- Frej Klem ThomsenFrej Klem ThomsenDepartment of Culture and Identity, Roskilde University
The conceptualization and moral analysis of discrimination constitutes a burgeoning theoretical field, with a number of open problems and a rapidly developing literature. A central problem is how to define discrimination, both in its most basic direct sense and in the most prominent variations. A plausible definition of the basic sense of the word understands discrimination as disadvantageous differential treatment of two groups that is in some respect caused by the properties that distinguish the groups, but open questions remain on whether discrimination should be restricted to concern only particular groups, as well as on whether it is best conceived as a descriptive or a moralized concept. Furthermore, since this understanding limits direct discrimination to cases of differential treatment, it requires that we be able to draw a clear distinction between equal and differential treatment, a task that is less simple than it may appear, but that is helpful in clarifying indirect discrimination and statistical discrimination. The second major problem in theorizing discrimination is explaining what makes discrimination morally wrong. On this issue, there are four dominant contemporary answers: the valuational and expressive disrespect accounts, which hold that discrimination is wrong when and if the discriminator misestimates or expresses a misestimate of the moral status of the discriminatee; the unfairness account, which holds that discrimination is wrong when and if the discriminator unfairly increases inequality of opportunity; and the harm account, which holds that discrimination is wrong when and if the discriminator harms the discriminatee. Each of these accounts, however, faces important challenges in simultaneously providing a persuasive theoretical account and matching our intuitions about cases of impermissible discrimination.
- Political Philosophy
Discrimination has been at the center of some of the most intense political debates in recent decades, ranging from affirmative action policies across labor market inequalities in hiring and promotion to police profiling. And yet, while there are clear and uncontroversial cases of discrimination, such as the disenfranchisement of women or the racial segregation of Blacks and Whites, there are many cases where it is unclear both whether a given act or practice constitutes discrimination; whether if it does so, it is morally wrong; and if so, in virtue of what. Is it discrimination, for example, when a religious organization refuses to hire nonbelievers as priests, or when employers in the retail and services industry give preference in hiring to applicants widely perceived as physically attractive?
Theorizing discrimination ideally aids us in answering such questions, and the concept and practices of discrimination have attracted increased scholarly attention in political theory and political philosophy, particularly over the past decade. This has served to clarify many previously underexplored aspects of the topic, but there remain important disagreements and ambiguities, about both how to best understand what discrimination is, and how to explain what (if anything) is morally wrong with it.
This article first explores how to define discrimination, illustrating some of the difficulties of deriving a satisfactory definition and distinguishing some of the various forms of discrimination. It proceeds to then present and discuss the four most prominent theories of what makes discrimination morally wrong: the valuational disrespect account, the expressive disrespect account, the unfairness account, and the harm account.
Discrimination is a surprisingly slippery concept, and much of the attention that it has received in recent years has focused on constructing a precise definition (Altman, 2015; Eidelson, 2015; Horta, 2015; Lippert-Rasmussen, 2007a, 2006a, 2013; Moreau, 2010; Heinrichs, 2007; Edmonds, 2006; Arneson, 2006; Halldenius, 2005).
Probably the most essential feature of paradigmatic cases of discrimination is the disadvantageous differential treatment that they embody, which is to say that they involve a person treating another person or group of persons badly, and in a way that she does not treat other persons (Halldenius, 2005; Lippert-Rasmussen, 2006a). Classical examples, such as a racist being rude or hostile toward members of a different race, or a sexist being paternalistic or demeaning toward members of the opposite sex, illustrate these two fundamental features. Paradigmatic discrimination is discrimination against these persons because they are treated in a way that others are not, and because the different treatment to which they are subjected is worse (Lippert-Rasmussen, 2007c).
Two important initial clarifications concern the comparative of discrimination and the dimension of treatment. With respect to the first, we should allow that the difference in treatment can be counterfactual—that is, that the discriminator treats the discriminatee differently than she would have treated someone else, even if she does not treat any such persons. We should be willing to say, for example, that Robinson Crusoe discriminated against Friday by treating him badly compared to how he would have treated a European person (assuming that we can characterize Crusoe’s treatment thus), had there been any European persons around to treat one way or the other. As for the second aspect, the badness of being discriminated against should be conceived of as tied to the specific respect in which the discriminatee is discriminated against. We would want to say, for instance, that a person who is fired for revealing her homosexuality at her workplace is discriminated against with respect to employment at that place, even if she is ultimately made all things considered better off by being fired because her being unemployed leads to her finding a superior position with a different employer (Lippert-Rasmussen, 2013).
Discrimination also requires a link between the property that defines the persons treated worse (and which distinguishes them from those treated differently) and the fact that they are treated worse. If the difference in treatment is accidental—suppose that it is decided by lottery, and it just so happens that every P-person is randomly assigned the worse treatment and every non-P-person is assigned the better treatment—it is intuitively odd to say that the agent discriminates against those treated badly, even if they happen to share some relevant property that those treated differently do not (Lippert-Rasmussen, 2006a, 2007c).
It is perhaps too restrictive to demand that the differential treatment must be motivated by the property that one group possesses (or, more precisely, the agent’s attitude toward this property), since it may well be enough, for instance, that the agent believes (perhaps mistakenly) that the persons treated differently possess that property. On the other hand, allowing any causal connection to suffice may be too inclusive. Consider an employer who desires to support his local community, and for that reason hires only graduates of a nearby college. Unbeknownst to her, the college is a religious institution that has a policy of accepting only students who profess faith in an established religion. As a result, the employer disadvantages atheists, and the difference in treatment is in at least one sense caused by the discriminatees being atheists. Set aside for present purposes the fact that the college’s policy discriminates against atheists, and that the employer’s policy discriminates against qualified applicants from other colleges—the question is whether the employer discriminates against atheists, and here the answer seems to be that although his actions meet the stated conditions, the causal link is not of the right kind to qualify (rather, she may be engaged in indirect discrimination, a variation that we shall discuss shortly). Although it is difficult to specify exactly which type or range of causal connections are at once necessary and sufficient, we can say at least that the difference in treatment must be in some suitable sense caused by the difference between those treated one way and those treated another (Lippert-Rasmussen, 2013).
Together, these considerations suggest that an agent engages in discrimination in the most basic sense iff:
She treats persons with property P (P-persons) differently than she treats or would have treated others (non-P-persons).
The treatment of P-persons is in some respect worse for P-persons than the treatment of non-P-persons is or would be for non-P-persons.
The difference in treatment is suitably caused by P-persons possessing P and non-P-persons not possessing P.
As illustrated here, paradigmatic cases of discrimination involve an agent treating a person or persons in a way that she does not treat others, which is worse than the way that she treats others, and are most naturally conceived as discrimination against the persons treated badly. This suggests that we should also recognize the possibility of discrimination in favor of a person or persons, which would involve treating them differently and better than the agent treats others, and discrimination between persons, which would involve the agent treating a person or persons differently than others, but neither better nor worse. Note that cases of discrimination that appear superficially to be neither bad nor good for either party, and therefore to constitute discrimination between the parties, may on more careful consideration constitute discrimination against one group, as in the notorious case of “separate but equal” discrimination between U.S. Blacks and Whites, which at once provided not equal but inferior conditions to U.S. Blacks and stigmatized them by suggesting that there were reasons to keep them separate from Whites. The existence of such cases of veiled discrimination against, however, does not rule out the possibility of actual cases of discrimination between groups.
At least three problems emerge from this basic sense of discrimination. First, should discrimination be target-specific? That is, should it be part of the definition of discrimination that it targets particular groups, and if so, how should we determine the list of those groups that can be targets of discrimination? Second, is discrimination a descriptive or a moral concept? That is, should it be part of the definition of discrimination that discrimination is morally wrong—if perhaps not all things considered, then at least pro tanto or prima facie? Third, paradigmatic discrimination involves differential treatment, but when is treatment equal and when is it differential in the sense at stake in discrimination? In addition to these issues, we need to distinguish prominent variations on the basic sense of discrimination defined here, including indirect discrimination and statistical discrimination. In the next sections, we will look at each of these problems and variations in turn, while reserving the issue of what makes discrimination wrong for later.
