Individualism is one of the oldest and most venerable topics in the philosophy of social science. But few topics can attract a high level of attention without being depleted, and as a consequence, until recently, the prospect of individualism seemed rather bleak. Surely everything that could possibly be said about it had already been said. As a research topic, individualism looked as dry as the Sahara desert.
The Ant Trap has made me revise these preconceptions. The Ant Trap is not only the best book on individualism of the last two decades; it is the best monograph on social ontology that I have read in a long time. It injects life in a moribund topic and introduces conceptual tools that will re-orient some of the debates in this field of philosophy. Part of its success is due to Brian Epstein’s deep knowledge of analytical metaphysics, an area of philosophy that has been bustling with activity during the last couple of decades. But borrowing tools from a neighbouring field is rarely sufficient to do interesting philosophy. The key is how the tools are used, and Epstein does a remarkable job, captivating even those philosophers who – like me – find much of contemporary metaphysics quite boring and pointless.
The Ant Trap is a very rich book, and my assignment for the symposium is to comment on the first part only. Before I proceed, I would like to put on record that I agree on many things that Epstein says. I agree with his diagnosis of the state of social ontology, and I endorse his attempt to shift the focus from collective intentionality onto other important topics. His heart is in the right place on many issues that I will not mention here for lack of space, and because our job is to focus on the controversial aspects of his monograph. As usual, I will spend most of the commentary focusing on those few parts of The Ant Trap that failed to convince me completely, or where I feel that Epstein could have told us more.
2 Two Kinds of Individualism
The first part of The Ant Trap is mainly devoted to illustrating and defending Epstein’s claim that individualism is not one but (at least) two positions: there is individualism about grounds, and individualism about anchors. The debate on ground individualism is basically the traditional debate on ontological individualism. The main issue at stake is what ‘high-level’ social facts (facts about groups or institutions, for example) are based upon. Are group-level facts entirely grounded in facts about individuals? Is the fact that the fourth conference of the European Network of Social Ontology is taking place (let us call it fact K) entirely grounded in the fact that some individuals are meeting in Palermo right now (let us call it fact F)? If not, what sort of facts ground K?
Ontological (ground) individualism however has limited scope. Epstein argues convincingly that the conceptual framework within which the debate has traditionally taken place is unable to capture several important issues that arise from recent work in social ontology. This work focuses on the way in which non-social (e.g. physical, or biological) objects, properties, and facts are transformed into social objects, properties and facts. The key question addressed in such debates thus is: what makes certain facts the grounds of certain Ks? For example: what makes the fact that my proposal has been selected by the Organizing Committee the ground for the fact that I have presented a paper at ENSO IV? (If the proposal of my paper had not passed the selection procedure, I could have still made a presentation in front of Palazzo Steri; and yet, in spite of its physical proximity with the other presentations, my performance would not have belonged to ENSO IV.) To mark the distinct character of these questions and to identify a new metaphysical relation, Epstein introduces the term ‘anchoring’. The central issue in many contemporary debates is individualism about anchors, and the answers to anchoring questions do not depend necessarily on the answers given to grounding questions.
Epstein discusses three theories of anchoring that have been, and still are, prominent in the literature. The first one is David Hume’s account, according to which social facts are anchored in social practices and in people’s beliefs that the practices are mutually beneficial. The second theory is John Searle’s account of institutions based on the acceptance of constitutive rules. And the third one is H.L.A. Hart’s theory of primary and secondary legal rules anchored in regular conformance, standards of behaviour, and active criticism by the members of the (legal) community.
Since Searle’s theory makes the ground-anchor distinction particularly transparent, I will use it as a reference point from now on. 1 According to Searle, the fundamental mechanism for the creation of institutions is the collective acceptance of a Constitutive Rule (CR). For example, let us take the following rule:
(CR) If a paper bill has been issued by the Bureau of Engraving and Printing, it counts as money in the United States.
‘Billy’ (a particular piece of paper) is money in the United States because it satisfies the conditions (C) stated in the CR, and we accept the CR. Epstein argues that any theory that fits this format (the Standard Model of Social Ontology) is implicitly committed to specifying the two metaphysical relations of grounding and anchoring. For example, in Searle’s account:
to be issued by the Bureau of Engraving and Printing is what grounds the fact that X is money;
collective acceptance of CR is what anchors the fact that these are the grounds for X being money.
