Thursday, May 28, 2020


Become a Patron! 



Excerpted From: Issa Kohler-Hausmann, Eddie Murphy and the Dangers of Counterfactual Causal Thinking about Detecting Racial Discrimination, 113 Northwestern University Law Review 1163 (2019) (153 Footnotes) (Full Document)

IssaKohler HausmannJudge Schroeder did not believe Dr. Lamberth could pick out Hispanic drivers by looking at them.

To be more specific, Judge Schroeder did not find the method Dr. Lamberth employed to create a “benchmark” of the objective rate at which Hispanic drivers violated traffic laws in North Carolina's Alamance County to be scientific.

To be even more specific, Judge Schroeder did not believe that Dr. Lamberth had deployed an objective, replicable, or verifiable method to detect the rate at which Hispanic drivers violated North Carolina traffic laws by hiring two auditors, Mr. Rivera and Mr. Valdez, to sit in parked cars on select roadways, observe passing cars, count which were violating North Carolina traffic laws, and look at drivers to see “who 'appeared to be’ or 'looked’ Hispanic.”

Judge Schroeder noted that other cases and other peer-reviewed studies relied upon a similar observational methodology to construct a benchmark of the rate at which a designated demographic group violated certain laws. However, he pointed out that those studies “utilized more reliable methods of observation,” and were comparing “African-American drivers, not Hispanic drivers, to non-African-American drivers,” which, his reasoning implied, presented obvious and unproblematic indicia of racial status. Judge Schroeder concluded that because “no control, standard, or description was used to identify Hispanics [,] ... Dr. Lamberth offered no information on what, if any, standard [the auditors] used,” and “Dr. Lamberth's study thus relies entirely on the subjective views of Rivera and Valdez and their personal, totally subjective say-so of who should be considered 'Hispanic.”’ Judge Schroeder, therefore, excluded Dr. Lamberth's expert report and testimony as failing to meet the Daubert standards for admissibility, including testability, known error rates, peer review, and general acceptance in the scientific community.

Based on his conclusion that this and another study could not prove discrimination, Judge Schroeder ruled in 2015 that the United States Department of Justice failed in its more-than-three-year effort to show that the Alamance County Sheriff engaged in “a pattern or practice of discriminatory law enforcement activities directed against Latinos in Alamance County” in violation of the Fourth and Fourteenth Amendments.

How do we know when a particular act, practice, or policy is an instance of discrimination? What precisely do we mean when we identify discrimination as an act, practice, or policy taken “because of” race or ethnicity?

This Article will probe these questions in one arena of social life in which this author just happens to have experience and interest--police and prosecutorial racial discrimination. The conceptual analysis offered here is applicable to other arenas of social life--from employment, to housing, to credit--and has modified implications for how to conceptualize other categories of discrimination--from sex to sexual orientation. However, I will stick to the example of race or ethnicity in criminal justice simply to focus the discussion and because these debates have immediate political salience as recent killings of unarmed black persons have pushed the question of discriminatory police violence into mainstream public debates. So how do we detect discrimination when questions of great importance hang in the balance?

This Article argues that animating the most common approaches to detecting discrimination in both law and social science is a model of discrimination that is, well, wrong. I term this model the “counterfactual causal model” of race discrimination. Discrimination, on this account, is detected by measuring the “treatment effect of race,” where treatment is conceptualized as manipulating the raced status of otherwise identical units (e.g., a person, a neighborhood, a school). Discrimination is present when an adverse outcome occurs in the world in which a unit is “treated” by being raced--for example, black--and not in the world in which the otherwise identical unit is “treated” by being, for example, raced white. The counterfactual model has the allure of precision and the security of seemingly obvious divisions or natural facts. Despite notable objections, this remains the leading conception of discrimination in both law and social science. But I contend that this model is wrong. It is wrong because it is based on a flawed theory of (1) what the concept RACE references and how it produces effects in the world, and (2) what we mean when we say it is bad to make important decisions “because of race.”

