Missing Mechanisms, Arbitrary Assignments, and Counterfactual Conundrums—Existing Accounts of Social Functionalism and Their Problems

Abstract

This chapter presents the main existing ways of spelling out the function of social institutions: (a) historical social functionalism, (b) structural functionalism, (c) Pettit’s recent virtual social functionalism, and (d) intentionalist social functionalism. In each case, the chapter then presents, updates, and analyzes the most compelling objections to these traditional forms of social functionalism. The upshot of the chapter is the conclusion that a new approach to the topic of the function of social institutions is needed.

2.1 Introduction

Understanding the function of social institutions is important both from the perspective of conducting social science (as noted in the previous chapter) and when it comes to the diagnosis and alleviation of major contemporary social problems. In particular, many of these problems appear to center on the extent to which social institutions do or do not function as they are meant to do (Thompson 1995, 2018; Lessig 2013; Miller 2017; Den Nieuwenboer and Kaptein 2008; Treviño et al. 2014). For example, several of the institutions meant to provide important social goods in many contemporary societies—such as the ability of every citizen to be equally able to influence political decision making, to be well informed about what is going on in the wider society, or to be secure—often seem to fail in their task. Newspapers around the world are full of reports of voting being made difficult for certain groups of people, of factual reporting being deliberately intertwined with opinion stating, or of security being unevenly enforced across a society (Thompson 1995; Lessig 2013; Miller 2017; Satz 2013). Determining the functions of social institutions can provide a fulcrum with which to understand, evaluate, and respond to this these issues: knowing more about what voting, the media, and policing are for can tell us something about what is wrong with some contemporary situations, and point us in the direction of rectifying these issues.

However, while it therefore seems clear that determining what social institutions are for is important both for social scientific and social policy-related reasons, it is not clear exactly how to justify a particular functional ascription to a given social institution. Put differently: the fact that functional ascription is social scientifically and social politically important does not mean that it is clear what the most compelling theory is of the nature of social functions. Several different answers to this question have been proposed in the literature, but they also have seen much criticism and discussion (Pettit 1996; Bigelow 1998). In turn, this threatens to pull the rug out from all attempts to analyze social phenomena using the tools of functional analysis, with some scholars concluding that there is very little of value in them (see e.g. Elster 1979).

It is the goal of this chapter to make this worry more precise. The chapter will consider the four key theories of social functionalism in the literature to date: the historical selectionist account, structural functionalism, virtual selection-based functionalism, and intentionalist social functionalism. (As noted in the previous chapter, the teleological account of social institutions has an in-built theory of social functionalism. The present chapter, though, looks at separate accounts of social functionalism. For more on the teleological theory, see Chaps. 1 and 5.) In each of these cases, it will present the relevant theories and evaluate them in light of some of the major criticisms that have been raised against them. The latter is important also due to the fact that recent work in evolutionary biology, anthropology, and philosophy of biology have sharpened the foundations of some of these criticisms considerably, requiring a novel look at them. While, as will become clearer, it remains true that existing work on this topic leaves many questions unanswered, it also becomes clearer (a) which questions, exactly are still left open, and (b) the kind of progress on this topic. In turn, these insights can then be used to formulate set of desiderata that a compelling account of social functionalism needs to satisfy. For this reason, the goal of this chapter is more constructive than destructive: the aim is to use the existing work on the underpinnings of social functionalism as a building block for the formulation of a novel, fully compelling form of social functionalism.

The chapter is structured as follows. Section 2.2 lays out the historical form of functionalism in the social sciences, updates and revises some of the key objections to it, and shows why it therefore cannot be considered fully compelling. In Sect. 2.3, I present an alternative form of social functionalism: structural functionalism, and show why it cannot be seen as making for a fully compelling theory of social functionalism either. In Sect. 2.4, I show why Pettit’s recent attempt to overcome this objection also fails. In Sect. 2.5, I briefly consider the intentionalist form of social functionalism (though I postpone a more detailed discussion of it until Chaps. 4 and 6). In Sect. 2.6, I summarize the discussion and use it to formulate desiderata on a compelling account of social functionalism.

2.2 Historical Social Functionalism and the Missing Mechanisms Argument

On the standard account of social institutions, a given institution N (in a certain society) prescribes a set of behaviors B across a set of situations S (Parsons 1951; North 1990). Members of the society know that, to the extent that they are part of institution N, they ought to do B in S (at least ceteris paribus), and they expect others to know this, too. This knowledge and expectation then leads members of the society to often (though not always) do B in situation S, and to often (though not always) censure others that fail to do B in S. These general facts about social institutions raise several further questions, though.

