What are social groups? Their metaphysics and how to classify them

Abstract

This paper presents a systematic approach for analyzing and explaining the nature of social groups. I argue against prominent views that attempt to unify all social groups or to divide them into simple typologies. Instead I argue that social groups are enormously diverse, but show how we can investigate their natures nonetheless. I analyze social groups from a bottom-up perspective, constructing profiles of the metaphysical features of groups of specific kinds. We can characterize any given kind of social group with four complementary profiles: its “construction” profile, its “extra essentials” profile, its “anchor” profile, and its “accident” profile. Together these provide a framework for understanding the nature of groups, help classify and categorize groups, and shed light on group agency.

When we catalog the innovations that make contemporary life possible, social groups are hardly the first things to come to mind. We think of artifacts, technologies, and infrastructure—things like wheels, forms of shelter, the printing press, networks of paths and roads, mechanized agriculture, antibiotics, refrigeration, microprocessors. No doubt, these are nice to have. But at least as crucial to the modern world is how we design and set up groups of people. Social groups have innumerable functions and purposes. They arise from a combination of conscious invention, unconscious habits, repeated patterns, routines, practices, and environmental features. And they come in staggering variety. Among the kinds of groups are sports teams, baseball teams, major league baseball teams, minor league baseball teams, college baseball teams, intramural baseball teams, pickup baseball teams, research groups, musical groups, pop bands, symphony orchestras, marching bands, social classes, races, genders, demographic cohorts, psychographic cohorts, geographic cohorts, corporate marketing groups, corporate HR groups, boards of directors, rioting mobs, marching platoons, processions of mourners,... we could go on and on, listing kinds and sub-kinds.

A chaotic list like this cries out for order and explanation. What, if anything, do social groups have in common? What sorts of entities are social groups, how are they individuated, and what are their persistence and identity conditions? How are they set up, and what do they do for us? How can we best construct a classification or typology of groups?

In approaching these questions in this article, I have two aims. First is to challenge the idea that they have simple answers. There seems to be a powerful drive among theorists to unify and simplify the endless diversity and variation among kinds of groups. Many people, for instance, try to draw simple lines between social groups and mere collections of people, or else to divide groups into a few fundamental varieties.¹ I hope to show that this is a non-starter. Too much faith in parsimony misleads an investigation of social groups from the outset.

The second aim is to present a practical and systematic approach for analyzing and explaining the nature of groups. Instead of thinking mainly about the umbrella category of social groups, I propose to investigate their details from a “micro” or “bottom-up” perspective, constructing profiles of the metaphysical features of groups of specific kinds. I propose that we can characterize any given kind of social group with four complementary profiles:

• 1. Its “construction” profile. This profile characterizes how groups of a given kind are built out of their members, how they persist over time and can be identified across worlds, and when they exist in the first place.

• 2. Its “extra essentials” profile. This characterizes other essential properties of groups of a given kind, in addition to their construction features. Typically it includes abilities, powers, rights, responsibilities, and norms that we set up or “anchor” groups of a given kind to have.

• 3. Its “anchor” profile. This gives the facts that answer the question, why does a group of a given kind have the properties it does? What makes the conditions for membership, identity conditions, and so on, what they are? The anchors are the facts that metaphysically set up, or put in place, various properties of groups of a given kind.

• 4. Its “accident” profile. When it comes to explaining and classifying groups, we are not only interested in their essential properties. This profile gives salient accidental properties of groups of a given kind, which can be equally or more important to understanding what groups are, and to classifying them or developing typologies.

I begin the paper by considering a prominent approach to social groups in the literature. Subsequently, I introduce several examples of groups, and show how the four profiles can be assembled for each. In the course of characterizing the profiles, I also discuss how questions of group agency—groups that can take actions or have intentional states—fit into the analysis of groups. Finally, I turn to the question of how we can classify groups or develop typologies, and what good it does to analyze the metaphysics of groups in the first place.

Despite my best efforts, this is a long paper. My central goal is to organize and explain, in four profiles, how to systematically analyze the metaphysics of any given kind of group. Social groups, at first blush, seem so simple that it may come as an unpleasant shock that a proper analysis turns out to be complex and multi-faceted. But on reflection this should be no surprise at all. We have had thousands of years to develop and elaborate kinds of social groups, for innumerable purposes. The shock would be if, after all that time, every social group fell into one of a few categories.

Yet for generations theorists have relentlessly oversimplified the analysis of groups. Even today, this hobbles our understanding of such things as group agency, the composition of groups, how they are generated, and how we should model them. The good news is that—despite the initial complexity—it is within our powers to tame and systematize their analysis.

1 Some troubles with simple typologies

To motivate the approach I am taking, I will briefly discuss Katherine Ritchie’s influential theory of social groups. Ritchie divides social groups into two categories.² Type 1 groups (or “organized groups”) have structures that reflect functional organization, with replaceable people filling in roles at the “nodes” of the structures. Groups of this kind also have collective intentionality, and members choose to be members—membership is “volitional.” Type 2 groups (or “feature groups”), in contrast, are unstructured. People are members in virtue of possessing some feature, not because they occupy nodes of a structure. Feature groups lack collective intentionality, and people may or may not choose to be members.

With this distinction, Ritchie refines a contrast that many theorists have found intuitive and developed variations of.³ These categories, however, are problematic, even if they are meant to be ideal types. One problem is that it is unclear that these two categories are distinct at all: it all depends on how the notion of “feature” is elaborated, and this is not a simple matter. Second, if we limit the “feature groups” so that they do not include all of the organized ones, we find that even taken together the two categories leave out nearly all actual social groups.

A key challenge for this approach is how to understand a “feature” in the latter category. Which sorts of features that members possess count for such groups, and which are ruled out? Ritchie needs to balance this carefully: if we include all properties, including extrinsic ones, then even the property being a person filling in a node of such-and-such a structure counts, so all groups would be feature groups and the intended contrast between the categories would collapse. If, on the other hand, the “features” were restricted to only intrinsic properties, then we would leave out the archetypal groups Ritchie highlights, such as races and genders.⁴ So Ritchie proposes to understand feature groups as grouping people according to “socially constructed properties,” and characterizes such properties in a very accommodating way. One characterization she mentions, for instance, is Haslanger’s definition of constitutive social construction: “X is socially constructed constitutively as an F iff X is of a kind or sort F such that in defining what it is to be F we must make reference to social factors.”⁵

Unfortunately, while this is a plausible analysis of constitutive social construction, using it to characterize “feature groups” would mean that all the organized groups fall within the category of feature groups. Consider, for example, structured properties such as being on the Microsoft board,being a member of the Supreme Court, and working at the State House. All of these make reference to social factors in their definitions. And they all come along with norms and roles and structures, so the division between feature groups and organized groups breaks down.⁶

Even bigger troubles arise with “organized groups.” To make these distinctive, Ritchie proposes a set of criteria that fail to be possessed by most groups that seem as though they ought to belong to this category. On her characterization, for instance, organized groups must have collective intentionality. How one understands this, of course, is sensitive to one’s theory of collective intentions. But according to most prevailing theories, many organized groups do not have them.⁷ For instance, legislatures, corporate organizations, and many other groups are organized of people working at odds with one another, having divergent projects and failing to satisfy the conditions of collective intentionality. If prevailing accounts of collective intentions are correct, none of these counts as an organized group on Ritchie’s characterization. The same problematic results arise from the other proposed conditions on organized groups. Many structured groups, for instance, do not have unified functions: some have a variety of functions, complementary and competing, and some have none, or once had functions that are now defunct, even as the organization lives on. Even the volition condition is too strong. Groups of workers on a slave plantation (consisting of hierarchies of masters and slaves) are organized structured groups, yet individuals do not have volitional control over their own membership. We may also be members of social groups with “latent functions,” in which we play roles but are unaware even of the existence of the group to which we belong.⁸

It is not just that the lines are vague, or that marginal groups fail to fit neatly into these categories. Rather, categorizations of this sort focus our attention on two small and rather arbitrary clusters, and risk distracting us from the analysis of broader cross-sections of groups.

A key virtue of Ritchie’s approach is the precise formulation of a theory of organized groups in particular, going so far as to introduce a criterion of identity. This precision, however, also enables us to see where the proposal is not successful, and why even a refined version could not be: different kinds of groups have sharply different criteria of identity, so we need a much finer breakdown of kinds of groups, if we are to formulate such criteria at all.