Much of this Article is dedicated to making the negative case against the predominant counterfactual causal model of discrimination by arguing that it is incompatible with the constructivist theory of race, to which most (but not all) academics and judges say they subscribe. In the process, I propose a radically different way of conceptualizing discrimination that uses two concepts largely unfamiliar in debates about discrimination. Although these concepts might at first blush seem challenging, I contend that they are essential to any plausible approach to discrimination.

Objections to talking about race as a cause in the counterfactual framework are usually raised in terms of manipulability. Candidates for causes in the counterfactual framework are limited to viable treatments to which a unit could be subjected at the time the outcome of interest might occur. If one cannot manipulate a person's race at the moment of a police stop, traffic encounter, or prosecutorial charging decision, then it is impossible to detect if the person's race was the sole cause of an unfavorable outcome. But, as many have pointed out, we should not confuse empirical and theoretical objections. If one accepts that race or ethnicity is the type of thing that is properly conceptualized as an isolated manipulation on units that can otherwise remain the same units, then there are workarounds to the practical problems of actual manipulation. We can, for example, imagine presenting a police officer or prosecutorial decision-maker with candidates for an outcome bearing identical credentials and vary some indicia of the candidate's racial status in order to detect the treatment effect of race. This is the logic of what are called audit studies--a method illustrated brilliantly by Eddie Murphy in the 1984 Saturday Night Live skit “White Like Me,” where he applies white face makeup to see how he is treated as a white man in New York City. But my concerns with conceptualizing race as a treatment as a way to define discrimination are theoretical, not practical.

The problem with identifying discrimination with the treatment effect of race is that it misrepresents what race is and how it produces effects in the world, and concomitantly, what makes discrimination because of race a moral wrong. In the classic counterfactual causal inference framework, race can be a treatment on units only if manipulating it does not entail fundamental changes to other aspects of the unit. Thus, audit studies can be interpreted as detecting the treatment effect of race and race alone by changing some signifier of racial status on candidates only if the manipulation does not transform the unit into a different unit in ways necessarily relevant to interpreting an action as discriminatory. But if the signifiers of racial categories fundamentally structure the interpretation and relevance of other characteristics or traits of the unit, then it is a mistake to talk about identical units that differ only by raced status. Yet, to talk about race as an isolated treatment on units reduces it to some set of signifiers that elicits outcomes in the world only as a psychological trigger or stimulus to disfavor. And to do so is a gross category mistake, at least if you subscribe to the constructivist theory of race.

A constructivist rejects the notion that racial categories in the United States are constituted by genetic or biological facts, and instead holds that what now seem like obvious, taken-for-granted categories of racial difference were constructed over hundreds of years of historical practices starting with chattel slavery and colonization. Categories such as “whiteness” and “blackness” were forged through social relations such as forced labor, colonization, immigration, and Jim Crow; they were contested and policed through many institutions including immigration laws, housing and education segregation, violent regulation of social and intimate relations, and hoarding of occupational and economic opportunity. These processes made certain aspects of physical appearance salient markers of social difference and reinforced their relevance over many types of interactions. Just because we can trace the historical processes by which these categories were constructed does not make them any less real with real effects; a wink or hoodie can have fatal consequences because of the status of race. We continue to live in a world in which most important institutions are racialized, meaning they play an active role in reproducing the significance of these very categories.

This--with extreme brevity--is what is commonly termed the social constructivist theory of race, which holds that the racial categories as we currently know them are not inevitable distinctions that naturally arise from objective biological differences, but instead are the product of contingent historical social processes. Thus, for a constructivist, the term “race” cannot refer to an attribute, a genetically produced trait, or a signifier--level of melanin in skin, phenotype, distinctive names or speech-- that people just have and thereby obviously belong to a designated racial group. The term references a complexly constituted social fact, whereby material and dignitary opportunities are organized such that certain physical and cultural signifiers become the salient markers of consequential cultural categories, and those categories are constituted by a constellation of social relations and meanings with a definite content and organization. Race in America is, as Bonilla-Silva and Zuberi say, “a social system that uses skin color as the criterion for classification .... Racial stratification is real, but biology is not its root cause.” Although the constructivist view is now widely accepted in academic circles, its implications are not appreciated by the predominant legal and social scientific approaches for detecting discrimination.