The Targets of Discrimination
The basic sense of discrimination defined previously is neutral with respect to which properties define the groups at stake. Bluntly put, it labels differential treatment of persons born on a Tuesday, for example, as discrimination just as readily as it does differential treatment of homosexuals, disabled persons, women, Blacks, Muslims, or other minorities who traditionally lie at the core of discrimination concerns. This will strike some as counterintuitive and can motivate a restriction on the properties that can be targets of discrimination. A solution common in both antidiscrimination law and theoretical definitions is to adopt a group criterion, which adds a condition to the definition that P be a member of a predefined set of properties (e.g., Ezorsky, 2001; Nickel, 1998; Council of Europe, 1950; Council of European Union, 2000).
While the move has some intuitive appeal, proponents need to support it, specify which properties go on the list, and explain why these properties can form the basis of discrimination while others cannot (Halldenius, 2005; Heinrichs, 2007). In one influential account, the relevant properties are those that are socially salient, which is to say that they are properties that are important in a broad range of social interactions (Lippert-Rasmussen, 2006a, 2007c, 2013). This has the benefit of making the criterion contextual and flexible—different social settings will make different properties socially salient, thus allowing for discrimination of different groups. It also provides an explanation of why these are properties that can form the basis of discrimination: such properties are important by virtue of being socially central, thereby allowing differential treatment to be widespread in a way that will aggravate the badness of individual cases.
Critics can object that none of these points gives us a principled reason to delimit discrimination, and that supplying a principled reason by explaining the conceptual or moral relevance of the distinction the intuition tracks is difficult (Thomsen, 2013). Furthermore, they can point out that while, for example, an employer who refuses to hire persons born on a Tuesday is engaged in a much rarer and perhaps less grievous kind of differential treatment than one who refuses to hire women, it is similarly counterintuitive to insist that the first is not idiosyncratic discrimination, but something else than discrimination entirely. And to the extent that we want to condemn discriminatory practices, there is no obvious reason why idiosyncratic discrimination could not be wrong for the same reasons that paradigmatic discrimination is (at least sometimes) morally repugnant, even if as noted previously, it is likely in general to be less bad because it will be less common (Thomsen, 2013; Hellman, 2008; Heinrichs, 2007; Arneson, 2006).
A persuasive compromise position is to hold that we employ discrimination meaningfully in both a broader and more generic sense, and in the narrower sense, in which it is limited to certain groups. So long as we make sure to distinguish the two, recognizing both senses may in fact help us avoid confusion (Lippert-Rasmussen, 2013).
Moralized and Descriptive Discrimination
A similar issue concerns the fact that the basic sense of discrimination given here is descriptive. It therefore allows for permissible or even laudable forms of discrimination, which will again strike many as counterintuitive. Surely, an objector might say, when we speak of discrimination, we mean a policy or action that is in some respect morally problematic, or at the very least suspect? This can motivate shifting to a moralized definition of discrimination, which adds a condition that the differential treatment be in some sense morally wrong, even if only pro tanto or prima facie (Horta, 2010). Doing so will allow us, for example, to avoid labelling the criminal justice system’s treating convicted offenders badly (by imposing punishment) in a way that it does not treat others as a form of discrimination (Lippert-Rasmussen, 2013).
The potential arguments for and plausibility of moralizing the definition will depend to a large extent on what particular form of the moralizing condition is adopted. But a general concern with the move is that doing so robs the concept of a normative explanatory power that we take it to have in at least some contexts. Thus, it is not uncommon for critics of a social practice to respond to an explanatory challenge—“Why do you think this practice is wrong?”—by claiming that the practice is wrong because it is discriminatory. But on a moralized definition, this move is dialectically vacuous: The follow-up challenge of “Why, then, do you think it is discrimination?” simply returns the critic to the task of showing that the practice is wrong, since only if this is the case can the practice be labelled discrimination.
Consider for comparison the idea that we ought to moralize the definition of lying. Certainly, just like discrimination, charges of lying are ordinarily raised when those raising the charge consider the act to be in some respect morally wrong. And yet we might hesitate to build into the definition the condition that something is an act of lying only if it is in some respect morally wrong. Any wrongness of lying is arguably better conceived as a consequence of at least some of the conditions that define something as an act of lying, along, perhaps, with certain conditions that are normally but not necessarily fulfilled in cases of lying, than as a requirement of something being lying in the first place (although the view that we should accept a moralized definition of lying—so-called moral deceptionism—also has defenders).
Regardless of whether one prefers a descriptive or moralized definition, though, it is crucial to bear in mind that there is a further question of what makes discrimination wrong (when it is wrong). On a descriptive definition, this follows immediately, since labelling something as discrimination does not in and of itself imply that it is wrong. Any claim that a practice is wrongful will need to argue that it is one of the instances of wrongful discrimination. Obviously, this is compatible with holding that all (or, probably more plausibly, many) practices that meet the descriptive definition are wrong by virtue of the features that make them descriptively cases of discrimination. In a moralized definition, one will need to present a similar argument before one can conclude that a practice constitutes a case of discrimination in the first place. As such, while there may be conceptual reasons to prefer a moralized to a descriptive definition, or vice versa, little if anything of normative importance ultimately depends on which of the two definitions one adopts.
Equal and Differential Treatment
A final complication for the basic sense of discrimination that I have defined is that it rests on a distinction between equal and differential treatment that stands in need of clarification. When exactly is treatment differential and when equal?
Note that it is not the fact that treatment is worse for some that makes it differential, first because this is a separate element in the definition, and second because we can easily imagine intuitively plausible cases of equal treatment that is bad for some of those affected, and differential treatment that is not bad for any of those affected. Consider as an example of the first category serving allergenic food to a group, some but not all of whom are allergic; and as an example of the second, consider constructing sex-separated changing facilities in a gym, which while it provides different facilities to men and women does not in any obvious sense treat either group badly (it is possible, of course, that such sex-separation treats other groups badly, such as transgender and crossgender persons). In short, the distinction between differential and equal treatment cuts across the distinction between those badly and not badly affected by the treatment in question.
It is also clear that no treatment is ever equal in the most exacting sense of that term, simply because any treatment of two persons will differ in a vast number of details. If treatment needed to be identical to be equal, then all treatment would be differential.
One answer to the question of what distinguishes equal and differential treatment is to say that differentiation originates with the agent, so that if the agent distinguishes between those treated in order to perform different actions with respect to the two groups, then they are differentially treated, while if the agent makes no such distinction, or the distinction is inactive (i.e., is not employed to perform different actions), then treatment is equal (Thomsen, 2015).
Note also that we should resist the temptation to assume that the distinction between differential and equal treatment makes a moral difference in and of itself. In particular, as famously pointed out by Ronald Dworkin, equal treatment does not equate to treatment as equals, and it is the latter, if anything, that is morally important (Dworkin, 2005).
This concludes our examination of the three complications. While this article has suggested answers to each, it is probably safe to say that they remain open topics in the discussion of how to conceptualize discrimination. In the next sections, we will look at some of the most important variations on the basic sense of discrimination, each of which introduces new complications of its own.
Direct and Indirect Discrimination
One important variation is what is often referred to as indirect discrimination. This term is typically applied to cases that do not meet the conditions of basic (or, in this context, “direct”) discrimination (e.g., by not involving differential treatment), but that involve treatment of a kind that remains similar enough to warrant calling it discrimination, and that are arguably wrong (if they are) for similar reasons (Altman, 2015; Doyle, 2007; Gardner, 1996).
Consider, for example, an employer who refuses to hire female employees and is thereby engaged in direct discrimination against women, as compared to an employer who installs equipment that requires employees to be tall and thereby excludes most women from being able to work there. Strictly speaking, the latter differentially treats short persons rather than women, but there is a sense in which it remains arguably discrimination against women. The question, which has only recently begun to attract attention in the literature, is how to conceptualize the indirect discrimination at stake in such cases, including what makes it discrimination and how to draw the distinction between it and direct discrimination.
One approach to defining indirect discrimination relies on the mental states of the discriminator; for instance, discrimination is direct iff it is intentional, and indirect if it is not. Some may want to include as direct discrimination forms of discrimination caused by some other objectionable mental state such as bias or callous indifference (Altman, 2015; Lippert-Rasmussen, 2013). This approach has the merits of simplicity, of tracking an intuitively important distinction, and of fitting classical cases of what have been labelled direct and indirect discrimination.
