Discrimination Revised: Reviewing the Relationship between Social Groups, Disparate Treatment, and Disparate Impact

1 Introduction

It is typically thought that immorally discriminating against someone involves treating them disadvantageously because of their sex, race, religion, etc. (e.g. their group membership), while treating someone else of another sex, race, religion, etc. well in comparison. One form of immoral discrimination involves intentionally disadvantaging someone because they belong to some social group; call this disparate treatment. Another form of immoral discrimination involves accidentally yet substantially disadvantaging, or tending to disadvantage, some social group according to a group-neutral criterion specifying certain behaviors or vernaculars, particular kinds of work experience, or reading abilities, to name a few; call this disparate impact. ^[1] There is some debate regarding disparate impact’s status as a form of immoral discrimination. Authors such as Richard Arneson, Sophia Moreau, and Kasper Lippert-Rasmussen have argued that disparate impact can be a form of immoral discrimination. ^[2] On the other hand, there are authors who dispute its status as a form of discrimination or question its political usefulness: Matthew Cavanagh, Iris Marion Young, and Michael Selmi, to name a few. ^[3] However, both sides agree that disparate treatment and disparate impact are significantly different from one another. For instance, a disparate treatment discriminator intentionally treats certain individuals differently just because of their group membership. On the other hand, a disparate impact discriminator only treats certain individuals differently because of non-group traits they have, and not at all because of their group membership. These seem like fairly important differences, hence why disparate treatment and disparate impact are regarded as substantially different from one another.

Although I think disparate impact can be a form of immoral discrimination, I believe that the standard analysis of its relationship with disparate treatment has fallen short in two ways: significant similarities between them have gone unnoticed, and group identity as a social phenomenon has been given too limited a role in understanding either action. For instance, although it is commonly said that disparate treatment discriminators are primarily motivated by the victim’s group membership, this is not the full story. Oftentimes it is not just the victim’s group membership they care about: they are also consciously motivated by some significant, non-group trait that is believed to be inextricably associated with the group (e.g. Muslims are violent, homosexuals are immoral, Jews are greedy). If that’s the case, then disparate treatment discriminators intentionally treat someone differently because of some non-group trait, much like disparate impact discriminators. In this paper, I argue that disparate treatment and disparate impact are structurally similar to one another, and present a definition of discrimination that captures their structural similarities while conserving our basic intuitions about what cases count as discrimination.

In Section II, I provide an overview and analysis of what others have said about the differences between disparate treatment and disparate impact. It is said that they differ regarding how the discriminator deliberates about her actions. It is also said that they are about group membership differently. In virtue of these two facts, they are thought to be dissimilar at a fundamental level to a significant degree, and are therefore significantly different actions (i.e. structurally dissimilar). In Section III, I’ll analyze two different cases of discrimination in order to conclude that disparate treatment and disparate impact actually have similar deliberative processes. ^[4] In Section IV, I’ll analyze the relationship between one’s group membership and possessing certain socially significant traits. This analysis allows us to conclude that disparate treatment and disparate impact are about group membership in similar ways. Therefore, contrary to the current discourse, Sections III and IV allow us to conclude that disparate treatment and disparate impact are structurally similar. To clarify their structural similarities, I provide a definition of disparate treatment and disparate impact in Section V based on the information provided in Sections III and IV. In Section VI, I show that understanding disparate treatment and disparate impact as being structurally similar is compatible with our basic intuitions about what is and is not discriminatory. In Section VII, I summarize my argument and reflect on the disparate treatment/disparate impact distinction.

2 The current discourse

There are two elements which compose the structure of either disparate treatment or disparate impact. First, there is their deliberative process, or how the discriminating agent makes her decisions (i.e. what factors she takes into account). Second, there is how they are about group membership. To say that discrimination is about group membership means the victim’s group identity is relevant to the fact that the resulting disparities are group-oriented; the fact that the person is part of some group helps explain why she was disadvantaged. In this section, I give a broad overview and analysis of how Kasper Lippert-Rasmussen, Sophia Moreau, and Andrew Altman have characterized these elements in the context of disparate treatment and disparate impact.

Let’s start with the deliberative differences between disparate treatment and disparate impact. The most obvious way to characterize the deliberations of a disparate treatment discriminator is to say that she uses the victim’s group membership as a reason to engage in disparaging treatment of the victim. Rasmussen makes this idea a necessary condition for disparate treatment: such discrimination can only happen if the agent treats someone disadvantageously because they supposedly belong to some socially salient group. ^[5] Andrew Altman also captures this notion in a very straightforward manner: “Without the intent to disadvantage persons based on their [group membership], there is no [disparate treatment]”. ^[6] Sophia Moreau captures the notion of what it means to treat someone differently “based on their group membership” with her example of a discriminating landlord. This imaginary landlord refuses “to lease an apartment to someone because he is of a particular culture”. ^[7] As she points out, “because of” refers to the fact that the victim’s belonging to a particular culture is the discriminator’s reason for refusing to lease him a room. ^[8] In each case, we can see that disparate treatment is thought to involve targeting some person for differential treatment just in virtue of their group membership, and that this fact is distinctive of disparate treatment.

