Affirmative Action in the Political Domain

This article has two parts. First, I argue that three prominent arguments in favour of affirmative action – the mitigating discrimination argument, the equality of opportunity argument and the diversity argument – may be based on a relational egalitarian theory of justice, as opposed to a distributive understanding of justice. Second, I argue that basing these arguments in favour of affirmative action on relational egalitarianism has an interesting implication when it comes to the site(s) of affirmative action. Whereas affirmative action is usually discussed and pursued in university admissions and in hiring for jobs, the relational egalitarian–based arguments entail, all else equal, that we have more reason to pursue affirmative action in the political domain than in the traditional sites.


Introduction
Affirmative action continues to receive attention in public and scholarly debates.Defenders of affirmative action contend that in the non-ideal circumstances in which we find ourselves, affirmative action is a requirement of justice (e.g.Beauchamp, 2002;Nagel, 2003;Sterba in Cohen and Sterba, 2003).Opponents of affirmative action, on the contrary, contend that affirmative action is unjust, for one because it is effectively a form of reverse discrimination (e.g.Cohen in Cohen and Sterba, 2003;Hook, 2002;Pojman, 2014).In this article, I will be concerned with three of the most prominent arguments in favour of affirmative action -the mitigating discrimination argument, the equality of opportunity argument and the diversity argument.These arguments are usually based on a distributive understanding of justice.Instead, I will argue that these arguments may be based on a relational egalitarian understanding of justice, respectively that non-discrimination may be a necessary part of relating as equals; that equality of opportunity may be a necessary part of relating as equals; and that diversity is beneficial when and because it leads to less inequality of power because an unequal distribution of power is incompatible with relating as equals.
Basing these arguments in favour of affirmative action on relational egalitarianism has an interesting implication when it comes to the site(s) of affirmative action, that is, with regard to where we should pursue affirmative action.Whereas affirmative action is usually discussed and pursued in university admissions and in hiring for jobs, the relational egalitarian-based arguments entail that, all else equal, we have more reason to pursue affirmative action in the political domain because the political domain is particularly important for relational (in)equality.This is a surprising conclusion in the sense that I take it that most people prima facie find affirmative action more appealing in university admissions or in hiring for jobs than in the political domain and because affirmative action is, in most countries, usually not pursued (to the same degree) in the political domain. 1 Thus, this calls for a change of political practice.
My arguments in this article speak to defenders of affirmative action in the sense that I show them how the underlying justification matters to the site(s) of affirmative action.I also urge them to choose whether they want to base their arguments in favour of affirmative action on a distributive or a relational conception of justice, or a combination of the two, because it makes a difference to when and where affirmative action is justifiable.Distributive-based arguments justify affirmative action in some instances in which relational-based arguments do not, and vice versa.My arguments speak to opponents of affirmative action in the sense that I raise their burden of proof for showing that affirmative action is unjustifiable.Inasmuch as these arguments in favour of affirmative action may be based on a relational egalitarian conception of justice, arguing that these arguments in favour of affirmative action are unjustifiable given a distributive understanding of justice fails to show that affirmative action is unjustifiable tout court.
A preliminary remark before moving on.I am not seeking to justify affirmative action in this article, partly for reasons of space, partly because others have already done so in an insightful manner (see, for example, Cahn, 2002;Cohen and Sterba, 2003;Lippert-Rasmussen, 2020).Instead, what I will do is to show that prominent arguments in favour of affirmative action may be based on a relational egalitarian theory of justice and what this implies for the site(s) of affirmative action.Whether relational egalitarian-based affirmative action is ultimately justifiable is a question that I will leave for future research. 2