There are, however, two challenges for this way of drawing the distinction, in both the narrow intentionality and the broader mental state version. The first is that the distinction between intentional and unintentional discrimination seems to at least occasionally cut across the distinction between cases that we might want to classify as direct and indirect. The second challenge is that drawing the distinction this way counterintuitively leaves us unable to classify cases as direct or indirect until we ascertain the mental state of the discriminator—a task that will often be difficult or even impossible in practice.
Consider a policy of providing free primary education to children, but insisting that they attend schools within a certain distance of their home which, combined with patterns of neighborhood separation along ethnic, racial, or religious lines and correlated differences in the quality of schools, causes minority children to receive inferior education. Because it does not differentiate on the grounds of minority/majority membership but rather on geographical grounds, the policy does not treat minority and majority children differentially, and as such, it does not qualify as discrimination against minorities in the most basic sense. On the other hand, many would presumably want to say that the policy involves some form of discrimination against minority children. And importantly, we would want to say that it does so without knowing what the mental states of those drafting and enforcing the policy were. In fact, we might be willing to insist that it is not a form of direct discrimination, exactly because it does not differentially treat the two groups in the sense of distinguishing between them on the basis of the properties that define them as groups, even if it turned out that the policy was intended to provide minority children with inferior education. The first consideration suggests that mental states are irrelevant to deciding whether a case involves direct or indirect discrimination; the second suggests that the distinction between intentional/mental state and unintentional discrimination cuts across the direct and indirect discrimination distinction. Note that even if this is the case, of course, it does not entail that the distinction between intentional and unintentional discrimination is not relevant in other respects. It may, for example, make a difference to the wrongness of the discrimination (see the discussion of the valuational disrespect account of wrongness later in this article).
Another way of distinguishing direct and indirect discrimination is to say that indirect discrimination against a group is characterized by the group being badly affected by an action or policy that does not differentially treat the group (Thomsen, 2015). The badness of equal treatment that thus defines indirect discrimination can be conceived of in two different ways. The first is to say that a larger proportion of P-persons than of non-P-persons are affected by being treated in the bad way that the agent treats persons. The second is to say that P-persons are, at least on average or prospectively, affected worse than non-P-persons. On this understanding, an agent engages in indirect discrimination against P-persons iff:
She treats P-persons equally to how she treats or would have treated non-P-persons.
The treatment is bad for P-persons, in that (a) a greater proportion of P-persons than of non-P-persons are badly affected by the treatment, or (b) P-persons are affected worse than non-P-persons by the treatment.
The difference in effect is caused by some persons possessing P and other persons not possessing P.
Note, as a brief prelude to the discussion of the wrongness of discrimination, that we should resist the temptation to label indirect discrimination as wrong simply because it affects more P-persons than non-P-persons. While there can be, and may in practice often be, instrumental reasons to be concerned with unequally distributed effects, and although egalitarians will maintain that the distribution of goods across individuals is of normative importance, it seems hard to account for why it should in and of itself carry moral importance how effects are distributed across groups (Lippert-Rasmussen, 2008; Alexander, 1992). This in turn suggests that if there is an important sense of indirect discrimination, it is the one where one group is more badly affected than the other (i.e., 2b). Also, note that it is consistent with this condition’s requirement of treatment being worse for P-persons that the treatment is in some respects good for P-persons and that it is not bad for non-P-persons.
This might be taken to be one sense of what is sometimes labelled a condition of disproportionality, but a different version of disproportionality is sometimes at stake in definitions of indirect discrimination, where the act must be sufficiently bad to be discrimination, as in the condition common to discrimination law that it fails to meet a standard of justification. Proportionality is here conceived as a relation between the justification of the action, such as any benefits produced, and the badness imposed on those subjected to treatment. Clearly, this requires that one can specify how to assign weights to the justifications that count in favor of the permissibility of the action, as well as to the disadvantage imposed on discriminates; determine the threshold as being sufficiently bad; and defend introducing a threshold in the first place. But beyond these challenges, applying a disproportionality condition in this sense to indirect discrimination implausibly entails that indirect discrimination is worse than direct discrimination, since there can be cases of proportionate direct but not indirect discrimination (Lippert-Rasmussen, 2013). As such, we should arguably avoid conceiving of indirect discrimination as limited by disproportionality in this sense.
A second variant of discrimination is so-called statistical discrimination—that is, the use of statistical information about correlations between the properties that define a group and other properties that the discriminator intends to target. The discriminator here targets a proxy-property in order to differentiate between those who do and do not possess another property. As such, statistical discrimination constitutes a case of direct discrimination in order to effect indirect discrimination on the definitions proposed in this article, whose distinguishing feature is the use of statistical knowledge to motivate and inform the discrimination.
Statistical discrimination is common; prominent examples include the use of statistical information about racial, gender, or class differences in probabilities of health and illness by both health care providers and insurers, and the use of information about gender differences in the probability of taking parental leave by employers in hiring. Perhaps the type of statistical discrimination that has drawn the most intense public and political attention in both Europe and the United States over the past decades, however, is the use of statistical stereotypes in deciding whom to subject to law-enforcement practices such as stop-searches and arrest (i.e., police profiling), and a brief consideration of it can clarify the issues that pertain to statistical discrimination more broadly.
When police profile, they are attempting to identify criminal offenders, and to focus their efforts at those individuals most likely to be offenders (set aside here the fact that profiling may involve other forms of discrimination) (Eidelson, 2015; Lippert-Rasmussen, 2006b; Gross & Livingston, 2002; Applbaum, 1996; Wasserman, 1996). To do so, they rely on correlations between other properties, such as gender or age, and criminal offending, so that a young man is ordinarily statistically more likely to be an offender than a woman or the elderly (and much more likely than an elderly woman). We can say loosely that police profile for property P if and only if police give P evidential weight in their decisions on whether to subject a person to a given police practice (a definition that could be broadened to encompass other forms of statistical discrimination) (Thomsen, 2011). In many situations, including police profiling, the use of many such proxies to inform decisions about how to treat individuals is both uncontroversial and rational (Risse & Zeckhauser, 2004). This is obvious once we consider that ultimately even properties not typically thought of as statistical indicators, such as how a person dresses or behaves, often work as proxies in our decisions of how to treat others, even if the statistics at stake are intuitive and unreflective (Schauer, 2003).
Obvious challenges to police profiling are that it can be and plausibly often is exaggerated, based on inaccurate or inadequate statistical knowledge, and counterproductive (Thomsen, 2011; Harcourt, 2004). That is, police may attribute far greater importance to the statistical knowledge than it warrants, and as a result subject far more persons with the proxy to practices than the statistics justify. They may also profile in the absence of adequate statistical knowledge since accurate statistics are difficult if not impossible to come by. And finally, even when profiling is proportionate and well founded, it may do more harm than good, if, for instance, the benefits of a slightly increased probability of apprehending offenders are outweighed by the social costs of implementing a program of police profiling. The latter is particularly likely when profiling ignores the profiled group’s deterrence-elasticity and focuses exclusively on probability of offending, since some groups with a high offending frequency may respond little even to substantial changes in the deterrent threat, while other groups with a low frequency of offending may respond more, including by dramatically increased crime rates, when profiling shifts the deterrent threat to another group.
However, many may feel that these considerations do not exhaust the scope of legitimate concerns that we should feel about police profiling, or statistical discrimination more generally, or even that they miss the most important concern. Is there not a more fundamental problem with treating a person on the basis of statistical facts for which she may bear no responsibility? Treating a person differently in a way which leaves her badly off on the basis of her being a member of a particular group may strike some as patently unfair, even if there is both reason to differentiate on the basis of a different property and a strong correlation between membership of the group to which she belongs and possession of the trait (Lever, 2005, 2007, 2010).
While intuitively appealing, the fact that, as previously mentioned, intuitively permissible statistical discrimination is widespread suggests that this view is mistaken (Lippert-Rasmussen, 2007b, 2011). An employer who prefers applicants with formal qualifications from a recognized educational institution can and presumably will often do so because she relies on the correlation between possessing such qualifications and possessing certain skills that make applicants more suited for the job. This is true even when such qualifications are not a legal requirement (e.g., when the employer is a publisher looking for a skilled proofreader who gives preference to persons with relevant degrees in the humanities). The employer in such cases is therefore treating applicants as members of certain groups—those with and without certain formal qualifications—rather than as individuals. But it seems implausible that she is for that reason wronging persons, including prospective applicants who lack these qualifications, even if she does not interview a person who possesses not the formal qualifications but the skills for which she is ultimately looking. The wrongness of statistical discrimination is therefore arguably better accounted for in terms of whatever makes discrimination more generally wrong (when it is) (Eidelson, 2015; Lippert-Rasmussen, 2013). This is the issue to which we now turn.
