Sex Discrimination, Normativity, and Begging the Causal Question

I. A CAUSAL ACCOUNT OF SEX DISCRIMINATION

Philosophical analyses of the concept of (non-moralized) discrimination typically feature three successive elements. First, discrimination is an essentially comparative notion. At issue in a discrimination claim is the fact that one person is treated one way, whereas “another person”—I use quotations to denote that this person may not be actual—is, or would be in some appropriate hypothetical circumstance, treated another way. Hence, differential treatment, actual or hypothetical, is a necessary feature of discrimination.⁹ Second, to qualify as discrimination, this differential treatment must be keyed to some feature, which is the basis or grounds of discrimination. Differential treatment with no basis whatsoever, i.e., entirely arbitrary differential treatment, cannot be discrimination. The third element describes what this keying amounts to, that is, what the link must be between the grounds of differential treatment and the treatment itself for it to constitute discrimination on those grounds.¹⁰

Philosophers of discrimination are in broad agreement about how to characterize the first two elements, but the third is a point of active scholarly debate. One way of filling it in is to claim that discrimination on the basis of X requires that X be the agent’s reason for her differential treatment. Another way claims that it suffices that X merely motivates her differential treatment for her action to be discriminatory on the basis of X. Yet another claims that the differential treatment is keyed to X in the right way if X causes (in the right way) the different treatment.¹¹ I’ll call this last way of filling in the above analysis a causal account of discrimination. By and large, advocates of the causal account adopt a difference-making analysis of causation: that causes make a difference to whether their effect obtains; so no cause, no effect. On this view, what it is for X to cause the differential treatment at issue—and thus for the treatment to be discriminatory on the basis of X—is for it to be the case that, were it that –X, that treatment would not obtain.

The causal difference-making idea is thought to easily explain why cases of exclusion due to overt interpersonal racism and sexism are discriminatory. But its advantage over alternative analyses really lies in its ability to make sense of those cases of discrimination that manifest in more subtle, non-one-to-one, non-explicit expressions of animus. In particular, hostile workplace dynamics, longstanding social practices with differential group impacts, and implicit cognitive biases are now broadly taken to be potential mechanisms of discrimination. Yet many theories struggle to explain why. The causal account dispenses with this difficulty. Conscious or unconscious, interpersonal or structural, what unifies these cases is the presence of a causal connection between an individual’s belonging to (or perceived belonging to) the relevant social category and the disadvantage they suffer at the hands of the workplace culture, the social practice, their employer’s biases, and so on. The following cases are illustrative. Granting for the sake of argument that Sexual Harassment, Appearance Guidelines, and Implicit Bias are bona fide cases of sex discrimination, the causal account successfully explains why.

Sexual Harassment

Gloria’s place of work has a culture of sexual harassment against women. She refuses her harasser-coworkers’ and -managers’ sexual advances. Were Gloria not perceived to be sexed female, she would not be so targeted, and would not be fired for refusing her boss’s advances.

Appearance Guidelines

It is standard practice at Sasha’s place of work for women to wear a dress and heels. She does not do so. Her employer cites her violation of the company’s appearance guidelines as the reason for her being let go. Were Sasha not a woman, she would not be in violation of the appearance guidelines and would not have been let go.

Implicit Bias

Ameera’s responses in her performance review were perceived by her manager as brusque and bossy. The manager was not consciously aware of the role that Ameera’s being a woman played in his evaluation of her performance. He would not have evaluated her so poorly and denied her the promotion were it not for her being a woman.

Recent discussion of the causal account of discrimination, however, has revolved not around these cases but another, trickier set of cases, which proponents of the causal account also claim to be able to explain. I will here group them together as Sex+ Cases.

Sex+ Cases

Alex is a transgender girl who is prohibited from using the girls’ bathroom at school. Were Alex not assigned “male” at birth, she would have been free to use the girls’ bathroom.

Bill, a man, is fired from his work upon his boss’s finding out that his partner Andrew is a man. Were Bill a woman and married to Andrew, “she” would not have been fired.

