Purity and Practical Reason: On Pragmatic Genealogy

1. The Basic Structure of a Pragmatic Genealogy

In his recent book, and in a series of papers, Mathieu Queloz has helpfully systematized pragmatic genealogy, giving it the argumentative clarity that it has often lacked (Queloz 2021). This, I should stress, is an enormous service to anyone interested in the topic. Williams and Craig were less than ideally forthcoming about the assumptions and movements of thought which drove their own genealogical projects, and Queloz is to be commended for making everything relatively clear. So, while I’ll focus my critique on his articulation of the method, I should be clear that this is only because he has done us the service of telling us what the method actually is.

Since I will be examining the method in detail, it will be useful to lay out the entire argument-form as Queloz presents it. Its point is to discover reasons for us to continue to support or engage in a given practice. To begin, we have the idealized development of that practice:

• (P1)

  In a prototypical group G, a set of root needs RN under root conditions RC generates a practical problem.

• (P2)

  This generates a practical pressure on G to solve the problem: the target need TN.

• (P3)

  Prototypical conceptual practice CP would meet the target need TN by serving point P.

• (P4)

  CP could develop quite naturally, i.e., out of the capacities we are prepared to grant G anyway.

• (C1)

  Therefore, circumstances permitting, CP would be highly likely to develop in G.

Next, we have a claim about the practical normativity of the practice, couched in terms of what would be rational in the idealized scenario:

• (C2)

  Therefore, it is rational for G to engage in CP in order for P to be served in G (in the sense that people with these needs under these circumstances would welcome and, if they could do so, aim for engagement in CP with a view to securing P).

Third, we have a similarity claim, which serves as a bridge between the idealized model and our actual world:

• (P5)

  In our actual group G*, there are close analogues to RN and RC.

Fourth, we have a claim about the practical normativity of the practice as it applies to us right now:

• (C3)

  Therefore, it is also rational for us (G*) to engage in CP*, the closest analogue to CP in G*, in order for P to be served.

Fifth, and finally, we discover that the practice’s current function is to serve point P, and that this means that we have reason to continue to engage in it:

• (C4)

  Therefore, the best explanation for why we go in for CP* is that it serves point P.

• (C5)

  Therefore, there is a prima facie reason for G* to continue to engage in CP*, and CP* is to that extent vindicated. (Queloz 2021)

In sum: a practice develops in a model because it satisfies root needs, and so it is rational for the inhabitants in that model to continue the practice. And since the model is relevantly similar to our actual situation, we can conclude that the function of the practice in our world is to satisfy roughly the same needs. Therefore, we also have reasons to engage in the practice. I will take each of these clusters in turn, pausing only to note when one of the continuity, collapse, or reductio objections has been triggered.

2. The Idealized Development of a Practice

To begin, we have the development of a practice, CP, in a group, G:

• (P1)

  In a prototypical group G, a set of root needs RN under root conditions RC generates a practical problem.

• (P2)

  This generates a practical pressure on G to solve the problem: the target need TN.

• (P3)

  Prototypical conceptual practice CP would meet the target need TN by serving point P.

• (P4)

  CP could develop quite naturally, i.e., out of the capacities we are prepared to grant G anyway. (Queloz 2021: 229)

I should explain that a “root need” is basically a kind of structural need which can be derived from first-order needs that all human beings have; that is, the need for shelter, food, water and so on (Queloz 2021: 92). A division of labor is a root need, because while it is not itself a basic need, it can be derived from such needs in ordinary human circumstances. You don’t get food, water and shelter very efficiently without one. We can therefore portray the human-like beings in this fictional scenario as feeling a kind of pressure: they need to solve this problem, and some new conceptual innovation, CP, as embedded in some new practice or institution, solves it for them. Moreover, this solution is one we would not be surprised to see enacted in any similar group of beings, given their cognitive and social capacities.

Queloz continues:

• (C1)

  Therefore, circumstances permitting, CP would be highly likely to develop in G.

This is the first substantive conclusion: the specified combination of needs, capacities and circumstances makes the practice likely to develop. While I won’t dwell too much on this point, I will have to note that in order for the practice to actually be likely to develop, a great deal of information will have to go into the model. As Queloz knows, it is not true that any feasible practice that satisfies our root needs is necessarily likely to develop. Several other conditions must obtain; at the most basic level, there must not be any other existing practice which works against the development of CP. Gender equality is almost certainly a huge boon for human societies, but humanity refused to even entertain the idea for thousands of years because other practices steadfastly militated against it. More basically, CP simply might fail to occur to anyone, and the fact that it could doesn’t mean that it will.

While some defenders of the model have argued that its simplified structure is similar to idealized modeling in the sciences, the contrast with the physical sciences is actually very unflattering to the pragmatic genealogist (Kusch & McKenna 2020). Certain models of the solar system dramatically simplify it, leaving out all small objects and proceeding as though every moon and planet is a perfect sphere. Yet, the model works because we are in possession of a set of physical laws (concerning such things as gravity, mass and motion) which are basically perfect in their predictive accuracy, and because those very laws entail that the idealization won’t matter. Pragmatic-genealogical world G is not like this. There is no background set of sociological laws against which the idealization makes sense, and according to which (C1) can be obviously true, given the extraordinary degree of simplification. As a matter of fact, it is very difficult to know how to evaluate the probabilistic conclusion (C1).