Disparate impact is distinguishable from disparate treatment because the discriminators are not concerned with the victim’s group membership. As Atlman explains, the primary difference between disparate treatment and disparate impact is that the latter sort of discriminators “do not aim to disadvantage persons for being members of a certain group”. ^[9] Alternatively, Rasmussen states that disparate impact is characterized by the fact that the discriminator does not bear any bias against the relevant social group. ^[10] Another way of putting the issue is to state that a disparate impact discriminator cares about the victim’s non-group/neutral traits. ^[11] Moreau provides an example that illustrates this point well. Suppose there’s a public pool, and the individuals who run the pool have restricted access to only those who live in the immediate neighborhood. The residents of the immediate neighborhood are well-off white people, and the people beyond the neighborhood tend to be poorer black people. ^[12] Although we cannot say that the black people were excluded just in virtue of their race, we can say that they were excluded “because of [their] possession of some extraneous [i.e. irrelevant] trait” (viz. where they lived). ^[13] Thus, in contrast to disparate treatment, we can say that disparate impact involves targeting someone for differential treatment on the basis of some irrelevant, non-group trait that they possess.

So much for the deliberative differences between disparate treatment and disparate impact. Now we can move on to discussing how they are said to be about group membership. Each of the authors mentioned above would agree that disparate treatment is about the victim’s group membership because the discriminator purposefully targeted the relevant group. As for disparate impact, one might say that the victim’s group membership is thought to be indirectly relevant to the disparities. Moreau seems to suggest something like this in the context of her public pool scenario. The fact that the black people lived in poorer neighborhoods can be explained by the lack of quality “education and opportunities, which [would] only partially [be] the result of other people’s assumptions about these individuals’ race”. ^[14] Altman suggests something similar when discussing the Griggs vs. Duke Power court case. The fact that the victims were black, combined with the poor state of education for black people in North Carolina at the time, explains why Duke Power’s black employees were consistently disadvantaged by the company’s promotional practices. ^[15] Said otherwise, the victims’ “identity as blacks was a key factor” in explaining their disparities. ^[16] The idea seems to be that the victim’s group membership is relevant to the resultant disparities in disparate impact, insofar as pre-existing social conditions typically cause members of the victim’s group to possess the neutral traits specified by the disparaging policy. In other words, group disparities obtain just because certain social conditions surrounding the relevant group cause the victims to possess certain traits.

The current discourse leads to three conclusions. First, disparate treatment and disparate impact have different deliberative processes: the former is concerned with the victim’s group membership and the latter is concerned with the victim’s non-group traits. Second, disparate treatment and disparate impact are about group membership differently. In disparate treatment, the victim’s group membership is relevant to the outcomes because of the discriminator’s intentions. In disparate impact, the victim’s group membership is relevant to the outcomes because of certain social conditions surrounding the victim’s group. In virtue of these two facts, and because their deliberative processes and relationships with group membership constitute their structures as actions, the current discourse leads to the conclusion that disparate treatment and disparate impact are structurally dissimilar (i.e. fundamentally different to a significant degree).

3 Overlooked similarities

One might be convinced that the current discourse adequately characterizes disparate treatment and disparate impact as being structurally dissimilar. However, there are good reasons to doubt that conclusion. In order to demonstrate that disparate treatment and disparate impact are structurally similar, let’s start by analyzing two instances of discrimination. Consider a popular case of disparate treatment: Dred Scott vs. Sandford. ^[17] The basic details of the case are as follows: Dred Scott was a slave owned by John Emerson. Scott filed a lawsuit to obtain his freedom, both in the state courts and in the federal courts, but was unsuccessful. The federal courts concluded that Dred Scott had to remain a slave because he was a “negro.” However, Justice Robert Taney also explicitly mentioned that black people were considered to be morally inferior to white people, and therefore had no rights or freedoms to cling to. ^[18]

Before analyzing Dred Scott’s case, let’s consider an instance of disparate impact: Nuala Crilly vs. Ballymagroarty Hazelbank Community Partnership. ^[19] In Northern Ireland, Nuala Crilly attempted to gain employment at the previously named partnership. As part of their employment requirements, applicants had to have 2 years’ paid experience in community development within the last five years. The partnership justified this requirement on the grounds that their employees had to “perform with minimal supervision; and do the job without the need for extensive training.” ^[20] Nuala Crilly, due to her childcare responsibilities, took six years off from paid work. However, she had a consistent voluntary role in the relevant field during that time period. Regardless, the partnership rejected her application. The judges noted that 90.6% of women were more likely to take time off for childcare duties than men were, thus putting them at a serious disadvantage. ^[21] The court ruled that her volunteer experience would be sufficient for employment, and that she could have been sufficiently trained during the partnership’s two month induction period.

Summarizing the first case by saying “Scott was intentionally mistreated because of his group membership” would fail to capture all the details of the judges’ reasoning. Scott was not mistreated simply because he was a “negro”. Rather, it’s because “negroes” were considered morally inferior that he and others were mistreated. We could even say that it’s because of their group membership that they were considered to be morally inferior, or that being morally inferior was a trait considered to have been intrinsic to “negroes”. Furthermore, it doesn’t seem like the judges would treat Scott in the way they did unless they had those beliefs. It would be inappropriate, in other words, to say that the judges were only motivated by Scott’s group membership. Rather, they were motivated to engage in intentional mistreatment by a significant trait that they believed to be inextricably associated with Scott’s social group; that is, both elements factored into their reasoning.

