/groups, and the available epistemic and non-epistemic resources), and accommodating the political and emancipatory dimension of social practice. Finally, in Sects. 6–8 we tackle three objections to RFM: the challenge from frame-necessitarianism, the challenge from the possibility of anchorless worlds, and the challenge from essentialism.

2 Social kind realism

The discussion surrounding the realism/anti-realism distinction is vast (see Miller, 2021). We will not attempt to review that discussion here. Rather, we want to frame our account in terms of two recent and powerful proposals concerning the realism/anti-realism distinction when applied to the social domain: the causal and the principle-based account.

In doing this we go beyond existence-based accounts of realism (i.e., views of the form ‘realism about K holds iff K exists’).¹ Existence is a necessary condition for realism, but it isn’t sufficient. Intuitively, there are things that exist, but we are not realists about them. Money exists but realism about money isn’t obviously true. Also, scientific anti-realists take electrons to exist but argue that our theories about them are constitutive of them in some way or another (for discussion see Boyd, 1991: 143-4, 144; see also Mason, 2020: Sect. 2.1.).

This reflects a broader meta-ontological stance according to which ontology should be concerned not merely with what exists but, more crucially, with its mode or way of existence (Schaffer’s, 2009 view that reality has an ‘ordered structure’ reflects this broad stance in contemporary post-modal metaphysics). In this sense, the question shouldn’t be whether money exists but about the way in which money exists: e.g., does money exist in a way that warrants being a realist about it? For this reason, proposed accounts build upon this existence criterion. Take the following toy causal account:

(CAUSAL) Realism about P is true iff:

• (i) P exists.

• (ii) P’s causal powers are entrenched.

There are various ways of explicating clause (ii). Khalidi (2018) and Ereshefsky (2018) have recently proposed, roughly, the following: let P be causally entrenched when < P > is a projectible predicate (i.e., we can appeal to P to make predictions).

This interpretation of CAUSAL is supposed to be a response to a naïve mind-dependence account of anti-realism: the view that anti-realism about P is true iff P’s instantiation depends on certain mental states. Synthetic chemical kinds like roentgenium are typical counterexamples to such accounts: their instances cannot exist unless some agents bring them into existence (see Khalidi, 2016). But under CAUSAL this is irrelevant: roentgenium is causally entrenched and thus realism about roentgenium is true (see also Mason, 2016: 846).

The same story is supposed to apply to social kinds like money. Under some accounts, instances of money come to exist only via the intentional act of certain agents. Still, money is causally entrenched. But to our minds it isn’t obviously the case that realism about money is true in the same way that realism about roentgenium is true. Of course, it is open to the proponent of CAUSAL to bite the bullet and accept that there is no deep difference between these two cases. Still, prima facie, an account that differentiates between these two cases is better than one that doesn’t. If anything, even if realism about money is true, it could be that it is true in a qualitatively different way than the type of realism that applies to synthetic chemical kinds.

Tahko (2022) has recently attempted to capture that fine-grained distinction by appealing to what he calls ‘unification principles’ (UPs). Roughly, the causal profile of a property (or a set of properties) is sustained or underwritten by a unification principle. In turn, UPs could be understood as laws of nature, facts involving essences, or by appeal to one’s preferred metaphysical posit. A UP explains the causal profile of the relevant properties. Still, not every UP is created equal: some UPs are mind-dependent whereas others are not. UPs that are not mind-dependent plausibly include natural laws, non-natural moral principles, metaphysical laws of modality, logic, and so on. Mind-dependent UPs, on the other hand, are supposed to be conventional and highly contingent in nature (although see Khalidi, 2016).

Consider Tahko’s ‘C-rock’ example (2022: 11). Tahko argues that he could incentivise people to collect rocks from a specific beach in Cornwall by giving them beer as a reward. But there is nothing special about these rocks (‘C-rocks’ henceforth) other than they come from that specific beach. Still, due to the beer-incentive, a social regularity emerges that makes the kind C-ROCKS causally stable.² Therefore, C-ROCKS is causally entrenched because of the socially constructed UP (i.e., the makeshift C-rocks market exchange). But such a UP is clearly mind-dependent: it cannot exist without the intervention of agents. In other words, it seems that the way in which the social powers of a given kind are entrenched is a philosophically relevant factor. So, the resulting account is this:

(PRINCIPLE) Realism about P is true iff:

• (i) P exists.

• (ii) P’s causal powers are entrenched.

• (iii) P’s causal powers are explained by a mind-independent UP.

We take Tahko’s account to be a step towards the right direction. PRINCIPLE correctly focuses on the grounds of the causal powers of a putative kind, and also correctly locates the mind-independence criterion at that level.

Still, we think that PRINCIPLE is substantially incomplete. Specifically, it is not clear what the mind-dependence clause really amounts to. What does it mean for a UP to be mind-independent rather than mind-dependent? Tahko has correctly noted that the mind-independence of UPs is what is at stake in the realism/anti-realism debate but has left the nature of that mind-independence unexplained. This is a problem in its own right: we want a characterization of realism to be as informative as possible.

Relatedly, given that clause (iii) is left unspecified, we worry that PRINCIPLE doesn’t fare any better than CAUSAL in terms of distinguishing social kinds like MONEY from synthetic chemical kinds like ROENTGENIUM. Even if we grant that the C-rock UP (UP_c−rock) is plausibly mind-dependent in contrast to the roentgenium UP (UP_roentgenium), it isn’t self-evident (once again) that the same applies to the principle governing money (UP_money). Clearly, MONEY is more causally entrenched than C-ROCKS and this is explained by the fact that certain robust non-accidental generalizations about economics are true (see Mäki, 2021).

In turn, UP_roentgenium is plausibly governed by the laws of chemistry. But, according to PRINCIPLE, realism about MONEY would imply that UP_money is metaphysically akin to the laws of chemistry. This is certainly a view that one could have, but we take it to be controversial and uncharitable towards the realist about money (Mäki, 2021). Therefore, we hold that UP_money is substantially different from UP_roentgenium.

Specifically, we hold that the metaphysical difference between UP_money and UP_roentgenium holds even if one is (broadly speaking) a realist about both entities: we can change or alter the principles governing MONEY, but not the principles governing ROENTGENIUM. In this sense, we hold that the mind-independence involved in UP_money is of a different sort than the mind-independence involved in UP_roentgenium.