What Is Morally Wrong with Discrimination?
The second central issue in the academic debate on discrimination concerns what makes wrongful discrimination morally wrong.
A range of classical accounts, such as that discrimination is wrong because it treats persons on the basis of properties for which they are not responsible (Singer, 1978), that it is wrong because it is arbitrary (Schauer, 2003; Frankfurt, 1997), or that it is wrong because it fails to treat persons according to merit (Miller, 1992), are now widely considered mistaken. The first is seen as such because there seem to be both impermissible cases of discrimination against groups defined by properties for which they are responsible (consider discrimination against religious converts, or disabled persons who caused their own injury by an activity such as reckless driving), permissible discrimination against groups defined by properties for which they are not responsible (consider refusing driver’s licenses to persons with congenital blindness), and because in many cases, it is the fact and effects of differential treatment, not lack of responsibility for the property that it targets, which is intuitively the crux of the problem (Heinrichs, 2007; Campbell, 1991). Similarly, discrimination need not be arbitrary—indeed, many cases of paradigmatic and intuitively wrongful discrimination are instrumentally rational, such as the employer who refuses to hire minorities because she knows that doing so will alienate her customers (Halldenius, 2005; Radcliffe Richards, 2000; Nickel, 1972, 1998; Wertheimer, 1983; Woodruff, 1976). And finally, we frequently do not require that persons treat others according to merit, particularly when doing so would clash with other moral concerns, which suggests that it cannot by itself explain what is involved in wrongful discrimination, and it can occasionally be discriminatory to treat persons according to merit (consider again so-called reaction qualifications, as in the example of the employer with biased customers) (Lippert-Rasmussen, 2009, 2013; Cavanagh, 2002)
Debate over the past decade has focused on four more promising accounts. First, wrongful discrimination involves the discriminator misestimating the moral status of the discriminatee, and second, wrongful discrimination involves the discriminator publicly misrepresenting the moral status of the discriminatee. These two accounts both identify disrespect as the explanation of the wrongness of discrimination, albeit the first is a valuational and the second an expressive variant. A third account holds that wrongful discrimination is unfair to the discriminatee, and a fourth that wrongful discrimination harms the discriminatee. We will review the accounts in this order, but before we do so, it is necessary to look briefly at how to identify the particular wrongness of discrimination.
Note also that for simplicity, only the descriptive conception of discrimination is employed throughout the following discussion, but as noted earlier, the issues are essentially the same if one adopts a moralized conception instead.
The Extrinsic, Intrinsic, and Uniqueness of the Wrongness of Discrimination
One complication in analyzing what makes discrimination morally wrong is how to delineate the particular wrongness that is attributable to discrimination. Paradigmatic cases of discrimination often involve actions that are intuitively wrong irrespective of discrimination. Consider, for example, a police officer who subjects minority citizens to humiliating and unjustified strip-searches, or a store clerk who insults and refuses to serve minority customers. Should the wrongness of humiliating and insulting in these cases be considered what makes the discrimination wrong, or the wrongness of the actions that also partially constitute discrimination? More generally, how should we separate the distinctive wrongness of discrimination from any other wrongness involved?
This separation is important for at least two reasons. First, by including the wrongness of the actions involved in discrimination in our analysis of the wrongness of discrimination, we risk double-counting reasons; and second, by doing so, we risk overlooking whatever is distinctively wrong with discrimination (Hellman, 2008). This suggests that we should distinguish between the intrinsic and extrinsic wrongness of discrimination, where extrinsic wrongness is any wrongness attached to a case of discrimination that stems not from the discrimination, but from the nondiscriminatory acts involved, and which would therefore remain even if the agent were to perform a similar action nondiscriminatorily. The intrinsic wrongness of discrimination, by contrast, is any wrongness attached to a case of discrimination that stems from the discriminatory treatment rather than any of the composite actions, and which would therefore disappear if the agent performed a similar action nondiscriminatorily.
Consider for illustration again the police officer who subjects minority citizens to humiliating and unjustified strip-searches. At least part of what is intuitively wrong with her actions stems, it seems, from the humiliating and unjustified strip-search itself, and that aspect would remain even if the police officer were to subject all citizens equally to humiliating and unjustified strip-searches. But arguably, there is something additionally and particularly wrong with subjecting only minority members to such treatment, which would be alleviated by indiscriminately treating all citizens thus.
A further question that follows on the heels of this point is whether there must be a unique wrong of discrimination. Much of the literature on the wrongness of discrimination implicitly assumes that there is such a unique wrong, but even if one accepts the distinction between intrinsic and extrinsic wrongness, it remains theoretically possible that there could be more than one distinctive, intrinsic wrong involved in cases of discrimination. Obviously, for this to be true, one must accept a background theory of normative pluralism; normative monists, such as consequentialists, will rule out the possibility simply on the grounds that all wrongness can be explained by reference to a single moral factor. Normative pluralists might prefer there to be a single wrong involved, since the presence of multiple wrongs complicates the picture and raises the possibility of conflicting reasons, which might require settling on a hierarchy of wrongs (Segall, 2012). Against this argument, one can object that that the theoretical convenience of a unique wrong provides no guarantee that there is not a multitude of wrongs involved in discrimination.
Having noted the issue, however, we shall set it aside and focus on unique accounts of wrongness when we turn to the discussion of the most prominent theories of the wrongness of discrimination in the next section.
Arguably the most influential contemporary account of the wrongness of discrimination is the idea that wrongful discrimination is discrimination that is based on the discriminator misestimating the moral worth of the discriminatee. In the seminal statement of the idea, Larry Alexander claims that: “[Nazi biases against Jews] were intrinsically morally wrong because Jews are clearly not of lesser moral worth than Aryans. When a person is judged incorrectly to be of lesser moral worth and is treated accordingly, that treatment is morally wrong regardless of the gravity of its effects. It represents a failure to show the moral respect due to the recipient, a failure which is by itself sufficient to be judged immoral” (Alexander, 1992, p. 159).
The idea will be intuitively appealing to many, is an extension of widely credited work in moral philosophy more broadly (Carter, 2011; Darwall, 1977, 2006; Frankfurt, 1997), fits our judgments in a wide range of cases, and can account for our misgivings about some otherwise tricky cases of discrimination in favor of a group. Consider Richard Arneson’s example of a bigot who is timely in paying back debts only to her Jewish friends because she believes Jews to be pushy and aggressive in economic matters (Arneson, 2006). While this is arguably discrimination in favor of Jews, since they enjoy the benefit of timely repayment compared to others, it is intuitively still a form of wrongful discrimination, and a form of discrimination that wrongs Jews, not those who are in one respect treated worse because their loans are not paid back in timely fashion (although that tardiness can still be wrong separately).
At least three versions of the account must be distinguished, although they can often be difficult to keep apart, and it is not always clear which version defenders of the account have in mind. In the first version, disrespect is tied to the mental state of the discriminator in that the discriminatee is represented as falsely having a comparatively or absolutely false moral status (Alexander, 1992; Glasgow, 2009). In the second, disrespect is tied to the action of the discriminator in that she acts in a way that would be permissible only if the discriminatee did in fact have a different moral status than she does (Eidelson, 2015; Glasgow, 2009). In the third, disrespect is tied to the culpable formation of and failure to revise a false representation of the discriminatee’s moral status (Arneson, 2006). While all versions have roughly the same arguments in their favor, they face different challenges; let us look at the challenges that each encounters in turn.