Something similar happens in Crilly’s case: she was treated unfairly on the basis of a trait that she possessed due to her group membership (e.g. woman, mother). Obviously, the hiring managers were motivated by a trait that they perceived to be significant relative to the goals of their company. Immediately, it seems like we have found something common between disparate treatment and disparate impact: the discriminators are motivated by a trait that is (supposedly) particularly associated with the victim’s group identity and is thought to be significant either in itself or relative to some goal. ^[22] If that’s the case, it’s false to claim that disparate treatment discriminators are only motivated by the victim’s group membership. ^[23] Rather, the discriminators are also motivated by what are perceived to be the victim’s relevant traits, much like in disparate impact.

If either sort of discriminator is concerned with significant traits that are, in some sense, markedly associated with particular groups, then either kind of discrimination involves a similar deliberative process. We can break down their deliberative processes thusly: The agent (a) acts according to some plan or policy that indicates some particular individual, set of individuals, or anonymous individuals for differential treatment; and (b) she acts differentially toward the relevant person(s) because they (i) belong to some social group and/or (ii) (supposedly) possess some significant non-group property. ^[24] As established in the preceding paragraph, the disparate treatment discriminator obviously has some social group in mind along with some non-group property (i.e. (i) and (ii)). Meanwhile, in disparate impact, the discriminator has some individual with some particular traits in mind (i.e. (ii)). Even if the disparate impact discriminator does not put a name to the traits (i.e. the parties are anonymous), she obviously knows that she is dealing with people who possess particular traits. Thus, even when the discriminator addresses anonymous individuals, she is committed to addressing individuals with particular traits in a loose sense (i.e. (ii)). Having said this, it seems that either sort of discriminating agent is concerned with some non-group trait that the targeted person possesses. In other words, contrary to what others have claimed, it seems that either sort of discrimination involves a similar, albeit non-identical, decision-making process.

What I have said is not completely at odds with what other philosophers have posited. For instance, Rasmussen suggests that what we really mean when we say that disparate impact doesn’t involve intentional decisions is that the discriminator and/or the policy doesn’t reflect any bias “against members of [some group]” just in virtue of their being in that group. ^[25] I agree with the no-bias condition, and there’s no reason to suppose that it contradicts the above analysis. A bias, roughly speaking, is a subjective tendency towards treating some particular person or group favorably or unfavorably, whether the person does so intentionally or not. This can be combined with the notion that disparate treatment involves the deliberative process identified above. We’d merely say that a disparate treatment discriminator has a subjective tendency to treat some particular group unfavorably because of some trait they supposedly possess, or that the policy reflects such a bias. Furthermore, saying that disparate impact discriminators focus on particular traits does not mean that they are biased in the relevant sense. Even if they are biased against the neutral traits, it’s still the case that they are not biased against the person qua group member. Hence, my way of understanding the deliberative process of disparate impact and disparate treatment is compatible with Rasmussen’s no-bias condition.

The above account is also compatible with the fact that disparate treatment and disparate impact policies do not always directly specify non-group traits, such as in cases involving policies that require employees to wear special uniforms while not making any further specifications. ^[26] In the context of disparate impact, the idea is that such policies might be disadvantageous to certain religious groups that require their members to constantly wear special adornment. In the case of disparate impact, the discriminator in question would have to make observations about what traits the victims possess, and make inferences regarding their compatibility with the requirements. In the case of disparate treatment, even if beliefs about the victim’s traits or character are not explicitly formed, the discriminator’s attitudes would imply or entail her having such beliefs. Alternatively, even if the discriminator doesn’t hold such beliefs herself – e.g. someone fires an employee due to having racist clientele – it’s still the case that she would be aware of the beliefs that people have. In other words, even if the policy is created or enforced on behalf of someone else, the person who is enacting the policy would have to at least consider why the person desiring the policy is prejudiced against the relevant group. Hence, even if a given policy doesn’t make explicit reference to some particular trait, it’s still the case that the person employing the policy or the person responsible for it holds certain (implicit) beliefs about what sort of characteristics the victims have qua group members. Thus, policies like the “special uniform rule” or reluctantly enacted policies do not contravene the deliberative process laid out above.

This section allows us to conclude that both disparate treatment and disparate impact involve treating someone differently on the basis of a trait that is either actually or supposedly possessed by the victim in virtue of their group membership. Subsequently, both have similar deliberative processes. However, we cannot yet conclude that disparate treatment and disparate impact are structurally similar. Supposedly, disparate treatment is still about group membership insofar as the discriminator includes the victim’s group membership in her deliberations. However, if we analyze how one’s group membership bears upon their identity, we might get a different story.