So, to recapitulate, RFM is trying to capture two important features that remain elusive under CAUSAL and PRINCIPLE: RFM aims to offer an informative characterization of the realist/anti-realist distinction by providing a characterization of the nature of mind-dependence, while also accommodating the intuitively plausible metaphysical difference between UPs like UP_roentgenium and UPs like UP_money.

3 Social kind realism as relative Frame Manipulability

Our proposal defines mind-dependence in terms of the ability of conscious agents to alter, manipulate, or intervene upon (we shall use these terms interchangeably) the frame under which a given social kind P operates. An initial characterization would be the following:

(RFM) Realism about P is true iff:

• (i) P exists.³

• (ii) The frame principle under which P operates is relatively difficult (or impossible) to shift.

There is a lot of ideology here to unpack. First, RFM builds upon Epstein’s (2015) powerful and influential grounding/anchoring framework.⁴ According to that framework, a given instance of a social kind K has certain grounding conditions: these are the conditions upon which instances of K are grounded. A frame principle governing K specifies the set of K’s grounding conditions. For example, a plausible first-pass characterization of the grounding conditions of money would involve having such-and-such physical features and being printed by a government body. So, these facts ground the fact that an instance of K exists. Epstein calls this the grounding project (see Epstein, 2019a: Sect. 1).

But what makes it the case that K has the grounding conditions it does? To illustrate, consider the naïve Searle (1995) view according to which various social facts are set-up via collective acceptance. So, it could be that the frame under which money operates has been set up via the fact that we have collectively accepted that money has such-and-such grounding conditions. In this example, the collective-acceptance facts are the anchors of K’s grounding conditions. Epstein calls this, the anchoring project.

We situate RFM in the context of the grounding-anchoring framework by focusing on the frames under which social kinds operate. We argue that if the frame principle governing K’s grounding conditions is difficult (or impossible) to shift, then realism about K is true. To say that a frame can shift means that at some possible world w the frame for K involves some set of grounds G₁ … G_n, whereas in some world w* the frame for K involves different grounds G₁* … G_n* in virtue of the relevant agents’ actions.

More concretely, consider the C-ROCKS kind once more (Tahko’s paradigmatic anti-realist case). As mentioned, the grounding-conditions for C-ROCKS are given by the following frame-principle:

(C-ROCKS-FRAME-1) Instances of C-ROCKS are grounded by the fact that they are shiny rocks, have a particular shape, and can be found on a specific beach in Cornwall.

Why does C-ROCKS-FRAME hold? In other words: what makes it the case that C-ROCKS has the grounding-conditions it does? Answer: because Tahko decided to make it so. Tahko’s mental states anchored C-ROCKS-FRAME-1:

(C-ROCKS-ANCHOR) C-ROCKS has the grounding-conditions it does because of Tuomas Tahko’s individual decision.

In Tahko’s example, the frame principle took a particular form. But given the nature of C-ROCKS-ANCHOR (i.e. Tahko’s individual decision), there are many other equally arbitrary ways in which the frame-principle could have been specified. For example, in some other possible world, Tahko’s mental states make it the case that instances of C-ROCKS are grounded differently:

(C-ROCKS-FRAME-2) Instances of C-ROCKS are grounded by the fact that they are not shiny rocks, have a particular shape, and can be found in beaches that are not in Cornwall.

In this sense, instances of C-ROCKS can operate under many different frame-principles (e.g. C-ROCKS-FRAME-1, C-ROCKS-FRAME-2, etc.).

To circle back to the realism/anti-realism distinction, RFM correctly predicts that realism about C-ROCKS does not hold: it is not the case that the frame principle under which instances of C-ROCK operate are relatively difficult or impossible to manipulate.⁵ Quite the opposite: moving from one frame to another (e.g. from C-ROCKS-FRAME-1 to C-ROCKS-FRAME-2) is as simple as a shift in Tahko’s mental states (we will consider more examples shortly).⁶

Two remaining notions of RFM require more elucidation: the notion of frame-shiftness and the notion of difficulty. What does it mean to say that a frame shifts or changes to another frame? To that end we appeal to a manipulability-based theory of causation (which we take to be both widely accepted and independently plausible). According to such a theory, roughly, an event C causes an event E insofar as there is a possible intervention upon the C-variable which results in a change in the E-variable. As per Woodward (2003), according to the interventionist framework causes are understood as ‘handles’ for manipulating effects. This tool gives us a neat characterization of frame-shiftiness:

(FRAME SHIFT) Let F be a frame principle specifying the grounding conditions G₁ … G_n of a social kind K. It is possible for F to change, iff, there is a possible intervention upon F which would result in F specifying a set of different grounding conditions G*₁ … G*_n.

On the standard Menzies and Price (1993) characterization, the notion of intervention is characterized in terms of facts about human agency. On such a view, saying that there is a possible intervention on some variable P is the same as saying that it is possible for an agent to intervene on P. As we will argue later, we take this to be a feature of our view.

What about the appeal to the relative difficulty of frame shifting employed in RFM? The idea is simple and was already hinted on during the C-ROCKS example. Some frame-interventions are easy to implement, whereas others are not. This is a direct consequence of the fact that the notion of intervention in FRAME SHIFT is understood in terms of human agency: clearly, it is easy for people to do certain things in contrast to other things.

More concretely, consider two possible frame-interventions that are nevertheless different in terms of their relative difficulty: frame-interventions about C-ROCKS and frame-interventions about MONEY. The grounding-conditions of C-ROCKS are, ex hypothesi, very easy to change: it would suffice for Tuomas Tahko to change his mind concerning the set of grounds of C-ROCKS (e.g. by taking non-shiny rocks to count as grounds for C-ROCKS).

This is plausibly not the case, however, for the MONEY-kind. Let us suppose that an instance of MONEY holds in virtue of the fact that it is printed by X-institution.

(MONEY-FRAME-1) Instances of MONEY are grounded by the fact that they are printed by X-institution.

What anchors that fact? According to one (simplistic) state-centred theory: the fact that the state made it the case that X-institution grounds instances of MONEY.

(MONEY-ANCHOR) MONEY has the grounding-conditions it does because a given state made it so.