One implication of the social constructivist theory is that race cannot be conceptualized as an isolated treatment in the counterfactual causal model, and accordingly, racial discrimination cannot be defined as the treatment effect of race. If we accept the constructivist theory of race, then we must reject attempts to detect racial discrimination that seek to isolate the causal effect of race alone because it rests on a sociologically incoherent conception of what race references and how it can cause a distinctive form of action called discrimination.

The first conceptual tool I use to proffer an alternative account of discrimination that is compatible with the constructivist theory of race is borrowed from moral philosopher Bernard Williams, who coined the phrase “thick ethical concept” for terms that simultaneously describe and evaluate the object to which they are applied. Thin ethical concepts, such as BAD, OUGHT, or RIGHT, do not require “institutional and cultural presuppositions” in order to impart judgment. To apply the terms properly, you do not need access to complex social facts, and to say that an action is BAD or RIGHT does not convey more information about the evaluated action beyond the moral valuation. Thick ethical concepts, on the other hand, such as RESPECTABILITY, CHIVALRY, or PIETY, do require complex social knowledge in order to be used and decoded. To invoke the term is to simultaneously represent the evaluated action as a particular kind of action--one that is only classifiable as such using a cultural repertoire and understandings about the functioning of a particular social world--and to impart judgment. That is, to morally evaluate an action with a thick ethical concept communicates information about the way in which the action is bad that relies on institutional and cultural facts.

Discrimination is not a thin ethical concept that can be represented as “choosing + bad,” “arresting + mean,” or “prosecuting + irrational,” because we know that the Fourteenth Amendment, Title VII, Title IX, and countless other state and federal statutes are not about outlawing all bad, mean, or irrational forms of state or private action. Discrimination is a thick ethical concept that can only be comprehended with access to situated cultural knowledge about the relevant categories that make up a particular society's system of stratification and a normative critique of how those categories operate. In order for something to be discriminatory--instead of merely mean, random, or irrational--the act or policy must rely on meanings or facts that constitute the social category in ways that we morally disavow. Therefore, any discrimination-detecting exercise must proceed from some moral theory--often implicit--of what is fair or just in the face of how a particular society's stratification works through meanings and relations of its social types.

This brings me to the second conceptual tool I use to build an alternative to the counterfactual causal model of discrimination, that of constitutive explanation, which I argue accurately captures the type of claim made when something is labeled discriminatory. A constitutive claim accounts for the capacities of complex systems by reference to their constitutive elements: the parts and organization that make the system what it is. To identify something as discrimination when it happened “because of” race or ethnicity is not to name a relation of counterfactual dependence defined as an outcome triggered by isolating and manipulating an individual trait. To identify something as discrimination when it happened because of race or ethnicity is to offer a constitutive claim that explains how an action or practice can be morally objectionable by virtue of the complex of social meanings and relations that constitute the social category. A constitutive claim unifies a set of disparate practices (choosing, excluding, promoting, demoting, arresting, jailing, beating, humiliating, killing) as morally problematic in the same way, namely by reference to how the action or policy engages the content of the socially constructed category.

Combining these two conceptual components yields a definition of discrimination as an action or practice that acts on or reproduces an aspect of the category in a way that is morally objectionable. It is a thick ethical concept that--to express the distinctive wrongfulness of the action vis-à-vis the category-- must rest upon an account of the system of social meanings or practices that constitute the categories at issue.

The definition contains empirical and normative elements, both of which are black-boxed in this Article. The first black box must contain social ontology or practical anthropology, requiring us to identify and define the stratifying social types in a given society; the second black box must contain political and moral philosophy, requiring us to decide what is fair and just in various state and private arenas given what the categories are. The point of this Article is not to fill in those black boxes, but to explain why both elements are fundamental to any discrimination-detecting endeavor.