That said, in what follows I’ll charitably assume that these details can be filled in and that (C1) can be rendered plausible. Now, according to the method, we can say that in this model society, the function of CP is to fulfill the relevant target need. And so we proceed to the first substantive normative conclusion.

3. Idealized Practical Normativity

• (C2)

  Therefore, it is rational for G to engage in CP in order for P to be served in G (in the sense that people with these needs under these circumstances would welcome and, if they could do so, aim for engagement in CP with a view to securing P). (Queloz 2021: 230)

Here, our first major problem is looming. For G is a group, not an agent. Neither Williams nor Queloz subscribes to the idea of group agency or to any related conception of group rationality.² Queloz clearly means ‘rational for each individual in G’. But in the actual world, different agents will have very different relations to the genealogical story, and we will have to ask difficult questions about their practical reasons. This argument must come later, but as we will see, the fact that only some agents might have their target needs on-balance satisfied by CP will make it very difficult to say that this genealogy has any vindicatory force for them at all. And since the only way to investigate this possibility is to simply perform nonideal empirical genealogy, it is entirely unclear why we should not just dispense with idealization at the outset and embrace the nonideal method, which can correctly identify who has the relevant reasons (and who does not).

In addition, it is worth emphasizing that everything thus far has been entirely fictional, by design. That is, in this first half of a pragmatic genealogy, we intentionally avoid talking directly about any group of beings that has ever actually existed, constructing a model of human practice that idealizes away from any actual situation. This, from Craig through Queloz, has always been a core feature of the method. We are not talking about descent here: the members of G are not related to us in any causal or genetic sense. Model-construction is ubiquitous in the sciences, and the pragmatic genealogist is of course free to make use of it. But there remains the important point: while (for example) evolutionary biologists often construct models in order to explain such things as the emergence of altruism, they nonetheless mean to be modeling our evolutionary ancestors. A pragmatic genealogist has no such aim.

Yet, there is a relation between us and the inhabitants of the state of nature, namely, the relation of similarity. A pragmatic genealogy is supposed to speak to us because the world it describes is relevantly similar to our own. This is the next major claim in the argumentative chain.

4. The Similarity Claim

• (P5)

  In our actual group G*, there are close analogues to RN and RC.

We are now meant to observe that in our own society, there are similar problems faced by similar creatures in structurally similar circumstances.

Now P5, in my view, has been and remains the most fundamentally under-theorized idea in all of functionalist genealogy, whether it starts with an idealization or with our hunter-gatherer ancestors (Kitcher 2011). Notice that not only do we need an answer to the question of who falls under the variable G, this is just the beginning of the problem, for the phrase “close analogues” hides an unbelievable amount of ambiguity.

This unclarity triggers the continuity objection to pragmatic genealogy, which this humble author outlined (Smyth 2016), but which takes its inspiration directly from Nietzsche.³ In order for us to have reasons to support and engage in the practice CP, there cannot be practically relevant disanalogies between G and G*, failures of continuity which in fact render the practice less obviously helpful or useful than it is in the ideal model. Very often, pragmatic genealogists simply assert that there are relevant similarities between G and G*, and indeed there often are, but we must also take enormous care to think through the ways in which dissimilarities might render the genealogy practically pointless.⁴

Consider an analogous genealogy of hunting animals, one which inquires into the point and value of hunting in modern Western societies. People in G, a postulated state of nature, obviously need food. So, their circumstances are easy to specify: they need a reliable food source, or they will starve. Given that they clearly have the capacity to do so, we can expect P1–P4 to be satisfied. Moreover, the root needs seem pretty similar in our society, and we (i.e., modern Western societies) have basically the same practice. But are the circumstances relevantly similar? Well, they had better not be, because otherwise we will have to conclude that the function of hunting in modern societies is to supply reliable food sources. And the answer here might seem obvious: there is no close analogue of RC, because we live in societies characterized by massive resource-abundance and (for many of us) significant safety nets. Hunting is not a necessity for almost any of us.

But notice: at a certain level of description, our needs and circumstances are identical to those in the state of nature. We are humans, we need food, if we don’t get it very regularly, we will starve. Hunting can supply that source of food. More similarities could be listed: we live in groups, we communicate with each other, we can construct tools, most of us live within half a day’s walk of animal populations . . . So why isn’t our situation closely analogous?

To his credit, Queloz has acknowledged this basic problem and tried to solve it (Queloz 2021). He rightly writes that his reconstruction reveals “that the soft underbelly of such genealogies is (P5), which assumes that the root needs and root circumstances in fact obtain in our present situation.” But he believes that he has a strategy for solving this problem:

the variables [RN and RC] are assigned to facts about human beings and their environment that stand a good chance of obtaining anyway, independently of the particulars of a given situation, because they are basic structural facts about the human situation picked out under highly general and abstract descriptions. A pragmatic genealogy thus aims to affect the space of reasons through an inference from a generic predicament to our local manifestation of it. (Queloz 2021: 231)

But this is not an answer to the original objection, because that objection explicitly acknowledged that even two radically different scenarios can be made “similar” by giving a suitably general description of them. If functionally relevant differences remain, then the fact that we can ascend towards superficial “similarity” by using more general descriptions doesn’t help.