Still, things could have been different. Due to the nature of MONEY-ANCHOR, in another possible world, the state makes it the case that Y-institution grounds instances of MONEY. If so, then MONEY would operate under a different frame:

(MONEY-FRAME-2) Instances of MONEY are grounded by the fact that they are printed by Y-institution.⁷

In the case of MONEY, in contrast to C-ROCKS, moving from one frame to another is significantly more difficult because of how entrenched money is in social relations. For example, consider Javier Milei’s plan to abolish the peso and adopt the US dollar. Doing so requires significant legislative and non-legislative work (e.g. devaluing the peso) not to mention that Milei needed to be elected first for it to be even considered moving towards a different frame. So even though C-ROCKS and MONEY are similar in the sense that some form of human intervention can shift the relevant grounding-conditions, they differ significantly in the amount of difficulty it takes to move from one frame to another.

4 Examples

Let’s apply RFM to even more cases.

ROENTGENIUM. First, consider a clear example of realism according to RFM. Instances of ROENTGENIUM, by belonging to a synthetic chemical kind, can only be brought about via the mediation of agents. Still, they plausibly operate under certain laws of chemistry. And, bracketing quixotic metaphysical theories, laws of chemistry are impossible to manipulate. This is the correct result: realism about ROENTGENIUM is intuitively true.

MONEY. Now consider a case of RFM-realism where the frame is relatively difficult to manipulate. We have already considered a state-centred theory of money in the previous section. While it is possible to change the frames under which instances of MONEY operate, it is very difficult to do so. This is because it seems that irrespective of one’s preferred theory, money will end up being deeply ingrained into the social fabric.

The same applies to other, more sophisticated, theories. In the literature about the nature and origin of money there is substantial disagreement over the relevant entrenchment of money to social institutions (Passinsky, 2024). One tradition understands money as corresponding to the development and evolution of human economies (e.g. Marx, 1992). Other traditions focus on the idea that money is contingently institutionalized, by invoking examples of communal control over particular forms of money like debt (Graeber 2011). Irrespective of which tradition fares better empirically, RFM makes sense of what is at stake in this debate: how easy it is for societies to control money.

To illustrate, consider that both traditions can agree on what grounds (and, for the most part, what could anchor) instances of money. In other words, they can be in agreement in terms of the grounding-project:

(MONEY-FRAME-1*) Instances of MONEY are grounded by the fact that they are instances of gold.

(MONEY-FRAME-2*) Instances of MONEY are grounded by the fact that they are instances of cryptocurrencies.

Still, there can be disagreement on what sets-up or anchors a given frame-principle. And, crucially, based on that disagreement there can disagreement on how difficult it is to shift from, e.g., MONEY-FRAME-1* to MONEY-FRAME-2*.

To illustrate, consider a functionalist theory according to which instances of MONEY play the ‘money-role’:

(MONEY-ANCHOR*) MONEY has the grounding-conditions it does in virtue of the fact that such conditions allow instances of MONEY to play to ‘money-role’: the role to generate patterns of behaviour which can be modelled as Nash-equilibria (Guala & Hindriks, 2015).

A Nash-equilibrium is a solution to a non-cooperative game in which every player has knowledge of the choice-set of the other players and has nothing to gain by changing their own strategy. The important point here is that, plausibly, there are many ways in which those kinds of solutions can be provided (i.e. there are many ways in which the MONEY-role can be fulfilled).

Perhaps in one economy that role is fulfilled by bars of gold and in another it is fulfilled by cryptocurrencies (although see Passinsky, 2020). If so, there are at least two possible frames for MONEY: MONEY-FRAME-1* and MONEY-FRAME-2*. And shifting from one frame to another, assuming MONEY-ANCHOR*, would take significant effort: moving from an economy regulated by bars of gold towards a decentralized economy based on cryptocurrencies is a non-trivial task (to say the least).

For comparison, consider another theory. The credit theory of money takes instances of MONEY to act as promises from someone to grant a product or a service to the holder of a token of MONEY (see de Bruin et al. 2023: Sect. 1; Passinsky, 2024: 41 − 3). Collective acceptance plays a crucial part in this process: we need to collectively agree (through a series of very complicated series of promises upon promises) that certain tokens ground instances of MONEY.

(MONEY-ANCHOR**) MONEY has the grounding-conditions it does in virtue of the fact that we have collectively accepted (roughly, in an environment of trustworthiness) that this is the case.

Again, this kind of anchor can plausibly set-up both MONEY-FRAME-1* and MONEY-FRAME-2*. And, as hinted, there are constraints at play. Credit relationships need to operate in an environment of trust: issuers must be ‘creditworthy’, and the credit itself must be transferable.

Note, however, that these constraints are less stringent than the ones that apply to MONEY-ANCHOR*. According to MONEY-ANCHOR*, a shift from a frame to another will require generating a social pattern which can be modelled as a Nash-equilibrium. And whether such an equilibrium is generated is not dependent on any collective acceptance from the part of the agents (Colombo & Guala, 2023).

But according to MONEY-ANCHOR**, there can be MONEY-frames which do not generate a Nash-equilibrium: what is required, instead, is a complicated process of collective acceptance based upon promise-keeping. So, the key difference between these theories lies at the level of difficulty it takes to move from one frame to another (depending on the relevant constraints at play).

GENDER. A good example of anti-realism according to RFM is Ásta’s account of gender. For Ásta (2018), gender is a conferred social property that varies substantially from context to context:

[Y]ou work as a coder in San Francisco. You go into your office where you are one of the guys. After work, you tag along with some friends at work to a bar. It is a very heteronormative space, and you are neither a guy nor a gal. You are an other. You walk up the street to another bar where you are a butch and expected to buy drinks for the femmes. Then you head home to your grandmother’s eightieth birthday party, where you help out in the kitchen with the other women while the men smoke cigars (Ásta, 2018: 73).

According to Ásta, a shift in context (in the appropriate way) entails a shift in one’s gender. Or, in RFM-lingo, a shift in the frame one operates under entails a shift in one’s gender.

(GENDER-ANCHOR) GENDER has the grounding-conditions it does in virtue of the fact that such conditions have been socially conferred.

As per Ásta’s example, in the office one is “one of the guys” whereas in the context of their grandmother’s birthday one is a woman. So, in this case, the nature of GENDER-ANCHOR allows for a great deal of frame-variability:

(GENDER-FRAME-1) Instances of GENDER are grounded by behaviour and bodily features X.

(GENDER-FRAME-2) Instances of GENDER are grounded by behaviour and bodily features Y.