The mental state version of the valuational disrespect account is perhaps the most obvious version of the account, and the one that ties most naturally to more foundational work in moral philosophy. In this version of the account, the agent acts wrongly because she acts for the wrong reasons—specifically, she acts on the basis of a misestimation of the moral worth of the discriminatee. Whatever the broader plausibility of such theories, in the context of discrimination, the mental state version of the disrespect account faces both a minor and a major challenge. The minor challenge is that it is incapable of accounting for the wrongness of intuitively wrongful cases of discrimination when these do not involve the discriminator misestimating the moral worth of the discriminatee. It thus appears committed, implausibly, to the view that there is nothing intrinsically wrong with an action such as discriminating against a vulnerable minority for instrumental reasons—say, because to do so will increase the profits of one’s business—so long as one does not also misestimate the moral worth of minority members. Proponents of the account might counter that the wrongness of such discrimination can be explained by referring to extrinsic wrongs, but it is an open question whether proponents of the account can credibly draw the intrinsic/extrinsic distinction in a way that will solve the problem.
Even if this response were successful, the account would still face a more serious challenge, in that it seems to turn our standard intuitions about the way badness is tied to beliefs on its head. Normally, it would seem to most of us that if the beliefs of an agent influence the permissibility/wrongness of an action at all (as opposed to the Scanlonian view that it affects the blameworthiness of the agent; cf. Scanlon, 2008), an impermissible act must be less bad if the agent performs it while holding a mistaken belief that implies that what she is doing is in fact permissible. Consider a case presented by Lippert-Rasmussen:
The Speciesist Scientist. A pharmaceutical company discriminates against animals by inflicting horrible pain on hundreds or even thousands of them to provide a very small benefit to a small group of humans; e.g., the ability to buy and wear a new perfume without suffering a small risk of having a mild allergic reaction. Researcher R1 misestimates the moral worth of animals, and therefore falsely believes that her actions are morally permissible. Researcher R2 correctly estimates the moral worth of animals, and therefore correctly believes that her actions are morally impermissible. She pursues them regardless.
As Lippert-Rasmussen argues, our intuitions about this case pose a serious problem for the disrespect account of discrimination because all else being equal, the actions of researcher R2 are intuitively worse than the actions of researcher R1, and: “if one’s discriminatory activities are less bad when they are accompanied by an underestimate of the moral status of the discriminatee than they are when they are accompanied by a correct estimate of the discriminatee’s status, it follows that discrimination cannot be bad simply because it reflects an incorrect judgement of moral status” (Lippert-Rasmussen, 2006a, p. 183).
Consider next the act version of the valuational disrespect account. With this version, recall that the agent acts in a way that would be permissible only if the discriminatee did in fact have a different moral status than she does. This is disrespectful because in so doing, she “[fails] to adequately recognize autonomous, independent, sensitive, morally significant creatures” (Glasgow, 2009). This version of the account faces an uncomfortable dilemma between relying on the recognition and acting components of the account.
If the act version of the account relies on the recognition part, then the agent acts wrongly because her actions demonstrate that she fails to act for the right reasons—specifically, those generated by the moral status of the discriminatee, which she fails to properly take into account. Although the two are subtly different, the similarity means that this version runs into problems similar to those of the mental state version, in that it fails to explain the intuitive wrongness of a discriminator who discriminates in spite of fully recognizing the moral status of the discriminatee. It also implausibly entails that intuitively benign acts of discrimination—say, handing out green caramels to girls and yellow caramels to boys—are wrong for the same reasons and to the same extent as paradigmatic cases of wrongful discrimination, so long as the agent fails to properly recognize the moral standing of the persons she interacts with.
If the act version relies instead on the acting component, then it holds that the action is wrong because it does not fit the actual moral status of the discriminatee. Proponents might say that regardless of her mental state and deliberations, the agent fails to “respond to [the discriminatees] status as a bearer of interests with presumptively equal moral weight” (Eidelson, 2015), where “response” is to be understood not as recognizing the moral status of the person (that is the recognition element mentioned previously), but as conforming one’s action to how one ought to act in the light of that moral status. A benefit of this is that it allows agents to act permissibly without recognizing or deliberating in any particular way, so long as their actions are compatible with the moral status of the persons they treat one way or another. Assuming plausibly that the moral status of a patient affects what one ought to do through giving rise to reasons for action, this version of the disrespect account appears to hold, roughly, that an act (of discrimination) is wrong (when it is) because it is contrary to the reasons available to the agent generated by the discriminatee’s status as a moral patient. While this is a plausible view, it appears to lose the explanatory power of a distinctive account: all plausible moral theories subscribe to the purely formal notion that a wrongful act can be described as one that is in some sense inappropriate to the moral status of the patients (Lippert-Rasmussen, 2013). Even act-consequentialism can require respect in this sense since in an act-consequentialist scheme, the act that impartially maximizes the good will be one that properly takes account of the moral status of those involved and the reasons for action that this generates (Pettit, 1989).
Consider, finally, the culpable false belief version of the valuational disrespect account (Arneson, 2006). In this version, discrimination is wrong only when the discriminator misestimates the moral status of the discriminatee and is in some relevant sense responsible for arriving at and holding the false estimation. This will clearly be the case in many paradigmatic cases of invidious discrimination. Modern racists, for example, can hold false beliefs about the lower moral status of members of other races, in spite of clear, uncontroversial, and available reasons to the contrary. And it can arguably fend off cases like the Speciesist Scientist, where it will hold that the actions of researcher R1 are either equally wrong, if her beliefs are nonculpably derived, or worse than the actions of researcher R2, if she is responsible for holding her mistaken beliefs about the moral status of animals, a conclusion that may not now seem as counterintuitive as before.
Nonetheless, this version too faces strong challenges. First, it is unclear how we should separate the wrongness of culpably forming and maintaining false estimates of moral status from the wrongness of discrimination. Proponents of the account presumably will want to hold that the former is independently wrong for, if not, it is difficult to see how they could suddenly make a difference when involved in acts of discrimination, but to do so invites charges of double-counting when the culpably false beliefs are simultaneously supposed to explain the wrongness of discrimination. Second, the account faces serious difficulties in explaining the moral difference that it must insist exists between otherwise identical cases of intuitively wrongful discrimination that differ only with respect to the existence of culpably false beliefs.
Suppose, for example, that two Nazi bureaucrats both discriminate against Jews, but that only one holds the mistaken belief that Jews have lesser moral status, and that she is culpable for doing so. Suppose further, however, that the culpable false belief plays no causal part in explaining his actions (e.g., because both bureaucrats act solely on the orders of their superiors and are motivated only to further their bureaucratic careers). It seems strongly counterintuitive to hold, as this version of the account must, that one bureaucrat is engaged in wrongful discrimination in a way that the other is not. Nor is it the function of the false belief that explains the wrongness, since we can easily imagine similar cases where two agents intuitively discriminate equally while culpably holding identical false beliefs, although the false belief have a causal effect for only one of them (Lippert-Rasmussen, 2013). Jointly, the two challenges suggest that the critic will say that we would do better to hold agents accountable for their biases and false beliefs independently, and explain the wrongness of discrimination some other way.
A second account, related to the previous one, which has attracted attention in recent years is the idea that wrongful discrimination involves not necessarily the discriminator misestimating the moral worth of the discriminatees, but rather the discriminator expressing a false estimation of their moral worth (Shin, 2009; Hellman, 2008). This expression need not be a verbal statement, and it need not accurately represent the discriminator’s beliefs. It suffices that the discriminator treats persons as if they had lower moral worth, in the sense that the act is best interpreted as reflecting this belief.
Importantly, therefore, the account does not claim that whether an action expresses disrespect, or whether doing so is morally bad, depends in any way on the effect of the action. It is opposed, for example, to the idea that the wrongness of discrimination is explained by the fact that expressions of disrespect can cause humiliation and loss of self-esteem or reinforce biases and negative stereotypes. Proponents of the account need not deny that expressions of disrespect can have such effects, nor that expressing disrespect is in one respect wrong when it does have such effects. But they insist that these considerations fail to capture all that is wrong with either discrimination or expressions of disrespect, since otherwise the expressive disrespect account would collapse into a variant of the harm account. And they might further claim that given the contingency of the effects—some expressions of disrespect will not have any harmful effects—causing these effects are at most extrinsic wrongs of discrimination, while the expression of disrespect is what is intrinsically wrong with discrimination (Hellman, 2008).
The expressive disrespect account has the advantage that the account arguably does not generate strongly counterintuitive answers in cases like the Speciesist Scientist. Regardless of what beliefs the scientists hold, they might be held to equally demean animals by equally treating them as if they have lower moral worth than in fact they do.