Searle is an individualist about anchors, but not about grounds. The facts that ensure that Billy is money in the United States are not facts about individuals – they are not even facts about human beings, actually. Human beings and their intentions only play a role at the level of anchors: they are what makes facts like being issued by the BEP the ground for being money in the United States.
Epstein uses the ground-anchor distinction to argue that individualists have a hard task because they have to defend two claims at once (individualism about grounds and about anchors). So a good reason to keep the latter separate, from an individualist’s perspective, is that at least one of them has a better chance to survive (although Epstein is sceptical about both).
The ground-anchor distinction is a very important contribution, which allows us to see that what seemed to be one project actually consists of two separate projects. It prevents us from mixing separate issues and helps us to focus on the right level of analysis depending on which of the two metaphysical relations is problematic from an individualistic point of view.
In the remaining part of this commentary, however, I would like to focus on a claim that Epstein makes in the course of the book but that is left partly unexplored and unjustified, namely that there are not only different grounds for the same types of entities (for example, different grounds for being money, in different circumstances), but also different anchors. Epstein in other words endorses pluralism about anchors and grounds, a position that I find problematic and that I think deserves more discussion than what we find in The Ant Trap.
Pluralism about grounds seems quite natural: Searle’s theory makes it seem obvious, for example, because the conditions of satisfaction that are part of a constitutive rule are conventional: the condition to be money in the United States is to be issued by the Bureau of Engraving and Printing; in the European Monetary Union, the condition is to be issued by the Central European Bank; in ancient Rome, the condition is to be minted by the Emperor. Since in different circumstances different conditions and different constitutive rules apply, there are different grounds for being money.
Pluralism about anchors is less obvious. Searle, Hume, Hart, and pretty much everyone else, have looked for the unique mechanism through which social facts are grounded in a particular kind of (non-social) facts. They have proposed general theories of social ontology that compete with one another. Although these philosophers have disagreed about anchors, in other words, they have agreed on monism about anchors: they have assumed or argued that there is only ‘one glue that holds social kinds together’, to use a metaphor that is dear to Epstein.
theorists have insisted that there can be only one schema by which anchoring works. Even a quick look at the last chapter shows that this is unlikely to be right. […] A theory of anchoring needs to confront the frame principles for many kinds of social facts. (p. 105)
Epstein does not say much more than this in The Ant Trap, but similar claims are repeated in an article (Epstein 2014, p. 42, 54) the content of which has not been included in the book. That article, unlike the book, is full of interesting examples and yet does not make a general, convincing case for pluralism. It is a pity, because anchor pluralism sets Epstein apart from all the philosophers who have written on social ontology in the recent past. So the question whether pluralism holds or not is extremely important and worth probing.
To trigger the discussion, I will argue that pluralism results from overlooking a simple distinction. Once the distinction is taken into account, pluralism loses much of its appeal.
4 Tokens and Types
The distinction that I am referring to is the familiar one between tokens and types. 2 Tokens are particular entities located in space-time. For example: the laptop I am working on is a token with certain unique characteristics (its serial number, for example, and a little scratch on its keyboard), a specific history (it came into being sometimes in 2013) and location (it is usually in Milan, although it has travelled widely across Europe). Nonetheless, it is not totally unlike other laptop computers. It is a computer of a certain kind (a MacBook Air), with characteristics shared with other similar machines. A MacBook Air is a type of computer, and as such it is not spatio-temporally restricted. In a sense, it has many locations – as many as its instantiations. But the type itself, as distinct from its instantiations, is neither here nor there.
The mode of existence of types is a difficult question that metaphysicians have debated for more than 2000 years. According to philosophers with a nominalistic attitude, types are concepts existing only in the minds of theorists. According to realists, they are (abstract) entities that exist independently of anyone’s theoretical imagination. Be that as it may, the type-token distinction is what interests us here, and the distinction holds independently of one’s metaphysical inclinations.
Now, the type-token distinction can be applied to social entities as well. Take a couple of familiar institutions: money and marriage do not have a specific history and geographical location, but they can become instantiated in innumerable ways. They are institution-types. Marriage as codified in Italian law in contrast is a specific form of marriage, characteristic of a particular society during a particular period of time. Similarly the euro is a particular instantiation of money, characteristic of a specific time and place. The euro and marriage-as-codified-in-Italian-law have a beginning, a geographical location, and inevitably will come to an end.