In particular, one might wonder if there are some behaviors in B that are more central than others. For example, being a police officer might require responding to reports of a vehicle theft as well as directing tourists to local sites, but the latter may not seem a central element of the institution of the police force. Why is that though? What makes some behaviors B₁ through B_n (such as investigating reports of property theft and destruction) central parts of the institution of N, and others (such as providing information for tourists) not?

Similarly, we might wonder if social institution N is a stable part of this society, or if the behaviors it describes are mere fleeting elements that, like fashion trends, are unlikely to be repeated much. Is policing something that is a key building block of the society in question or is it something that can be eliminated (or defunded) without loss? What about corporate offices? Are these a passing fad, soon to be replaced with widespread working from home or on a gig basis? What about publicly funded, obligatory schooling? As noted above, answering these questions is not just important for purposes of predicting or explaining the structural features of a given society, but also for designing successful interventions.

Underneath these questions is the function of a social institution. To determine the nature and stability of social institutions, it can help know what these institutions are for. (Of course, this presumes that it is the case that social institutions have a purpose—at least often. However, it turns out that, as will be made clearer below, this presumption is quite reasonable.) As Pettit (1996, p. 300) notes:

The tradition of thinking associated with the likes of Durkheim in the last century and Parsons in this is shot through with the desire to separate out the necessary and the reliable from the contingent and the ephemeral. The idea in every case is to look for the core features of a society and to distinguish them from the marginal and peripheral. Functionalist method is cast throughout the tradition as a means of providing “a basis—albeit an assumptive basis—for sorting out ‘important’ from unimportant social processes (Turner and Maryanski 1979, p. 135”).

However, exactly what is it that grounds the function of a social institution? How can we determine what policing, voting, corporations, etc. are for?

Traditionally, the main way of answering these questions is constituted by an appeal to some form of bio-cultural evolution (Bigelow 1998; Rosenberg 2012; Kincaid 1990; Elster 1979).¹ In the background of this idea is the fact that appealing to the selective history of the trait in question is one of the major ways of grounding functional ascription in the biological and cognitive sciences. So, on this view, the function of the human heart is to pump blood and not to make a certain kind of noise, because pumping blood is what the heart was selected for.² Humans with hearts that pumped blood (or whose hearts pumped blood more reliably or efficiently) had a greater expected reproductive success than those whose hearts did not pump blood (or as reliably or efficiently)—irrespective of the noise the hearts made. Hence, the fact that hearts pump blood (reliably or efficiently) supports the spread of hearts. This should thus be taken for their function.³

Several key traditional versions of functionalism in the social sciences propose to apply this same reasoning to the social realm (Elster 1979; Bigelow 1998; Kincaid 1990).⁴ This yields the classic, historical form of social functionalism: a given social institution N has the function F if past tokens of N were biologically or culturally selected for to do F. That is, if those past tokens of N that did F had a higher chance to reproduce tokens of N than those tokens of N that did not, then N (now) has the function to do F.

A classic example of this kind of functionalism is Rappaport’s (1968) analysis of ritualistic pig slaughter among the Tsembaga Maring of New Guinea. The Tsembaga Maring raise pigs in abandoned gardens and consume them during ritualistic slaughter festivals. According to Rappaport (1968), the function of the ritualistic slaughter is to avoid damaging the soil in the gardens so that planting crops becomes harder: while having a small number of pigs is helpful for the planting of crops (as the pigs clear the grounds in the gardens), too many pigs are damaging. Hence, the regular killing of pigs helps keep their number down, while still allowing them to have their beneficial effects of preparing the soil for planting (as well as providing protein for human consumption).⁵

However, this classic, historical form of social functionalism faces an equally classic objection: the so-called “missing mechanisms argument” (Elster 1979; Pettit 1996; Bigelow 1998). According to this argument, there is hardly ever any reason to suppose that actual social institutions have the kind of selective history to ground functional ascription in the above manner. While this is a familiar argument in the literature, it turns out that there are many more complexities here than are typically recognized—many of which are the result of the recent work in (the philosophy of) evolutionary biology. Noticing these complexities is important, as they help in the formulation of the desiderata of what a compelling account of social functionalism looks like.

The first point to note here is that, in its most common form, the “missing mechanisms argument” relies on the standard view of the nature of selective processes. According to this view, for a process of change concerning a particular set of entities to be genuinely a case of “natural selection,” the following presuppositions need to be met (Godfrey-Smith 2009; Brandon 1990; see also Schulz 2020):

• (a) There needs to be variation in the population of the entities in question.