However, this approach faces challenges of its own, the first of which is that it is not clear why it should be wrong, in and of itself, to express a false belief. After all, most persons express false beliefs at least semifrequently, and we do not ordinarily consider it to constitute a moral wrong if the speaker’s mental state is not itself problematic (e.g., wilfully malicious or deceptive) and the expression does no harm (e.g., humiliates or deceives). The expressive disrespect account must insist that discrimination that expresses disrespect is morally wrong even when these two conditions occur, since they represent alternative accounts (the valuational disrespect and harm accounts, respectively), but it is not easy to see how it can explain the alleged difference between intuitively innocent statements of false beliefs and the wrongness of expressing this particular false belief (Eidelson, 2015).
Second, the fact that the account limits wrongful discrimination to situations where differential treatment objectively demeans means that it at once fails to cover some cases of intuitively wrongful discrimination and that it allows an implausible remedy. The account fails to cover some cases because many forms of intuitively problematic discrimination do not appear to involve expressing that discriminatees have lower moral status. Conventional labor market discrimination against women, for example, is clearly demeaning in some cases, but it seems plausible that there are or could be cases where it need not be based on or express any beliefs about the worth of women either as persons or employees, but rather reflect accurate statistical beliefs about the probability of prospective employees taking parental leave. In such cases, the expressive account is counterintuitively incapable of condemning the discrimination. The account also allows an implausible remedy since for any case of wrongful discrimination, the expressive disrespect account allows us to avoid wronging the discriminatee simply by changing the cultural environment in such a way that the treatment no longer expresses disrespect.
Suppose, for example, that a state disenfranchised women, but could somehow (through advanced brainwashing, perhaps) create a cultural environment where this was not seen as in any way demeaning. On the expressive disrespect account, this disenfranchisement would then not be morally wrong qua discrimination (although it could be condemned for reasons unrelated to discrimination), which seems counterintuitive to say the least. A different way of framing the problem is to see that when choosing between the two solutions to the problem of discriminating by disenfranchising women, considerations of discrimination alone would give no reason to prefer giving women the right to vote over changing the cultural environment to make their disenfranchisement hold no disrespectful meaning. While the dramatic cultural and psychological changes necessary to avoid discrimination being demeaning may in practice often be impossible, in many cases of intuitively problematic discrimination, it is implausible that we could make any substantial difference to the wrongness of the action simply through altering the contextually dependent meaning; as such, the implication counts against the expressive disrespect account nonetheless (Lippert-Rasmussen, 2013).
Another recent account holds that discrimination is wrong (when it is) because it reduces equality of opportunity, and this is unfair (Segall, 2012). Opportunities should here be understood in the wide sense that contemporary egalitarians, particularly in the luck-egalitarian tradition, have taken the term such that they concern, for example, individuals’ whole-life opportunities for welfare. Importantly, some cases that might seem initially troubling for the account (such as racial segregation that offers different but apparently equal opportunities) can thus be accommodated by adopting the proper egalitarian currency; i.e., because such segregation may, as previously noted, deny discriminatees the opportunity for respect or self-respect. Notably, however, as an egalitarian account concerned with fairness, it is the inequality of opportunities that explains the badness of discrimination, not the fact that some person’s opportunities are bad or worse than they could or should have been. The latter views are, rather, versions of the harm account, to which we turn next.
The account will appeal to many, both because it draws on ideas that are prominent in contemporary political theory and because it fits with our intuitions in a range of paradigmatic cases. Thus, it is obvious that intuitively wrongful discrimination, such as denying minorities equal access to positions in the labor market or setting lower standards of guilt for and imposing higher sentences on minorities in the criminal justice system, deprives discriminatees of opportunities that privileged groups enjoy, and it is natural to describe the wrongness of such treatment in terms of it being unfair.
However, because it is an egalitarian theory, the account also becomes susceptible to familiar criticisms. Thus, as an egalitarian account holding that increased inequality of opportunity is morally bad (because unfair), it appears that the account must also be committed to the view that decreased inequality of opportunity is necessarily, in at least one respect, good (because it is fairer). This arguably makes the account susceptible to a modified version of Derek Parfit’s levelling-down objection: It is intuitively implausible that there is any respect in which simply decreasing the opportunities of those with more or better opportunities is morally good if doing so in no way benefits those with worse opportunities, even though to do so increases equality of opportunity (Parfit, 2002; Cavanagh, 2002). While this challenge is unlikely to sway devoted egalitarians, it is likely to limit the appeal of the unfairness account for those not already persuaded by egalitarianism.
Proponents of the unfairness account can avoid the implication by denying the commitment to the moral goodness of increased equality of opportunity suggested previously, but doing so requires that they can provide an explanation for the discrepancy between the moral badness of increasing inequality of opportunity and the moral neutrality of decreasing it. That is, if the former is unfair, and morally bad for that reason, why does the latter not promote fairness?
Furthermore, even the narrow version of the account, which holds only that increasing inequality of opportunity is unfair, faces the problem that it is unable to explain the intuitive wrongness of cases that do not involve increased inequality of opportunity. Consider as an example racial or ethnic discrimination against a better-off minority, such as Burmese Gurkhas, Syrian Alawites, or Malaysian Han Chinese, in a way that does not create better opportunities for the worse-off racial majority. Since the minority is better off, discriminating against them can promote equality of opportunity, and if it does that, the discrimination cannot, on the current account, be morally wrong. This will strike many as strongly counterintuitive.
Two responses to this explanatory deficit might be to rely either on the intrinsic/extrinsic wrongness distinction and accept that such cases are wrong, but insist that they are so for reasons extrinsic to the discrimination involved, or to adopt a pluralist account of the wrongness of discrimination. However, such responses remain unexplored in the literature, and it is therefore presently unclear whether they can be argued persuasively.
Finally, perhaps the simplest account of the wrongness of discrimination may be the idea that discrimination is wrong (when it is) simply because it harms the discriminatees (Moreau, 2010; Lippert-Rasmussen, 2006a). The notion that causing harm is in some sense morally bad is perhaps the least controversial view in moral philosophy—it is all but universally accepted by deontologists and consequentialists alike—so the task for the proponent of the harm account is not to defend this point, but rather to show that it applies to discrimination, that is, to explain how and when discrimination causes harm in a way that meshes with our intuitions about wrongful discrimination. A modest version of the argument for the account needs to show only that discrimination can be harmful by virtue of being discrimination, and that, since causing harm is pro tanto wrong, harm is therefore at least part of the explanation of what is wrong with discrimination. A more ambitious monist version will need to further argue against alternative accounts of wrongness, as well as to explain away the intuitive wrongness of certain harmless examples, to build the case that discrimination is wrong only when it causes harm.
Showing that discrimination causes harm involves answering two questions: Harming a person involves making her worse off, but worse off than what, and worse off in what respects? These are the baseline and currency questions, respectively, and both are subject to controversy in the broader debate on how to conceptualize causing harm. While this might appear to be a flaw, friends of the harm account can argue that this open-ended nature should be considered an advantage since it allows thinkers who disagree about the baseline and currency questions to slot in their preferred views. Regardless, it is worth sketching how the two questions might be answered, and I shall present two answers to each in the following, before discussing a few prominent objections to the overall account.
The most obvious answer to the baseline question may be the counterfactual view that a person is harmed by a discriminatory act if she is made worse off than she would have otherwise been. This suggestion fits with what is arguably the most plausible general theory of harm (Holtug, 2002). However, this suggestion faces the objection that it is forced to say that acts of discrimination are not wrong if they turn out to eventually and inadvertently leave the discriminatee better off than they would otherwise have been. Hence, acts of Nazi discrimination against Jews that caused these Jews to emigrate prior to the Holocaust, and thereby prevented their suffering and likely death, were not harmful even if they involved violence, humiliation, and the deprivation of important goods and opportunities, and therefore not morally wrong. But many would want to hold that such treatment constitutes a paradigmatic case of wrongful discrimination, and that at least part of the reason why it was wrong was that it harmed Jews by making them worse off in certain respects, despite their being made totally better off compared with the counterfactual baseline.