This application of the type-token distinction mirrors a similar distinction between species and higher taxa in biological classification. According to a tradition that goes back to Ghiselin (1974) and Hull (1976), species are individuals (tokens) in virtue of their historical contingency, while higher taxa are classes or types. An important motivation behind this distinction is that types are not involved in causal processes in the same way as individuals do. They do not interact with each other, for example, or with other entities. So tokens seem to enjoy a kind of existence that types lack. While marriage as a type is causally inert, marriage-according-to-Italian-law affects the behaviour of Italians and is (arguably) affected by it.
Whether the analogy holds perfectly, of course, may be disputed. However, nothing hinges on this particular choice of classification: the main point is that social entities are stratified at different ontological levels, and that we should better keep these levels separate. Although I prefer to use the token-type distinction in the same way as biologists do, a classification based on different levels of types will do equally well for the point that I intend to make here.
When we identify the grounds and anchors of a social entity we must be careful to specify the ontological level it belongs to. For example: Searle’s theory of constitutive rules is essentially a theory of tokens. 3 The conditions in Searle’s constitutive rule for ‘money’ for instance specify the grounds for being a currency, not for being money. Being issued by the Bureau of Engraving and Printing is the ground for being a dollar, and dollars are particular instantiations of the type money. That’s why the bills in my pocket are money even though they have not been issued by the BEP.
In this case we have an established terminology to mark the distinction: a currency (e.g. dollar) is a token institution, whereas money is an institution-type. In other cases unfortunately we are not so lucky, and we must be very careful not to mix things up. We must be careful because type-token confusions can have important practical consequences: they have been used by conservatives for example to argue that contractual unions between same-sex couples are not marriages. But once the distinction between the two levels has been clarified, we can see that the claim made by conservatives holds for marriage-as-codified-in-Italian-law, not for marriage in general. A contract for same-sex unions may well be a marriage even if it does not satisfy the conditions that are currently specified in the law. 4
The type-token distinction is important also because it calls pluralism about anchors and grounds into question. First of all, it makes us realise that the grounds of tokens differ from the grounds of types. Money (the institution-type) as we have seen is not grounded in conditions such as being printed by the Bureau of Engraving and Printing. Standard economic theories define money in a functionalist manner: money is instantiated in whatever entity fulfils the classic functions of money (medium of exchange, store of value, and unit of accounting). Money is what money does, and collective acceptance of conditions such as being printed by the BEP is not a requirement for something to function as money. Cigarettes can work as money, in some circumstances, just as cowry shells and animal pelts have fulfilled the function of money in specific places for some time.
What does this imply for ground pluralism? Both sestertii and dollars are money for the same reason, i.e. because they fulfil the same functions. So the grounds for money (the type) are the same in all cases: to be money is to be a medium of exchange, a store of value, a unit of accounting. Ground pluralism does not hold at the level of types.
At the level of tokens, clearly the grounds for Billy to be a dollar are not the same as the grounds for Romy to be a sestertius. Each one has its own grounds, and only one set of grounds. Therefore, ground pluralism does not hold at the level of tokens.
So much for ground pluralism. I will address pluralism about anchors later, after I have explored some implications of the ground-anchor and token-type distinctions for The Ant Trap. Some of the issues are small and only concern specific arguments made by Epstein. Others are bigger and concern the very nature of social kinds. In the next couple of sections I will look at both, starting with the small ones first.
5 Tricky Counterfactuals
Pluralism about grounds conflates different levels of analysis and different entities. A sestertius and a dollar are different social entities at the right level of description. But if we describe them at a higher level (as money) they misleadingly appear to be the same entity, endowed with different grounds. This conflation of levels surfaces unexpectedly at various points in Epstein’s arguments. Here’s an example from Chapter 9 of The Ant Trap: Epstein is looking at some tricky counterfactual claims that he uses to attack ‘conjunctivism’. 5 A crucial premise of his argument is that ‘In assessing the grounding of social facts across possibilities, we take the grounding conditions to be fixed’ (p. 123), separating them from the anchors. Thus, for example,
We can look back at ancient societies, and evaluate whether there are classes or castes, aristocrats or serfs. We might look for baristas in the Ottoman Empire or in Seventeen Century England, and variable annuities among the ancient Egyptians. (ibid.)