• (b) The entities in question need to be able to reproduce (Brandon 1990; but see also Godfrey-Smith 2009), and the offspring of the entities in question need to be able to (somehow) inherit features from their parents (though the exact nature of the inheritance processes—including any mutational elements—can be left open; see e.g. Boyd and Richerson 2005; Godfrey-Smith 2009; Sober 2014).⁶

• (c) The features that are able to be inherited by the offspring of the entities in question need to drive differences in the reproductive success of these entities.

Now, as will be made clearer momentarily, there is room to question some or all of these conditions. However, it also needs to be noted that these conditions are widely accepted and that they have been foundational for a lot of work in the biological, cognitive and social sciences. Hence, these conditions are not an unreasonable foundation for the “missing mechanisms argument.” This means a version of this argument built on conditions (a)–(c) cannot be brushed aside simply for being built on these conditions. This argument needs to be tackled head on.

At the core of the “missing mechanisms argument”—in this standard version—is the claim that, when it comes to most social institutions, conditions (a)–(c) are seldom if ever all met. First, many social institutions lack historical variation: they are not “tried out,” in different versions, in one or several cultures. Rather, there was often only ever one such institution that was present. For example, in the case of ritualistic pig slaughter among the Tsembaga Maring, it is not like different rituals—some involving pig slaughter, some not—were tried out. Rather, there was ever only one kind of ritualistic slaughter.

Second, it is far from clear that social institutions reproduce, rather than merely persist. That is, it is often not clear (to say the least) that a given token social institution gives rise to one or more separate offspring token social institutions, which then go on to reproduce independently of their parent. Rather, this case may often be better described as simply seeing the social institution as persisting. So, it is not that different tokens of qwerty keyboards gave rise to more offspring qwerty keyboards compared to tokens of Dvorak keyboards and their offspring. Rather, this case seems better seen as involving the continued persistence of qwerty keyboards, and the gradual disappearance of the Dvorak keyboards. This matters, as genuine selection requires reproduction. If there is merely differential persistence, then this might qualify as a case of “sorting” (Schulz 2020; Vrba 1984)—but it would not be selection in the strict sense.

Third and finally, even where there is variation, even where this kind of selection exists, it is often purely extrinsic. That is, the features that determine which institutions persist or reproduce are chance-based. This makes the evolution more drift-like than selective. A familiar example of this is the adoption of the “qwerty” keyboard design, which was not adopted because of its inherent advantages compared to rival designs like the Dvorak keyboard, but because it happened to become popular early on, thus making it costly to switch away to alternative keyboard designs (David 1986; Lewin 2001).

In short: it is often thought that the traditional forms of social functionalism suffer from the fact that the kind of selection process they posit as grounding the function of social institutions did not in fact take place. Hence, traditional social functionalism is often thought to be empirically implausible.

Now, it needs to be noted that this classic criticism of traditional social functionalism may be somewhat overstated. It turns out that the bio-cultural selective history of a number of social institutions does satisfy assumptions (a)–(c) (see e.g. Boyd and Richerson 2005; Henrich and McElreath 2007; Wilson and Gowdy 2013). For example, different ways of pronouncing words (e.g. by dropping a leading “h”) may well evolve by a process of selective copying, as they can provide individuals ways of identifying useful role models (Boyd and Richerson 2005). Similarly, some corporations may well evolve in a selective manner in competitive markets (Schulz 2020). Related points can be made about different forms of music, different moral frameworks, or different political systems (Henrich 2015; Heyes 2018; Nichols 2004).

However, what remains true is that there are many social institutions for which this is not the case (Pettit 1996; see also Schulz 2020). This is important, as for social functionalism to be a compelling and useful approach in the social sciences, it needs to be able to ground functional ascription in a wide set of cases. It may be defensible that some religious practices, some corporations, or some social mores have no function. However, functionalism as a research program loses much of its epistemic value if most religious practices, corporations, or social mores have no function. If only Dunkin Donuts has a function—as it has the right bio-cultural selective history—but Krispy Kreme, Starbucks, Walmart, and so on do not—as they lack this needed history—the functional analysis of corporations has little to add to the social sciences. For the latter, it needs to be the case that corporations typically have a function. Otherwise, the approach has too little methodological hold. The same goes for other social institutions (both types and tokens).

The upshot of this is therefore that the problem with the traditional, historically-focused version of social functionalism is not that it is never plausible. Rather, it is that it is not general enough. It may work in some cases, but as a general account of functional ascriptions in the social sciences, it cannot do the kind of work we ask it to do.