Ritchie’s criterion of identity for organized groups is this: take any pair of organized groups $G_{1}$ and $G_{2}$. For all times t and worlds w, if $G_{1}$ has the same structure as $G_{2}$ at t and w, and if $G_{1}$ has the same people occupying the structure in the same roles as $G_{2}$ has at t and w, then that implies that $G_{1} =G_{2}$.⁹ That is, if $G_{1}$ and $G_{2}$ have the same structure and same structure-ordered-members at all times and worlds,¹⁰ then $G_{1} =G_{2}$.

Unfortunately, this condition is too weak for many organized groups and too strong for many others. In the following sections I will discuss several criteria of identity, but to illustrate the problem here are a couple of examples:

Too weak: Many kinds of groups plausibly have essential origins.¹¹ Consider, for instance, the Senate, or the Supreme Court, or a corporate board, or a sports team. Groups like these are formed with a particular action; and the property having been formed by that particular action is essential to the group. In addition, for many groups, it is essential to them that their memberships too have specific historical origins—i.e., that their members are elected or initiated or appointed in a particular way. It is not enough that two groups have the same members playing the same roles, in order for them to be the same group.

Too strong: For many kinds of groups, we can get identity with weaker conditions. Typically, we understand criteria of identity as giving minimal conditions to guarantee the identity of a pair of objects of a given kind, but Ritchie’s proposed criterion is far from minimal. Some types of groups, for instance, have their members essentially. Such groups do not persist through changes of members. Such a group may be organized, with roles, structures, and the rest—but there are no substitutions. For kinds of groups like these, a minimal criterion of identity is much weaker: two groups of this kind that have different members at any time or world are different groups.

Despite the present criticism, Ritchie’s approach does ask the right questions. We need to give clear characterizations of the essential properties of kinds of groups, analyze distinctions between kinds, and formulate criteria of identity among other characterizations. The underlying problem, however, is the idea that the highly structured, organized, and voluntaristic groups are a basic paradigm rather than a minor variant in the vast ecosystem of social groups.

2 Approaching the question

It might turn out that all the various kinds of social groups share key characteristics, or can be organized into a simple typology. But there are so many different kinds of social groups—committees, boards, legislatures, classes, among others I mentioned at the outset—that we should at least be open to the possibility that social group is just a generic umbrella, and that the real interest lies in the details of particular kinds of social groups.

Let me kick off with a small but important point: the topic of our inquiry and classification is kinds of social groups, not just particular groups in the amorphous category of social groups overall. In analyzing groups, that is, we should focus on the kind board of directors, not just the Microsoft board; the kind gender groups, not just the group women; the kind social classes, not just the bourgeoisie and the proletariat. And equally for other kinds of groups. It seriously damages an analysis of social groups to overlook this.

For one thing, criteria of identity are criteria for kinds. This is what a criterion of identity does: it starts with a kind K and instances x and y of K. Then it gives minimal conditions for x and y to be identical. So a given criterion is a criterion for a kind K. This is clear when we think about the familiar search for criteria of identity for persons: we are looking for a criterion for the kind, not one specific to Brian Epstein. Similarly for kinds of groups versus an individual group.

A second reason this is important is that many of the kinds we are interested in are functional kinds. Analyzing the function of a kind, as we will see, can be of importance for analyzing the metaphysics of the kind. But a well-known feature of functional kinds is that their instances can malfunction.¹² In fact, something can be an instance of a functional kind even if it never properly functions. Only if we explicitly analyze the kinds, not just the individuals, can we discern the place of the function, the conditions for membership in the kind, the identity conditions, and other features.

We do, of course, want to analyze particular groups and investigate their properties. We want to understand the Microsoft board, the group women, and the bourgeoisie and the proletariat in particular. For those individual groups, we may want to know when they come to exist and when they cease to, whether they have intentions or take action, whether they have norms essentially associated with them, or whether they are deserving of praise or blame.

To do so, however, we cannot avoid thinking about the kinds of which the particular groups are instances. Often it is not even determinate which group we are interested in until we hone in on a kind. Consider, for instance, the intuitive idea of a “feature group.” It seems that for some groups, people are members in virtue of having some property, and that is all there is to the group. The group is just made up of people who have that property. In discussing Ritchie, I pointed out that we run into trouble distinguishing feature groups from organized groups. But what if we abandon that aim of marking off the feature groups from other kinds of groups? What if we just allow any property—being a woman, being middle-income, being an adult—to mark off a group? Why not simply analyze that group, without worrying about kinds of groups?

The problem is this: suppose we choose a particular property $P_{g}$, the possession of which is necessary and sufficient for membership in group g. Even given that property, that does not suffice to determine the other essential properties of group g. Under what conditions, for instance, does g persist? Suppose ten people have property $P_{g}$ from April to June, then no one has the property from June to September, then eleven people have the property from September to November. Do the ten members in the spring belong to the same group as the eleven in the autumn? Was that one group, or two distinct groups? Furthermore, suppose there is another property Q that has the same extension as $P_{g}$. Maybe even the same extension in all possible worlds. Is the Q-group distinct from g, or identical to it? Even though property $P_{g}$ determines how the group is constituted—that is, how its membership is fixed—it does not answer these questions and others. Rather, their answers depend on the kind of group g is.

Here, instead, is a more satisfactory treatment of groups like these. We can agree that the dominant characteristic of certain groups is that they are constituted by and only by people having property $P_{g.}$ But that dominant characteristic is not their only characteristic, and there is not just one way of filling out the rest. Not all such groups, therefore, will fall into one single kind. Instead, there is a family of kinds of groups, all of which have that dominant characteristic, but whose answers to the additional questions vary. I will call this family the “constitution-dominated” kinds. The groups in some of the kinds, within this family, are essentially continuous: these groups cease to exist as soon as no one has the relevant property. Others admit discontinuity: these groups persist even if there are breaks in the property’s exemplification. Other kinds in this family have different persistence conditions: kinds that persist through three breaks but not four, or that persist through breaks so long as they are brief. And persistence is not the only issue: kinds within the constitution-dominated family vary along different dimensions as well.

In short, there is no single default kind of group that is “features only.” In a moment I will introduce a particular kind in the constitution-dominated family as one of the working examples for us to profile.¹³

Once we focus on specific kinds of groups, it is easy to get off and running. To analyze a kind of group K, we need to answer a variety of questions. When and how does an instance of K come to exist? Given a group g of kind K, under what conditions is a collection of people the membership of g at a given time? Under what conditions are instances of kind K identical to one another? What are the rights, obligations, and hierarchies associated with groups of kind K? Where do these come from, and for that matter, where do the conditions come from pertaining to the existence, membership, and identity of groups of that kind? And finally, once we have these characterizations of many kinds of groups, what kinds of kinds are there? How should they be classified?

To begin, let us consider several kinds of groups as working examples. If we are to get a sense of the inadequacies of simple models, and see how to analyze diverse kinds, we cannot start with just one or two. So here is a range of kinds:

K1:

Groups of street musicians. A group of kind K1 is formed when musicians gather together on the street, standing or sitting relatively close to one another, and start playing. Players can join the group or leave the group, with membership dependent on their being in close proximity to the others, joining in, and being responded to appropriately. A group of this kind terminates when it stops playing for more than a few minutes.

K2:

Tufts University College of Arts, Sciences and Engineering elected standing faculty committees. There are about fifteen actual committees instantiating kind K2. Groups of this kind are created by a process of voting and setup by the faculty, with members nominated and voted on by the faculty. The terms are staggered so that each year only a fraction of the members rotate out and are replaced; replacements are nominated by the faculty and voted in by the faculty. The structure and functions of the committees are dynamic over time, with proposed changes made by the dean or faculty members and voted on by the faculty. Sometimes committees perform their intended functions, and sometimes the members are at odds with one another for long stretches of time.

K3:

Capitalist social classes. Groups of kind K3 are formed when a world’s economic system becomes structured in a particular way, with groups of this kind playing particular roles. The groups exist so long as that role is played at all. Membership in the group involves ownership and deployment of capital in certain ways, and standing in certain social relations to other people.