Statements of this kind are ambiguous: what are we looking for when we engage in this kind of inquiry? Identical tokens or similar institutions belonging to the same type? It seems that Epstein is referring to tokens: he suggests that
One is a war criminal if one has committed or has conspired to commit any of a long list of crimes in association with armed conflict. We can sensibly ask whether Caligula was a war criminal, or whether Genghis Khan was. (p. 124)
The suggestion is that when we ask questions of this sort we implicitly apply the conditions that hold now for social roles like ‘barista’ or ‘war criminal’, and check if they were satisfied in the past. When we do this, we ignore the anchoring conditions; that is, even though there was no International Court or no Geneva Convention in the thirteenth century, we can still say that Genghis Khan was a war criminal. ‘This is simply how we use social facts’ (ibid.).
This is controversial terrain of course. Many constructivist, relativist, historicist philosophers and social scientists would find Epstein’s claim disturbing. They would say that to classify Gengis Kahn as a war criminal would be a misapplication of this category, outside of its appropriate social context.
Although in general I am not very sympathetic to constructivism, I think that there is a genuine ambiguity in this case: ‘war criminal’, ‘barista’, ‘serf’, can be used in different ways. These terms may refer to whoever satisfies certain conditions (e.g. working in a bar under certain contractual obligations), or more generally to those individuals who fulfil a certain role in society, broadly conceived (e.g. serving food and drinks). When we apply Epstein’s counterfactual we may either focus on the specific conditions (the conditions that identify the token), or on the general role (the role that defines the type). We can ask:
was there any token X fulfilling conditions C (as we define them now)?
was there any institution similar to the one that we have now?
The first one is a question about a token, the second is a question about a type. Whichever question we focus on, it is important that we clarify which one we are interested in.
For example: as Epstein says, we can apply to famous villains of the past the conditions that we take to define a war criminal now, in 2015. If we apply contemporary criteria, we can identify as a war criminal even someone who lived in the thirteenth century. The idea presumably is this: we take the nearest possible world to our own where there is a guy with Genghis Khan’s CV. Then we ask whether he fulfils the conditions that a war criminal must satisfy in 2015. Provided we specify which token institutional role we are taking as a reference point (war criminal as defined today, say), we can ask and answer such questions.
But another possibility is that we are asking: was Genghis Khan a war criminal according to the conditions that applied in 1200? In this case we would be looking at whether an institutional role of that kind existed in Genghis Khan’s society. The way to ask this question is not to apply our conditions for being a war criminal, but to ask whether in Genghis Khan’s world there was an institutional role that fulfilled roughly the same function as the role ‘war criminal’ does in ours. This is a question about the institution type.
This distinction suggests that the decoupling of anchors and grounds may be deceptive. The reason we can ask questions of the first kind is not that the grounds for war criminal are different from its anchors. Rather, we can ask such questions because tokens have different grounds from types. When we look for tokens, we look for people who have committed horrible crimes during a war. When we look for types, we look for international courts or Geneva-style conventions.
The role of a war criminal functions mainly as a deterrent: since bad things are supposed to happen to war criminals when they get caught (martial court, imprisonment, capital punishment), the main point of having the institutional role of war criminal is to reduce the level of gratuitous violence in a war. If this is the question we are asking, then it turns out that there were no war criminals in the thirteenth century, because no such institution existed.
This does not mean that cross-temporal questions are meaningless, or that the answer must be always negative. Moving a couple of centuries forward one can find some plausible candidates for the institutional role of war criminal. In 1474 commander Peter von Hagenbach (the inventor of the ‘Lanzknecht’) was convicted by an ad hoc court of the Holy Roman Empire for atrocities committed by his troops during the occupation of Breisach. His troops murdered, raped, and pillaged the city, crimes that ‘he as a knight was deemed to have a duty to prevent’. According to contemporary commentators such acts went against the laws of God and nature. Von Hagenbach claimed – not very imaginatively, in retrospect – that he was just following orders, but he did not convince the court and was ‘drawn and quartered’ after the trial. Although the notion of war crime was refined in the Hague (1899) and Geneva (1949) conventions, the institution that convicted von Hagenbach was similar enough to its contemporary counterpart that it makes sense to say that there were war criminals in the fifteenth century, both in the token and in the type senses of the term.