However, as noted earlier, all of this is premised on the fact that conditions (a)–(c) are accepted as grounding genuine selective processes. Recently, though, some philosophers and biologists have questioned the need for some or all of (a)–(c) as presuppositions of selective processes. For example, some authors have questioned the need for reproduction and allow even mere differential persistence as counting for selection—at least in some contexts (see e.g. Papineau and Garson 2019; Doolittle 2014; Schulz 2020; James et al. 2023; Price 1995; Fresco et al. 2018). The most radical version of this recent alternative approach towards selection is Doolittle’s (2014) revised Gaia hypothesis.

According to Doolittle (2014), it is plausible to see selection occurring even in cases where there is only one entity that does not reproduce—such as the biosphere as a whole. The core idea behind Doolittle’s argument is that, due to the second law of thermodynamics, we should expect any entity to sooner or later die out. The longer an entity persists, therefore, the more we have reason to think that the entity has features that prevent the extinction of that entity. Importantly, this is independent of how many entities there are in the population. In turn (Doolittle argues) this implies that there can be evolution without (a)–(c). Doolittle’s argument rests on the idea that, in at least some contexts, there are many important similarities in these three scenarios (Figs. 2.1, 2.2 and 2.3):

Fig. 2.1

Scenario A—standard natural selection

Fig. 2.2

Scenario B—differential persistence

Fig. 2.3

Scenario C—mere persistence

In all of these cases, we see one entity—c—taking over the population over time. In scenario A, this happens because of traditional natural selection. In scenario B, this happens because of differential persistence. In scenario C, this happens because the entity in question successfully manages to stay in existence over the entire time period in question. While the underlying mechanisms are thus different here, the patterns these mechanisms give rise to appear quite similar.

Applied to the present case, this therefore means that we could see an institution N that was not tried out in different versions, and which does not genuinely replicate, as still having function F in much the same way as envisioned by the classic, historical form of social functionalism. We just need to read the claim that N was “selected for” doing F as the claim that N has feature set F, which caused it to have an increased chance of persistence, compared to what would have been the case purely by chance. For this reason, one might be inclined to say that the “missing mechanisms argument” fails to get off the ground here: there is no missing mechanism here at all—we just need to accept a broader set of mechanisms of selection.

However, as it turns out, this conclusion would be too quick. It is true that, especially when it comes to social institutions, it is not necessary to focus strictly on reproduction, and that differential persistence can also be seen as a source of a selection process in the relevant sense. This is an important insight to which I return in Sect. 2.6 below, and it will also be a key element in the discussion of the next chapter. However, this does not mean that the classic, historical forms of social functionalism are now vindicated. This is not just because the other aspects of Doolittle’s (2014) argument are quite controversial. In particular, there may well often be reasons to distinguish cases of differential reproduction with heritability from cases involving mere persistence, especially in a population of one (Godfrey-Smith 2009; Brandon 1990). However, even if we leave this worry aside, there are two further reasons why the appeal to Doolittle’s alternative account of the nature of selection cannot, in fact, rescue the historical approaches towards social functionalism.

On the one hand, there is still a version of the extrinsicality worry here. In particular, Doolittle’s (2014) argument for seeing single, merely persistent entities (like the biosphere) as being selected depends on the assumption that the relevant entities survive for a considerable period of time. Especially for shorter periods of time, random, drift-based persistence is entirely plausible. That is to say, it is true that the longer an entity E persists for, the more plausible it becomes to see E as having a feature G that increases its chances of survival. However, for short periods, this means that it is entirely plausible to see the persistence of E as having been mostly driven by chance. Importantly, exactly this is the situation of many social institutions: many social institutions—like the qwerty keyboard even ritualistic pig slaughter have only been around for very short periods. Indeed, given that human culture, as a whole, is relatively young, there is every reason to think that the persistence of many social institutions is heavily based on chance—or, at least, Doolittle’s (2014) argument provides no reason to think otherwise. That argument may well apply to something like the persistence of our entire biosphere (which is also what it was designed to explain) but its applicability to social institutions is much less clear.

On the other hand, seeing the function of an institution as those of its features that increased its survival in the past is likely to yield a large and quite heterogenous set that is not particularly meaningful. Over time, social institutions may persist for many different reasons. The qwerty initially persisted because of successful marketing and then because it was already widely adopted. Ritualistic pig slaughter similarly may have persisted partly because it increased crop yields but also because children were taught that this is “the thing to do” at certain times, and then, once it got established, simply because it was embedded in the life of the community. More generally, it is likely that there is some set of features F₁ of an institution N that increased its chances of survival in some point in the past t₁, some other set of features F₂ that increased its chances of survival at t₂, and so on. We would then be forced to say that N has functions F₁ and F₂—even though they may be very different, and even though they may have nothing much to do with how the institution has recently operated.