Crucially, according to GENDER-ANCHOR, frame-shifting is easy: it merely takes a shift in one’s context (e.g. from GENDER-FRAME-1 to GENDER-FRAME-2).

By contrast, let us consider a realist theory of gender under RFM. Here’s Bach’s (2012) historical essentialist theory:

According to historical essentialism, men form a natural kind and women form a natural kind. The essential property of women, in virtue of which an individual is a member of the kind “women,” is participation in a lineage of women. In order to exemplify this relational property, an individual must be a reproduction of ancestral women, in which case she must have undergone the ontogenetic processes through which a historical gender system replicates women. (2012: 271)

Belonging to the kind WOMAN, according to Bach, requires participating in the lineage of women:

(WOMAN-FRAME) Instances of WOMAN are grounded by the fact that they belong in the ‘lineage of women’ (i.e. by the fact they participate in a very specific ontogenetic historical process involving features such as being structurally oppressed, etc.).

Why does WOMAN-FRAME hold? Given Bach’s theory, and the fact that the kind WOMAN is supposed to be metaphysically on par with biological natural kinds, the relevant-grounding conditions are determined by purely evolutionary mechanisms.

(WOMAN-ANCHOR) The grounding-conditions of WOMAN are determined by specific evolutionary mechanisms.

And, importantly, given the nature of WOMEN-ANCHOR, frame-shifting is impossible. One does not get to choose in what lineage they participate. And, for this reason, one does not get to choose whether they are a woman or not (for criticism see Mikkola, 2016). In this sense, there is no amount of human intervention that could ever bring a difference in the frame-principle governing the kind WOMAN.

SOCIAL CLASS. A final example. Consider two different theories concerning the metaphysics of social class. Take a broadly Marxist theory of class according to which one’s social identity is ultimately determined by the structural features of an economic system (see Wright, 1997). Depending on the kind of Marxist theory one adopts, such structural economic features will be determined, in one way or another, by the development of economic laws. In RFM-terminology, this rough characterization gives us the frame principle and its anchor (take the WORKING CLASS-kind as an example):

(WORKING-CLASS-FRAME) Being a member of WORKING CLASS is grounded by (among other things) being the subject of exploitation by the capitalists via the appropriation of surplus value.

(WORKING-CLASS-ANCHOR) The grounding-conditions of WORKING CLASS are determined by purely structural economic features which are entailed by economic laws.

The nature of WORKING-CLASS-ANCHOR makes it the case that there cannot be variation in the frame governing WORKING CLASS. This is so because one cannot manipulate the economic laws that anchor WORKING-CLASS-FRAME.⁸

Now consider Bourdieu’s theory of class. For Bourdieu (2010), class is ultimately understood in terms of a multiplicity of historical processes involving both economic and socio-cultural features. Certain kinship relations (that might be purely arbitrary) create different spaces (each space being called a ‘habitus’) that can be individuated in terms of aesthetic elements such as ‘commonality in taste’, as well as cultural similarity. So, for Bourdieu, being a member of WORKING CLASS is grounded in terms of one’s participation in the appropriate kinship network:

(WORKING-CLASS-FRAME*) Being a member of WORKING CLASS is grounded by (among other things) in one’s participation in a particular kind of kinship network.

But the cultural and aesthetic elements that we associate with WORKING CLASS are anchored by largely contingent regularities:

(WORKING-CLASS-ANCHOR*) The grounding-conditions of WORKING-CLASS are determined by the contingent cultural mechanisms underlying the working-class kinship network.

In other words, there are possible worlds whether the social kinship network we have attributed (let us suppose) to the working-class is of a different sort that the actual one. Or, in RFM-terminology, there can be different frames under which the WORKING-CLASS kind can operate.

To be fair, it is very difficult to bring about a frame-shift under Bourdieu’s theory. Doing so would be a long historical process trying to bring about cultural change. So we are inclined to say that some realist intuitions are going to be accommodated under Bourdieu’s theory. Still, what we are trying to highlight here is that even though the Marxist theory and Bourdieu’s theory can be both understood as realist theories of class, nevertheless they are importantly different (i.e. to the degree in which there can be frame-variability). This is the sort of fine-grained metaphysical insights that RFM can provide.

5 Motivating RFM

According to RFM, realism about a social kind K is true iff K exists and it is relatively difficult (or impossible) to manipulate the K-frame. In this section, we will further motivate our account by noting various interesting consequences of RFM.

We noted that an account concerning the realism/anti-realism distinction should be informative but also able to distinguish between different types of social kind realism and anti-realism. RFM accommodates both desiderata. First, RFM replaces talk of ‘mind-independence’ with the more intricate notion of ‘frame-manipulability’ by drawing inspiration from the causal interventionist framework. Secondly, the notion of frame-manipulability allows us to make fine-grained distinctions between different types of theories. For example, as discussed, the kind of anti-realism that applies to MONEY is different from one that applies to C-ROCKS: it is significantly more difficult (though not impossible) to manipulate the MONEY-frame compared to the C-ROCKS-frame. And it is a feature of RFM that it can capture that difference.

An interesting implication of RFM is that it is importantly agent-relative. We take this to be a feature of our account. First, it makes our account particularly fine-grained in a metaphysical sense. Under RFM the social world has an intricate structure that involves a full range of entities that fall on the realism/anti-realism spectrum. In other words, the world involves entities that fall under frames that are very difficult (or impossible) to manipulate, entities that fall under frames that are very easy to manipulate, and all the entities that fall between those two ends. But also, given the agent-relative nature of RFM, every relative frame-manipulability claim would be relativized in terms of a particular individual or group of individuals thus allowing us to identify the intricate structure of the social world.

For this reason, RFM is particularly fitting for our emancipatory practices. It highlights the idea that different social kinds have a different metaphysical status for different individuals or groups of individuals. Consider the monetary system: imagine how difficult if not impossible it is for ordinary individuals to change the value of goods. By contrast, corporations can manipulate prices in the stock market to make a profit. Trade unions can negotiate wages and hence change the value of at least one important commodity: labour power. Finally, a government has the power to change prices at will. Of course, a government does not have absolute power in this decision as it is also determined by external social forces (i.e., socio-economic laws). At any rate, the difference between the kind of influence that different types of agents can exhibit is evident.