4 “Being” in a social group

The particular account of group identity I present is inspired by Iris Marion Young’s analysis of said subject in Justice and the Politics of Difference. ^[27] The account of group identity that follows is meant to help us see that disparate treatment and disparate impact are similarly about group membership

Social groups can be identified along the lines of two kinds of traits: profound traits and/or superficial traits. Profound traits are those that have some bearing on the individual’s comportment, ways of thinking and reasoning, etc. Superficial traits generally do not bear on the individual’s way of being, but are rather those traits associated with the general appearance of the relevant group’s members. Profound traits include customs, beliefs, worldviews, and shared experiences. Superficial traits are easily accessible to others: skin color, eye shape, height, and hairstyle. ^[28] These traits are identifiers if and only if they are qualifiers for membership in Group N. For a trait to be a qualifier means that having that trait is a necessary condition for being part of a social group. Similarly, having a trait as an identifier means that the trait is necessary for identifying someone’s group membership. However, simply having a relevant qualifying trait is not sufficient for being in a group. The individual must have all or a sufficient number of the relevant qualifying traits, some of which may be more important than others, in order to be considered a member of the group.

Traits become qualifiers and identifiers through the members’ interactions with each other, other social groups, shared experiences, etc. ^[29] The group members share in the experiences, or are brought up on stories, and these give rise to certain shared beliefs, norms, and interpretations of certain superficial traits. ^[30] Since the shared history of the group and the common experiences of its members determine which profound and superficial traits identify the group, it follows that the necessary trait sets will vary depending upon historical and social circumstances (i.e. they are contingent). Regardless of the contingency of the traits, their possession remains a necessary condition for group membership in that particular historical and social context. In other words, one’s possessing the historically relevant yet necessary traits constitutes their belonging to a particular social group. If someone has all or a sufficient number of the relevant traits, then we can say two things. First, all the traits together jointly constitute the person’s group membership. Second, all the traits individually help constitute the person’s group membership.

For the sake of simplicity, we could say that we’ve identified one primary kind of trait that is relevant to group identity: constitutive traits, with profound and superficial traits included under that general category. However, we can specify another category of traits that are directly relevant to one’s group membership, although they do not play a constitutive role: derivative traits. Derivative traits are those traits that tend to obtain for a particular group because of pre-existing constitutive traits, the person’s group membership itself, some social circumstance relevant to the social group, or some combination thereof. ^[31] One can accept that such traits do in fact obtain in virtue of the following argument. ^[32] First, it is self-evident that one’s personal history, experiences, and culture profoundly shape their identity. Second and subsequently, it seems reasonable to conclude that the shared history of the group, common experiences of its members, and the relevant constitutive traits give the individuals involved a shared identity: they identify with the group in a fundamental way. ^[33] Said otherwise, one’s group identity influences their sense of self or personal identity, which is the individual’s way of understanding and interacting with the world, herself, and others. ^[34] This helps explain why certain traits will tend to obtain, in a derivative sense, for the various members of a given social group. Insofar as group identity is shared, it will influence each member’s sense of self in similar, though not necessarily identical, ways. It is self evident that one’s identity or sense of self mediates their interactions with the world and how they respond to events. Subsequently, one’s identity mediates which traits will obtain for that person, at least to some degree, aside from the ones that they already have. Insofar as group members will have fairly similar self-understandings, certain traits will tend to obtain, in a derivative sense, for a particular group due in part to how the members of said group understand themselves qua group members. In other words, it’s in virtue of the relationship between one’s group identity and personal identity that certain derivative traits will tend to obtain for particular social groups.

We can see this in action in Nuala Crilly’s case. Nuala Crilly undoubtedly took into account the fact that she was a mother when she decided to take time off for the sake of her children. As a result of her decision, she lacked the relevant kind of experience within the timeframe required by the community partnership. Here, the derivative trait is the particular kind and amount of work experience Nuala Crilly had, and it obtained in virtue of her decisions qua mother. Since the trait obtained because of how she took her own group membership into account while acting, we can say that her particular kind of work experience obtained in virtue of her group membership.

For ease of reference, the above information can be summarized thusly:

Group membership is a complex of commonly shared, constitutive traits and experiences from whence further traits can be derived, and either of which informs individual identity and comportment; certain traits and experiences constitute one’s group membership, whereas other traits are derivatives thereof; as an identity-informing complex, group identity cannot be understood independently the relevant constitutive traits; when derivative traits obtain from one’s group membership, they must be understood as being relevant to one’s group membership. ^[35]

Having laid out this account of the relationship between one’s group membership and personal identity, I can explain how disparate treatment and disparate impact are similarly about group membership. When disparate treatment occurs, the discriminating agent treats the victim differently because she believes, or the persons responsible for the policy believe, the victim has some significant trait that is associated with the relevant group. Since the discriminator, or the person(s) responsible for the discriminatory policy, believes that the relevant traits are inextricably linked with the victim’s group membership, she implicitly believes that every member of the relevant group has those traits qua group member. The above account provides at least two plausible explanations for what it means for someone to have particular traits qua group member. Either the person has a trait qua group member because it constitutes her being in that group, or it is derivative relative to her group membership. Subsequently, a disparate treatment discriminator, or the person(s) responsible for such policies, is concerned with traits that she at least implicitly believes to be constitutive or derivative relative to the victim’s group membership. ^[36] For instance, if an employer refuses to hire a female applicant because she plans to become a mother, we can say that she is discriminated against because of a trait that constitutes motherhood (viz. “having a child”) or a trait that is derivable from her being a woman (viz. “becoming pregnant”).