I’m afraid that Queloz does not really see the challenge here, which is to give an informative account of “relevant similarity”, one that tells us when P5 is and is not satisfied. No pragmatic genealogist has, to my knowledge, even made a start on this problem.

Now, I think we actually can give such an account: we must require that the genealogist preserve in G* the level of description that is necessarily operative in G, the level of description at which the practice CP would actually be “likely” to develop. If further generality of description would undermine the functional claim about the hypothetical idealized world, then we cannot go on to describe our own world in those general terms without equivocation.

Thus, a proper specification of the circumstances in a fictional pre-hunting world is not just the very general: they need food and there are animals. That description is “highly general and abstract” but the case under scrutiny shows that the genealogical inference falls apart. Rather, the proper contextual specification is more like this: they need food, and there is resource-scarcity, and if an individual doesn’t secure food for themselves, no other social structure exists to provide it for them, and there are edible huntable animals. These are the conditions that make it likely that hunting will develop in G. Absent these circumstances, we would not expect hunting to develop in G, but they are exactly the circumstances that fail to obtain in our world, G*. This is why our world is not closely analogous to G here: because the conditions which make CP likely to develop in G do not obtain in our world. So (phew!) we do not have to conclude that the function of hunting in modern industrial societies is to secure reliable food sources for the population in general.

So P5 should dispense with loose talk of “analogues” and say this:

• (P5)

  In our actual group G*, RN and RC also obtain, at the level of description that is necessary to make our causal-functional claims about G true.

This establishes genealogical continuity, which really is just a kind of similarity that allows the desired functional inferences to go through. At this point, I have no objection to the method. However, it is worth noting two things. First, it has become much more empirically demanding. We cannot follow Queloz in merely generalizing our descriptions to make sure that G and G* match. We have to check whether suitably specified descriptions match, and this involves doing a great deal of empirical inquiry into our own society, into the ways it matches (or fails to match) the precise descriptions we needed to give of the ideal world.

But once we are doing this, the collapse objection rears its head. For at this point, it is just not at all clear why we should start with the idealized model at all, or why we shouldn’t just do nonideal empirical genealogy. The argument here is simple: if G and G* are made to match in all functionally relevant respects, then any and all functional conclusions derived from G can also be derived from G*.⁵ As I will stress at the end of this paper, the nonidealized form of genealogy consists in exactly this sort of inquiry, since it gathers present-day and recent-historical data in order to form functional conclusions of its own, and has no need to confirm or disconfirm speculative hypotheses derived from idealized models.

The continuity objection is particularly pressing for one very important reason. The fact is that the modern world is in so many ways radically discontinuous from any heavily simplified social world, be it the world of our actual evolutionary ancestors or of some idealized proto-social persons. For any given functional or pragmatic story we wish to tell, there is a very good chance that the relevant underlying conditions don’t match, such that P5 is not satisfied. As I read her, this is a primary motivation for Miranda Fricker’s insisting that we build power relations into the state of nature (Fricker 2007: 177). Any genealogy of this sort which needs the state of nature to be egalitarian will fail to satisfy P5, and it will license no conclusions whatsoever about our own society. Just as an ancient society’s emergent hunting practices can teach us nothing about the function of our own.

One crucial variable is population size. Pragmatic genealogies nearly always feature smallish groups of people interacting in a face-to-face manner. Yet, such a description spectacularly fails to describe a modern social context, and the invention of so-called “mass society” has brought about a vast array of social forces which simply do not exist in small groups. According to Queloz, Hume is a pragmatic genealogist, so it is worth noting that Hume shows a pretty clear awareness of this problem while constructing his genealogy of property rights.

Hume argues that a fictional group of humans without such rights would be beset with difficulties. So they have a structural need for a system that regulates property. Their solution is what Hume calls “justice”, the propensity of individuals to “bestow stability on the possession of those external goods” and to put “these goods, as far as possible, on the same footing with the fix’d and constant advantages of the mind and body” (Hume 1739: 3.2.2.9). The fictional persons invent a concept and a practice to satisfy a root need in certain circumstances. But they are only able to accomplish all of this because they live in small societies. This is what allows each person to express the “general sense of the common interest” to each other person, it is what allows each person to believe that the “sense of interest has become common to all our fellows, and gives us a confidence of the future regularity of their conduct” (Hume 1739: 3.2.2.10).

But Hume explicitly acknowledges that in larger societies, these explanatory elements fade from view, and so the explanation has to introduce a new element: a government. He concedes (to Mandeville) that this gap must be at least partly filled by “the artifice of politicians, who, in order to govern men more easily, and preserve peace in human society, have endeavored to produce an esteem for justice, and an abhorrence of injustice” (Hume 1739: 3.2.2.12).

So Hume does not stop with the all-too-easy and highly general “we live in a society that needs property rights, and so do they”. He secures explanatory power and continuity by introducing into later genealogical stages a governmental structure that can create and enforce moral norms. This is just a way of obeying my revamped P5: Hume makes his model more like our society in order to ensure that his desired functional and normative conclusions can apply to us.

But the problem is that those conclusions now must change, something Hume does not acknowledge. Once we see that a top-down power structure is necessary to complete the explanation and secure continuity, it is no longer clear that we have a rosy, egalitarian picture of human beings mutually satisfying their needs. We have something else: a group of powerful agents inculcating a respect for property in a large, diverse population for a variety of reasons. Might this not lead to some negative or harmful states of affairs? Can anyone with what Nietzsche called historical sense fail to understand that, in fact, it did?