B. What to Expect and Why It Matters

Before proceeding, let me be clear that my aim here is not to criticize quantitative methods, audit studies, or legal strategy from any case or research program. My aim is to make a set of sociological and analytic points concerning the meaning of those studies. My conceptual points lead to political-strategic ones. At the risk of being disowned by my materialist intellectual family, I will say that ideas matter. They especially matter in the legal field, where the way in which powerful legal actors conceptualize an issue has profound implications for what they do with their power.

Currently, many courts, experts, and commentators approach detecting discrimination as an exercise measuring the counterfactual causal effect of race-qua-treatment, looking for complex methods to strip away confounding variables to get at a solid state of race and race alone. But what we are arguing about when we argue about whether or not statistical evidence provides proof of discrimination is precisely what we mean by the concept discrimination. We are arguing about the social meaning of race and how it structures outcomes of interest. Similarly, what we are arguing about when we debate what variables ought to be controlled for or balanced on in a quantitative exercise to detect discrimination is what are the fair or just grounds for decision-making or resource allocation in light of what race is and how it operates. We ought to be clear about the nature of the debates we are having such that the driving issues are not obfuscated by claims of methodological rigor or objective scientism.

Because thick ethical concepts and constitutive explanations are significantly less familiar ways of approaching discrimination than the counterfactual causal model, Part I offers a primer so that the reader may sense that there is an alternative way of thinking about this issue before I proceed to my negative case.

Part II turns, briefly, to showing just how prevalent the counterfactual causal conceptualization is in both law and social science. To fully comprehend my critique of the model, it is essential to present it with formal rigor, and Part III does so.

The first two Sections of Part III explore critiques of talking about race as a treatment in the counterfactual causal inference literature and lay out the most widely accepted retort that has been offered to support the counterfactual causal model of discrimination. The third Section of Part III fleshes out my sociological objection to race-qua-treatment by exploring how attempting to isolate the treatment effect of race is at odds with a constructivist account of race, which rejects the view that racial categories “reflect natural, stable differences between human groups.”

My arguments are pitched at a fairly high level of abstraction in Parts I through III, and therefore, Part IV explores the cash value of these theoretical points by thinking through the nitty-gritty of design and interpretation of audit or correspondence studies, which are usually touted as the gold standard for causal inference. In this Part, I argue that audit studies certainly can produce evidence of discrimination, but they don't do so by virtue of isolating the treatment effect of race. Audit studies are often recognized as compelling evidence of discrimination because they instantiate widely shared moral convictions, namely that, at a minimum, persons in the designated social groups with the given set of credentials ought to elicit the same treatment. But, properly understood, audit studies produce evidence of discrimination in the same way that analysis of observational data or an individual encounter do: by relying on a constitutive claim about what race is to ground a moral claim about what is distinctively wrong about the act or practice.

The final Part of this Article does not offer a new set of clean, determinate doctrinal formulations of discrimination, nor list magic-bullet methods to detect it. As will become clear, one of my main claims throughout this Article is that it is impossible to do so without a prior moral-political philosophy of what justice requires in private and public domains in light of what racial and ethnic stratification is in America today, a substantial project for a different paper (or book). However, I suggest that an upshot of the arguments I advance about the counterfactual causal model is that distinctions between disparate treatment and impact that have been advanced in terms of the former being caused exclusively by race and the latter being caused by something that is not-race, but correlated with race, are not conceptually tenable distinctions. They are not tenable because these formulations only make sense if one defines race as the visual or social cues associated with the category: race is skin color, or is phenotype, or is one of the physical or social signifiers of the category. If one subscribes to the constructivist notion of race--in which signifiers come to be indicative of a status only through entrenched social practices--then it is nonsensical to talk about constitutive practices as somehow being race-neutral things. Of course, we can advance other distinctions between what should or should not be legally actionable discrimination, but we can't do so by relying on value-free notions of counterfactual causality.