K4:

For this example, I will choose one member of the “constitution-dominated” family of group kinds. To some extent, this choice is reverse-engineered: only by thinking through the profiles does it become clear what options there are, so the nature of this group will become clearer as we examine the profiles. I will assign this group-kind a name based on its characteristics: it is the discontinuous intensionally-individuated constitution-dominated (DICD) groups. A horrendous name, but it does highlight that this is one among the family of “constitution-dominated” kinds. (I will explain the “discontinuous intensionally-individuated” part below.) And despite its name, it is likely to be a fairly common kind of group. It may, for instance, be reasonable to regard groups like middle-income people and adults—inasmuch as they are determinate groups at all—as being groups of this kind.

These four examples are nothing more than a sampling of kinds of groups. In the following sections, I use them to explain and fill out the “profiles” for characterizing kinds of social groups.

3 The construction profile

The first task for understanding the nature of groups is to analyze how they are built out of their members, how they persist over time and can be identified across worlds, and when they exist in the first place. These are characterized in what I will call the “construction profile.”

3.1 Analyzing groups in terms of their “stages”

To treat the constitution and identity of groups, I will center the discussion on “stages” of groups—how stages constitute groups at a given time and how they are related to one another.¹⁴ Talk of stages is familiar from the metaphysics of ordinary objects, and of persons. A stage of a person, for instance, is an instantaneous snapshot of material that typically includes a head, a torso, arms, hands, legs, feet, etc. Similarly, a stage of a group is an instantaneous snapshot of that material that constitutes the group. That is, it is a collection of people in a world at an instant in time.¹⁵

A given stage s exists at and only at a moment in time and in a particular world. I will not assume that a given stage must be a stage of a group of any kind. Stages are merely snapshots of collections of people, and might be able to exist on their own without being constituents of groups.¹⁶

Throughout this paper, I will speak of groups being constituted by stages at a given time and in a given world.¹⁷ In Sect. 4 I discuss the constitution view of groups directly, but my aim in analyzing and profiling social groups is to be fairly ecumenical about what groups “really are.” Essentially the same profiles can be constructed, with slight variation in terminology, whether groups are real or fictional, whether they are continuants that persist in time or are abstract objects, or whatever else they might be. If, for instance, a group is best understood as a mathematical object like a set, then we could translate talk of stages constituting groups into talk of ordered pairs of sets and times being the elements of a group.¹⁸ I do find it helpful to learn from the ways we analyze ordinary objects, in order to clarify features of groups. But I hope that most of the results we develop will apply regardless of one’s view on the appropriateness of that analogy.¹⁹

To analyze a kind of group K, then, a central task is to find generalizations about how K-groups are constituted by their stages. (I will use the term “K-group” to abbreviate “group of kind K.”) For instance, take a particular stage s. What conditions does s need to satisfy, in order to be a stage of a K-group? Or take two stages, $s_{1}$ and $s_{2}$. Suppose that both $s_{1}$ and $s_{2}$ are stages of K-groups. What additional conditions do $s_{1}$ and $s_{2}$ need to satisfy in order for them to be stages of the same K-group?

We can use stages to analyze the constitution or membership of groups of a given kind K, as well as to formulate criteria of identity. But we also will need one more bit that does not always involve stages: formulating the conditions under which a group of kind K comes to exist at all, and the conditions under which it continues to exist. Sometimes these are a function of the group’s stages, but often they involve other factors.

3.2 The four parts of a construction profile

Thus there are four formulas to fill in, in order to characterize how groups of a given kind are built—at a given time, over time, and across possibilities.²⁰

• Coming to exist in a world: A new K-group comes to exist at t in w if and only if...

• Continuing to exist in a world: Given a K-group g that came to exist at $t_{0}$ in world w and a time $t > t_{0}$. Then, g exists at t in w if and only if...

• Constitution in a world: Given a K-group g and a time t and world w. Then, stage s constitutes g in w at t if and only if...

• Identity: Given K-groups $g_{1}$ and $g_{2}$, and given that $s_{1}$ constitutes $g_{1}$ in $w_{1}$, and $s_{2}$ constitutes $g_{2}$ in $w_{2}$. Then, a minimal requirement to guarantee that $g_{1}=g_{2}$ is that $s_{1}$ and $s_{2}$ stand in relation...²¹

In certain cases, these can be slightly redundant with one another, but mostly they are not.²² And in any case, it is almost always illuminating to fill in all four. Without being excessively detailed, let us give at least approximate answers for the four groups K1–K4 I profiled above.

As we work through these sixteen questions—four questions for four kinds of groups—it becomes obvious that kinds of groups vary enormously along many dimensions. Existence conditions are all over the map. Constitution conditions are all over the map. Identity conditions are all over the map. Even if you disagree with the answers I give to specific questions, you will probably find that your improved answers are even more textured and varied than mine. The dimensions along which kinds of groups differ will give us rich material for classifying groups and building taxonomies.

This, however, is not the only value of working through the examples. With the profiles—construction, extra essentials, anchor, and accident—I hope to provide and illustrate a template for analyzing any given kind of group one is interested in. Without several examples, it would be hard to see how to apply these profiles to new cases. And there are practical benefits to fully profiling a kind of group one is interested in; it is not just a curiosity for metaphysicians. If, for instance, one wants to model the decisions of K2-groups (Tufts faculty committees), then one may want to model their creation and dissolution, how they gain and lose members, and ensure that distinct committees are modeled as distinct. For building models, that is, the construction profile matters. Similarly for the other profiles. We may want to model their rights and obligations, or how existence and membership conditions can be changed, or how rights and obligations come to be acquired, or perhaps even various accidental properties that the groups have over time. My aim in working through the detailed examples is largely to help illustrate the parts and applications of the profiles.

3.2.1 The conditions for a stage to constitute a given K-group

Within the construction profile, I will start with the constitution conditions—i.e., the conditions for a stage s to be a stage of a given group g at time t in world w. Those are often the most interesting and important ones for understanding the makeup of a group.²³

To work them out, we need to recall that stage s is an instantaneous snapshot of a collection of people. So the constitution conditions will largely be a matter of the people in that collection having the right characteristics at time t. It can be useful to think separately about the synchronic and the diachronic characteristics of the people. What do the people need to be doing at t, and what history must they have had, in order to be part of a stage of the group at t? Further, we need to think not only about what it takes for stage s to be a stage of some group of that kind, but to be a stage of group g in particular. All this needs to be included in the constitution conditions.

Consider, for instance, some examples of K1-groups (i.e., street musician groups). Suppose group a plays at 500 Boylston Street on Monday from 10 a.m.–2 p.m., and group b plays at 500 Boylston Street on Tuesday from 10 a.m.–2 p.m. Consider some stage s—e.g., a snapshot of a collection of people playing on Tuesday at noon. What conditions does s need to satisfy in order to be a stage of group b? Some of the conditions on s are synchronic: the people in s need to be standing and playing together at 500 Boylston, and perhaps also have collective intentions regarding their performance. But that much only guarantees that s is a stage of some K1-group, not that it is a stage of group b in particular. (As opposed, for instance, to being a stage of the distinct group a.) To ensure that, it is also necessary that s be part of an unbroken sequence of stages stretching back to the origin of b.

Already with this example we can start to see why this detail is crucial for understanding groups, and why highly idealized categorizations are inadequate. Are all groups held together by collective intentions? Should we divide groups up into those that are held together by collective intentions and those that are not? The reality is more complicated than this. Here is at least a tentative analysis of the constitution conditions for groups K1–K4:

Suppose we have a K-group g, a time t, a world w, and a stage s that is a snapshot of a collection of people at t in w. Then, s constitutes g at t in w if and only if...

• K1 (street musician): g exists at t in w, and s is a snapshot of a collection of people performing with one another and with the relevant collective intentions, and s is part of an unbroken sequence of stages with those characteristics going back to and including the time of origin of g in w.

• K2 (faculty committee): g exists at t in w, and s has gone through the legislated rotations and processes for its members, and s is part of an unbroken sequence of stages with those characteristics going back to and including the time of origin of g in w.

• K3 (social class): g exists at t in w, and s is a snapshot of the collection of all the people at t who have performed the functional role with which g was formed (e.g., the role of the bourgeoisie) within some reasonable period of time before and after t.

• K4 (DICD): s is a snapshot of the collection of all the people that have property $P_{g}$ at t in w (where $P_{g}$ is the property whose first instantiation at $t_{0}$ formed g).