In order to see where this goes, let’s return to Queloz in order to uncover the first normative conclusion for us.

5. Practical Normativity for Us

• (C3)

  Therefore, it is also rational for us (G*) to engage in CP*, the closest analogue to CP in G*, in order for P to be served.

Here we meet again the question of rationality, and our earlier problem has come home to roost, now that we have rightly introduced the idea of power-relations. Societies, abstractly idealized, need something like Humean property rights in order to satisfy a basic root need for peace and stability. And we can grant, for the sake of argument, that P5 is secured, and that in general our group needs property rights in the same way and for the same reasons. But societies don’t have reasons, people do; or, if there is such a thing as group rationality, there are still very pressing questions about each concrete, embodied individual’s relation to that abstract rationality.

Yet, it is clearly possible that a practice which satisfies the needs of all in an abstract model fails to do so in the real world. So in an actual society where, say, a minority class is brutally enslaved by that government Hume was forced to talk about, do the enslaved persons have reason to support property rights in any meaningful way? The answer is probably a negative one, no matter which model of practical rationality one adopts. This is because, in any pragmatic genealogical model, the relevant reasons are grounded in the satisfaction of root needs.

The slave thus has no subjective reason to support the institution of property rights, because that very institution is a precondition for the frustration of their root needs. We could describe this in an internalist mood, as Queloz does, noting that the slave will not welcome the persistence of property rights under full information. Or, we can follow an alternative, externalist model of practical rationality, and simply say that the frustration of root needs is itself rationally decisive, whether or not the slave is motivated by that frustration (Railton 1986). Either way, the enslaved person will have more of their root needs satisfied in Hume’s chaotic pre-property state than in their actual state, and so they have no reason to promote or support the institution of property. The genealogy of property has answered the continuity objection only by encouraging us to notice that its own normative conclusions may be called into serious question.

I should stress that the objection here is quite general: it is not merely that the genealogy will only fail to speak to particularly unfortunate or oppressed persons, though that would be bad enough, since the genealogy could easily constitute an ideological justification for their continued domination. It is also that each and every one of us does not know if we are in a situation that renders the genealogy practically pointless for us unless we look at our actual world, in the spirit of Rousseau or Nietzsche, to see if newer, darker, less obvious functions have grown up around CP (Rousseau 1755). And again, once we are already doing that, the collapse objection returns with a vengeance.

It might be said that Queloz’s genealogy just generates a pro tanto reason for the enslaved to value property rights, one that is easily defeated. It is not clear that such persons have any reasons whatsoever to promote or value property rights, because they are much worse-off under the convention than they would be if it were abolished. If we try to say that they are enjoying at least some of the shared stability enabled by property rights, this will ring pretty hollow for someone eating scraps and sleeping on the ground in chains. To reply that such a person would welcome property rights if they were not enslaved misses the point; remember, we must supply them with practical reasons given their actual situation, and as such we cannot simply abstract away from that situation. This genealogical exercise is not supposed to just be Rawlsian reasoning conducted from behind a veil of ignorance; if the reader will forgive some gatekeeping, any putative genealogy which collapses into that form of argument is not a genealogy.

However, I am willing to grant that a beefed-up pragmatic genealogy of property rights may provide a certain kind of very weak pro tanto practical reason even to someone who has become property. But we should note that the concept of a pro tanto reason in this sense may be deeply uninteresting. While philosophers often describe such reasons as those which ‘weigh’ in favor of an action (Broome 2013), it is unclear whether this entails that all such reasons ought to be weighed by an agent, on pain of irrationality. I can allow that a pragmatic genealogy of property might supply a consideration which ‘weighs’ in favor of their promoting property rights, in a very abstract and weak sense, but not in the sense that they would be irrational if they ignored it. It is not my impression that Craig, Williams or Queloz mean to supply us with inert practical reasons of this sort. Queloz’s vindicatory genealogies are meant to, in his own words, show that a given practice is “an indispensable instrument to the satisfaction of an important concern” (Queloz 2022b).

We have now seen that the continuity and collapse objections make serious trouble for the pragmatic genealogical method. But it is at the final stage, with all of the desired conclusions laid out before us, that the equally serious reductio objection will finally appear.

6. The Practice’s Real-World Function

Having deployed the model and (hopefully) secured continuity with our actual world, the pragmatic genealogist concludes:

• (C4)

  Therefore, the best explanation for why we go in for CP* is that it serves point P.

• (C5)

  Therefore, there is a prima facie reason for G* to continue to engage in CP*, and CP* is to that extent vindicated.

In order to properly critique these final ideas, I’ll have to switch gears and do some storytelling of my own.

Here is a story that explains where racial concepts came from and what value they have for us.

Imagine a fictional society made up of some two hundred elves. You might wonder why I am saying that they are elves—in fact, as we will see, I am doing this for a reason. But for now, let us just assume that there are some elves living in a small society together. The elves are basically like humans in all relevant respects, and they live in a resource-scarce area without technology or writing. Moreover, the elves come in five colors. Some of them have blue-tinged skin, others are reddish, some are green, some are brown, and some are sort of a greyish mauve.