Under the standard discourse, part of how disparate treatment is about group membership has to do with the discriminator’s intentions. Accepting that premise allows us to say the following: Since a disparate treatment discriminator makes her decisions on the basis of traits that are supposedly constitutive of or derivable from the victim’s group identity, this must factor into how we understand disparate treatment to be about group membership. Such traits factor into our explanation about the relevance of group membership because under the account presented above, group membership cannot be understood apart from the relevant traits. Therefore, under the account I have presented, we would say that disparate treatment is about group membership insofar as the discriminator is concerned with significant traits that are either constitutive or derivative relative to the victim’s group membership.

As for disparate impact, it occurs when the discriminator targets some supposedly neutral trait that is factually associated with the victim’s group membership. With this account of social groups in view, the targeted traits could be those that are constitutive or derivative relative to the victim’s group membership, similar to disparate treatment. Supposing that the targeted traits are constitutive or derivative fits well with the facts of disparate impact and sufficiently explains why the disadvantages obtain. To see this, let’s return to Moreau’s public pool scenario. Some of the relevant traits that the excluded black people had were “living in a particular neighborhood”, “being poor”, “lacking quality education and opportunities”, and “having faced a history of oppression”. It seems fairly safe to say that “having faced a history of oppression” is something that constitutes their being part of the modern day racial group, relative to the society they find themselves in. ^[37] The other traits mentioned above can be thought of as derivative: their history of oppression led to a lack of opportunities, which led to their poverty, which led to their living in a particular neighborhood. Therefore, we can say that their residence derived from some trait that constituted their group membership. Subsequently, when explaining how this scenario is about group membership, we would say this: their group membership was relevant because some trait constituting their group identity led to their possessing the trait used to exclude them. ^[38]

In more general terms, we would say that disparate impact is about group membership insofar as the discriminator targets traits that are either constitutive or derivative relative to the victim’s group membership. Obviously, the disparate impact discriminator doesn’t necessarily think of the relevant traits as being constitutive or derivative. Regardless, it wouldn’t make sense to say that the disparate impact discriminator doesn’t treat the victim differently on the basis of those sorts of traits: the traits just are constitutive or derivative. However, the discriminator does think of the traits as being important or significant. Subsequently, we can say that disparate impact is about group membership insofar as the discriminating agent targets traits that are both (believed to be) significant and just are constitutive or derivative relative to the victim’s group membership.

In addition to the similar deliberative processes of disparate treatment and disparate impact laid out in Section III, we can now say that they are about group membership similarly: ^[39] either discriminator is concerned with a trait that is both believed to be significant in some way and is (believed to be) a constitutive or derivative trait for some social group. Said otherwise, group membership is relevant to either sort of discrimination insofar as the discriminator is concerned with traits that are (supposedly) inextricably linked with the victim’s group membership. Subsequently, we can conclude that disparate treatment and disparate impact are structurally similar (i.e. fundamentally similar to a significant degree). I will now provide a definition of disparate treatment and disparate impact that captures the full scope of their structural similarities.

5 Criterion for disparate treatment and disparate impact

Before presenting my criterion, there are some crucial notes that must be made. This criterion is meant to capture the structural similarities that exist between disparate treatment and disparate impact; it is not meant to point out the elements that make a particular instance of either immoral. Although this criterion does not provide the exact conditions for discrimination’s immorality (e.g. disrespect, harm, loss of freedom), it can be assumed that meeting this criterion is a necessary, though not sufficient, condition for immoral discrimination. ^[40] My criterion includes the elements most commonly identified as being morally relevant to discrimination’s wrongness (e.g. group disparities, the actor’s intentions, the justifiability of the policies). If one of these elements must be present for immorality to obtain, and these elements are present in my criterion, we can assume that my criterion has the necessary conditions for immoral discrimination. ^[41] As such, my criterion only applies to instances of disparate treatment/disparate impact that have the potential to be immoral.

One final caveat about my criterion: it is only meant to cover instances of discrimination that are concerned with social groups and their members. ^[42] There might be other forms of discrimination that are concerned only with particular traits; there might be no good reason to restrict the concept of discrimination to just social groups. ^[43] For instance, it might be impossible to argue that mistreating someone just because they wear glasses doesn’t constitute discrimination. However, we can at least agree that there is a kind of discrimination that is just about social groups, and it is this sort of discrimination I am concerned with.

With these provisos in mind, I propose thinking of discrimination in this way:

S1 discriminates against S2 when all of the following obtain: i) S1 (attempts to) treat S2 differently relative to S3; S2 is (supposedly) a member of socially significant Group N and S3 is (supposedly) a member of socially significant Group M. ii) S1‘s (attempted) treatment of S2 (would have) caused or maintained some kind of disparity between S2 and S3. iii) S1’s (attempted) treatment of S2 can be causally explained by the fact that S2 (supposedly) possesses some socially significant property P and is (supposedly) a member of Group N; property P is associated with membership in Group N insofar as it is either (supposedly) constitutive or derivative relative to membership in Group N; Property P is the contrast of Property Q, which is associated with members of Group M. iv) S1’s differential treatment of S2 relative to S3 is not sufficiently justified on the basis of Property P (or Group Membership N). ^[44]