What makes a stage constitute a group varies enormously from group to group. For some groups, s must have collective intentions, or play a functional role, and for others not. Sometimes s being a stage of g is a synchronic matter: it only depends on what is going on at that time t. Much of the time it is a diachronic matter: the membership of a group depends on what precedes or follows t.

For many groups, being a stage of g depends on the stage’s relation to other stages in the path stretching back to the group’s origin. This is why many of these analyses of constitution conditions make reference to the origin or formation of the group. (How K-groups are formed is discussed just below, in Sect. 3.2.2). The distinction between synchronic and diachronic conditions does not line up with any distinction between “organizations” and “features”: some groups with no organization have historical membership conditions, while others have synchronic ones. And the same options apply to functional-role groups, committees, and groups that join together to play music on the street.

There is only one thing we can say in general: for s to be a stage of g at t, it is always required that g exist at t. This is not at all an empty condition. In fact, for two groups $g_{1}$ and $g_{2}$ of kind K that coincide with one another—that is, that have the same members at the same time—then how those groups originally come to exist is often the most important thing for distinguishing them from one another.

3.2.2 The conditions for a K-group to come to exist

I divide the existence conditions into two parts: the conditions for a K-group to come to exist at time $t_{0}$, and the conditions for it to continue to exist at time t. Here is a tentative analysis of the conditions for K1- to K4-groups to come to exist:

A new K-group comes to exist at $t_{0}$ in w if and only if...

• K1 (street musician): a collection of people playing music together on the street begins at $t_{0}$ in w. (That is, playing did not continuously occur at that place prior and leading up to $t_{0}$.)

• K2 (faculty committee): the appropriate process of voting and setup by the faculty takes place leading up to $t_{0}$ in w.

• K3 (social class): a nation’s economic system becomes structured in a particular way at $t_{0}$ in w, with this group playing one of the relevant functional roles.

• K4 (DICD): there is a property P that is, in w, instantiated for the first time at $t_{0}$. If there are n properties $P_{1}, {\ldots }, P_{n}$ with distinct intensions that are simultaneously instantiated for the first time at $t_{0}$, then n DICD-groups $g_{1}, {\ldots }, g_{n}$ come to exist in w at $t_{0}$.

In these answers, we can see that groups of certain kinds come to exist as soon as an activity takes place. Groups of other kinds come to exist when a functional role is filled by a collection of people, or else when a collection is assigned to be members.²⁴ With K4-groups, a new group comes to exist in a world whenever a property is instantiated for the first time in that world.

We can also see, in these answers, how existence conditions like these help us explain the possibility or impossibility of coinciding groups of a given kind. Many street-musician groups can exist at a given time, but only in separate locations. Many faculty groups can be created and coincide with one another, if the faculty has gone through the appropriate setup repeatedly. There are several kinds of K3-groups, and once they are formed they do not get formed again. And distinct K4-groups can coincide in a given world, but not in all worlds.

3.2.3 The conditions for a given K-group to continue to exist

In general, this other part of the existence of K-groups is simpler. Here is a tentative analysis for K1- to K4-groups:

Given a K-group g that came to exist at $t_{0}$ in world w and a time $t > t_{0}$. Then, g exists at t in w if and only if...

• K1 (street musician): musical performing has continuously occurred at that place from $t_{0}$ to t in w.

• K2 (faculty committee): from $t_{0}$ to t in w, the college has continued to exist and the faculty has not disbanded g.

• K3 (social class): the global economic system is structured in the relevant way at t in w.

• K4 (DICD): there is a time $t^{\prime } \ge t$ such that some person has property $P_{g}$ at $t'$ in w (where $P_{g}$ is the property whose first instantiation at $t_{0}$ formed g).

Some groups continue to exist only while an activity continues without break, some allow breaks, and some exist in perpetuity or until they are expressly disbanded.

I have labeled K4-groups “discontinuous”: this is because a group g of this kind persists even during times when no one has property $P_{g}$. Once it is formed, it continues to exist until $P_{g}$ is exemplified for the last time in that world, and then ceases to exist. So even if there are times when the DICD-group of people with the top 1% of wealth does not exist at all (for instance, if everyone in the world is economically equal for some period), nonetheless people who are in the 1% before and after that period are members of that same one group. A different kind in the same family is continuous constitution-dominated groups: groups of this kind must be continuously constituted, so the before and after groups of the top 1% would be distinct groups.

3.2.4 The criterion of identity for groups of kind K

The fourth part of the “construction profile” of kind K is its criterion of identity. We should note, however, that criteria of identity often add less information than one might suppose, even for complicated kinds of groups. The idea of a criterion of identity is to give a minimal relation R between two groups that guarantees that if they are both K-groups and stand in that relation R, then they are the same group. (This is a “one-level” criterion of identity. A “two-level criterion” gives a minimal relation R between two stages that guarantees that if they are both stages of K-groups and stand in that relation R, then they are stages of the same group.)²⁵ The reason criteria of identity are often not very informative is that so much work is done in those antecedents. The criterion only gives the additional requirements to guarantee that $g_{1}=g_{2}$, once we are given that that $g_{1}$ and $g_{2}$ are both K-groups. (And “two-level criteria” only give the additional requirements once we are also given that $s_{1}$ and $s_{2}$ are stages of $g_{1}$ and $g_{2.}$) If there are not very many groups of kind K, then it need not take much to guarantee that $g_{1}=g_{2}$.

Here are rough criteria of identity for the four examples. For K1, K2, and K4, one-level criteria are simpler, and for K3, a two-level criterion is simpler:

Given K-groups $g_{1}$and $g_{2}$. Then, a minimal requirement to guarantee that $g_{1}=g_{2}$is that...

• K1 (street musician): the time and place of origin of $g_{1}$ is similar (within some reasonable tolerance) to the time and place of origin of $g_{2}$.

• K2 (faculty committee): $g_{1}$ and $g_{2}$ originate in the same formation act.

• K4 (DICD): for all worlds w and all times $t, g_{1}$ and $g_{2}$ have the same members in w at t.

Given K-groups $g_{1}$and $g_{2}$, and given that$s_{1}$constitutes$g_{1}$at$t_{1}$in$w_{1}$, and$s_{2}$constitutes$g_{2}$at$t_{2}$in$w_{2}$. Then, a minimal requirement to guarantee that$g_{1} =g_{2}$is that...

• K3 (social class): $s_{1}$ and $s_{2}$ both play the same functional role (among the roles of various social classes).

Criteria of identity only give us interesting information about the nature of K when there are many instances of K that need to be discriminated from one another. If we were to choose a kind K for which there was only one K-group, then when we start with two stages of K-groups, it is already guaranteed that they are stages of the same K-group. So in that case, the “criterial relation” between the stages to guarantee identity can be completely empty. How much work the criterial relation must do depends on how finely individuated K’s instances are. For faculty committees, the faculty can form new ones each time it performs the appropriate formation act. Therefore, to discriminate them from/identify them with one another we need to trace the stages back to those acts. For street music groups, they are distinguished from one another roughly based on their placement in space and time.²⁶

K4-groups are individuated according to their “intensions”: given K4-groups $g_{1}$ and $g_{2}, g_{1}$ is identical to $g_{2}$ just in case $g_{1}$ and $g_{2}$ have the same members in all worlds. But DICD-groups are just one in a family of constitution-dominated kinds. Other kinds in that family individuate groups more finely or more coarsely.

The construction profile of a kind K is only part of the profile of the nature of K. But it alone answers many of the questions we started with. It gives us the persistence and identity conditions for groups of kind K, and it tells us how such groups get created. Comparing construction profiles for various kinds of groups expands our understanding of the variety of ways groups can be constructed. And it gives us ideas about the various cross-cutting ways we might classify kinds of groups.

4 Constitution and social groups in general

It is worth pausing to consider whether there is anything we can say about the nature of social groups in general. In Epstein (2015) I propose this characterization: x is a social group if and only if x is an entity constituted by and only by people.²⁷ On this understanding, social groups are a broad and inclusive category. This characterization does not accommodate every social entity; not by a longshot. In fact, most social entities are not social groups—not money, not corporations, not universities, not economies, not nations, not borders, not laws, not languages, not artifacts. Still, my characterization is inclusive, rejecting a particular “mark of sociality” for something to count as a social group.