However, the elves face a practical problem. Supplies continually run low, and members of the community often starve or suffer because of this. Fundamentally, this is because they lack a division of labor (Hume 1739: 3.2.2.3). Their practice is to simply rotate all adult members through all the major tasks: fishing, childrearing, hunting, healing, and shelter-construction. This means that none of the elves ever develops any expertise in any of these areas, and moreover, coordinated action between persons assigned to the same task is very difficult, since your ‘teammate’ today won’t be your teammate tomorrow, and the habits, dispositions and preferences of each new teammate can vary quite widely.

One evening, a large group of them are huddling around a badly constructed fire (tonight’s wood-splitter has never really learned how to split wood properly, but in all fairness he hasn’t done it in months). And one of them says: “Hey all, wouldn’t it be better if each of us specialized in one form of labor?” The group is initially puzzled, but she has their attention. The innovator leaps to her feet, and cries:

. . . if one will wholly apply himself to the making of Bows and Arrows, whilst another provides Food, a third builds Huts, a fourth makes Garments, and a fifth Utensils, they not only become useful to one another, but the Callings and Employments themselves will in the same Number of Years receive much greater Improvements, than if all had been promiscuously followed by every one of the Five! (Mandeville 1714/1997: 182)

But her enthusiasm is soon quenched, for a questioner shouts: “How are we to decide who does what?” The group’s excitement fades; each elf knows that there are many ways of accomplishing this which will prove difficult. Fortunately, another innovator cries out: “The Greens shall do the fishing and hunting, the Reds shall be childrearers, the Browns shall be healers, the Greys can manage tools and the Blues can maintain our shelters!” The elves quickly see the sense in this; as new generations grow into adulthood, they will simply be sorted into training by their skin tone. This decisively solves their problem, and as a result the elves grow much more prosperous, healthy and happy. Moreover, such identity labels such as “red” and “blue” start to acquire a great deal of social significance, since they become bound up with norms for behaviour and performance, and also with particular perspectives and lived experiences.

And that’s the end of the story, How The Elves Came Up With The Concept of Race. What relevance does it have for us? Well, note that each of us can easily look around us and see that there are structural similarities between our world and Elfworld. Real human beings do come in various skin tones, and real human beings do need to generate and sustain collective resources. Moreover, modern societies desperately need a division of labor in order to generate and sustain these resources. Every day, we rely on countless thousands of others to do something that we could not do ourselves. So there are corresponding specialized forms of labor: managerial, and worker-level, technological and manual, et cetera.

And, of course, it should not escape our notice that these roles are correlated with skin color: in many Western countries, whites are more likely to occupy managerial roles, whites and Asians are more likely to occupy leading roles in technological enterprise, and so on. These divisions aren’t as clean as they are in the Elfworld, but the correlations are nonetheless striking.

And so, by merely substituting other values in for his variables, I conclude with Queloz’s C4 and C5: the point of racial concepts is to solve the human division of labor problem by assigning people different roles in this way. This is the best explanation for why there is race at all. And we should all be very thankful for racial concepts, because they solve such an absolutely critical problem for us.

Now, this is quite obviously a silly story, and those are silly conclusions. In fact, given what we know about racial concepts, it is downright offensive. While no-one in the contemporary literature would ever accept it, the story is, in fact, structurally similar to stories that were long told about the justification for racial hierarchies. No person should ever take this story to provide anyone with any practical reasons whatsoever, let alone reasons to bolster or support our actual racial concepts. But the question is: can the method of pragmatic genealogy, as outlined by its ablest defenders, deliver this conclusion? After all, the Elfworld story just involves filling in the variables in the model we’ve just refined, and so counts as a reductio ad absurdum of the method.

Elfworld satisfies P1–P4, because there is a prototypical human-like society with a problem, and readily available materials (i.e., visible skin color) help them to solve that problem. We can therefore expect a color-based division of labor to develop. This gives us C1, and since the elves are so much better off under this practice, it seems rational for them to continue to engage in it. That’s C2.

Then, we have the key P5:

• (P5)

  In our actual group G*, RN and RC also obtain, at the level of description that is necessary to make our causal-functional claims about G true.

This condition at least seems to be satisfied. After all, in Elfworld, there is a powerful need for a division of labor, and we share that need. Group members come in varying skin tones. There is a resultant practice of sorting, a clear potential for using variable skin color as a sorting mechanism, and even (this is a bonus) evidence that this sorting mechanism is actually operative, given the predominance of certain skin tones in certain sectors of our economy. Obviously, our circumstances differ, but do they differ in any functionally relevant respects?

We could, at this point, try to explore some disanalogies, but this would not be wise. In fact, I think it would be obtuse if we explored the possibility that the problem with Elfworld might be something to do with whether it is properly analogous to our own world. That isn’t the problem here. The problem is that we already know how racial concepts came about, and we know how labor is divided in our society. In fact, we know all of this because of the scholarly work done by—nonideal empirical genealogists of morals.