Another way frame-intervention claims can be relativized is in terms of the availability of various epistemic and non-epistemic resources at a given time. It is plausible that the manipulability of various frames depends on the material and theoretical resources that are available at a given historical moment (see Boyd, 1988). For example, the working class, the queer movement, racially oppressed groups, etc. systematically produce sociological, economic, political, and philosophical tools to understand what restricts their emancipation. The actions of the social movements come alongside the public justification of their aims in the form of brochures and manifestos. In such initiation pamphlets we do not only observe the goals and demands of these movements but also ideas on how they understand the world they live in and fight against (de Sousa Santos, 2018).

It is precisely for this reason that an integral function of various social movements involves looking into the nature of oppressive power structures and identifying whether they can be manipulated given the available resources (and, subsequently, how difficult it is to do so). For example, consider the case of the nature of mental health (i.e., the frame governing mental disorders and well-being) and its relation to political action.

If mental disorders operate under a frame that is impossible to change, perhaps because these frames are anchored by unalterable laws of psychiatry, then this would shape our political demands appropriately. In such a case, we would demand more resources to be allocated to scientific research about the origin, prevention, and treatment of these disorders, etc.⁹ On the other hand, if as the anti-psychiatry movement argues (see Szasz, 1962; Foucault 2006/ 1961) the relevant frames are anchored by oppressive and power-structures, then our political goals would be to scrutinize, criticise and eventually abolish these structures (note that doing so is no easy task for the oppressed).

It could be argued that the inherent agential aspect of RFM introduces an objectionable kind of relativism into the picture. The kind of (anti-realism) that applies to social kind K is, according to RFM, agent-relative and relative to the epistemic and non-epistemic resources available to those agents.

But we find this kind of relativism metaphysically unobjectionable. It is not the case that ‘anything goes’ in the sense that we can shift around the grounding conditions of a kind in whatever way we see fit. Rather, it is an objective (or, in RFM-terminology: un-manipulable) matter whether realism or anti-realism is true about a kind K for agent A under conditions C. In other words, we take for granted that the world has a fixed metaphysical structure. And it is that very structure that determines whether realism about K (relative to agent A and conditions C) is the case or not.

There is a useful precedent here. Boyd (1999) calls his theory of natural kinds (the so-called ‘accommodationist’ model) ‘bicameral’. Roughly, he takes a kind K to be a natural kind iff (i) the members of K satisfy some disciplinary demands, and, (ii) the causal structure of the world actually accommodates those demands. This model is bicameral in the sense that both the world and our interests need to come together in order for a kind to be understood as natural.

RFM is bicameral in a similar sense. Whether realism about K is the case or not is, firstly, determined by the nature of K. Is K anchored in way that allows for frame-variability? For example, in the case of Bach’s theory of gender, it is impossible to move from one frame to another given the fact that they are ultimately underwritten by evolutionary forces which are not prone to human intervention. In the case of a state-centred theory of MONEY, however, moving from one frame to another is possible but relatively difficult.

This brings us to the second consideration concerning RFM’s bicameralism. In cases where frame-variability is at least in principle possible, the degree of difficulty in which these frames can be manipulated will be different depending on one’s social circumstances. As mentioned, one’s social position as well as one’s access to the relevant epistemic and non-epistemic resources determines, together with considerations about the nature of K, whether realism about K is the case or not (relative to that position). So, in this sense, even though there is undoubtedly an agential dimension to RFM, we find it not only palatable but also theoretically fruitful.

Here is a related worry.¹⁰ Let us assume that for a kind K and its frame F it is very difficult for group A to manipulate F but it is easy to do so for group B. But if group A realizes that group B have it in their power to change F, then shouldn’t this prompt group A to become anti-realists about K even though they cannot themselves change F? In other words, isn’t it enough that it is in principle possible for F to be manipulated, to conclude that anti-realism about K is true?

We take it that the objector wishes to relativize claims about (anti-)realism in terms of cognizant agents in general instead of groups and their related circumstances (as we did in our earlier discussion). But note that the definition of RFM (see Sect. 3) is pluralistic: it makes no explicit mention of groups and, thus, it can accommodate various types of relativizations depending on one’s theoretical demands.¹¹ In some contexts, it might be theoretically useful to consider whether realism about K is true or not, in a coarse-grained manner. In such a case we would consider whether cognizant agents in general have it in their power to manipulate F (the frame under which K operates). But in other contexts (presumably in the politically relevant ones we’ve sketched earlier) where more fine-grained metaphysical work is needed, we would adopt a relativization in terms of groups and their circumstances.

Perhaps the suggestion is that one should always relativize in terms of cognizant agents instead of some subset of those agents. But we find this view implausibly stringent. In fact, we have provided two types of considerations which illustrate that relativizing in terms of groups and their circumstances is theoretically fruitful. Doing so allows us to identify the highly fine-grained structure of the social realm (see, e.g., the examples on Sect. 4). Also, identifying the intricacies of that structure is politically useful: as mentioned, knowing the difficult by which a frame can shift for group A is necessary in determining A’s political action and agenda (see, e.g., the examples about money on Sect. 4 or the anti-psychiatry case mentioned earlier).

Let’s circle back to the objector’s case. We agree that there is a general sense in which anti-realism about K is the case (i.e. for both groups A and B). Specifically, such a result is delivered once we adopt what we’ve called a coarse-grained relativization of RFM: a relativization of social (anti-)realism claims in terms of the ability of cognizant agents (in general) to manipulate the frame under which K operates. But, crucially, RFM’s inherent pluralism allows us to adopt multiple types of relativizations as part of our philosophical toolkit. One such relativization is a fine-grained relativization in terms of groups and their circumstances. And in terms of that assumption, realism about K is true for group A and anti-realism about K is true for group B.

Note that the worry here cannot be that different claims about realism (relativized in terms of different groups of agents) are mutually contradictory (in the same sense that certain kinds of moral relativism account for moral discourse). There is nothing contradictory in saying that a given frame F is difficult to change for group A but easy to do so for group B. And the fact that group A is aware of group B’s ability to change F does not affect their own ability to bring about such a change. We thus conclude that RFM does not have any objectionable relativistic consequences.