Let me now turn to some larger issues raised by Epstein’s pluralism. They are fundamental metaphysical issues, concerning the very nature of social types. What are social types, and what do they do for us? Do social types fulfil just one function or do they do several things at once? In general: can the realm of the social be explicated by a single overarching anchoring theory, or do we need a patchwork of more specific accounts, each with its narrow domain of application?
As a naturalist, I think we should take seriously the hypothesis that the natural and social world are held together by the same ‘glue’. This is admittedly a scientistic approach: the idea is that the classifications that matter are those found in our most advanced scientific disciplines, because the latter provide the most powerful tools to investigate the fundamental structure of the natural and social world. If the social sciences licence a unified account, then we should be monists about social kinds.
One way to develop this idea is to follow the tradition of ‘real kinds’ pioneered in the Nineteenth Century by John Stuart Mill, John Venn and William Whewell, and updated recently by Richard Boyd. The kinds identified by our best scientific theories are clusters of properties held together by ‘homeostatic’ causal mechanisms. 6 Because the properties tend to be co-instantiated, real kinds support inductive inference. The inferences are useful for pragmatic purposes, but can also be used to test theoretical hypotheses and to find out about the way the world is structured.
How can these ideas be extended to social kinds? Again, the basic idea is that social entities are grouped in virtue of objective correlations that exist among their properties. At the high level, the level of types, social kinds are grouped according to their functional properties. At the low level, the level of tokens, there are objective correlations between the grounding conditions and the behaviour of people who are engaged in social interaction.
For example: the functions of money (the type) are correlated in interesting ways. Being a store of value is a precondition for being used as a medium of exchange, for the trivial reason that trading takes place over time: if the value of a currency decays too quickly, then that currency cannot solve the problem of coincidence of wants. 7 Or, to take another example: investment in kids’ welfare and education is costly, so it is not a mere coincidence that most marriage contracts regulate both child-rearing and financial support. The functions are correlated because marriage institutions solve a cluster of problems that are objectively related to one another.
Let us now move to the level of tokens. The problems that institutions solve are typically problems of coordination and cooperation. Since coordination requires predicting each other’s behaviour, token institutions usually facilitate inductive inferences. Consider currencies again: if being printed by the Bureau of Engraving and Printing did not support the prediction that people will exchange commodities for paper bills printed by the BEP in the future, then the conditions that ground the kind ‘dollar’ would cease to influence people’s behaviour. Or, similarly, if the fact that they said ‘Sì, lo voglio’ in front of an official did not help predict the behaviour of two spouses, then that ceremonial formula would cease to glue the kind marriage-according-to-Italian-law together.
This gives us a unified theory of institutions. The theory is monist, in the sense that it tells a single grounding and anchoring story for all the social entities that belong to the same level (a story for tokens and a story for types). The scheme of Table 1 may help clarify the relationship between Epstein’s metaphysical categories and the theory of institutions shaped on the Mill–Venn–Whewell–Boyd approach.
To defend this account in detail would require much more space and would be beyond the scope of this commentary. All I want to suggest is that monism is a defensible position, and that we should think twice before we give it up (metaphysicians should be looking for theories that are as general as possible, after all). There are various reasons why one may dislike the real kinds approach. John Dupré (1993), for example, has argued that the scientific taxonomies of biology do not and should not be privileged with respect to other kinds of classifications. Grouping onions and lilies together makes sense for the purposes of science, but not for the purposes of cooking. So perhaps Epstein’s liberalism on anchors and grounds is rooted in the same sort of pragmatism that informs Dupré’s ‘disunified’ metaphysics.
One way to test monism would be to ask: are there any social entities, anchored in different ways, that are not used primarily to make inductions? Speaking of biology and psychology, Paul Griffiths (2004) has proposed a distinction between epistemic and normative kinds. Scientific kinds on the one hand are used primarily for prediction and explanation; normative social kinds on the other are used to evaluate, blame, punish and praise. This fits the case of social kinds rather neatly: while the notion of ‘first-price sequential auction’ is used primarily to denote a mechanism that matches buyers and sellers in a market, the term ‘war criminal’ is used to deter soldiers from committing cruel acts during an armed conflict.