For example, in giving an actual genealogy of race, Charles Mills argued, on the basis of a great deal of primary and secondary source material, that racial concepts were a kind of post-hoc rationalization of economic and social exploitation conducted by Europeans (Mills 1998). In other words, there was a set of powerful economic incentives driving such things as colonial domination and the slave trade, and a biologically loaded concept of race provided the much-needed rationalization for these practices, particularly in the context of a religious culture which seemed to prohibit domination and slavery. The incentives themselves were not conscious, explicit motivators; no agent actually reasoned: “In order to feel better about enslaving these people, I’ll come to believe that they belong to an inferior biological type.” Yet, Mills shows, in the spirit of critical theory, how such an ideological belief could become prominent because it rationalized economic advantage. This sort of story has become incredibly influential over the years, informing a huge amount of work in the social sciences.⁶

In addition, historically-minded philosopher Kwame Anthony Appiah argues that during recent history racial concepts have become a locus of personal identification, such that many persons organize their lives and values around them in spite of their highly questionable history (Appiah 1998; 2020). So now we contend with this dual legacy; we are wondering what to do with race given that it seems to be functionally bound up in power and exploitation but also connected to cultural growth and personal meaning (Jeffers 2013). We understand all of this precisely because we are not thinking about the division of labor, or about Elfworld, or any idealized model at all. We are thinking correctly about our practical situation because we have discovered some empirical facts about the arrival of race on the human scene and about what it has been doing for people in recent cultural history. We know these facts because genealogists of morals inquired into them, and because they did not allow themselves to become distracted by idealized reconstructions. Moreover, divisions of labor also have recorded histories; we can see why they were created and how they are presently maintained (Levy & Murnane 2012; Durkheim 2014). These are nonideal-empirical genealogies of morals.

The elves have nothing at all to teach us about either race or the division of labor. This is not because they aren’t adequately described. It is because they aren’t real, and because we have a huge amount of evidence concerning the actual history and functioning of racial concepts. So models of this sort are not relevant to explanation or normative assessment.

This, by the way, is why I used elves and not the abstractly described “people”. Remember, we are just discussing a group G, and there is in fact no reason that its members should be human at all; they need only be basically similar to us. My suspicion is that by subtly inviting the reader to think of the denizens of these fictional worlds as our distant ancestors—or even as models of our ancestors—pragmatic-genealogical inferences can be made to seem more sound than they are. After all, if a practice really is a historical ancestor of our own, then there is at least a reason to think that their functional profiles will be similar. As Nietzsche pointed out, it is not even a very strong reason, but at least it is a reason (Nietzsche 1885/2011: II:12). But this reason should never bolster a pragmatic-genealogical inference: we must always bear in mind that we are being asked to draw substantive, practically important conclusions on the basis of a fiction plus some analogies to our present situation. But, to return to the collapse objection, how could such an inference ever be stronger than one which includes actual descent?

Let’s look again at Queloz’s key conclusion, with two key words highlighted:

• (C4)

  Therefore, the best explanation for why we go in for CP* is that it serves point P.

Queloz must say that this is the best explanation, because actual members of our own group have no reason to promote CP on these grounds unless they have good reason to think that it actually serves point P. Note that this is an explicitly comparative claim: this mode of explanation is supposed to be the best available.

It’s worth pausing for a minute and reflecting on what we are being asked to believe here: that a highly idealized model, in combination with some purported similarities between the model and our own world, will constitute the best explanation for real, lived social phenomena, even by comparison to detailed, recorded histories of those same phenomena. That is, by all the normal standards of explanation in the social sciences, this model is going to win. I cannot see how this could be so, and if it is so, we are surely owed some kind of argument for this. There is no adjustment to Elfworld that is going to make it a better explanation of racial concepts than Mills’ explanation; the only adjustments that will be even remotely helpful will simply make that world look suspiciously like the world in 1491; in other words, the model must actually start to resemble real history. Collapse threatens all over again.

Thus, it seems to me that C4 is never going to be the object of justified belief, not in the case of any social practice which has a history we can actually study. Pragmatic genealogy could only help us understand our reasons with respect to social concepts and practices with histories that we know fairly little about. I’m not sure which concepts and practices those are, but they may be very rare indeed. We have burgeoning, detailed histories of property rights, of divisions of labor, of gender, economic class and government; what is left for this allegedly helpful research program to study?

7. Objections and Replies

Now, a reader might object here, noting that it has been common in the pragmatic genealogical literature to distinguish, with Williams, between the essential core of a concept and its cultural elaboration (Williams 2002). For Williams, the concept of truthfulness is a universal with a stable function across all cultures, but in some societies (such as ours) it has taken on a particular form, for example our obsession with personal authenticity. Craig believed that he could identify the essential, culturally invariant core of the concept of a knower by constructing a fictional state of nature, and he arrived at the idea that it has the social function of labeling reliable informants (Craig 1990). But this, he thought, was consistent with its being realized in many different forms in many different places. Isn’t this enough empirical history? Shouldn’t we be satisfied with this two-stage process, where the main functional inference is bolstered with observations from specific cultural situations?

The first thing to say here is that this elaborative stage is not in the model as Queloz presents it. It is therefore very difficult to evaluate its role in the overall structure, because we do not know which conclusions it is meant to bolster and how it is meant to bolster them.

So let’s distinguish between two forms this second stage might take, between an elaboration and an interrogation. An empirical elaboration of a pragmatic genealogy takes the functional conclusion in (C4) for granted and merely looks for different forms that the relevant practice might take. This might be an interesting exercise, but it is not an answer to my challenge, since I am specifically looking for reasons to believe (C4), and no exercise which assumes its truth can provide such reasons.