Many scholars have compellingly argued that the law of equal protection ought to be interpreted as a principle of anti-subordination, or that the purpose of antidiscrimination law more broadly ought to be understood as a project of remaking social meanings of historically marginalized groups. I agree with their arguments. But I approach this debate from a new angle, backing out a theory of the category of race from the prevailing methods used to detect discrimination. I conclude that if one subscribes to the constructivist theory of the category of race, then it is incoherent to understand the legal proscription against discrimination as anything but a project to remake the very meanings of social categories (unless you don't want to distinguish discrimination from mere irrationality or idiosyncrasy, but then you have a different problem--which is to explain why the state's heavy coercive machinery should be concerned with some forms of classification and not others).

Insofar as we (and I use the first-person plural pronoun to indicate I understand this Article as a part of conversation with activists on this front) are interested in transforming the social structures that systematically oppress and disadvantage minority communities, we must be incredibly attentive to the way discourses about race and ethnicity circulate and settle in the halls of power. In presenting quantitative evidence to courts, it is a mistake to talk as if we have gotten at the true effect of race by modeling it as a counterfactual treatment. In fact, to do so cements an already predominant and problematic understanding about race in public and legal discourse: one that is distressingly dehistoricized and desocialized. Discourses have effects. Folks in positions of power--namely judges considering discrimination cases--make important decisions because they understand words and concepts in a particular way. And we should train our sights on trying to make them understand things in what, I contend, is the right way. Doing so will bring to the fore difficult political and moral judgments that are at the heart of debates about discrimination broadly.

At a minimum, I hope to start a dialogue with the community that provides evidentiary expert statistical services to litigants in discrimination cases about how that material is consumed and given meaning in courts. Social scientists can do more in the fight against discrimination than provide technical skills; they can also offer careful reflexive thinking that rejects folk, commonsense-sounding concepts of race. We must be vigilant to “uncover the hidden assumptions in [our] own scientific unconscious ....”

[. . .]

Bertrand Russell famously admonished philosophy to abandon the notion of causality, alleging that it survives, “like the monarchy, only because it is erroneously supposed to do no harm.” Nancy Cartwright responds that causal notions are essential to differentiate between effective and ineffective strategies. In the context of police and prosecutorial actions, we often observe an association between negative policing or case outcomes and the raced status of individuals or aggregations like neighborhoods. But in order to know what is to be done about these associations, we must inquire into causal relationships.

The question of detecting discrimination could be posed in terms of effective strategies--What sort of causal understanding helps us to know if the practices at issue in a discrimination case should be changed? Conceptualizing race as treatment does not help us distinguish between effective and ineffective strategies for dismantling discrimination because it essentially asks what the effects of a racial signifier would be if the social facts of race were not what they are today in the United States.

The ideal thought experiment that captures the treatment effect of race and race alone asks the following: produce for the decision-maker's consideration two units with identical credentials (What if the entrenched systems of racial stratification were not so?) and purge the racialized meanings that inflect different values to similar credentials or attributes (What if racial identification did not actually change the way decision-makers perceived or evaluated formally similar things about individuals or neighborhoods?). The question is, Does information about a causal link of that variety pick out practices that the Constitution prohibits? I say no because that thought experiment would produce information about a particular decision-maker's idiosyncratic distaste for floating signifiers.

I contend that the familiar refrain that defines discrimination as an action or practice that happens “because of race” does not identify a relation of counterfactual causality. Instead, it identifies a constitutive relation that grounds a thick moral evaluation, which means we can only identify the distinctive wrongfulness of the action or practice by reference to what social types such as BLACK, HISPANIC, or WHITE culturally reference, in what it consists to name someone such a type, and other ways of identifying what the categories are.

To illustrate how constitutive, and not counterfactual causal, explanations are at work in identifying discrimination, let us return to the visiting anthropologist on the island nation stratified by Royals and non-Royals. Asking about counterfactual causal dependence is just an unhelpful way of figuring out whether stepping off the street when Royals approach is properly described by a thick ethical term such as “non-Royal debasement.” And it is an equally unhelpful way of identifying effective strategies, namely what needs to be changed in order to dismantle Royal-based stratification.