Effingham (2010) worries about accounts of groups that “overgenerate”: that is, accounts that mistakenly count “mere collections of people” to be social groups. But I regard “undergeneration” as far more serious: it is more common for theorists to overlook a variety of kinds of social groups in their attempts to circumscribe the overall category. As I have argued above, when we analyze the kinds of social groups, we find that they are constructed in heterogeneous ways. Different kinds of social groups have little to unify them aside from their being built of people. In the following sections, a similar heterogeneous range will be seen in other characteristics—in what powers and norms they have, in what anchors them, and in their accidental properties. If we are to make room for this heterogeneity, we need to avoid an overly restrictive understanding of social groups.²⁸

Analyzing groups in terms of constitution also gives us resources to talk about groups that coincide with one another, using some of the same machinery we use for talking about coinciding material objects. Much like the widely discussed statue and lump of clay that constitutes it, a social group may be distinct from a collection of people that coincides with it, and two distinct social groups may have the same members over the duration of their existence. The constitution relation is not the only way to address these phenomena, but regarding groups as ordinary objects in the world lets us draw on a useful toolkit.

Effingham (2010), Ritchie (2015) and Thomasson (2016) explicitly consider the “constitution” view of social groups, and raise three objections that I should respond to. First, Thomasson cites Ritchie as raising the following objection to my analysis: there are other ways groups could be constituted than just by people. A team could be constituted by teams, for instance.²⁹ If so, then that appears to contradict my condition that groups be constituted “by and only by people.” However, this objection misreads my proposal. When we say that a sample of water is constituted by and only by water molecules, or a bunch of asparagus is constituted by and only by stalks of asparagus, we know what that means. It is true, of course, that a bunch of asparagus is also constituted by larger things, such as three-stalk-collections of asparagus. And it is true that a sample of water is also constituted by smaller things, such as electrons, protons, and neutrons. But the fact that many things stand in the constitution relation does not negate the point about a sample of water being exhaustively constituted by (that is, by and only by) water molecules, and a bunch of asparagus by and only by stalks of asparagus. To be perfectly explicit, we can write it this way:

x is a social group if and only if:

• (1) for each world w and time t, if there is any object y such that y constitutes x in w at t, then there is a collection z consisting of and only of people such that z constitutes x in w at t, and

• (2) there is some world w, time t, and object y such that y constitutes x in w at t.

Condition 1 states that wherever x is constituted, it is (also) constituted by a collection consisting only of people. Condition 2 just ensures that x is constituted by something somewhere, so as to keep immaterial (or otherwise non-constituted) objects from vacuously satisfying the definition.³⁰

Second, Effingham and Ritchie criticize the constitution view as taking groups to be “sui generis” entities.³¹ They object—mainly on grounds of parsimony—to Uzquiano’s proposal that groups are a “relatively unfamiliar category” distinct from sets or mereological fusions (Uzquiano 2004, p. 147). While Effingham and Ritchie may be right that Uzquiano’s proposal does not maximize parsimony, I am not persuaded that his or other constitution views are nearly so mysterious as the accusation suggests. The term ‘sui generis’ is typically applied to objects that are fundamental or ungrounded, and in the literature on social entities, it seems to be connected to a sort of dualistic holism about the social world. But the refusal to identify groups with sets, fusions, or similar kinds of entities does not entail any of that.³² My own view, certainly, is not that groups are ontologically fundamental.

We do need to explain the essential and accidental properties of groups, and also how groups are set up to have these properties. This is what we have begun in the preceding discussion, and will continue shortly. But none of it entails that we identify groups with some other “more familiar” kind of object. To be sure, contemporary metaphysics remains embroiled in controversies regarding the nature of ordinary kinds (like statues and lumps of clay) quite generally.³³ At some level, we will not fully understand social groups until these problems are solved. Perhaps the study of social groups will even contribute to unlocking the solution more generally. But this is a general problem of metaphysics, not one that is any more urgent for social groups than for any other kind of object.

Third, Thomasson objects that the constitution view of social groups does not do very much for us—it does little to illuminate the nature of groups, and certainly does not solve the “overgeneration problem.” On this point I agree: merely to say that social groups are constituted by people does not constrain them much. But that is what we should expect of a generic umbrella category like social groups. That category is dynamic: inasmuch as there is a line between social groups and mere collections, it is a product of the kinds of social groups that have been set up at a given time. As I discuss in Sect. 7 below, kinds of social groups are set up or “anchored” in a variety of ways. Interestingly, as societies and environments change, we do not only set up new kinds of social groups, but we also develop new ways of setting up kinds of social groups. Before we had formal legal systems, for instance, new kinds of groups did not arise via legal enactments. Now they can. The introduction of such “anchoring” methods, together with the kinds introduced using those methods, expands the category of social groups. The line between social groups and mere collections of people is not eternal and universal, and so we should not expect too informative a characterization of the umbrella category.

5 The extra essentials profile

I began this paper with the observation that today’s social groups are the beneficiaries of thousands of years of innovation and experimentation. Much of that innovation has been coming up with new ways for social groups to be constructed. How do we originate groups in different ways, so that they accomplish our goals? How do we set up membership conditions so that group members are responsive to other people? However, perhaps the most innovative elements of group design are not a matter of what makes a group come to exist or to be constituted as it is. In addition to innovation in construction, we have also developed and evolved ways to set up, or “anchor,” groups to have other kinds of properties: abilities, powers, rights, responsibilities, norms. We anchor these features to be essential to groups of a given kind, separately from the construction features treated in the last section. (I present anchoring in Sect. 7 below. And in Sect. 5.3, I discuss why the “extra essentials” are not part of the “construction” of a kind of group.)

Just as important as the powers, abilities, norms, etc. that we set up groups to have are their limitations. We anchor groups to have abilities, for instance, only under very limited conditions. The creative ways we limit those conditions affect group actions nearly as much as do the abilities they have.

5.1 Varieties of “extras”

A Tufts faculty committee (i.e., a group of kind K2) may not seem to have a particularly complex design—just appoint some people to it and let it go—but that is deceiving. Tacitly and explicitly built into that kind are abilities, rights and obligations, and relations to people, other groups, and things in the world. These are the product of generations of evolution, experimentation, and weeding out failures. All of these abilities and powers have limitations and constraints. It is the texture of these, and how they are triggered and circumscribed, that largely determines how and when a faculty committee acts, and what effect it has when it does.

Some but not all of these features are deontic. Many groups have rights, obligations, ways the group or members should behave, or ways they should be treated. But not all groups have deontic properties, nor are all these extra essentials deontic. For instance, certain groups are set up to have abilities—to do things like call meetings, create subgroups, and enact rules. (There is a difference between being able to do something and having the right or obligation to do it.)³⁴

Another powerful innovation is to anchor powers, abilities, rights, and obligations not to the group as a whole, but to individual members. These can be equal for all members (such as the right to assert one’s opinion in a faculty committee), or they can apply differentially for different members (such as the responsibilities of the secretary versus the responsibilities of the treasurer). A particular kind of differential treatment of members is to arrange them into hierarchies or reporting relationships (such as when certain members have the obligation to obey the orders of others).

Make no mistake, these powers, abilities, norms, etc. do not need to be beneficial to society or to a group’s members. Such properties can be innovative in their ability to oppress or promote injustice. Consider, for instance, the powers and norms anchored for different racial and gender categories.

The innovative ways we attach powers, norms, etc. to social groups make them potent and effective. But they also create complexity when it comes to understanding or analyzing kinds of social groups. Perhaps the biggest complication is this: sometimes they are included among the construction conditions, and sometimes they are not. For instance, it may be a condition for being a member of the group women that one is subject to certain oppressive norms (Haslanger 2000). In that case, the norms are among the constitution conditions for the group. Alternatively, there may be norms essentially attached to groups that are not included among any of the construction conditions.

Think, for instance, about the conditions for a pair of people constituting a married couple: signing papers, or going through a ceremony, or whatever. The satisfaction of those conditions is what it takes to make the married couple exist and for those people to constitute the couple. At the same time, however, marriage brings along with it many other powers and limitations, rights and obligations. These other essential properties are separate from and supplementary to the construction conditions.³⁵

This point can be puzzling, especially to those of us in metaphysics who are trained to analyze objects mostly or entirely in terms of how they are constituted. It also seems puzzling to philosophers who assume that all the essential properties of a kind must figure into the kind’s criterion of identity. No doubt, the idea that there are such “extra essentials” might seem radical: these are essential properties—i.e., properties that a group necessarily has at all times and in all worlds. Even so, these essential properties are neither part of the constitution conditions nor the existence conditions nor the criterion of identity.³⁶ In Sect. 5.3, I clarify this point further: how there can be extra essential properties that are not included in the construction properties. But first let us fill out some examples.