The second and more directly relevant kind of genealogy interrogates the functional conclusion (C4), looking for specific historical reasons to confirm or disconfirm it. For example, it might be discovered that once massive corporate entities emerge in modern societies, the function of property rights changes, becoming more about the protection of socially destabilizing corporate interests than about the reduction of strife and conflict. Or, taking Elfworld as a starting point, the pragmatic genealogist might look around for ways in which racial concepts have been (or are) responsive to very different pressures, such that they cannot be said to play the role that the original model says they play. The pragmatic genealogist might take this route, looking for just these important discontinuities which might disrupt the simpler functional story.

But again, if you are already doing this, then aren’t you precisely already inquiring, in an empirical-genealogical way, into the present function of the practice? So why not simply start with that method and abandon pragmatic genealogy altogether? The case of Elfworld makes this problem even more stark, since any attention to that genealogy will positively distort our subsequent interrogative inquiry, giving us a prior confidence in a hypothesis that deserves absolutely no confidence at all. Why risk this sort of distorting effect at all?

After all, here’s another model of genealogical inquiry, which I have been calling nonideal empirical genealogy:

• (P1)

  In our actual group, G, there is some practice CP into which I want to inquire.

• (P2)

  By looking at its contemporary shape, its recent history, and potentially its ancient ancestor-practices, I find good evidence that CP exists largely because it produces effect E, and I am able to specify, in some general terms, the feedback mechanism that reproduces CP on the basis of its having effect E.⁷

• (C1)

  Therefore, I have good reason to believe that one major social function of CP is to produce E.

• (C2)

  Therefore, to the extent that I would value E under conditions of full information, I have reason to promote or foster CP. And, to the extent that I would disvalue E under conditions of full information, I have reason to frustrate or oppose CP.

It is very easy to read Mills as following this model, where CP is our stock of racial concepts and E, basically, is European colonial domination.⁸ Yet, it is notable that Mills is not (to my knowledge) included in any contemporary overview or list of prominent genealogists in English, despite the fact that his method is plainly genealogical, avowedly a descendent of Rousseau’s own (Mills 1998: 5–7).⁹ Mills gives us a two-stage genealogy of race, beginning with a race-as-colonial-construction model and illustrating how its supporting structures persist into the second stage, where de jure racial oppression is officially ended (Mills 1997: 39–61). There is much to contest in his story, but The Racial Contract is a naturalistic, nonideal history of how prominent social concepts arrived on the scene and evolved, grounded in textual sources and in sophisticated social analysis. By comparison, Elfworld, like any pragmatic genealogy of race, is at best a waste of time, at worst downright distorting.

Next, it might be objected that I am overgeneralizing, that my three objections won’t have force against all manifestations of pragmatic genealogy. Indeed, Queloz has argued that the method is particularly helpful only with respect to certain kinds of concepts and practices. Might it be that a peace treaty is in order here, with pragmatic-genealogy taking care of some cases, and nonideal empirical genealogy handling others? Unfortunately, a close examination of these types of cases reveals that pragmatic genealogy continues to underperform relative to its nonideal cousin.

Some practices are, Queloz notes, self-effacingly functional, and are, in his view, particularly amenable to a pragmatic-genealogical reconstruction (Queloz 2018; 2021). These are practices which are mainly reproduced because they serve a certain function, but this function is only discharged if the participants are motivated by considerations that are very distinct from the functional effect itself. Many practices are like this: they are sustained by an intrinsic motivation to perform some action-type, but the actual explanation for this widely-shared intrinsic motivation is that it has some kind of extrinsic value.

Queloz argues that pragmatic-genealogies are particularly well-placed to describe such self-effacingly functional practices because they can isolate both the best explanation for the practice’s persistence (i.e., its actual function) and can show, through cultural elaboration, how that explanation fades from view and is replaced with a more intrinsic motivation.

But unfortunately, this comparative methodological advantage is only possessed with respect to his actual target, which is Fricker’s alternative paradigm-based model of functional explanation (Fricker 2016). There is simply no reason to think that a nonideal empirical genealogy will fail to capture self-effacing functionality. This is something we’ve already seen, since Mills’ genealogy describes a process where social function and motivating psychology are quite different. Many participants in the slave trade genuinely believed that Africans were of an inferior type, and they didn’t justify such beliefs in terms of the functional effect produced by everyone else sharing them (economic advantage). The social practice required genuine belief, and not merely a wink-and-a-nod, “let’s believe this so we all get rich.” Clearly, a nonideal empirical genealogy can fully capture cases where motive and function come apart, and no detour into an idealized version of social life is necessary.

And, as you might have guessed, exactly the same is true of the second kind of case where pragmatic genealogy is said to be particularly helpful, the case where “a multiplicity of functions have been layered on to a practice,” such that it fails to even display a single core function or paradigm instance (Queloz 2021: 61). Here, Queloz’s model may fare better than one which assumes the existence of a paradigm instance, as Fricker’s does, but it actually fares worse than a nonideal empirical genealogy.