Before applying this criterion, I will explain its various components. The verb “treats” is not meant to indicate that the agent is consciously trying to put the other person at a disadvantage; the same may be said for “attempts”. Similarly, the agent may either account for the victim’s group membership and traits (consciously/unconsciously), or be solely concerned with their particular non-group traits. However, in either case, the discriminator necessarily targets socially significant traits that are associated with some socially significant group, whether consciously or unconsciously. As for “disparity,” this term covers a broad range of differential outcomes, both harmful and innocuous. For instance, this allows my criterion to cover scenarios such as this: A hiring manager only hires members of particular races for particular jobs and only those jobs, although she pays all of them equally well. ^[45] Intuitively, this seems discriminatory but not harmful. Since “disparity” is being used broadly, my criterion is able to include odd cases such as this one.

Under my criterion, both the affected groups and the targeted traits must be socially significant. A socially significant property or social group has (or is believed to have) relevance in several social dimensions; or it substantially influences the comportment (e.g. appearance, mannerisms) or capacities of those who possess it. An example of a socially significant property with relevance in several social dimensions would be one’s writing capabilities: Being unable to write in clear English disqualifies people from certain work in the U.S. and attaches a stigma of unintelligence to them. An example of a socially significant group with multi-dimensional significance would be something like ex-convicts or prior sexual offenders. Having that status has a wide array of effects on their social lives, including general interpersonal relations, family ties, employment opportunities, etc. An example of a trait that substantially influences the agent’s comportment would be one’s religious beliefs. For instance, someone who believes that the Christian Bible is authoritative on ethics will try to live her life in a particular way. An example of a social group that bears on the comportment of the individual would be something like one’s cultural origin: people from different societies typically act and think differently from one another to some degree.

The third clause of my criterion captures the most significant similarities that exist between disparate treatment and disparate impact. As previously explained, the correlative relationship between socially significant traits and group membership is one of (supposed) constitution or derivation. That is, S2 possesses a trait that either helps constitute her being in Group N, or is derivative relative to her being a member of Group N. Alternatively, the discriminator or the person responsible for the policy believes, explicitly or implicitly, that such a relationship obtains. Insofar as the individual’s group membership is constituted by having certain traits and is relevant to the attainment of further derivative traits, we can say that disparate treatment and disparate impact are alike since they focus on those sorts of traits.

The third clause also captures the deliberative similarities that exist between disparate treatment and disparate impact. Namely, the fact that either discriminator treats S2 differently on the basis of a socially significant trait that is consistently associated with her group membership. The deliberative involvement of the victim’s group membership and particular traits can come in degrees. For instance, an employer who fires an employee because her clientele is prejudiced against the employee’s social group does not give a great deal of weight to the victim’s group membership and relevant traits. However, she must take stock of the strength of her clientele’s prejudice, and it seems unreasonable to say that she would be unaware of the reasons for their biases and subsequently take them into account. Hence, even if there may be differences in degree when it comes to the deliberative weight of the victim’s traits and group membership, we can say that the discriminator employs some criterion specifying a trait that is (supposedly) consistently associated with the victim’s group membership.

The fourth element of my criterion requires that the policy must be insufficiently justified. As a prima facie, non-exhaustive characterization, insufficient justification obtains when S1’s reasoning for excluding S2 on the basis of property P (and her group membership) is actually irrelevant to any given legitimate goal. ^[46] Regarding legitimacy, we can say that legitimate goals advocate morally permissible ends (this is not an exhaustive characterization). Additionally, although relevance might vary from case to case, there are at least some cases where a criterion is irrelevant because it is not possible to produce a logically sound argument that can establish its relevance to achieving a legitimate goal. For another sense in which irrelevance might obtain, consider the following example. In America, medical schools typically deny admittance to students who went to community college prior to attending a four year university “even after GPA and MCAT scores [are] taken into consideration”; the differences between the GPAs and MCAT scores for four-year students and First-CC students was modest, with the former group only slightly exceeding the latter. ^[47] This policy of not admitting First-CC students not only has a significant impact on economically disadvantaged students and racial minorities, but seems to lack proper justification. Thus, in addition to what was already said about irrelevance, it seems we can say that a poorly justified criterion either fails to take full account of all the facts related to achieving the legitimate goals of the individual or organization making the decision, or they overestimate the relevance of certain traits.

Having set up disparate treatment and disparate impact as being structurally similar, one might wonder if such an understanding is incompatible with our intuitions about certain real-world cases. The following section demonstrates that understanding disparate treatment and disparate impact as fundamentally similar is compatible with our basic intuitions about certain real world cases. Showing that my criterion has this compatibility would not definitively demonstrate that disparate treatment and disparate impact are actually structurally similar. Rather, by demonstrating that my criterion captures our basic intuitions about discrimination, I eliminate the possibility that the conclusion of Section III and Section IV (viz. disparate treatment and disparate impact are structurally similar) can be falsified by discrediting my criterion. Doing so allows us to say that it is at least plausible to understand disparate treatment and disparate impact as being structurally similar to one another.