We can address those questions with constitutive explanations. One would need to ask how it is possible for non-Royals to feel compelled to step off the sidewalk by reference to the constitutive aspects of the socially constructed category Royal, namely by detailing the structure and content of the social meanings and relations that make the category what-it-is. One would have to ask if the fact of non-Royals deferring sidewalk access is conceptually or logically dependent on the very structure of the social kind Royal as it currently exists. Said yet another way, if a researcher were able to make a person on the street of this society perceive two identical walkers but for purple-cape-wearing and stick-carrying--and significant other meanings about this person's actions, credentials, or behavior were left unaffected by this manipulation--then there would not be such a thing as ROYAL in the way this society currently knows it. There would not be a morally salient issue called “non-Royal debasement” to be addressed.

The constitutive explanation grounds the thick ethical evaluation of the act. To interpret the act of stepping off the sidewalk as non-Royal debasement (or conversely, respectful Royal obedience)--as opposed to a spontaneous adjustment to scarce sidewalk space or expression of a preference for road-walking--we need access to sociological and anthropological knowledge about what constitutes the relevant social kinds in this society. And describing it as non-Royal debasement (or conversely, respectful royal obedience) is not merely disapproving or approving of the act. It is invoking a thick ethical concept, which simultaneously describes, with textured, system-level information, and evaluates the object to which it is applied. These two facets cannot be separated because the evaluative aspect--the expression of judgment about the act--can only be activated using the descriptive component--the constellation of situated social meanings and cultural constructs referenced by the concept.

We can describe something as discrimination only if it implicates social meanings in a way that constitutes some social kinds as degraded or disfavored, over many domains and times. Race does not have effects in the world by triggering mere affectual dislike for random physical signifiers. In fact, I contend that our culture's signifiers of racial groups are just not available to be the objects of thin preferences (or “tastes”) the way that other aspects of physicality would be, such as freckles or bunions, because of the history of racial group construction. And that same process that constructed racial categories explains why so many people seem to have the same affectual response to the same signifiers--why racial discrimination is a pervasive practice in many domains in the way that freckle or bunion discrimination is not.

We can still seek to detect discrimination using audit studies, regression techniques with observational data, and many of the same methods folks have long used in social science and legal challenges. But we should be very clear what we are doing with those methods. The argument I have advanced in the preceding 29,807 words is that what we are not doing with those methods is detecting the treatment effect of race in the counterfactual causal sense.

Other careful thinkers committed to the counterfactual causal definition of discrimination have thoughtfully engaged questions regarding the design and interpretation of quantitative measures of discrimination, arguing that researchers ought to be both reflective and explicit about which variables are included and excluded when trying to isolate the causal effect of race. But central to all counterfactual causal accounts of racial discrimination is the notion that there is a solid state race in units (individuals, neighborhoods, etc.), an objective fact about the units that can be isolated after stripping away all confounders. For something to be a treatment, there must be a way to pick out what THE-TREATMENT is--distinct and apart from all of the things that are NOT-THE-TREATMENT so that we are sure we are talking about identical units that differ only on the-treatment. If we cannot pick apart THE-TREATMENT from NOT-THE-TREATMENT, then we are not estimating a treatment effect of race and race alone when we compare the outcomes of candidates with some list of similar credentials and signals for different racial categories. We are doing something else.

I believe that what we are doing with both observational and audit studies of discrimination is building a case, collecting evidence to support that case, and otherwise “vouching” for a particular constitutive claim with moral dimensions: that a specific action, practice, or policy is possible because of the social fact of race (or ethnicity) in a manner that implicates constitutive aspects of the category that we would like to change. Because racial discrimination is a thick ethical concept, the way we figure out if a specific action, practice, or policy is possible because of the social fact of race is inextricably intertwined with the grounds for the moral evaluation of whether or not it ought to be tolerated.

An implication of my argument is that the disparate treatment versus disparate impact binary, so central to so much of antidiscrimination law and literature, is not a tenable distinction along the lines it has often been advanced. We cannot define the former as an outcome caused by race (or where race was a substantial motivating factor, or other “close enough” formulations) and the latter as an outcome caused by a facially neutral consideration that just happens to affect racial groups unequally.