5.2 Organizing the “extra essentials” profile

The “extra essentials” are a catch-all category for those essential properties that groups are set up to have, but that are not included among the construction properties. Many groups have none of these supplementary properties. There are groups with no abilities, rights, obligations, or norms beyond those of the collection of people constituting them. The constitution-dominated groups (such as K4, the DICD-groups) are among these. And many kinds of groups have abilities, rights, and so on, but these properties are not supplementary: instead, they are included among their construction properties. Still, many kinds of groups do have extra essential properties.

As we did above, we can write a profile to characterize group-kinds with respect to these features. To organize the profile, I will characterize the various powers, limitations, and so on, that apply to the group as a whole, and then separately those that apply to the members of the group. Among the ones that apply to the members are those that apply equally among all members, those that apply differentially among members, and those that apply to specific people.

Those essential to any K-group g:

• K1 (street musician): These are governed by certain norms: a K1-group g owes respect to its surroundings in certain ways and is owed respect in certain ways. For instance, g should not block traffic, play too loudly, or disrupt businesses or pedestrians excessively. Conversely, listeners may ignore g, continue their conversations nearby, but if standing near g may clap or participate at certain points and not others, and should not disrupt the performance. (These norms, I should point out, are not derivable from moral norms that we owe individuals on the street. Instead, they are anchored by the intersection of practices regarding musical performance and practices regarding behavior on the street, as well as historical practices and perhaps laws regarding street performance in particular.)

• K2 (faculty committee): A group of this kind has a number of abilities, such as creating subcommittees, issuing rulings, and conducting meetings. It also has the ability to choose some of its own functions and to set up certain rules and norms for itself and for other people and events. It also has limitations and restrictions on its abilities: these committees are often held to specific timetables and activation dates, require approvals, and so on. And it has obligations to perform certain actions, to conduct meetings in certain ways, and so on.

• K3 (social class): Rights and powers involving control over the deployment of capital and influence over its regulation. (A topic to investigate is the extent to which these rights and powers are supplementary, or whether they are among the construction conditions for capitalist social classes.)

• K4 (DICD): None.

In addition to those applying to the group, there are also extra essentials that apply to members of a group, either all equally or differentially to different members.

Those applying to members of a K-group g:

• K1 (street musician): Applying equally: there are norms applying to all members regarding behavior on entering and leaving the group, responding to others appropriately, meshing play with that of others, how and when to take control for a solo, not dominating the performance, and so on.

  Applying differentially: norms differ according to the type of instrument being played, and if there is a leader then certain rights and obligations apply to the leader, and others to non-leaders.

• K2 (faculty committee): Applying equally: voting rights, rights to speak, call meetings, rights to information, obligations to attend.

  Applying differentially: based on the assignment of distinct roles and duties, for the chair (setting agenda, calling meetings, reporting responsibilities), for people who are not the chair.

• K3 (social class): Applying equally: individual members of the class have rights and powers involving control over the deployment of their own capital.

  Applying differentially: special rights may accrue to the wealthiest members.

• K4 (DICD): None.

Ritchie points out that many groups are essentially structured. Her notion of structure can be understood as the differential possession of functional role properties by individuals and binary relations between members of the group.³⁷ But group structure can involve more than what Ritchie’s node-edge structure suggests: the properties can include more than functional-roles, and the relations can be multi-place, not just binary.³⁸ Hierarchies are a special subcase of this sort of structure, involving certain relational powers and norms among members. And again, while some groups have structure like this, many do not. Many kinds of “organizations,” for instance, do not involve differential powers and abilities.

5.3 Distinguishing the extra essentials from the construction profile

How is it possible that a group’s essential properties are not already included in the criterion of identity or other parts of the construction profile? And how do we tell when an essential property is among the construction conditions and when it is not?

To answer the first question, it is helpful to clarify exactly what a criterion of identity for a kind K accomplishes. Consider the canonical form of a one-level criterion: for all x and y such that x and y are both of kind K, xRy iff x   =   y. To satisfy this formula, the criterial relation R needs to be reflexive when applied to objects of kind K, and also to be a minimal guarantee of identity if we already are given that x and y are of kind K. That means that the criterial relation does not need to include all the essential properties of K. After all, it is already given that x and y are of kind K. So R just needs to include enough texture to distinguish groups in K from one another. It is true that x and y must have the extra essentials in order to be groups of kind K. But those are not part of K’s criterion of identity.

Nor are the extra essentials captured in the other components of K’s construction profile. To see this, consider the following example. Suppose we have two group-kinds K5 and K6. Let us set up these kinds to be very simple. Suppose there is only one group of kind K5—call it G5—and one group of kind K6—call it G6. Suppose further that groups of both kinds have exactly three members, Alice, Bob, and Carol, and exist from January 1 to January 31, 2017. The groups exist in and only in worlds where Alice, Bob, and Carol all exist at least for that duration. Let the only difference between the group of kind K5 and the group of kind K6 be an extra deontic power that the respective groups have. The group of kind K5 has the right to assign spaces in the faculty parking lot, whereas the group of kind K6 has the right to assign faculty mailboxes. These essential characteristics are summarized in Table 1.

Table 1 Two kinds with the same construction profile but different extra essentials

In this example, G5 is the only member of K5 and G6 the only member of K6. Groups G5 and G6 exist at the same times and worlds as one another and are always constituted by the same people. The criterion of identity for K5 groups is the same as the criterion for K6 groups, as are their construction profiles. Still, they are distinct kinds, and G5 and G6 are distinct groups having different essential properties from one another.

Though this example is artificial, it is not just a weird thought-experiment. In fact, there are real examples of this in the actual world. In Epstein (2015) I discuss two actual coinciding groups—the board of the Massachusetts Department of Transportation and the board of the Massachusetts Bay Transportation Authority. These two groups are set up such that they coincide with one another. Yet they are distinct boards with distinct powers.³⁹

Turning to the second question: How do we tell which profile a given property belongs to? Take a given group power, member power, deontic property, norm, structure, or hierarchy. Is it among the construction conditions? Or is it an extra essential? The answer will vary from case to case. We might set up a group-kind such that a condition for Alice’s membership in a group of that kind requires that she have a certain deontic power. Or as we did with K5 and K6, we might set up a group-kind such that the possession of a deontic power is separate from the constitution conditions. The way to tell where the properties belong is to work through the profiles.

In both the construction profile and the extra essentials profile of a K-group, we are enumerating essential properties of groups of that kind. Together these profiles are simply a way to characterize all the essential properties of K-groups. Still, it is valuable to keep these profiles separate. There is more to the analysis of essential properties of groups than how they come to exist and are constituted, and certainly more than criteria of identity. Explicitly separating the construction profile from the extra essentials helps us analyze the properties of group-kinds accurately and comprehensively.

Moreover, keeping these profiles separate can also be helpful in analyzing categories such as races, sexes, and genders. Membership in such groups can sometimes be a matter of possessing a physiological marker, and yet the category essentially carries with it social norms as well.

6 A comment on group agency

Central to many people’s interest in social groups is the question of whether and how groups can perform actions and have intentional states, such as beliefs, desires, intentions, plans, knowledge, and reasoning. Some theorists also identify sociality with agency: the only kinds of groups that count as social groups are those that are group agents. As I argued earlier, this is too strong a condition for most real-world groups. Nonetheless, even though the category of group agents is only a small subset of groups, it is an interesting one.

Surprisingly, however, most discussions of the topic skip over a distinction that has long been central to work on individual agency: should we assess the agency of a thing by considering only the characteristics of that thing itself? Or rather, should we understand the agency of a thing in terms of the characteristics of the kind of which it is an instance?