Now, it may seem tempting to follow a common move in the literature on functions in the biological and cognitive sciences, and focus just on the last case of selection (Godfrey-Smith 2009; Piccinini and Garson 2014; Birch 2017). The trouble with this move is that this is still very ambiguous: since time is extended, it is not clear what the latest instance of selection is. If we focus our attention on the immediate past—last month, say—an institution may have survived for some reason R₁, but if we focus on the last five years, for some reason R₂, and so on. What is the “latest” instance of selection here? Note that this is particularly pressing here, since in the case of Doolittle’s alternative account of selection, there are no generations and no reproduction, so selection is much less clearly defended than in the standard treatments. In short: it remains the case that it is not clear that appealing to the features that did increase an institution’s chances of survival yields unambiguous assignments of functions. That said, the idea of appealing to the features of an institution that enhance an institution’s chances of survival is important and is something to which I will come back. It is specifically the historical focus that is problematic here.

Because of these concerns, the classic, historical form of social functionalism still has to be seen to be problematic and subject to the “missing mechanism argument.” Whether we see the relevant mechanism as a standard form of selection—with differential reproduction and heritability—or in a more heterodox manner—as merely involving persistence—it is not clear, at the very least, that most social institutions satisfy the conditions for the existence of the relevant mechanism, or that this would yield clear functional ascriptions. Hence, it is not clear that the classic, historical approach is a good way to ground the function of social institutions.

There are three alternative accounts of social functionalism that have been proposed that can avoid these challenges.⁷ One appeals to the social structure of which a given institution is part, another to a firm of virtual selection, and a third to the intentions of a designer. However, as the next three sections make clear, none of these forms are fully compelling either.

2.3 Structural-Functionalism and Its Problems

The first way to avoid the issues with historical forms of social functionalism is based on appealing to facts about the structure a given social institution is part of. That is, some classic works in the structuralist functionalist tradition in the social sciences—such as that of Parsons (1951)—see social institutions as akin to elements in a large social system, and ground their functions in the causal roles they play in this larger social system. This is thus a social science-focused version of the more general causal role-based view of functional ascription in the biological and cognitive sciences (Cummins 1975), according to which functional ascription is analyzed in terms of the causal roles a trait or component plays in a larger causal system.⁸

So, for example, the function of a police force may be to safeguard law and order, as this is a key causal role it might play in a larger social system: the existence and behavior of the police force causes the safeguarding of law and order—which, in turn, contributes to the existence of social stability and therefore enables calm and deliberative political decision making, which is a key cause of the creation of a police force focusing on safeguarding law and order in the first place (Fig. 2.4).

Fig. 2.4

A structuralist-functionalist analysis of policing

Of course, this is not necessarily the only causal role the police force might play. If we focus our attention on the specifics of who is in power, the police force could also suppress dissent, thus maintaining the monopoly on power of a ruling elite—and thus keeping the police force on task to suppressing dissent (Fig. 2.5).

Fig. 2.5

An alternative structuralist-functionalist analysis of policing

This sort of view is a foundational theory in many social sciences, has therefore seen much discussion and development (for overviews, see e.g. Vincent 2015; Potts et al. 2016; Bigelow 1998). However, for present purposes, it is sufficient to restrict the discussion to two key features of the view—one positive, and one negative. This is due to the fact that the goal here is not a detailed treatment of structuralist functionalism per se, but just to provide the motivation and foundation for the development of a new account of functionalism (to be laid out in the next chapter).

First, on the positive side, the view is “presentist:” it is focused on the present features of the institution in question. This is important, as it means that functional ascription does not need information beyond the social structure in question—in particular, the history of the institution or the social structure as a whole is not relevant. As just noted, this is a major benefit of the account that deserves to be stressed. In particular, it avoids the problems for the historical functionalist account just laid out. Indeed, it avoids some classic issues with historical analyses in general (see e.g. Turner 2007): historical processes—again in virtue of the second law of thermodynamics—tend to erase evidential traces of the past, making the justification of historical claims often very difficult. Of course, this is not to say that we can never find out anything about the past. Rather, the key point to note here is that by not making functional ascription dependent on historical claims, we are able to ascribe functions to institutions even in the absence of a detailed understanding of their history, and even if they do not have much history, as they have just been created. This is a major benefit of the account, as it enables it to have a much wider reach than historical forms of social functionalism. In turn, this gives this form of social functionalism a more compelling theoretical basis.