Sex+ Cases will play a central role in the remainder of this article—but not as a “test case” for whether the causal account can or cannot deliver the “right” verdicts about these cases. Rather, I will be setting to the side whether cases of discrimination on the basis of transgender status and sexual orientation are or are not cases of sex discrimination, in order to ask whether a causal account of discrimination can deliver conclusory verdicts on these cases at all—be they affirmative or negative. I will argue that it cannot. Sex+ Cases thus serves as my entry point into diagnosing a set of issues that afflict the causal account of discrimination on the basis of sex more broadly. Working through how the account struggles to deliver causal verdicts and thereby discrimination verdicts in these cases will preview the difficulties it faces in explaining other, far less controversial and supposedly far easier to analyze, cases of sex discrimination. These will be the cases in my argument that really endanger the viability of a causal account of discrimination.

Two final notes before proceeding. First, this article focuses on the normative phenomenon of “sex discrimination” or “discrimination on the basis of sex”; I take these to be terms of art that for various historical and legal reasons use the term “sex” rather than the term “gender,” even though the latter may seem to readers today to be more apt. The matter of what the conceptual scope of “sex” in “sex discrimination” is will be a central question in this paper. I do not mean that question to be settled or even considerably informed by the use of the word “sex” in this terminology.

Second, in the wake of Bostock, many scholars have offered their own critiques of Sex+ Cases. But thus far, these arguments reject Sex+ Cases on the grounds that these particular causal analyses are flawed, incomplete, or draw problematically on undefended normative choices. I do not wish to deny these claims. But I take these responses to issue only a partial dissent to the causal account of discrimination. They do not argue against causal analyses of discrimination in toto, only that the specific analyses offered in Sex+ Cases fail.¹² My aim in the next two sections is to show why all such causal accounts of what constitutes sex discrimination necessarily fail to resolve these cases. This will require attending more closely to the inner-workings of causal analyses than previous discussions have thus far, and focusing, in particular, on the question of what it is for sex to be a cause.

II. BEGGING THE CAUSAL QUESTION

Sex+ Cases brings out the difficulty of carrying out a difference-making analysis that bears witness to the causal relevance of sex to the outcomes in question, rather than other factors that may confound the inquiry at hand. So while proponents of the causal analyses in Sex+ Cases use the cases to show that the treatment in question is indeed caused by sex, their opponents retort that the comparators used to generate these causal verdicts suffer from causal confounding. The charge is this: Changing Alex’s and Bill’s sex status makes changes to other sex-related facts about them: whether they are transgender or cisgender, and whether they are gay or straight. To allow these factors to vary across comparators is to allow their causal influence to be “picked up” as the causal relevance of sex and as a result, to count as discriminatory on the basis of sex. But whether that is so—whether the causal operation of these factors are instances of the causal operation of sex—is precisely what the causal analysis is supposed to test for. The test just is a test of whether treatment caused by gender identity and sexual orientation is treatment caused by sex. Thus, to set up the casual analysis in this way is to rig the game by presupposing the answer that the analysis is itself meant to prove.¹³

What is more, these opponents claim, alternative ways of running the analyses suggest that the adverse outcomes suffered by Alex and Bill are positively not caused by their sex status. According to the analyses in Sex+ Cases*, causation does not vindicate these cases as instances of sex discrimination.

Sex+ Cases*

Alex is a transgender girl who is prohibited from using the girls’ bathroom at school. Were Alex assigned “female” at birth and still transgender and seeking use of “his” preferred bathroom at school, “he” would be similarly prohibited.

Bill, a gay man, is fired from his work upon his boss finding out that his partner Andrew is a man. Were Bill a woman and gay, “she” would have still been fired.

But does Sex+ Cases* really manage to avoid begging the question? Its comparators are constructed by changing sex and holding fixed Alex’s being transgender and Bill’s being gay. But what justifies holding these factors fixed? Holding them fixed is only methodologically licensed, after all, if they might confound sex’s causal effects. Otherwise, doing so will yield comparators that do not fully vary in the target variable of interest, which will in turn generate the wrong causal conclusions. To presume, then, that these factors indeed risk confounding is to presume that their effects are distinct from sex’s and so must be struck out of any observed differences across the comparators. But now we are back at the question precisely at issue: whether these effects—effects of gender identity and sexual orientation—are effects of sex? So, Sex+ Cases* begs the question just as Sex+ Cases does.