A common approach to thinking about group agency is from the perspective of functionalism. We start with humans as a template for how the functions of an agent are performed. Humans can be understood to implement a system of practical activity, in which we employ our beliefs, desires, knowledge, intentions, plans, reasoning, etc., as parts of a system of acting in the world. We can describe that system as an abstracted or high-level set of interacting modules, each of which plays a functional role in a larger system. There may be many ways of implementing such systems. Dogs, Martians, and robots could each implement a system of practical activity in different ways. If one of these performs actions radically differently than humans do, we might not consider it to be the same functional system. But for the ones that have the same modules arranged in the same functional structure, we consider those different implementations of a single functional role. That role is complex, involving many modules and interactions and sub-roles. But we can think of it abstractly as the functional role of agency.

Call that functional role F. Many people have argued that it is not only humans, Martians, dogs, and robots that can implement such a system, but groups as well. Thus a number of theorists have worked to cash out what it is for a group to implement part or all of F. What it is, for instance, for a group to have a group intention, or group reasoning.

Missing from much of this discussion, however, is a distinction at the center of functionalism: between role-kinds and realizer-kinds.⁴⁰ The first is a kind defined by the role itself: g is an instance of kind K if and only if g performs role F. And there is another kind defined by having essential properties P such that groups having P normally perform role F, or such that the properties are chosen for that reason. That is, being a kind that actually realizes the function (within certain tolerances). For this sort of kind, g is an instance of kind K if and only if g has property P (even if g itself is not performing F).

It is rare for us to classify objects into pure role-kinds. We typically consider something an instance of a functional kind at times when it is not performing that function, or not able to perform that function. I, for instance, am an agent even when I am asleep. Even if we are functionalists about human agency and regard agency as essential to humans, that does not mean that we must perform (or be able to) the functional role at all times.

Moreover, we count humans as agents even if they can never implement function F. Consider Alice, who is fully mentally capable but unable to move, and hence to act. Or Bob, who has beliefs and desires, but is incapable of forming any intentions whatsoever. Alice and Bob are nonetheless agents, since they are humans and realize at least certain relevant states. Alice’s intentions and Bob’s beliefs are genuine intentional states, despite their not ever being able to play the relevant roles in the system of practical activity. The reason is that this is the way these states are implemented in humans, and human is a realizer-kind of F.⁴¹

How does this apply to group agents? If we want to preserve the analogy between a functionalist approach to individual agency and a functionalist approach to group agency, then we should not understand group agency to be the performance of functional role F by a group on its own. Rather, we might better understand it in terms of realizer kinds. This means that we do not, at first, assess groups one-by-one for whether they are group agents. What we assess is group kinds, to see if they are realizer-kinds of F. We find, for instance, that groups of kind K2 normally realize F in the actual world. Tufts faculty committees, that is, may have the right kind of constitution and powers so that they normally realize a modular system of practical activity. Now we find a particular K2-group g. Is it a group agent? To evaluate this we do not assess g on its own, but rather as a member of K2. Whether or not g is a group agent is parasitic on the fact that K2 is a realizer-kind of F.

We can find group agency in “functionally misfiring” groups just as we might in individual agents. Bob cannot form intentions, so his system of practical activity never works. Yet on a realizer-kind account of belief, Bob does have beliefs; he can be in one of the brain states realizing belief in humans. Being in that realizer-state suffices for him to have a belief, even though that belief will never play a role for Bob in performing F. Similarly, even if g cannot successfully perform all the interlocking functions of group agency, it still might be able to be in a “group belief” state or a “group intention” state. What that state is, for a K2-group, depends on how K2-groups realize those states. Since g is a K2-group, it may count as having a “group belief” just by being in the appropriate state, even if g is incapable of integrating that state into a system of practical activity.⁴²

The point is not to argue for a functionalist theory of individual minds or group minds. Rather, it is to highlight several things. First, in trying to analyze group agency, we make a mistake if we merely look for a special property common to individual groups that are group agents. Second, it highlights that when we begin to work on understanding and classifying social groups more generally, we need to think first about the nature of kinds of social groups. And third, it motivates looking into the profile I discuss next: profiles not just of the essential properties of groups, but also of what makes a group of a given kind have the essential properties it does.

7 The anchor profile

In the preceding sections, I have discussed essential properties of groups of various kinds. The conditions for a stage s to constitute a given street music group at t are very different from the conditions for s to constitute a given faculty committee at t, as are the powers and deontic properties of groups of these respective kinds.

But there is another fundamental difference between groups of these kinds: the source or metaphysical basis for these kinds of groups to have the properties they do. Think, for instance, about the conditions for membership in the faculty committee. We captured those conditions in the “construction profile.” But why are these the conditions for membership? What makes them the conditions?

One way to answer this question is with a causal explanation. We can give a history or genealogy for why the membership conditions were set up the way they were. Maybe the faculty senate at Tufts did a survey of universities nationwide, learned about the pros and cons of various membership strategies, and chose to follow this one. That tells a causal story about why these membership conditions were enacted.

But there is also a constitutive explanation.⁴³ Why are the membership conditions for K2 what they are? Because the faculty enacted those conditions. That enactment—consisting of intentions, speech acts, majority votes, and so on—is the metaphysical reason K2 has the membership conditions it does. K1’s membership conditions, on the other hand, have a different metaphysical explanation. Again, what the membership conditions are are captured in K1’s “construction profile.” But in the case of K1, there was no formal enactment. Instead, the category K1 and its associated membership conditions are set up in a more organic way. K1 and K2 differ from one another not only in what their membership conditions are, but also in what makes them have the membership conditions they respectively do. The membership conditions for K1 are, in other words, anchored by different facts than are the membership conditions for K2.⁴⁴ And the same goes for other properties of these groups.

The anchor profile of a kind of group is a list of facts that metaphysically put in place various properties of that group. Even for a given kind of group, some properties may be anchored in one way, while others are anchored in a different way. Take, for instance, kind K2, the Tufts faculty committee. For each of the components of the previous profiles, the anchors are those facts that set the conditions up as they are:

• Conditions for a K2-group coming to exist: anchored by the actions that enacted Chapter 1 of the Tufts Faculty Handbook

• Conditions for a K2-group g continuing to exist: anchored by the actions that enacted Chapter 1 of the Tufts Faculty Handbook

• Conditions for stage s constituting K2-group g at t: anchored by the enacting of the subsections of the Tufts Faculty Handbook dedicated to elections

• Criterion of identity for K2-groups: anchored by historical patterns regarding the individuation of formally established committees

• K2-group g having the power to create subcommittees: anchored by the enacting of subsections of the Tufts Faculty Handbook dedicated to the powers of standing committees

• K2-group g having quorum restrictions on taking certain actions: anchored by the enacting of the subsections of the Tufts Faculty Handbook dedicated to the powers of standing committees

• Etc.

As this example shows, even for a single kind of group, different properties may be anchored in different ways. Still, we may be able to generalize about how kinds of groups are anchored. For instance, both the construction profile and the extra essentials profile for K2 are anchored largely by explicit enactment. That may not be strictly true: some essential properties of K2 are at least partly anchored by habits, practices, and more. After all, explicit enactments leave out a lot of detail, which is filled in by patterns of historical practice and more. Even so, it may be fair to regard K2 as mostly anchored by enactments.

In contrast, the conditions for being a member of a street performance are not to be found in any bylaws, nor are they anchored by explicit choices or enactments. That does not mean they come out of nowhere. We have a long history of musical performances, activities on the street, and people joining together in crowds. There are thousands upon thousands of tokens of such events, and we also have attitudes toward these various events. It is facts like these that anchor the conditions for s being a stage of a street performance.

It is common to talk as though kinds of groups are consciously designed or chosen or set up. That is partly a consequence of the typical choice of examples. It is easy to talk about faculty committees, because they are largely set up by explicit choices and enactments, and their properties written down in the faculty handbook. But other kinds of groups are not like that: a much wider range of worldly resources figures into setting up their construction conditions, powers, limitations, and so on. With the caveat that these are rough, we can loosely generalize about the anchors for the construction and extra essentials of our four examples of group-kinds.