Second, though, on the negative side, this view is, in its nature, observer-dependent in a problematic manner: depending on what social system is chosen as the target of the analysis, the function of an institution will differ. Now, observer-dependence on its own is neither mysterious nor necessarily problematic. Observer interests determine what something is evidence for (whether a deck of cards is well shuffled or whether my hand is a winning hand) and observers can be part of a network of people that determine what something is (to be a 1-dollar bill is to be recognized as such). Here, though, we get a problematic form of observer-dependence: the same institution can be assigned different functions even by the same researcher and without changing the question that is asked (viz., what their function is). A dollar bill does not cease to be a dollar bill if a researcher chooses to conceptualize it as mere a piece of paper. However, exactly this is the case when it comes to the function of social institutions on the structural functionalist account.

This comes out clearly from the above examples: whether the function of the police force is the maintaining of law and order or the suppression of dissent depends on what overall system we are looking at. Indeed, this is a familiar worry with this kind of view (Kincaid 1990; Holmwood 2005; Vincent 2015; Ariew et al. 2002). Consider the human eye. On the one hand, when we focus on navigation and locomotion, its function could be taken to be to enable humans to see: after all, providing visual information is the key causal role eyes play in the human navigational-locomotive system. On the other hand, though, when we focus on social interaction, the function of the eye could be seen to consist in signaling mental states (intentions, say) to others: after all, a lot of mental state attribution depends on how and where someone looks, and given the importance of this to human social living (see e.g. Whiten and Byrne 1997; Spelke 2022), the function of the eye could be seen to rest in this.

This sort of observer-dependence is problematic, as it introduces an element of arbitrariness into the account. One of the main reasons why a functionalist approach has been attractive in the social sciences is precisely because it allows for an analysis of social dynamics that is non-arbitrary, but objective (Merton 1968; Pettit 1996; Kincaid 2009; Turner 2009b).⁹

One way to see the concern here is by noting that the view conflates cause and function (Millikan 1990; Papineau 1987). Lots of things can be caused by an institution or be the cause of it; however, we want to use functional ascriptions to single out some of these causes as being especially significant. For example, we may want to use functional ascriptions to predict and understand why certain institutions are more stable parts of society, and others less so (on which more in Chap. 4), not just what these institutions do. Similarly, we might want to use functional ascriptions to criticize certain actions as corrupting an institution (on which more in Chap. 5)—i.e. as preventing the institution from fulfilling its function, not just preventing it from doing something we happen to currently focus on. Finally, we might want to use functional ascriptions to pinpoint what an artifact is (on which more in Chap. 6)—not just what it is commonly used for, but what it ought to be used for. By contrast, it seems that the structuralist-functionalist account is forced to say that assignments of F being the function of an institution N is just a matter of our theoretical interest.

A different way of making this point is in terms of another common criticism of structuralist-functionalism: that it is inherently reactionary. The worry here is that since functional ascription is relativized to an existing social system, it is not clear how to make sense of subversive institutions (Rosenberg 2012). For example, how can we make sense of a political magazine, whose aim it is to disrupt the workings of the social system? If functional ascription is social system-relative, then it is not clear how it is possible to understand entities whose functions are subversive of that social system. Here, the structuralist-functionalist view faces the challenge that it seems to only be able to make sense of subversive institutions by focusing on the particular parts of the social system the latter are involved in. For example, the magazine above may be embedded in a causal-structural network involving specific political parties and voters, whose aims are to change the social system as a whole. In this way, the magazine turns out not to be subversive in and of itself; rather it is the causal structure that it is part of that is subversive of the larger social structure it is part of.

The trouble with this, though, is not just that it is oddly indirect—though it is that: it is not the magazine that is subversive, but the causal structure it helps to build up. More importantly, though, the trouble with this is that it defines away the possibility of inherently disruptive social institutions. Since the latter obtain their functional ascriptions due to the causal roles they play in maintaining some larger social structure, they cannot really go against this larger social structure.

For these reasons, while the structuralist-functionalist tradition certainly is an important social functionalist approach, it cannot be seen to be fully compelling either. Consider therefore the next non-historical form of social functionalism: the counterfactual one.

2.4 Virtual Social Functionalism and Counterfactual Challenges

A more recent, but philosophically very influential way to move social functionalist theorizing beyond the historically-focused versions is the account of Pettit (1996). While, as will be made clearer momentarily, this account faces some problems, it also has some key benefits: in particular, it has some important novel and interesting features.