As an example, Queloz cites Nietzsche’s description of punishment as a disunified and functionally layered practice, but the reason Nietzsche is even able to say that the practice is disunified is that, as Nietzsche explicitly says, he has looked into how the practice works at various times and in various places.¹⁰ Nonideal empirical genealogy is thus arguably a precondition for even discovering that a practice is functionally layered or disunified in the first place, whereas a pragmatic genealogy can only encourage us to “excavate” an entirely illusory unitary function by the construction of an idealized model which delivers Queloz’s “the best explanation for why we go in for CP* is that it serves point P.” Note that we could easily treat the practice of punishment precisely as the Elfword genealogy treated the practice of racializing. We could imagine a proto-society wracked with chaos which is stabilized by the invention of punishment, and conclude that social stability is the core function of punishment. But Nietzsche is right: punishment does not have a sole, definitive function, and so an idealized model can only encourage us to miss this fact.

To conclude, we lack a satisfying justification for the continued preference for idealized over nonideal empirical genealogy. Philosophers working on this topic regularly cite and discuss the more idealized, fictional genealogies of Hume, Williams, and more recently Philip Pettit (Pettit 2018), continuing to mainly ignore nonideal genealogists like Mills. Why, given these multiple methodological shortcomings, do so many philosophers continue to prefer idealized or even partly idealized models?

8. Purity and Practical Reason

There is one final incentive here that I must discuss, one which Queloz actually endorses and to which Williams was strongly attracted (Williams 2002: 39). A nice, ‘clean’ model that can deliver substantive normative conclusions on the basis of things that philosophers are supposedly particularly qualified to think about is understandably attractive. Such a model is the rationalist moral philosopher’s holy grail: a method which, like that of an early Rawls or Kant of the Groundwork, can be evaluated from the armchair but which does not require so much empirical input that we might have to ask someone in a different department for help in generating the relevant conclusions. But this drive for purity in practical philosophy does not have a particularly strong track record, as Rawls was famously forced to admit,¹¹ and it would be very surprising to find it operating in, of all places, a body of thought that lists Nietzsche as a godfather.

Well, surprise! Here is Queloz, responding to similar concerns about idealization:

Idealization is of value here already because considering simplified prototypes of our practices in an uncluttered state of nature helps us break through the veil of familiarity to gain a sense of the more generic human purposes they serve. Idealization also cuts out noise, such as infelicitous conditions in which practices are temporarily prevented from serving a point. . . . idealization helps us maintain a good overview by displaying the successive layers of practical significance in a neat and organized manner. . . . Uncluttered by the messiness of reality, the model sharpens our eye for certain patterns of pointfulness. (2021: 197, 225, emphasis added)

Rationalists have long deployed the language of purity and cleanliness in order to rhetorically motivate idealization, and this is just what Queloz does in these passages. Moreover, we should remember: for our enslaved person whose misery is predicated on property conventions, that “infelicitous condition” is not “noise”. It is a manifest reality that blocks them from enjoying the “pointfulness” of this allegedly vindicated practice. At the explanatory level, their condition is powerful evidence that the original genealogy of property is far too optimistic to be the whole story. That evidence disappears when we bracket them and their condition by describing an idealized society in very general terms and merely suggesting certain similarities to our own. It’s hard to avoid the conclusion that Mills was right: the move to idealization in normative theory is not just a theoretical device, it also functions politically, fixing our gaze on certain persons and causing us to forget about others.

In a related passage, Queloz attempts to ground the philosopher’s special right to this neat and organized subject-matter:

Considered as regular psychology, sociology, or historiography, the pragmatic genealogies of Hume, Nietzsche, Craig, Williams, and Fricker are odd creatures indeed. But their contours make perfect sense when considered as answers to philosophical concerns about the naturalistic credentials or the point and value of certain concepts. Here, idealizing and potentially distorting narrative devices like the state of nature serve a point, because the concerns of philosophy are not best served by a ‘Laplacean’ genealogy capturing every little detail in the meandering history of our practices. (2021: 246, emphasis added)

This passage, I suggest, is not helpful. First, genealogies do not establish anything particularly interesting about the ‘naturalistic credentials’ of their targets, since they by and large assume that such targets are amenable to naturalistic explanation and set about trying to construct those explanations. Second, it is not true that the philosopher alone seeks to understand “the point and value of certain concepts”, as the cases of Weber and Durkheim clearly show. Finally, even if this were true, what this paper has shown is that you cannot understand the point and value of just about any social practice in abstraction from a great deal of real social history.

Last, Queloz’s characterization of real history as ‘Laplacean’ is a wild overstatement, as I’m sure he knows. The reference here is to an omniscient demon who acts as a theoretical model of metaphysical determinism, because this demon, knowing literally everything that has happened and all the laws of nature, can predict the course of future events perfectly. But I am quite sure that Charles Mills was not an omniscient demon with perfect predictive powers. Neither for that matter is Colin Koopman, or Nietzsche, or Appiah, or Foucault, or the Hume of the Natural History of Religion (Koopman 2019; Nietzsche 1887/2011; Appiah & Gutmann 1998; Hume 1757). We do not face a forced choice between the airy abstractions of pragmatic genealogy and investigation which aims at knowing literally every truth about history.

Practical reasons are messy because practical reasons are grounded in messy social-psychological reality (or, as people less attached to purity norms sometimes call it: social-psychological reality). Genealogy, qua reason-giving practice that is also explanatory, necessarily gets its hands dirty. This is why it can actually help us to understand the reasons we actually have.