6 Against frame-necessitarianism

For RFM to be a useful philosophical tool, it should be that at least some frames can be intervened upon. What we will call Frame Necessitarianism (FN) goes against this view. According to FN, every frame is metaphysically necessary. For example, if according to the war-criminal frame being a war-criminal is grounded by the conditions specified in the Geneva Conventions, then this is metaphysically necessarily so. So, the idea is that if FN is true, then we cannot make sense of frame-interventions given that metaphysically necessary entities cannot be intervened upon.¹²

FN is the social analogue of certain necessitarian views about the metaphysics of lawhood which take laws of nature to hold with metaphysical necessity (e.g., Bird, 2005). FN is also entailed by essentialist views of social kinds (Mason, 2021; see also Epstein, 2015: 124, 128; 2018: 6; 2019a: 777-8). In this sense, FN is a view that deserves serious consideration.

Consider the positive case for FN. Epstein (2019a: 772) has recently argued for the following thesis:

(UNIVERSALITY) At a possible world w the fact that x is K ([x is K]) has anchors [A₁ … A_n]. At w* [x is K] can hold in the absence of [A₁ … A_n] (and without there being any substitute anchors at w*).

A plausible argument (via inference to best explanation) for FN would go like this:

(1) UNIVERSALITY is independently plausible.

(2) UNIVERSALITY is best explained by FN.

(C) So, we have good reasons to accept FN.

The idea is that the reason why there can be social facts like [x is K] in the absence of underlying anchors at w, is that w still involves an unbounded frame specifying the grounding conditions for K at w. This is because, as per FN, if a social kind falls under a frame, then the same holds in every possible world. In our response, we take issue with (2). Specifically, we will argue that there is a better way of explaining UNIVERSALITY which is also compatible with RFM.

The main motivation of UNIVERSALITY comes from the fact that it allows us to track kinds in (historical) places where the relevant anchor is absent. For example, take the alleged fact that Genghis Khan was a war criminal. According to our intuitions, it is permissible to claim this even though there was no juridical system to set-up the anchors for being a war criminal back when Genghis Khan was alive.

The issue goes even beyond mere intuitions. So-called etic analyses employed by anthropologists precisely involve the application of certain social categories to contexts which are significantly foreign to ours (see Epstein, 2019a: 771). For example, Marxists talk of class struggle or exploitation in pre-capitalist societies (Marx, 1965) and feminists speak of sex/gender systems throughout history (Rubin, 1975). UNIVERSALITY would explain why such a theoretical move is warranted. We follow Schaffer (2019) and call the intuitions generated by cases that export certain social categories to distant possible worlds, exportation intuitions.

Still, we argue that the appeal to etic analyses to defend FN should be resisted for at least two reasons. First, even though it is certainly true that etic analyses are used routinely in the social sciences, it is also true that there are clear-cut cases where etic analyses are unwarranted or, at the very least, controversial. We are referring to the charge of anachronism which occurs when a term is wrongly taken out of its historical context and is inappropriately applied at a different historical era. For example, consider problematic accounts which project ideas of national consciousness to ancient civilizations in what is called ‘retrospective nationalism’ (see Smith, 1998).

Or consider analyses that infer gender norms from certain perceived gender traits in art. For example, take the controversial case of Cycladic female figures: roughly, the existence of female marble figurines in Ancient Aegean civilization (which date back to 4500–4000 BCE) are taken to indicate the existence of a matriarchic period. This is based on certain observed features of these figures which are said to indicate fertility (e.g., exaggerated legs and buttocks). But this kind of analysis has been resisted in the literature for resting on sexist background beliefs (see Meskell, 2017). It is simply unclear if one can legitimately apply contemporary conceptions about gender to a culture that is so distant from our own. The charge of anachronism here concerns the fact that people falsely assume that the relevant frames were present even though they were not.

In response, the proponent of FN might argue that our exportation intuitions can be, in fact, accommodated. Schaffer (2019) has recently suggested that social properties are covert relational properties. According to what he calls ‘The Relations Reply’, (*) would be true because Genghis Khan bears the relational and extrinsic property of being a war criminal relative to the Lieber Code of 1862, or relative to the Geneva Conventions of 1949 (Schaffer, 2019: 764). In this sense, it seems that frames are metaphysically necessary, but they have a relational form: e.g. ‘If x performs such-and-such acts, then x is a war criminal relative to the Geneva Conventions of 1949’.

We find this kind of solution unsatisfying. We agree that one way of salvaging our exportation intuitions would be to relativize our frames appropriately. But there is a cost to this strategy. First, it would require adopting a revisionary interpretation concerning the use of social terms. It seems that when one ascribes a social predicate, at least in some cases we wish to ascribe that predicate in a non-relativized matter (for a full development of this line of response see Epstein 2019a).

More importantly, there are important theoretical benefits (other than being non-revisionist about predicate-use) in countenancing non-relativistic social properties. Let us suppose that there are, indeed, many genuine relativized properties of the form just mentioned (e.g. ‘being a war criminal according to the Lieber Code of 1862’, ‘being a war criminal according to the Geneva Conventions of 1949’, etc.).¹³ But what makes it the case that all these properties are about war criminals? In saying that there is such a thing as being a war criminal simpliciter, we can unify these properties under a single kind. And, in such a case, there would be a question of whether the frame involving the property of being a war criminal simpliciter is metaphysically necessary or not (thus generating the same issues we considered earlier).

Another option for the proponent of FN would be to deny the legitimacy of exportation intuitions. We admit, however, that we cannot think of a non-question begging way of taking every example of an etic analysis to be illegitimate. After all, as demonstrated, the history of the social sciences involves multiple plausible cases of both legitimate and non-legitimate etic analyses.

We should also add that putative legitimacy of certain etic analyses is not settled even within the context of social science. For example, we earlier sketched the view that class struggle can be coherently understood in pre-capitalist societies as per certain Marxist schools of thought. But, (in)famously, such views have also been resisted in the literature. One could insist that since in such societies there were no self-conscious classes, it is illegitimate to talk about class struggle (see Baudrillard 1973 who warns against the pitfalls of economism).

Secondly, there are ways of explaining seemingly plausible exportation intuitions on grounds that do not involve the adoption of FN. Consider the Genghis Khan case once more. The following claim has an air of plausibility:

(*) Genghis Khan was a war criminal.

One explanation for this is that the frame specifying the grounding conditions of being a war criminal is metaphysically necessary. On an alternative explanation, however, (*) is false and the apparent plausibility of (*) is explained by the fact that a similar statement is true:

(**) Genghis khan did morally abhorrent acts.