In sum, a causal account of sex discrimination determines whether an action constitutes discrimination on the basis of sex by analyzing sex’s causal relevance to the treatment in question. Doing so requires setting up causal contrasts that may tell the causal role that sex plays. But in the cases at hand, candidate causal factors are conceptually related to one another. As a result, there are multiple ways of setting up a causal contrast in which the target variable “sex” is different—with different ways of setting up the contrast bringing in different “other” changes into the analysis (e.g., sexual orientation, trans/cis gender status). The trouble is that settling on any particular contrast as the right one requires settling on that which is precisely in question: whether an effect of these “other” changes is or is not an effect of sex. So, the causal account cannot establish whether an action is discriminatory on the basis of sex, because defining what it is for sex to have a causal effect raises the question of what effects confound rather than constitute sex’s causal influence (in order to set up the right causal contrast). And to answer this is just to answer the central question at issue.

Note the contrast between this argument and the typical objections that scholars discussing Bostock have put forth. Those objections tend to revolve around the mere fact of there being multiple changes brought in tow by the change to sex status in both Sex+ Cases and Sex+ Cases* and the need therefore to find grounds on which to privilege one over the other.¹⁴ This framing sets up the problem as one of tie-breaking between two legitimate ways of determining sex’s status as a cause for the purposes of determining sex discrimination. By contrast, my contention is that each way of setting up the causal contrast begs the central question at issue: whether some factor’s influence constitutes or confounds the effect of sex. Thus, the source of the problem I draw out is internal to the causal analysis, rather than deriving externally from a need to out-compete another candidate analysis. The internal critique suggests that the problem with the causal account’s handling of these cases has deeper roots than prevailing explanations have noted.

In the next section, I will move to show how these worries about circularity indeed apply more widely than just the cases in Sex+ Cases. We will arrive at that broader challenge by way of first considering some possible lines of response to the argument above. This will serve to further elucidate how the argument works and lay the groundwork for seeing how it will extend.

III. EXTENDING THE CHALLENGE

My main claim thus far has been that the causal account cannot explain or establish whether the actions at issue in Sex+ Cases are or are not cases of sex discrimination, because running the causal analysis requires drawing on an account of what defines the causal effects of sex. Is an effect of a factor conceptually tied to sex—such as sexual orientation or trans/cis gender status—an effect of sex? Or distinct from it? Alas, this is precisely the question at hand: whether differential treatment caused by these factors is differential treatment caused by sex.

That argument may have seemed fast. So this section will further elaborate on it, by considering two attempts to sidestep the metaphysical circularity I diagnose above. These look to ground the choice between Sex+ Cases and Sex+ Cases* on some prior independent account that distinguishes sex and its causal effects from sexual orientation and gender identity and their causal effects (or shows that there is no way of distinguishing these). I consider two such strategies. One looks to ground the choice of causal comparator in a metaphysical account of the category sex and its causal behavior. The second looks to do so by considering what constitutes a mental representation of sex rather than a non-sex or merely sex-related representation.

IV. TOWARDS A NORMATIVE-CAUSAL ACCOUNT OF DISCRIMINATION

I have thus far sought to draw out the substantial hurdles that loom ahead of a causal account of discrimination. To have practical value, the analysis must be capable of deciding among competing causal contrasts on principled, non-circular grounds. The preceding arguments cast doubt on prevailing approaches’ prospects of clearing this bar.

One might be tempted at this juncture to take the off-ramp from causation entirely. I am not wholly unsympathetic to this response. But it is worth pausing to appreciate the extent to which dispensing with all causal analysis sends us back to the drawing board on several key parts of a theory of discrimination. For one, without causation, we will need to find another explanation for the trademark “because of” clauses in discrimination talk, which link the action at issue with the relevant statuses. This will also involve a major revision of several of the currently leading theories of discrimination, which rely on causation.²⁴ Meanwhile, a return to motivating reasons, the other contender account for the link condition, is also beset with substantial hurdles. To the extent that a unified account remains a key desideratum of philosophical work on discrimination, the motivating reasons approach is notably lacking and struggles to explain more subtle non-attitudinal or “structural” cases of discrimination and account for what ties these together with the “simpler” interpersonal cases. And while such cases as Sexual Harassment, Implicit Bias, and Appearance Guidelines were once seen as more marginal to the phenomenon of discrimination, they are now widely recognized to be as paradigmatic of discriminatory harms as are cases of deliberate exclusion. Failing to explain cases so established in the discrimination canon can no longer be taken as simply drawing out a theory’s consequences; ruling them out now risks disqualification. The question then is: Can an analysis of discrimination account for these alongside other strands of discrimination without at all appealing to sex, race, etc., causation?