I just do not see any difference between disparate impact and disparate treatment that can be gotten at with value-free notions of counterfactual causality, much less a distinction between classification on the basis of race as such in contrast to race-neutral factors that just happen to produce dissimilar racial impacts. Disparate treatment is often distinguished from disparate impact with reference to intentional discrimination, but the explanandum (discrimination) in the explanans (discriminatory intent) is not saved by reference to intent or motive (to discriminate) because the point of the explanatory endeavor is to specify which sort of purposive differentiating practices are discriminatory and which sort are permissible.

Furthermore, it is not clear why reliance on constitutive elements of a category (black-boxing how we determine in what those consist) can be coherently referenced as “race neutral” for purposes of deciding if use of such element is discriminatory. Calling these “race neutral” is coherent if one subscribes to a biological conception of race, in which the category consists in sharing some genetic or biological facts (but, as argued extensively above, then we have another problem, which is explaining why we need to super-size scrutiny when the state classifies on the basis of those facts). But if one subscribes to the constructivist theory of race, one must recognize some set of cultural performances, social practices, and institutions that constitute the system of social meanings of the racial or ethnic category. If the set of racially constitutive cultural performances, norms, meanings, social practices, or institutions were empty, then there would not be a salient category capable of producing discrimination. There certainly would be a group of people with certain physical traits (just like there is a group of people with bunions or freckles), but there would not be a complex of social meanings such that we could talk about groups being discriminated against in the thick sense. Again, someone could reject the constructivist theory of race and hold that it is a biological or genealogical fact. But such a view simply leaves the proponent no way to distinguish the thick meaning of discrimination--a morally problematic way of allocating benefits and burdens--from mere choosing based on idiosyncratic tastes or random meanness.

We often lose sight of the practices and meanings that constitute the very categories of race because one of the properties of this social category is to appear as a natural fact about bodies instead of the effect of persistent social stratification and meaning-making. But the categories of WHITENESS or BLACKNESS are only available as a basis for perceiving and acting upon in a discriminatory manner because of the system of social meanings and practices that bring about the very category; said another way, properties and structure do not exist independently of each other.

Many theorists and commentators have argued in distinct fashions that antidiscrimination law ought to be a project of cultural reconstruction. For example, Robert Post has argued that antidiscrimination law should not be thought of as obliterating salient differences of race or sex, but changing the meanings; with respect to sex, Post urges an interpretation that “would not require us to imagine a world of sexless individuals, but would instead challenge us to explore the precise ways in which Title VII should alter the norms by which sex is given social meaning.” Reva Siegel and Jack Balkin have characterized the antisubordination tradition as “[t]he moral insistence that the low be raised up--that the forces of subordination be named, accused, disestablished, and dissolved--is our story, our civil rights tradition.” Andrew Koppelman proposes that what he terms the “antidiscrimination project” is necessarily an endeavor in which the state actively undertakes the goal of cultural transformation that “seeks to reconstruct social reality to eliminate or marginalize the shared meanings, practices, and institutions that unjustifiably single out certain groups of citizens for stigma and disadvantage.”

I concur with the content of those accounts in terms of a prescriptive vision for antidiscrimination law. However, the conceptual points that I have argued above have two important implications for antisubordination theory. First, in order to defend a position that the Equal Protection Clause ought to be understood as remedying group inequality, one needs a theory of what constitutes GROUPNESS in the relevant respects. Only with this in hand can we account for why we care about members of specific groups occupying disadvantaged positions in the social hierarchy above and beyond caring about the existence and shape of the hierarchy. Second, if one accepts the constructivist account of racial groups, then transformation of the constitutive meanings of the relevant groups is all antidiscrimination law could coherently be about. Therefore, antisubordination can make a more forceful claim as the only sound interpretation of antidiscrimination norms for those that reject a biological definition of race. 

Associate Professor of Law and Sociology, Yale University.