According to the sort of explanation we are proposing, (*) is false because there was no appropriate frame present when Genghis Khan was alive. Still, (**) could be true in the sense that there is a general moral fact that doing such-and-such acts counts as an instance of moral wrongness (and such moral truths are metaphysically necessary). Or perhaps these exportation intuitions are seemingly true because (**) is also seemingly true. In such a moral error-theoretic scenario, due to various societal or evolutionary reasons we are inclined to have certain moral beliefs even though such beliefs are false (see Joyce, 2022).

At any rate, our claim is that there are ways of explaining the apparent plausibility of exportation intuitions which are less controversial in themselves, and they also accommodate both legitimate and non-legitimate etic analyses. The upshot is that FN cannot be motivated by appealing to UNIVERSALITY. The exportation intuitions of UNIVERSALITY can be explained more plausibly by the view that different possible worlds can exhibit different frames (for a given social kind). So, the claim that Genghis Khan is a war criminal is true in a world where the ‘war criminal’ frame applies and Genghis Khan fulfils the relevant grounding conditions, whereas the same claim is false in worlds where that frame is absent.¹⁴

7 Against UNIVERSALITY

Epstein presents UNIVERSALITY in the context of arguing that the relation of anchoring is fundamentally distinct from the relation of grounding (Epstein, 2019a; cf. Schaffer, 2019). One reason for this, according to Epstein, is that grounding is world-bound whereas anchoring is not. According to UNIVERSALITY, if a world w has some anchors [A₁ … A_n] then these facts also anchor social facts holding in other possible worlds. Call this type of anchoring, cross-world anchoring. Thankfully, our account is neutral towards the question of whether anchoring is a type of grounding.¹⁵

However, UNIVERSALITY by itself poses a problem to RFM. The reason for this is that UNIVERSALITY, if true, suggests that frame-variability is a mystery. Consider two possible worlds (w and w*) each involving a different frame (F and F* respectively) for a given social kind K. It is reasonable to ask what explains the fact that w involves F whereas w* involves F*. A natural answer to this involves noting that F and F* are anchored differently in respect to social kind K. But UNIVERSALITY threatens this picture since it allows for social facts to hold without anchors. This means that there can be a difference in frames without a corresponding difference in the underlying anchors.

In this section we argue that the fact that only certain exportation intuitions are probative (as per the previous section) is more naturally coupled with the claim that only certain social facts involve anchors (and, thus, frames) that are metaphysically necessary. For example, the fact that Genghis Khan was a war criminal is true because Genghis Khan lived in a world that did involve the appropriate anchors for the ‘war criminal’ frame.

It could be objected that this is implausible given that the Genghis Khan world doesn’t involve the existence of the Geneva Conventions. According to the objector, the Geneva Conventions are what make or set-up the ‘war criminal’ frame. But, as Epstein (2019a: 769) suggests, this is empirically wrong. Instead, the anchors for a social fact like being a war criminal involves a multitude of considerations that track an underlying reality about that fact. So, according to this picture, it is simply wrong to say that the law arbitrarily brings social facts into existence.

We agree with Epstein on this. However, we do not see how this suggests that there can be cross-world anchoring. Rather, we claim, for the relevant exportation intuitions to work, certain anchors must obtain in a multitude of possible worlds. So, if the war criminal frame is anchored in a way suggesting that being a war criminal belongs to a natural kind, then this entails that the relevant anchors obtain in many (if not every) possible worlds (perhaps in the same way Kripkean identities hold in every possible world). On the other hand, if the war criminal frame is anchored by some collective acceptance facts, and these facts don’t obtain at the time where Genghis Khan is alive, then the claim that Genghis Khan was a war criminal is false. But this view is predicated on a controversial understanding of the war criminal anchors, so it doesn’t pose a threat to those who take the relevant exportation intuitions to be powerful.

So, the legitimacy of certain exportation intuitions is compatible, if not naturally coupled with, the view that anchoring is world-bound. This is a good result for our view. RFM allows for the possibility of different possible worlds exhibiting different frames while having the same social kind. Assuming the world-boundedness of anchoring provides an explanation for these differences.

8 The challenge from essentialism

Mason (2021) has recently provided a compelling view concerning the metaphysical status of social kinds. According to her view, social kinds are essentially mind-dependent. This means that it is in the essence of social kinds that their instantiations occur because certain mental states exist (Mason, 2021: 7).

There is a tension between RFM and Mason’s essentialist framework in at least two senses. First, contra Mason, RFM allows for at least some social kinds to involve frames which are anchored in a way that doesn’t involve the existence of minds. Consider again the Marxist theory of the kind SOCIAL CLASS. As mentioned, according to such a theory, the relevant frame can exist in worlds where there is no class consciousness given that it is anchored by natural economic laws. In other words, there might be formations of groups (classes in themselves) without any collective acceptance concerning the existence of these groups (Wright, 1997). This is a live view in conceptual space (albeit not uncontroversial) and RFM can accommodate it (similar points apply to other realist theories we considered earlier, such as Bach’s essentialism about GENDER, or the functionalist theory of MONEY). Secondly, one might worry that RFM is incompatible with essentialism since essences, being metaphysically necessary, are unmanipulable.

In response to the second point of tension, we hold that even if it is true that social kinds have essences, it is reasonable to claim that there is a sense in which such essences are manipulable. Specifically, under an essentialist gloss of RFM, the essential properties of a social kind K can shift once we manipulate the frame under which K falls under. Mason (2021: 16) has recently presented a challenge for this sort of view:

[A]nyone committed to the idea that the very nature of a kind can change needs to provide an explanation of how a kind can persist through changes in its nature. Suppose that the nature of a social kind is constituted by properties p₁, p₂, and p₃ at time t₁ but that the kind’s instances change due to human intervention so that its nature is constituted by properties p₃, p₄ and p₅ at time t₂. Call the social kind at t₁ K₁ and the social kind at t₂ K₂. Those who wish to maintain that the same kind changed its nature must answer the following question: in virtue of what is K₁ identical to K₂? In other words, in virtue of what are K₁ and K₂ the very same kind?

In this section we take up Mason’s challenge. According to an essentialist version of RFM, the essence of a kind K at a world w is specified by a frame F at w. By intervening upon F we directly alter the essence of K. So, at world w K falls under F whereas at w*, that very same K falls under F*. The question now is this: why is it warranted to say that K is one and the same kind in both w and w*?