Without ruling out the possibility of a non-causal account of discrimination, and with full recognition that further elaboration and finessing of prevailing causal accounts may yet hold promise, I want to ask in this section: What different approach might there be? I want to propose that we might yet be able to preserve a key role for causation in an analysis of discrimination—without claiming to wholly overcome the charge of circularity. Circularity is, after all, not always defeating; to the contrary, it can be, as coherentists have liked to say, “virtuous, not vicious.” Might there be a way of elaborating a causal account of discrimination that contains a non-vicious circle? My aim in the remainder of this article is to sketch one possibility in this vein—call it a “recipe” if you will—and discuss what promise and pitfalls it might present for our theorizing of discrimination.

The idea is this: In the spirit of reflective equilibrium, we might draw on cases of sex causation exhibited in paradigmatic cases of sex discrimination as “data” to formulate a general theory of sex causation, which may then be used to establish the status of more uncertain cases. On this approach, paradigmatic cases of sex discrimination qua paradigmatic cases of sex causation sit alongside the standard stock of intuition-based cases that populate the causation literature, involving preemption, overdetermination, trumping, etc.²⁵ In fact, it is rather natural to think that a wider set of cases than the usual vignettes needs to be brought to bear on an account of causation featuring social statuses such as sex. But unlike in the standard approach to theorizing causation, the role of these cases is not primarily to serve as test cases that a plausible analysis of causation must be able to vindicate. Instead, they are inputs into an account of what it is for sex to be a cause, that is, an account of the line that separates those effects that constitute sex’s causal effects from those that are distinct from and thus may confound it. As I have argued, how exactly this line is to be drawn is the crux of dispute in controversial cases of sex discrimination. My proposal is that we may draw on our intuitions about paradigmatic cases of sex discrimination qua paradigmatic cases of sex causation—for example, our intuition that cases like Sexual Harassment, Implicit Bias, and Appearance Guidelines in addition to interpersonal animus-driven cases are bona fide cases of sex discrimination and thus cases of sex causation—to answer the causal delineation problem. From here, this account of the line that distinguishes sex’s causal effects from those might compete with it can be applied to deliver verdicts on cases of sex causation where our intuitions are less firm—for example, to decide between Sex+ Cases and Sex+ Cases*. Causal theorizing, on this approach, is shot through with normative theorizing insofar as the data from which an account of sex’s causal effects is developed contains normative inputs; thus call this a normative-causal approach to discrimination.

This approach to analyzing discrimination embraces the methodological view that our ethical judgments of what is discriminatory on the basis of sex and race gives us insight into what sex and race causation are. And if these notions of causation are themselves informed by our normative judgments about discrimination, then an account of what is discriminatory on these bases that makes use of such causal notions will certainly fail to be reductive. Still, it may not be viciously circular, if the causal judgments to which the account appeals in an explanation of whether a given case is discriminatory do not derive from the case in question. On this way of proceeding with a causal analysis of discrimination, what is needed is indeed information about discrimination and causation—but it is information about other cases to determine the status of this one. This contrasts with the vicious metaphysical circularity that besets extant causal accounts, which attempt to ground an answer to whether a given act constitutes sex discrimination or causation in an account of sex causation that presupposes the causal status of the very act in question.