According to Pettit (1996), functional social institutions should not be seen as having a particular bio-cultural selective history—whether this is understood in the traditional or Doolittle’s alternative sense—but as being virtually selected. In particular, for Pettit, for a social institution N to have the function F, it is not required that past tokens of N’s having F led, through a process of differential replication, to present tokens of N having F; neither is it required that N’s having F made it less likely for N to have gone extinct over the course of its existence. Rather, all that is required is that if the existence of N were threatened by some external factor, N’s having F would ensure it continued to exist.

That is, according to Pettit (1996), social functionalism should be counterfactually grounded: what matters is not how some social institution actually arose. Rather, what matters is how the institution would respond if its continued survival were called into question. The source of the survival pressures on N can be manifold and include the arrival of an alternative social institution or other changes to the society in question. What is important is just that, on the account of Pettit (1996), N’s having F means that it is due to N’s having F that it is a stable part of the relevant society. N’s having F ensures that N will continue to be part of the society even if its existence is otherwise threatened.

The example Pettit (1996) uses to illustrate this point is golf clubs. Assume, for the sake of the argument, that these clubs have not been bio-culturally selected for in the way required by the historically-focused form of social functionalism (in either its traditional or persistence-focused version). So, perhaps these clubs arose more or less accidentally when people regularly joined together at the same places to play the game and network. No alternative ways or arranging golf games or business networking opportunities were tried out, and the present clubs were therefore not the result of some selection or sorting among competing alternatives, and the short time of their existence makes the persistence-focused arguments non-applicable.

However, according to Pettit (1996), it may still be true that these clubs have the function to facilitate business relationships and networking opportunities. This would be so if these clubs would persist even if there were external shocks to their existence. So, assume further that it just so happened that people found it harder to learn to play golf (maybe changes in working hours limit the time available to spend on hobbies), or that the physical components needed to play golf such as the racquets or grassy plains became harder to obtain (e.g. because of climate change or materials shortages). If it then turns out that, even under these conditions, people would still join golf clubs so as to network with potential business partners (perhaps they would just take a directed, timed walk, or change the rules of the game in some way, such as just focusing on driving ranges), then golf clubs have the function to facilitate the latter, whatever may be true about the history of golf clubs.

Now, it needs to be noted that, in several respects, this response is a compelling way to handle the concerns of the traditional way of grounding functionalism in the social sciences. The step away from the historical focus of the traditional version of social functionalism is certainly moving in the right direction. As noted in the context of the discussion of structuralist-functionalism, moving away from a dependence on historical information enables functionalist theorizing to sidestep a number of worries. Similarly, the idea that the functions of social institutions should be seen as those of their features that make it more likely that they are part of future iterations of the relevant society is plausible, too: as noted in Sect. 2.1, this does get at a key motivation behind social functionalism. However, that said, the account of Pettit (1996) also faces several problems that prevent it from being fully compelling as it stands.

In the first place, the weight of the account rests entirely on the truth of the relevant counterfactuals. A social institution N has function F if N’s having F ensures that N would continue to exist even if it had to face threads to its existence. However, it can be quite difficult to evaluate counterfactuals like this: how do we know what would happen if people had a hard time getting to golf clubs, or buying the necessary equipment, etc.? Would they still join the clubs? How do we know? Note that the point is not that we never know how to evaluate counterfactuals (Stalnaker 1968; Fodor 1990)—as also noted below, there are many contexts where we do have a good sense of what would happen if so and so were the case. The point is just that there are also many counterfactuals that we do not know how to evaluate, and many of these are crucial to the account of Pettit (1996). I may be able to say, for some worlds W, whether N would persist in W; however, for many other worlds W’, I may not be able to determine this. (Would humans still play golf if they moved to Mars, or would they network differently? Given that social and technological living in the world where people live on Mars is bound to be quite different from what it is now, it is not clear how this question can be answered.) This matters, as it makes this account often difficult to apply. While we may sometimes be able to say if social institution N were to persist if so and so happened, we often would not be able to do so. In turn, this would leave it unclear whether N has a function—and if so, what that is.

Second and even more importantly, it is not clear which shocks social institution N is meant to be buffeted against. Is it all shocks? This seems overly strong: clearly, if a new social institution—urban hiking clubs, say—appeared that also facilitated business relationships and networking opportunities, but did so more easily than golf clubs, golf clubs might cease to exist. This, though, might not be seen to affect the fact that golf clubs have the function to facilitate business relationships. Indeed, it is precisely because urban hiking clubs coopt this feature that they can push golf clubs out of existence in this counterfactual scenario. However, if we are to limit the circumstances to consider when determining the function of N, how are we going to do this? On the face of it, it is not clear what a compelling answer to this question looks like. Without such an answer, though, functional ascription in the social sciences becomes arbitrary.