6 Applying the criterion ^[48]

Let’s start with the two court cases mentioned earlier, since it has yet to be shown that they meet all the requirements of my criterion. In Dred Scott’s case, he was being treated differently relative to whites. Additionally, the court’s decision left him disadvantaged relative to whites: he walked away from the courts without citizenship and its attached privileges. Additionally, he was treated differently due to supposedly lacking the proper amount of moral worth, which was either believed to be constitutive of or derivable from being black. ^[49] Finally, even assuming that having the proper amount of moral worth is a relevant consideration for being a citizen, there is no reason to suppose that it can be soundly demonstrated that black people possess a morally inferior nature. Hence, the judges’ decision was poorly justified. Thus, Dred Scott’s case meets my criterion for disparate treatment provided at the beginning of the previous section, and also lines up with our intuitions about this case – namely, that it’s an instance if disparate treatment.

As for disparate impact, recall Nuala Crilly’s case. In hindsight, according to the court’s findings regarding the likelihood of women to take time off work for family, Nuala Crilly was indeed treated differently by the partnership’s criterion relative to men. Furthermore, it is certainly the case that Nuala Crilly’s rejection was disadvantageous for her (i.e. her unemployment was perpetuated). It’s also clear that her differential treatment at the hands of the partnership can be explained by the fact that she lacked a certain amount of experience, and her having this property was derived from the fact that she belongs to particular social groups (viz. woman and parent). ^[50] However, the employers only cited her work experience, and were not directly concerned with her group membership. Finally, as the court pointed out, the partnership’s specific criterion for work experience was actually poorly justified relative to their aims as a company. Hence, Nuala Crilly’s case meets my criterion for disparate impact, and our basic intuitions about the scenario – namely, that it’s an instance of disparate impact.

Now I’ll demonstrate that my criterion captures our intuitions about what doesn’t count as discriminatory. Suppose that a hiring manager is looking for applicants with degrees in engineering for a highly technical job. One man and one woman apply for the job, but only the man has the relevant degree. Suppose further that due to certain social conditions surrounding gender socialization, men are more likely to have engineering degrees than women. Since the man had the degree, he was hired, but the woman was turned down. Obviously the female applicant was treated differently and is disadvantaged relative to the male applicant, and this differential treatment can be explained by the fact that she had a certain property/lacked a certain property. Furthermore, her having a certain property/lacking a certain property obtained in virtue of social conditions surrounding her group membership. However, this case cannot constitute an instance of disparate impact under my criterion: It seems like having an engineering degree is relevant to the job at hand, and thus fails to meet the fourth element of my criterion. This conclusion also lines up with our intuitions about this case, viz. that it’s not discriminatory.

Let’s take another example, this time regarding disparate treatment. Suppose two people apply for a pastoral position at a church, one of them an Atheist and the other one a Christian. The job description for the pastoral position requires the individual to be a practicing Christian, so the Atheist is obviously rejected for the position. It’s fairly obvious that she was rejected in virtue of her group membership, a necessary component of which is having the property of not believing in God or not believing in the Bible. Either of those properties could have been cited as a reason for rejecting the Atheist candidate if the people in charge of the church were pressed for further explanation. Furthermore, this kind of policy certainly has a disparate effect between Christians and Atheists. However, it seems like the criterion employed by the church is fairly relevant to their goals. Here’s one reason why the criterion might be considered relevant: As far as fundamentalist churches are concerned, the “no Atheists allowed” rule is relevant insofar as one of the roles of a pastor is to provide counseling to individuals who face subjective spiritual crises. Insofar as believing in God is a necessary part of helping people deal with such problems in the context of a fundamentalist church, being a Christian and not an Atheist is a relevant hiring criterion. In other words, since this scenario fails to meet the fourth element of my criterion, it therefore fails to count as an instance of disparate treatment. This conclusion also lines up with our pre-conceived intuitions about the case, viz. it’s not a case of disparate treatment.

Before concluding this section, let’s consider two scenarios that would initially appear to counter my criterion. Let’s say that a small business owner is considering some applicants, and she has a deep, irrational hatred for Germans. Subsequently, she refuses to hire anyone of German descent and fires anyone who is later exposed as being German. Suppose that someone of German descent applies to her business, and she knows that this person is German. On that basis alone she rejects the applicant. This seems to be a counter-example to my criterion: the business owner rejected the applicant purely based on her group membership, and not on the basis of some trait correlated with her group membership. Even though this scenario apparently does not meet my criterion, it’s obviously a case of discrimination.

I do not believe a case precisely like this could obtain in the actual world, and as such cannot constitute a legitimate counter-example to my criterion. I’ll explain the former assertion momentarily but justify the latter claim presently. If nothing precisely like this case could happen in our world, it is a counter-example on the assumption that we ought to give greater weight to conceptual possibility than to actual possibility when theorizing discrimination, at least when the two conflict. However, it seems like what is actually possible should be the first thing to determine how discrimination is conceptualized. Discrimination is an empirical event, and therefore must be conceptualized according to what actually happens, first and foremost. What actually happens depends on what can actually happen or “how the world really works”, so to speak. Hence, discrimination must be conceptualized according to what is actually possible before turning to what is logically possible. Therefore, when we seek to understand discrimination via examples, the kinds of examples that have priority must be based on what is actually possible. In other words, if we have a situation where logical possibility and actual possibility conflict, we ought to give priority to actual possibility. Thus, if the above scenario is not actually possible, we have good reason to think it does not undermine my criterion for discrimination.