Thus, the kind of circularity in the normative-causal account can be likened to that which lies in the leading interventionist theory of causation itself. James Woodward’s defense that such circularity is non-vicious might here be instructive. Despite the fact that an interventionist defines what it is for X to cause Y in terms of an intervention I, causal routes between X and Y, and other causes of Y—themselves all causal notions—the interventionist theory is “not viciously circular in the sense that the characterization of an intervention on X with respect to Y itself makes reference to the presence or absence of a causal relationship between X and Y”; rather, the interventionist theory characterizes the causal relationship “by appealing to facts about other causal relationships.”²⁶ All that is needed in order to accept such an account is that, as Woodward puts it, “[w]e begin in media res… with a stock of already known causal and correlational information and use this to reach new conclusions about other causal relationships perhaps using these new conclusions to revise other previously accepted causal beliefs.”²⁷ By analogy, all we need in my proposed normative-causal account of sex discrimination is that the account of how sex acts as a cause applied to a given case of disputed sex discrimination is developed from analysis of other cases of sex causation, which are less controversial.

This is, of course, only a sketch of how such an account might work; a fuller analysis will have to wait for future writings. But to give a better sense of why such a project might be worth pursuing, I want to turn to say more about the prospects for a normative-causal approach to discrimination and the hurdles that I take accounts developed in this vein to face.

I will start by noting that although the thought that causal facts might be sensitive to normative ones might seem at first downright heretical, it in fact accords well with views that have wide support in the philosophical literatures both on causation and on race and gender. Many prominent theories of causation to varying degrees make explicit room for normative notions. These include accounts which explicitly note that in some circumstances, causal facts are sensitive to moral and political ones.²⁸ Scholars of race and gender, meanwhile, commonly take an approach to theorizing that freely mixes the metaphysical and the ethical. In particular, substantive normative considerations have been argued to impinge even on our so-called descriptive projects when it comes to elaborating an account of, say, “racism” or “woman.”²⁹ More broadly, to the extent that normative phenomena like sex discrimination are among the key explananda that a good social metaphysics will be able to explain, it is not so far-fetched that our metaphysical theories of sex and of sex causation might themselves be laden with ethical considerations. This explains how a normative-causal account of discrimination may stay in keeping with a standard causal-explanatory interpretation of the hallmark “on the basis of” and “because of” locutions in discrimination talk—while also making sense of why actions “on the basis of” or “because of” sex are often morally wrongful. For, as it turns out, what it is for sex to causally explain some outcome is itself in part a normative matter.

A normative-causal account does, however, mark a substantial departure from going theorizing in discrimination theory itself. In particular, admitting a normatively-inflected analysis of sex’s causal effects results in a picture of the concept that is at odds with a principal tenet of philosophical theories of discrimination. Insofar as sex causation is an element of the concept of sex discrimination, that concept can only be so “non-moralized.” This narrows the gap somewhat between a theory of discrimination and a theory of wrongful discrimination as pertains to certain social statuses—a gap that most scholars of discrimination implicitly or explicitly endorse.³⁰ In particular, an analysis that allows first-order ethical judgments about racial or sex-based wrongs and harms to figure directly into one’s theory of race and sex causation and discrimination invites further debate about the kinds of first-order ethical judgments that are permitted to enter the normative-causal account. This will likely turn on one’s moral theory of discrimination. So, a view that takes the wrongness of discrimination to lie in the wrongness of, say, disrespect (Eidelson) may easily recognize a given case of disrespect as discriminatory and on that basis, take it to be a paradigmatic case of sex causation. Whereas a different moral theory which takes the wrongness of discrimination to inhere in, say, demeaningness (Hellman) might be less certain about the same case and so not include it among such core cases from which to draw out a broader theory of sex causation and discrimination. So, different accounts of what is wrong in wrongful discrimination may endorse different cases as “core” to an analysis of sex or race causation and discrimination, depending on the extent to which they exemplify the wrong in question.³¹ This means that different moral analyses will have wider consequences than has been typically presumed, as they lead also to different accounts of race or sex causation and discrimination more broadly. These connections between what discrimination on the basis of race or sex is, what makes such discrimination wrong, and what it is for race or sex to be causes challenges the predominant philosophical methodological approach to these topics, which by and large treats them separately.