This can motivate the adoption of a moralized baseline. One such suggested baseline is how well off the discriminatee would have been had she not been subjected to discrimination in this, or any future instance (Lippert-Rasmussen, 2013). But it is possible to imagine more comprehensively moralized baselines, such as the idea that the discriminatee is harmed if she is made worse off than she ought to be, all things considered (Lippert-Rasmussen, 2007c)
Consider next the currency question. A recent suggestion here is that discrimination is wrong (when it is) because it deprives discriminatees of deliberative freedoms that they ought to enjoy—that is, the freedom to make decisions in certain central spheres of their lives without having properties of their personality impose costs on some choices (Moreau, 2010). The idea is plausible insofar as that it often seems intuitively wrong because it is harmful for others to impose costs on our choices, such as where to live or what career to pursue, in extremis by simply denying us the possibility of living in certain places or pursuing certain careers. Our lives are to some degree impoverished thereby, even if we would not have chosen to live in that place or pursue that career had there been no cost attached, particularly if we are aware of these costs and the social conditions from which they sprang.
However, there are several reasons to be skeptical that this is ultimately the right currency. First, it implausibly privileges opportunity costs. Discrimination that simply imposes blanket costs on all of a person’s choices does not thereby change her deliberative situation, and as such cannot restrict or otherwise alter her deliberative freedom. And yet, imposing such blanket costs would seem, if anything, to be more harmful and morally worse than imposing costs on only some choices. And second, it seems clear that we can imagine situations in which persons are intuitively better off while suffering restrictions on their deliberative freedom, which suggests that they are not harmed by such restrictions (Lippert-Rasmussen, 2013).
A different suggestion powerfully argued in the recent literature is to adopt a desert-adjusted prioritarian welfare currency. This takes the well-being of individuals as its fundamental currency but suggests that the moral value of individual welfare levels should be adjusted both by a concave value function that gives welfare-diminishing marginal value the further above the neutral level a person is (prioritarianism), and by considerations of how well off the person deserves to be as a result of her past actions (Lippert-Rasmussen, 2007c, 2013). Although this approach is most commonly identified with consequentialism (explicitly so by Lippert-Rasmussen), there is, strictly speaking, no reason why it could not be adopted by nonconsequentialists as the answer to the currency question (e.g., in fleshing out a deontological constraint against causing harm).
Unsurprisingly, the harm account faces objections of its own. Perhaps the most prominent are that the account is trivial, that it fails to identify what is intrinsically wrong with discrimination, and that it is incapable of accounting for the wrongness of certain intuitively wrongful cases of discrimination.
The point of the first objection is that the account seems to simply apply a general moral principle, and by so doing fails to capture the special moral character of discrimination. Here, proponents of the account can respond that any difficulty is not unique—the harm account is no more the application of a generic and broad moral principle to the context of discrimination than competing accounts. And while it would perhaps be a problem if it picked out nothing particular or interesting about cases of discrimination, it does focus attention on particular features of these cases, and by doing so draws a significant distinction between permissible and impermissible cases that may at times cut across our prereflective evaluations.
Recall next that some theorists of discrimination hold that we should distinguish between intrinsic and extrinsic wrongs. These might point out that the most obvious harms in paradigmatic cases of discrimination are extrinsic. A bigoted educational institution that refuses to accept persons of a minority religion plausibly harms these persons, but the harm that they suffer through lack of access to education would largely remain intact if educational institutions also denied places to atheists and persons of other religions. These harms, therefore, are extrinsic to discrimination, and the harm-based account cannot rely on them to explain the wrongness of discrimination itself. Proponents of the harm account can respond either by challenging the intrinsic/extrinsic distinction, or by pointing out that there do seem to be particular harms that arise from the fact of differential treatment—notably loss of positional goods and psychological consequences such as stereotyping, stigmatization, and loss of self-respect. Thus, in the example of the bigoted educational institution given in this discussion, an education is presumably valuable both for a range of absolute goods that it provides and because it is a positional good with respect to employment. And while losing the former can be counted as an extrinsic wrong, loss of the latter is the result of differential treatment and would disappear were institutions to deny education to all prospective students. Furthermore, beyond the absolute and positional goods, the differential treatment of an openly discriminatory policy is likely to stigmatize and humiliate those discriminated against, both of which will count as harms on any plausible version of the harm account.
Last but not least, some will object that there are intuitively wrongful cases of discrimination that do not involve causing harm, such as denying equal civil liberties to men and women, and by so doing averting great welfare losses. Critics can even limit their claim to the modest observation that there seems to be something wrong with such cases of discrimination, rather than holding that they are all-things-considered wrong, since the harm account is incapable of admitting even this much on pain of contradiction. This is the most forceful criticism, but it remains an option for proponents to insist that such cases are in fact not morally wrong, preferably by offering error theories of why we might initially be inclined to think so, and to point out that in this respect too, the account fares no worse than competing accounts since, as we have previously seen, these were similarly incapable of meshing perfectly with intuitions about cases of wrongful discrimination.
As the preceding overview has hopefully made clear, the conceptualization and moral analysis of discrimination has become a burgeoning theoretical field, with a number of open problems and a rapidly developing literature. This article has suggested some of the main features of the field by setting out a definition of paradigmatic discrimination as disadvantageous differential treatment of two groups that is in some respect caused by the properties that distinguish the groups. In addition, it explored arguments for and against restricting discrimination to particular groups, as well as for and against a moralized definition. It has illustrated the complexities involved in drawing the distinction between equal and differential treatment, as well as clarifying important variations on discrimination, including indirect discrimination and statistical discrimination. Finally, it has presented and discussed the four arguably most prominent competing theories of what is morally wrong with discrimination (when it is wrong), illustrating that each faces important challenges in simultaneously providing a persuasive theoretical account and matching our intuitions.
Two important elements of the debate that this article has not addressed are the role and proper shape of discrimination law. This is not because these topics are irrelevant or theoretically impoverished; on the contrary, the literature on discrimination law is rich, complex, and rapidly evolving, and there are substantial overlaps between the conceptual and moral analysis of discrimination on the one hand and the legal analysis on the other (Khaitan, 2015; Segev, 2014; Hellman & Moreau, 2013; Fredman, 2011; O’Cinneide, 2011; Bamforth, Malik, & O’Cinneide, 2008; Connolly, 2006; Bagenstos, 2003; Réaume, 2003; Gardner, 1989, 1996; Rutherglen, 1995; Campbell, 1991; McCrudden, 1982; Wasserstrom, 1977; Brest, 1976)
There is not, however, any simple relation between conclusions in one field and in the other. Thus, one cannot assume, for example, that the best conceptual understanding of discrimination should be adopted as the legal definition of discrimination, or that acts of discrimination that are morally wrong ought for that reason to be prohibited. Nor can one, obviously, do the opposite and (at least in any direct and unqualified way) derive conclusions about how to conceptualize discrimination or explain what is morally wrong with it by referring to existing discrimination legislation. Discrimination law, in short, is a related but distinct topic within criminalization theory, or legisprudence more broadly, which merits separate scrutiny.
I would like to acknowledge Thomas Søbirk Petersen, Kasper Lippert-Rasmussen, Sune Lægaard, Jesper Ryberg, and two anonymous reviewers for their help and guidance in writing this chapter.
- Alexander, L. (1992). What makes wrongful discrimination wrong? Biases, preferences, stereotypes, and proxies. University of Pennsylvania Law Review, 141, 149–219.
- Altman, A. (2015). Discrimination. Stanford encyclopedia of philosophy. In E. N. Zalta (Ed.).
- Applbaum, A. I. (1996). Racial generalization, police discretion, and Bayesian contractualism. In J. Kleinig (Ed.), Handled with discretion (pp. 145–157). Lanham, MD: Rowman, & Littlefield Publishers.
- Arneson, R. J. (2006). What is wrongful discrimination? San Diego Law Review, 43, 775–808.
- Bagenstos, S. R. (2003). Rational discrimination, accomodation, and the politics of (disability) civil rights. Virginia Law Review, 89, 825–923.
- Bamforth, N., Malik, M., & O’Cinneide, C. (2008). Discrimination law: Theory and context, text and materials. Suffolk, U.K.: Sweet & Maxwell.
- Brest, P. (1976). Foreword: In defense of the antidiscrimination principle. Harvard Law Review, 90(1), 1–57.
- Campbell, T. D. (1991). Unlawful discrimination. In W. Sadurski (Ed.), Ethical dimensions of legal theory (pp. 153–169). Amsterdam: Rodopi.
- Carter, I. (2011). Respect and the basis of equality. Ethics, 121(3), 538–571.