Third and finally, Pettit (1996) account is not fully spelled out. So, it is not made clear whether functional ascription really involves virtual selection—i.e. whether it is based on the counterfactual replication of the relevant social institutions—or merely virtual sorting—i.e. whether it is just based on the (counterfactual) growth or survival of the relevant social institutions. This is an important issue that, as made clear in Sect. 2.2 above, should be made explicit.

For these reasons, while making a start towards a compelling grounding of functionalism in the social sciences, a further development of this account is needed. Consider therefore the final non-historical account of social functionalism.

2.5 The Intentional Design Based Account and Its Limitations

The final account of functional ascription to be mentioned here is the intentional design-based one. It will be discussed in more detail in Chap. 4 and especially Chap. 6, but a brief statement of it is still useful here. According to the intentional design-based account of functional ascription, the function of something X is determined by the intentions of X’s designer D (for versions of this kind of account, see e.g. Dipert 1993, 1995; Hilpinen 1993; McLaughlin 2000; Houkes and Vermaas 2004, 2010; Vermaas and Houkes 2003).¹⁰ So, if D built X for reasons R, R is the function of X. For example, the function of a pair of scissors is to cut things with, since that is what they were designed to do. They can also be used (more or less well) as paper weights, crowbars, or musical instruments, but that is not their function, since that is not what they were designed for. We might say the same thing about social institutions: a social institution N has function F if we instituted N to achieve F. For example, we may argue that the function of public universities is to create an informed and intellectually nimble citizenry because that that’s what they were designed to do. While universities also advance scientific knowledge or provide new technologies, that’s incidental to their mission, and thus not part of their function.

As will become clearer in Chaps. 4 and 6, there may well be cases where the intentions of designer influence the function of something (Millikan 1984, 2002). The problem is that this is not generally the case. Not all institutions have a clear intentional designer: the institution of marriage, for example, was not clearly designed by anybody. This means that this account of functional ascription is inherently limited and does not speak to many cases of interest. This makes it unsuitably as a general approach towards social functionalism. Still, as made clearer in Chap. 6, it does deserve mention here, as it is a central treatment in some parts of the literature.

2.6 Desiderata for a Compelling Account of Social Functionalism

Pulling all the strains of the discussion thus far together, we obtain four desiderata for a compelling account of social functionalism.

2.6.1 The Account Should Be Focused on the Present

If the account of social functionalism is based on historical facts, it is unlikely to be able to ground functional ascriptions to social institutions in general. Social institutions tend to be too recently created to allow appeal to a purely persistence-based form of selection and lack the kinds of features—variation, heritability, and reproduction—to make appeal to standard reproduction-based forms of selection compelling. By appealing to features of the present only, compelling accounts of social functionalism avoid the problems that befall historical analyses in general, in that functional ascription is no longer held hostage to historical information that we may not have, or which may not obtain.

2.6.2 The Account Should Be Non-arbitrary

A compelling account of social functionalism does not make functional ascription merely a matter of our individual interests. Of course, which social systems we want to study, and which questions we want to ask about these social systems is a matter of our theoretical interests, as well as our moral, political, and aesthetic commitments (Potochnik 2017; Fox Keller and Longino 1996; Douglas 2009; Longino 1990). We may not be as interested in determining the function of the institution of Pokémon trading card exchanges as that of kawaii culture more generally. However, what the function of kawaii culture is should not just be a matter of what social structures we happen to focus. We want to use functional ascriptions to make claims about the nature of social institutions that go beyond our personal or social commitments, but speak to these institutions in and of themselves.

2.6.3 The Account Should Be Actualist, Not Counterfactual

To be compelling an account of social functionalism should not just depend on counterfactual judgments per se. Of course, counterfactual reasoning is pervasive in science, social science, and beyond (Fodor 1990; Kusch 2005), and it is plausible that by saying anything about will or does happen, one implicit makes claims about what would happen under certain counterfactual circumstances (Stalnaker 1968; Lewis 1973). However, it is not compelling for functional ascriptions in the social sciences to just depend on what would happen in various counterfactual situations (such as the arrival of factors threatening the existence of a given social institution). These kinds of counterfactual judgements are often very hard to justify and assess, thus threatening to render functional ascriptions in the social sciences highly insecure. Instead, such functional ascriptions should be based on.¹¹

2.6.4 The Account Should Be General

An account of social functionalism should apply to all social institutions—at least in principle. It should not just be applicable to a small subset of the latter (e.g. ones with an intentional designer), but be useable across the board. This is meant to be a major approach towards social analysis, not a minor tool to be used in a select few cases only.¹²

In the next chapter, I shall lay out and defend an account that satisfies these desiderata.