Of course, this very entanglement comes with its own set of theoretical and practical risks.³² As a general matter, reflective equilibrium as a method for moral theorizing always runs the risk of overfitting to pre-existing intuitions. But it faces a distinctive methodological pitfall when used to develop a philosophical account of discrimination on the basis of sex (or race, etc.). For our intuitions about which cases are genuine instances of discrimination on the basis of sex may draw upon our views as to whether they ought to be legally deemed and thus legally prohibited as discriminatory on the basis of sex. But to the extent that these latter intuitions are influenced by what are presumably theoretically extraneous matters—matters such as how certain jurisdictions have come to structure their discrimination doctrine—they are suspect as intuitions about sex discrimination as such.³³ On the other side of things, when used to develop a legal analysis of what constitutes discrimination on the basis of sex, the normative-causal account’s open appeal to substantive judgments about the moral wrongness of cases runs up against legal ideals of neutrality and objectivity. This in turn raises questions of what exactly it is to uphold such ideals in judicial reasoning about discrimination and whether doing so is even possible let alone desirable.³⁴ Suffice to say, all this leads to an entirely different realm of highly complex and knotty issues.³⁵

Drawing out these risks and revisionist implications is a valuable exercise. For if, as I suspect, they are what drive theorists of discrimination away from a normative-causal account, they indicate theoretical pressure points in analyses of discrimination that have shaped how the literature has developed. In so doing, they clarify the stakes of analyses of discrimination, allowing us to better appreciate the lines of inquiry that these commitments enable as well as those they close off, and so in turn allowing us to better scrutinize whether we should continue to cling onto these features or give them up in exchange for a theory with other theoretical and practical benefits.

I want to now close this paper with some remarks on why I think the aforementioned theoretical costs are worth paying. They are, I claim, because what we get in exchange is an analysis of discrimination that better than prevailing accounts, successfully integrates the core social and ethical dimensions of the phenomenon.

V. CONCLUSION

A normative-causal approach to discrimination vindicates the longstanding thought that what connects a given act of discrimination to the grounds of discrimination is a causal link—and crucially, it does so in a way that explains why the cases raised in Sex+ Cases are genuinely troublesome (and not just due to “mistakes” in the discourse). Furthermore, it suggests that what we should say about charges of question-begging in the literature is that they are both apt, because true, and also inapt, insofar as they suggest falsely that there are wholly non-circular ways to a solution. I take these to be substantial theoretical virtues of an account of discrimination.

Still, in my view, the most important advantage of a normative-causal approach over extant analyses lies elsewhere: in its integration of the social and ethical dimensions of discrimination. The fact that the features of conventional discriminatory concern such as sex and race are markers of social difference is typically taken to be significant only for normative analyses of discrimination. But it seems to me that the social nature of these cases of discrimination has significance not just for our understanding of what makes them wrong but for our conceptual understanding of these kinds of discrimination. As markers of social difference, race and sex track not just differences in skin color, phenotype, chromosomal makeup, secondary sex characteristics, and so on, but also differences in the distribution of social goods: resources, norms, expectations. These differences are systematically bundled together such that the causal effects of race and sex are now broadly taken to include effects issuing from the latter set of factors, which lie outside of what might traditionally be taken to be the sine qua non of the category. To the extent that the boundaries of these categories and their causal effects are under contestation due to substantive ethical debate—Should sex include also gender conforming behavior? Which ones? Are these effects thereby effects of sex?—it seems only natural that what it takes to discriminate on the basis of these categories would also be a normative matter.

A normative-causal approach presents a promising avenue to further elaborate this thought. It is by now a truism that discrimination is a “social problem,” that it does not take place against a “background” of social inequality but actively produces and reproduces that inequality, and that we fail to understand it fully so long as we apply to it the framework of one-off animus-fueled interpersonal transactions. These features are, in my view, worth placing at the core of the discrimination concept. They guide us towards an analysis of what discrimination is that attends closely to how various practices reflect, make, and remake our unjust social reality—a story which will most certainly be both normative, to pick out that unjust social reality, and causal, to pick out those processes that make it.

ACKNOWLEDGEMENTS

This paper owes its existence to years of conversation on this and related topics in discrimination theory with Issa Kohler-Hausmann. It was subsequently rethought and honed through discussions with participants at the Princeton MAP talk series and the Philosophy Mountain Workshop. I am also grateful to the thorough and thoughtful comments I received from two anonymous reviewers at Political Philosophy, which undoubtedly greatly improved the final product that appears here.

COMPETING INTERESTS

The author declares that she has no competing interest.