In our response we appeal to the notion of a sub-kind. Let a kind K have a sub-kind K* iff K* fulfils some appropriately specified membership-conditions. Consider functionalism about the kind PAIN. According to that view, the property of being in pain is functionally realized by first-order (e.g., neural) properties that play the ‘pain-role’ (e.g., alerting for tissue damage, etc.). Now consider a plausible sub-kind for PAIN: MILD PAIN. Properties that fulfil the ‘mild pain-role’ (e.g. alerting for minor tissue damage, etc.) functionally realize the kind MILD PAIN. But, crucially, they also realize the kind PAIN. This is because the membership-conditions of properties that fall under the kind MILD PAIN also plausibly satisfy the membership-conditions of the kind PAIN: properties whose causal import is to alert for minor tissue damage are also properties that alert for tissue damage simpliciter.

The same functionalist construal can be plausibly applied to social kinds. Let the kind MONEY be a social kind whose members are functionally realized by properties that play the ‘money-role’ (for simplicity take that role to be: ‘acting as a medium of exchange and a store of value’). Social kinds like CAPITAL MONEY and BITCOIN are plausible candidates for being sub-kinds of MONEY since the membership-conditions of the former are plausibly a subset of the membership-conditions of the latter (in the same way MILD PAIN is a sub-kind of PAIN). For example, BITCOIN has instances that act as a medium of exchange in a decentralized economy (thus fulfilling the BITCOIN-role) but also acts as a medium of exchange simpliciter (thus fulfilling the MONEY-role).

Appealing to sub-kinds helps in the following sense. Let K₁ be a social kind that falls under frame F₁, and K₂ be a kind that falls under frame F₂. Our proposal is that K₁ is importantly connected to K₂ in the sense that both K₁ and K₂ are sub-kinds of the same kind K. So, for example, a possible world involving a frame-shift from CAPITAL MONEY to BITCOIN is a world where MONEY persists (even though there was a frame-shift in the sub-kinds of the kind MONEY). Under the essentialist framework, RFM would entail that a social kind can persist even though there could be underlying shifts in the sub-kinds it contains. This is because the essence of that social kind should be located at the functional role of that kind. To compare with PAIN, the essence of pain is the pain-role; not the properties of the first-level properties that realize that role.¹⁶

To be more precise about how exactly two different sub-kinds K₁ and K₂ fulfill the same membership-conditions for falling under K, consider the distinction between a nested and a master frame. The frames under which the sub-kinds K₁ and K₂ operate are nested frames: they are frames that operate within another frame (see Epstein, 2015: 96 − 7). To see this, consider again two different frames under which the kind MONEY can operate, under the functionalist theory (recall Sect. 4):

(MONEY-FRAME-1*) Instances of MONEY are grounded by the fact that they are instances of gold.

(MONEY-FRAME-2*) Instances of MONEY are grounded by the fact that they are instances of cryptocurrencies.

These frames are nested frames in the sense that they operate under a higher-order frame of the form:

(MONEY_M) For any possible frame principle F, if F makes it the case that instances of MONEY play the ‘money-role’, then F specifies the grounding-conditions for MONEY.

Call MONEY_M the master frame under which the nested frames (i.e. MONEY-FRAME-1* and MONEY-FRAME-2*) fall under. The master frame establishes the unity of MONEY by supplying the membership-conditions for falling under it. To be more precise, the kind GOLD (for example) is a sub-kind of MONEY, in virtue of the fact that its nested frame (MONEY-FRAME-1*) falls under MONEY_M.¹⁷

At this point it should be clear that any talk of frame-manipulability concerns the nested frames of a given kind. In this sense, we grant that a given master frame cannot be manipulated (for it is what secures the unity of the relevant kind) even though the frames of the underlying sub-kinds can be. To echo Mason (2016: 846), the master-frame of a kind is not (and could not be) up to “our discretion”. What could be to our discretion (depending on whether frame-manipulability is at least in principle possible) is the manipulation of the nested frames corresponding to the sub-kinds of a given kind.

One might worry that the appeal to sub-kinds is philosophically parochial in the following sense: it requires a functionalist understanding of social kinds, as per the MONEY example earlier. Thankfully, this is not the case. One doesn’t have to be a functionalist about social kinds in order to accept the idea that kinds can have sub-kinds. Earlier we appealed to functionalism about MONEY and PAIN for illustration purposes. But this is just one way in which we can understand the notion that two different sub-kinds K₁ and K₂ can both fulfill the membership-conditions for falling under K. In the functionalist case, the instances of both K₁ and K₂ are supposed to play the K-role.

Under a non-functionalist account, K₁ and K₂ would both fulfill the membership-conditions for falling under K because the instances of K₁ and K₂ are importantly similar to each other. For example, according to the state-centered theory of MONEY, K₁ and K₂ are both sub-kinds of MONEY because their instances share the extrinsic, relational property of being the result of state-decision. Or consider again the Marxist theory of social class. In this case the relevant membership-conditions are structuralist in nature (Soon 2021): SERFS and WORKING-CLASS would be both sub-kinds of the SUBORDINATED CLASS kind given that their instances share important structural similarities in terms of, for example, occupying a particular node in the social nexus.

9 Conclusion

In this paper we motivated and defended a novel theory of social kind realism. RFM is superior to rival accounts in terms of providing an informative and fine-grained conception of mind-independence. Also, it has important virtues such as recognizing the relativity of social kind realism claims (both in terms of the relevant agents, and the available epistemic and non-epistemic resources), as well as accommodating the emancipatory dimension of political practise.

We also argued that RFM is best coupled with the view that different possible worlds can exhibit different frames for a given social kind. In doing so, we retained the main intuition behind RFM: some frame principles can be manipulated (with various degrees of difficulty), whereas others cannot. We also provided reasons to reject UNIVERSALITY. The resulting picture is this: one can apply social category K to (perhaps distant) possible world w insofar as w has the appropriate anchors for K. This allows RFM to explain the variability of frames exhibited by different possible worlds. Finally, we tackled Mason’s essentialist objection: we argued that even under an essentialist construal of RFM, the unity of social kinds can be retained by appealing to the notion of a sub-kind (and sub-kinds can be further understood in terms of the nested/master-frame distinction).