- Cavanagh, M. (2002). Against equality of opportunity. Oxford: Oxford University Press.
- Connolly, M. (2006). Discrimination law. Suffolk, U.K.: Sweet & Maxwell.
- Council of Europe (1950). Convention for the protection of human rights and fundamental freedoms.
- Council of European Union (2000). Implementing the principle of equal treatment between persons irrespective of racial or ethnic origin. 2000/43/EC.
- Darwall, S. L. (1977). Two kinds of respect. Ethics, 88(1), 36–49.
- Darwall, S. L. (2006). The second-person standpoint: Morality, respect, and accountability. Cambridge, MA: Harvard University Press.
- Doyle, O. (2007). Direct discrimination, indirect discrimination, and autonomy. Oxford Journal of Legal Studies, 27(3), 537–553.
- Dworkin, R. (2005). Taking rights seriously. London: Gerald Duckworth & Co. Ltd.
- Edmonds, D. (2006). Caste wars—A philosophy of discrimination. Abingdon, U.K.: Routledge.
- Eidelson, B. (2015). Discrimination and disrespect. Oxford: Oxford University Press.
- Ezorsky, G. (2001). Discrimination. In C. B. Becker & L. C. Becker (Eds.), Encyclopedia of ethics (pp. 413–416). New York: Routledge.
- Frankfurt, H. (1997). Equality and respect. Social Research, 64(1), 3–15.
- Fredman, S. (2011). Discrimination law. Oxford: Oxford University Press.
- Gardner, J. (1989). Liberals and unlawful discrimination. Oxford Journal of Legal Studies, 9, 1–22.
- Gardner, J. (1996). Discrimination as injustice. Oxford Journal of Legal Studies, 16(3), 353–367.
- Glasgow, J. (2009). Racism as disrespect. Ethics, 120, 64–93.
- Gross, S. R., & Livingston, D. (2002). Racial profiling under attack. Columbia Law Review, 102(5), 1413–1438.
- Halldenius, L. (2005). Dissecting discrimination. Cambridge Quarterly of Healthcare Ethics, 14, 455–463.
- Harcourt, B. (2004). Rethinking racial profiling: A critique of the economics, civil liberties, and constitutional literature, and of criminal profiling more generally. University of Chicago Law Review, 71(4), 1275–1381.
- Heinrichs, B. (2007). What is discrimination and when is it morally wrong? Jahrbuch für Wissenschaft und Ethik, 12, 97–114.
- Hellman, D. (2008). When is discrimination wrong? Cambridge, MA: Harvard University Press.
- Hellman, D., & Moreau, S. (Eds.). (2013). Philosophical foundations of discrimination law. Oxford: Oxford University Press.
- Holtug, N. (2002). The harm principle. Ethical Theory and Moral Practice, 5(4), 357–389.
- Horta, O. (2010). Discrimination in terms of moral exclusion. Theoria, 76(4), 314–332.
- Horta, O. (2015). Does discrimination require disadvantage? Moral Philosophy and Politics, 2, 277.
- Khaitan, T. (2015). A theory of discrimination law. Oxford: Oxford University Press.
- Lever, A. (2005). Why racial profiling is hard to justify: A response to Risse and Zeckhauser. Philosophy & Public Affairs, 33(1), 94–110.
- Lever, A. (2007). What’s wrong with racial profiling? Another look at the problem. Criminal Justice Ethics, 26(1), 20–28.
- Lever, A. (2010). Treating people as equals: Ethical objections to racial profiling and the composition of juries. Journal of Ethics, 15(1), 61–78.
- Lippert-Rasmussen, K. (2006a). The badness of discrimination. Ethical Theory and Moral Practice, 9, 167–185.
- Lippert-Rasmussen, K. (2006b). Racial profiling versus community. Journal of Applied Philosophy, 23(2), 191–205.
- Lippert-Rasmussen, K. (2007a). Discrimination: What is it and what makes it morally wrong? In J. Ryberg, C. Wolf, & T. S. Petersen (Eds.), New Waves in Applied Ethics (pp. 51–72). Chippenham and Eastbourne, U.K.: Palgrave Macmillan.
- Lippert-Rasmussen, K. (2007b). Nothing personal: On statistical discrimination. Journal of Political Philosophy, 15, 385–403.
- Lippert-Rasmussen, K. (2007c). Private discrimination: A prioritarian desert-accommodating account. San Diego Law Review, 43, 817–856.
- Lippert-Rasmussen, K. (2008). Discrimination and the aim of proportional representation. Politics, Philosophy, & Economics, 7, 159–182.
- Lippert-Rasmussen, K. (2009). Reaction qualifications revisited. Social Theory and Practice, 35(3), 413–439.
- Lippert-Rasmussen, K. (2011). “We are all different”: Statistical discrimination and the right to be treated as an individual. Journal of Ethics, 15(1), 47–59.
- Lippert-Rasmussen, K. (2013). Born free and equal? Oxford: Oxford University Press.
- McCrudden, C. (1982). Institutional discrimination. Oxford Journal of Legal Studies, 2(3), 303–367.
- Miller, D. (1992). Deserving jobs. Philosophical Quarterly, 42, 161–181.
- Moreau, S. (2010). What is discrimination? Philosophy & Public Affairs, 38(2), 143–179.
- Nickel, J. W. (1972). Discrimination and morally relevant characteristics. Analysis, 32(4), 113–114.
- Nickel, J. W. (1998). Discrimination. In E. Craig (Ed.), Routledge encyclopedia of philosophy. London: Routledge.
- O’Cinneide, C. (2011). The uncertain foundations of contemporary anti-discrimination law. International Journal of Discrimination and the Law, 11(1–2), 7–28.
- Parfit, D. (2002). Equality or priority. In M. Clayton & A. Williams (Eds.), The ideal of equality (pp. 81–125). Basingstoke, U.K.: Palgrave Macmillan.
- Pettit, P. (1989). Consequentialism and respect for persons. Ethics, 100(1), 116–126.
- Radcliffe Richards, J. (2000). Practical reason and moral certainty—The case of discrimination. In E. Ullmann-Margalit (Ed.), Reasoning practically (pp. 151–163). Oxford: Oxford University Press.
- Réaume, D. G. (2003). Discrimination and dignity. Louisiana Law Review, 63(3), 645–695.
- Risse, M., & Zeckhauser, R. (2004). Racial profiling. Philosophy & Public Affairs, 32(2), 131–170.
- Rutherglen, G. (1995). Discrimination and its discontents. Virginia Law Review, 81, 117–147.
- Scanlon, T. (2008). Moral dimensions: Permissibility, meaning, and blame. Cambridge, MA: Belknap.
- Schauer, F. (2003). Profiles, probabilities, and stereotypes. Cambridge, MA: Belknap.
- Segall, S. (2012). What’s so bad about discrimination? Utilitas, 24(1), 82–100.
- Segev, R. (2014). Making sense of discrimination. Ratio Juris, 27(1), 47–78.
- Shin, P. (2009). The substantive principle of equal treatment. Legal Theory, 15(2), 149–172.
- Singer, P. (1978). Is racial discrimination arbitrary? Philosophia, 8, 185–203.
- Thomsen, F. K. (2011). The art of the unseen—Three challenges for racial profiling. Journal of Ethics, 15(1), 89–117.
- Thomsen, F. K. (2013). But some groups are more equal than others—A critical review of the group criterion in the concept of discrimination. Social Theory and Practice, 39(1), 120–146.
- Thomsen, F. K. (2015). Stealing bread and sleeping beneath bridges. Moral Philosophy and Politics, 2(2), 299–327.
- Wasserman, D. (1996). Racial generalizations and police discretion. In J. Kleinig (Ed.), Handled with discretion (pp. 115–130). Lanham, MD: Rowman & Littlefield Publishers.
- Wasserman, D. (1998). Discrimination, concept of. In R. Chadwick (Ed.), Encyclopedia of applied ethics (Vol. 1, pp. 805–814). London: Academic Press.
- Wasserstrom, R. A. (1977). Racism, sexism, and preferential treatment: An approach to the topics. UCLA Law Review, 24, 581–622.
- Wertheimer, A. (1983). Jobs, qualifications and preferences. Ethics, 94, 99–112.
- Woodruff, P. (1976). What’s wrong with discrimination? Analysis, 36(3), 158–160.