The question now is whether or not the above case is actually possible. If we mean to say that the business owner hates Germans for no reason whatsoever, this is implausible. It’s a truism that one’s prejudices develop over time through socialization, exposure to stereotypes, personal experiences, etc. Furthermore, the reasons for prejudice are usually trait or stereotype oriented, with cognition playing some role in bringing about or justifying the relevant biases. ^[51] In other words, prejudice is fueled by one’s perceptions or cognitive interpretations of the traits that certain groups (supposedly) possess. All of this is to say that I don’t think it’s plausible to suppose that our imaginary business owner hates Germans without believing, consciously or unconsciously, that there are certain traits they possess that warrant her ire. ^[52] I am willing to grant that she is consciously motivated by her hatred for Germans qua group members, but I don’t think the scenario, realistically constructed, stops there. Similar to what happens in unconscious disparate treatment, we can say that the business owner was unaware of the fact that she was motivated by particular traits, or at least they were not at the forefront of her mind as she was acting. ^[53]

We can say that this crude case of discrimination actually fails to counter my criterion for two reasons. First, given the way prejudices form in the actual world, nothing precisely like it could happen in our world. Second, were it presented realistically, it would actually meet every requirement of my criterion: the business owner treats the applicant differently; the business owner disadvantages the applicant; the business owner treats the applicant differently because of some trait associated with her group, and she cares about the trait at least unconsciously; her selection of the trait and the applicant’s group membership cannot be adequately justified. Hence, this cannot serve as a legitimate counter-example to my criterion.

The second scenario is supposedly a case of disparate impact. Let’s say that S2 works at a restaurant where all employees are required to wear a uniform. At the time she was hired she self-identified as a Christian, but later she decided to renounce her faith. However, she still wears a cross around her neck. Although her wearing the cross at work does not interfere with her basic duties, the restaurant has a policy stating that non-uniform accessories may not be worn sans a legitimate religious reason. Since she refuses to follow the policy, her manager had to let her go. Intuitively, this seems like a case of disparate impact: one could say that she was discriminated against on the basis of a group-neutral trait (i.e. Christians don’t have a “monopoly” on that symbol) that is still consistently associated with a particular social group; she is also at a disadvantage relative to her former co-workers, religious and non-religious; and since her wearing the cross does not interfere with her basic duties, and given how commonplace the cross qua symbol has become (e.g. fashion), it doesn’t seem relevant to the goals of the restaurant. Intuitively, we seem to have a case of disparate impact, but the problem is that it doesn’t meet the first element of my criterion: S2 is no longer a member of the relevant social group. If this is a case of disparate impact, it seems to counter my criterion.

The simplest way to deal with this scenario is to reiterate that my criterion is concerned strictly with group-oriented discrimination. That is, my criterion is meant to address instances of discrimination that affect social groups and their members. In this case, S2 is no longer a member of the relevant social group. Therefore, my criterion doesn’t apply to this scenario in the first place. That’s not to say this example doesn’t involve some sort of discriminatory behavior. She was discriminated on the basis of some trait she possessed, so we can admit that this is an instance of trait-oriented discrimination. Since this is actually a case of trait-oriented discrimination, and my criterion only addresses group-oriented discrimination, this does not constitute a legitimate counter-example to my criterion.

7 Conclusion

The overarching goal of my paper has been to demonstrate that group-oriented disparate treatment and disparate impact are structurally similar to one another, and that this notion is compatible with our basic intuitions about discrimination. Since both kinds of discrimination involve focusing on socially significant traits, this reveals that they involve similar deliberative processes. Furthermore, given the relationship between how being in a social group affects an individual and the deliberative process of disparate impact and disparate treatment, they turn out to be about group membership in fundamentally similar ways. Not only does the criterion provided capture these similarities, but it conforms to our intuitions about what does and does not count as discriminatory. Since my criterion accurately captures which cases are discriminatory and which ones are not, it seems like an acceptable way of understanding disparate treatment and disparate impact. As such, we may conclude that disparate treatment and disparate impact are structurally similar to one another without sacrificing our basic intuitions about discrimination.

One might wonder the following: If disparate treatment and disparate impact are as similar as I have claimed, are we warranted in distinguishing them as separate kinds of discrimination or are they fundamentally the same? There is still at least one difference between them: a disparate treatment discriminator still accounts for the victim’s group membership. Subsequently, there are at least three possible conclusions regarding the distinction between disparate treatment and disparate impact. If the former difference is substantial, as is commonly believed, we’d still be merited in saying that disparate treatment and disparate impact are distinct kinds of discrimination. Alternatively, the fact that a disparate treatment discriminator accounts for the victim’s group membership might be insignificant in the grand scheme of things. If so, disparate treatment and disparate impact might be practically identical given all their other similarities. However, one could argue that since the two forms of discrimination do not share all the same properties, it would be disingenuous to claim that they are virtually identical to one another, regardless of how insignificant their differences are. For my part, the second alternative has a certain intuitive persuasiveness, although the third alternative seems to carry more logical force. Regardless, these are implications best left to another paper.