• K1 (street musician): anchored by functional intentions, practices, regularities in the environment

• K2 (faculty committee): anchored by agreement, enactment

• K3 (social class): anchored by functional niches, practices, regularities in the environment

• K4 (DICD): anchored by practices and regularities

In short, the essential properties of group kinds may be anchored in diverse ways. Some may be anchored by enactment or legislation, some by a functional role that they realize, some by the fact that they work in inductions, some by patterns. Presumably there are other ways of anchoring properties of group kinds as well.⁴⁵

In thinking about anchors, it is crucial to keep them separate from the causal reasons for setting up a kind as we do. Thomasson (2016) points out that in order to understand groups, we need to understand why we have the group categories we do. This is surely correct—I would only add that this question needs to be systematized and clarified, that it will have different answers for different kinds of groups, and that we especially need to distinguish the metaphysics of how group categories are set up from the causal histories of why we have set them up the way we have.

8 The accident profile

When it comes to explaining and classifying groups, we are not only interested in their essential properties. The accidental properties of groups can be equally or more important to understanding what groups are, and to classifying them or developing typologies.

To understand social groups, we want insight into how they actually work, what they are for, what characteristics they have. This is part of what it is to understand any kind of thing: for instance, if we are in the business of analyzing kinds of birds, we may be able to give a full account of their constitution and identity conditions without singling out that they fly. But we have hardly given an illuminating answer to the question of what birds are if we leave flight out of it.

Profiling the accidental properties of a kind of group might include anything at all. They can include properties that groups of the kind actually have in all or most cases, properties that members have, historical properties, size, location, and so on. Among the accidental properties are also various causal properties: the causes by which they came to exist, the causes for them to have the actual memberships they do, the causes for exercising various powers.

There are also the causes for the anchors to be in place. We might construct a typology of groups according to whether their anchors were influenced by the U.S. Constitution, or by the institution of slavery, or by bribes from interest groups.

There is no reason for a classification based on accidental properties to be of less typological interest than one based on essential properties. The question is the practical aims toward which we are putting the typology. Writing an accident profile for a given kind of group involves a choice of the properties of interest, and empirical investigation into the exemplification of those properties by actual groups. That means that the categorization will be tied to the actual world. But that’s fine, since we live in the actual world, and are categorizing groups for actual purposes.

To keep track of the elements of the profiles, it may be helpful to summarize them in one place. Table 2 lists all the components of the four profiles, which can be filled in for any given kind K of social group. This template can be a tool for analyzing a group kind, as well as for assessing how to place it in various systems of classification.

Table 2 Profile template for group kind K

9 How to classify kinds of groups

How should social groups be categorized, or organized into taxonomies? The profiles we have developed give us many options. They can be classified according to the components of their construction profile: according to how they are brought into existence, or what sorts of conditions people must satisfy in order to be members, or according to their criteria of identity. Social groups can be classified according to their extra essentials, such as the norms or abilities they have that are not part of their construction conditions. Or we can classify groups without regard for whether the properties that interest us are part of the construction conditions or the extra essentials. For instance, we might classify groups according to whether they are organized in hierarchical structures, regardless of where the norms and abilities fit into the profiles.

We might conceive of typologies of social groups much as we would a typology of foods or recipes. There is a long list of dimensions—cuisine, nutritional values, taste, cost, appearance, calories, skill to make, number of ingredients, types of ingredients, etc.—from which to choose. A useful typology can divide dishes up according to a single dimension, a set of dimensions, or cut across all the dimensions in one way or another, depending on the purpose.

We classify foods according to accidental properties, as well as to essential ones. We might, for instance, classify kinds of foods according to the caloric content of their actual instances. Similarly with social groups: we can classify them according to how long they tend to persist, or how effective they are at achieving their aims, or how ethnically diverse their memberships are.

It is also interesting to classify kinds of groups according to how their properties are anchored. We can classify group-kinds, for instance, according to how their existence conditions or their construction conditions are anchored. Sometimes, as I pointed out in Sect. 7, we can loosely generalize about the anchors for many of the essential properties of a given kind of group. For some group-kinds, their construction and extra essentials are anchored largely by agreement or legislation. For others, they are anchored largely by functional roles in a context, or by patterns of practices.

Yet another basis for classifying kinds of groups is according to the accidental properties of their anchors. For instance, we might classify kinds according to whether their anchors are caused by historical injustice, or by certain economic conditions.

There are so many ways to construct typologies of groups that it is difficult to say much in general. But let me provide a bit of detail to show how we might get started. I will look at the construction profile for K2-groups (Tufts faculty committees).

To begin, we can distill or tabulate the elements of a profile for a given kind of group. Table 3 depicts a tabulated construction profile of K2. The columns, labeled along the top, separate the four components of the construction profile, and the rows are one (somewhat arbitrary) way of dividing properties into different kinds. The cells indicate whether a given kind of property is part of a given component of the construction profile, and whether what is included is the property manifested synchronically (marked “s”), diachronically (marked “d”), or both (marked “s&d”). In columns B, C, and D, the notation “$\rightarrow +$” in the cells is to remind us that the earlier columns are included in the later ones. The factors listed in column C, for instance, are the additional ones in virtue of which s constitutes g at t, but only once it is given that g exists at t, which involves the factors listed in columns A and B.

To go through an example, consider row 8, the attitudes of other people (that is, people other than group members). The attitudes of other faculty members at $t_{0}$ and preceding $t_{0}$ (such as their intentions in the course of enacting the committee’s existence) figure into creating a faculty committee g at time $t_{0}$. In addition to that, historical and current attitudes of others are involved in g’s existence at a subsequent time t (for Tufts to continue to exist, for instance). For a stage s to constitute g at t, the appropriate rotations need to have taken place prior to t, which involves attitudes prior to t. And the identity of $g_{1}$ and $g_{2}$ can be guaranteed by their respective stages tracing back to one originating act, which involves historical attitudes.

Table 3 Tabulating elements of the construction profile for K2

To extract patterns for classifying kinds of groups, we would need to tabulate many different kinds, not just K2. Then we could see whether it is preferable to classify groups according to whether intrinsic or extrinsic properties are involved in various construction conditions, or whether synchronic or diachronic properties are involved, or according to some other component of their existence, constitution, or identity conditions.

Even looking at this one kind of group, though, makes it clear that no simple typology is likely to be too informative. And that there is not likely to be any simple way of circumscribing the social groups. Social groups are just too heterogeneous, along too many dimensions.

10 Conclusion

When we analyze kinds of social groups in their detail, we see the many spectra along which they lie. Even the membership conditions for different kinds of groups vary enormously. Some involve intrinsic properties of members, some involve extrinsic properties, and some involve properties that have little to do with the members at all. Equally diverse are the conditions under which groups of various kinds come to exist, and continue to exist, and the ways groups are individuated. Kinds of groups may have a range of norms, powers, and abilities, some of which are among their construction conditions and some supplementary. And they are carved out the way they are for a variety of reasons as well, both metaphysical and causal.

At some level, we should have expected this variation. After all, human societies have been around for as long as there have been humans; and from the beginning, the ways we arrange ourselves and get arranged—consciously and unconsciously, accidentally and purposefully, beneficially and maliciously—have evolved nonstop. I suppose there was a remote possibility that all the various kinds of groups would neatly divide into vast kingdoms, as biological kinds divide (even if imperfectly) into plants, animals, fungi, and the others. But as it turns out, there is not the slightest evidence that this is so.

Careful analysis of group kinds can be immensely profitable. We can construct models with confidence and precision about their building blocks. We can design new kinds of groups with a fuller understanding of all the options and choices we have developed through the years, rather than revisiting the same small subset of design options over and over. And we can approach thorny issues like group agency and responsibility without flailing atop a muddy understanding of the objects whose agency and responsibility we are assessing.

Even with this framework, there remains a great deal of work to be done in understanding social groups. Each of the four profiles needs more exploration: the construction profile may be reasonably complete, but the others are only a start. These profiles need to be filled out, and patterns assessed, for many kinds of groups beyond the few examples I have sketched. Useful ways of classifying group kinds deserve more investigation. And more work needs to be done on interesting families of group-kinds. The family of “constitution-dominated” kinds, for instance, deserves attention, and groups that have traditionally been considered “feature groups” need to be re-examined.

It may also be fruitful to consider whether these four profiles, and the general approach to the metaphysics of groups, can be applied to other sorts of objects. This inquiry has been informed by insightful work by metaphysicians on constitution and ordinary objects. Perhaps we can return the favor. The results in this paper, derived from scrutinizing groups, in some ways mirror current treatments of ordinary objects, in some ways extend them, and in some ways depart from them. Perhaps these extensions and departures can help inform a better understanding of objects more generally.