My own inclination however is to resist the proliferation of kinds: normative kinds are also tools that we use to predict and control behaviour. If the ‘war criminal’ label sticks to General Mladic, then we can make all sort of predictions about his behaviour. We can predict that he will probably not visit New York City, he will not run for president, he will hide somewhere where he cannot be arrested, and so on. Using the same concept we can also issue predictions about the behaviour of other people (Serbian voters and politicians, UN troops). 8 I think that this point can be generalised to most other social kinds, including religious and artistic kinds. Fundamentally, the main point of introducing institutions is to control and predict behaviour.
Epstein’s pluralism – i.e. the idea that there are several possible anchors and grounds for social entities – is not the only game in town. I have argued that there appear to be several grounds for the same kinds only if we mix the levels of description. (For example: there seem to be several grounds for money, only if we mistake currencies for money.) But there is no plurality of grounds when we describe social entitites at the right level: there is only one set of grounds for dollars or for sestertii (tokens); and one set of grounds for money (type). One set of grounds for ‘servo della gleba’ or for ‘muzhik’ (tokens); and one set of grounds for ‘serf’ (type).
A similar moral applies to anchors: once social entities belonging to different taxonomic levels are distinguished, we can tell a unified story for each level. In the last part of the commentary I have tried to elaborate some consequences of the monism-pluralism contraposition. Although I am aware that we are still far from reaching a consensus, we should be grateful to Epstein for having provided a conceptual framework within which these questions can be asked.
Boyd, R. (1991): “Realism, Anti-foundationalism, and the Enthusiasm for Natural Kinds”. In: Philosophical Studies 61, p. 127–148.Google Scholar
Dupré, J. (1993): The Disorder of Things. Cambridge, Mass.: Harvard University Press.Google Scholar
Epstein, B. (2014): “How Many Kinds of Glue Hold the Social World Together?”. In: M. Gallotti and J. Michael (Eds.): Perspectives on Social Ontology and Social Cognition. Dordrecht: Springer, p. 41–55.Google Scholar
Epstein, B. (2015a): The Ant Trap: Rebuilding the Foundations of Social Science. Oxford: Oxford University Press.Google Scholar
Epstein, B. (2015b): “What Alabama’s Roy Moore Got Right”, In: The New York Times, February 16, 2015. http://opinionator.blogs.nytimes.com/2015/02/16/what-alabamas-roy-moore-gets-right/.
Ghiselin, M. T. (1974): “A Radical Solution to the Species Problem”. In: Systematic Zoology 23, p. 536–544.Google Scholar
Griffiths, P. (2004): “Emotions as Natural and Normative Kinds”. In: Philosophy of Science 71, p. 901–911.Google Scholar
Guala, F. (2016): Understanding Institutions. Princeton: Princeton University Press.Google Scholar
Hull, D. L. (1976): “Are Species Really Individuals?”. In: Systematic Zoology 25, p. 174–191.Google Scholar
The same point can be cashed out in terms of particulars and universals, individuals and classes, or similar notions. There are subtle differences between these concepts that, however, do not matter for our purposes.
Conjunctivists reject the ground-anchor distinction, and hold that anchors should be included among the grounds of a social fact. The refutation of conjunctivism thus is a crucial step in Epstein’s overall argument.
A mechanism is homeostatic if it has the tendency to bring a system back to its equilibrium state when there is a small perturbation. The idea is that the entities that possess the standard properties of the cluster tend to persist and to proliferate, while those that lack some properties tend to acquire the missing property or disappear. See e.g. Boyd (1991).
There is a problem of coincidence when Ann wants to trade fruit for vegetables, Bob wants to trade vegetables for meat, and Carol wants to trade meat for fruit. Currencies solve problems of this kind by working as media of exchange.
This is why it does not seem terribly interesting to ask counterfactuals like ‘would Genghis Khan be a war criminal according to our standards?’ Why would we want to ask that question? Not to blame him, presumably, but neither to predict his behaviour.
About the article
Published Online: 2016-03-23
Citation Information: Journal of Social Ontology, ISSN (Online) 2196-9663, ISSN (Print) 2196-9655, DOI: https://doi.org/10.1515/jso-2016-0003.
©2016, Francesco Guala, published by De Gruyter. This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License. BY-NC-ND 3.0