Defining Affirmative Action
For the purposes of this article, I will follow Anderson (2010: 135) and understand affirmative action as 'any policy that aims to increase the participation of a disadvantaged social group in mainstream institutions, either through "outreach" (targeting the group for publicity and invitations to participate) or "preference" (using group membership as criteria for selecting participants)'. 3,4For instance, a policy which reserves seats for members of a disadvantaged group for admission into prestigious universities would qualify as affirmative action on this definition.A few remarks about this definition.
First, the definition refers to 'any policy'.If we assume that when an individual acts independently of the political domain, this act cannot count as a policy, it follows that an act by a wealthy individual to incorporate some disadvantaged individuals into his networks to increase their participation in mainstream institutions cannot count as affirmative action (cf.Lippert-Rasmussen, 2020: 9, n. 15).More generally, if only political agents can implement any policy, it follows that only political agents can be the authors of affirmative action.Second, the definition does not specify whether 'disadvantaged' must be understood locally or globally.For instance, men may be disadvantaged in a local setting, say with regard to being nurses, but be advantaged in a global sense.I assume that Anderson intends disadvantaged to be understood in a global sense; at least she does not speak of affirmative action in favour of men in local settings. 5In any case, I will be focused on affirmative action for groups who are disadvantaged in the global sense.This means that a policy that ultimately seeks to benefit men in a local setting cannot count as affirmative action given the assumption that men are advantaged in the global sense.Third, Anderson does not specify what counts as a mainstream institution.I take it that the following at least count as mainstream institutions: educational institutions, political institutions and the workplace.Fourth, the immediate target of affirmative action does not have to be a disadvantaged group.A policy, which gives preference to an advantaged group in non-mainstream institutions with the aim of increasing the participation of a disadvantaged group in mainstream institutions, may count as affirmative action on this definition (cf.Lippert-Rasmussen, 2020: 11). 6Finally, the definition does not imply that any policy, which counts as affirmative action, is thereby justified.That is, a policy may not be justified even if it is affirmative action.This is an advantage as we do not want to settle the justifiability of affirmative action by definitional means.Having thus defined affirmative action, let us turn to three prominent arguments in favour of affirmative action.I will show how they, contrary to what is usually the case, may be based on a relational egalitarian conception of justice.

The Mitigating Discrimination Argument in Favour of Affirmative Action
One prominent argument in favour of affirmative action is the mitigating discrimination argument.According to this argument, justice requires that we eliminate unjust discrimination, and affirmative action may be, in many cases, the best way of securing that we eliminate unjust discrimination. 7Thus, justice requires affirmative action (Lippert-Rasmussen, 2020: 51-71).As Lippert-Rasmussen does in his discussion of this argument, I will assume the following definition of discrimination: X discriminates against Y relative to Z if, and only if, (1) X treats Y worse than Z; (2) this is (a) due to the fact that X believes that Y is a member of a certain socially salient group and believes that Z is not, or (b) explained in a certain way by the fact that Y is a member of a certain socially salient group and Z is not (Lippert-Rasmussen, 2020: 55).
Suppose that an employer treats his female employee worse than his male employee because she is a woman.This would count as discrimination on this definition. 8 The mitigating discrimination argument is expressed by Scanlon who says: Some policies of affirmative action for female and minority candidates may be justifiable in a similar way, and therefore compatible with formal equality of opportunity as I am understanding and defending it.Discrimination as I have defined it occurs when there are widespread beliefs about the inferiority of members of certain groups, and this leads to their being excluded from positions of authority and expertise, on grounds of their supposed unsuitability or lack of ability.Because people's beliefs about who is capable of performing well in positions of a certain kind is heavily dependent on who, in their experience, has generally done this, one important way of combatting discrimination is to place individuals from previously excluded groups into positions of authority where they can be seen to perform as well as anyone else (Scanlon, 2018: 48;see This argument justifies affirmative action in admissions into prestigious jobs when seeing individuals from previously excluded groups at such jobs perform as well as individuals from other groups leads to less discrimination against them. In discussions of affirmative action, this mitigating discrimination argument is usually put forward assuming a distributive ideal of justice where justice requires realizing a certain pattern of distribution (Lippert-Rasmussen, 2020: 144, 146).One such distributive theory of justice is luck egalitarianism, 9 according to which individuals' distributive positions should reflect only their relative exercise of responsibility (Lippert-Rasmussen, 2015, 2020: 149). 10On this view, justice requires anti-discrimination because discrimination against (in favour of) an individual leaves the individual worse-off (better off) through no fault or choice of her own.We thus have the first part of the mitigating discrimination argument, that is, that justice requires anti-discrimination.However, we do not have to base the mitigating discrimination argument on a distributive theory of justice.Indeed, we may base the argument on a relational theory of justice.One such theory, which has gained widespread support in recent years, is relational egalitarianism.
According to relational egalitarianism, justice does not require realizing a certain pattern of distribution.Instead, justice requires that people relate as equals, as opposed to inferiors and superiors (Anderson, 1999;Fourie et al., 2015;Højlund, 2022;Kolodny, 2014;Lippert-Rasmussen, 2018a;Nath, 2020;Scheffler, 2003Scheffler, , 2005Scheffler, , 2015;;Schemmel, 2011a;Wolff, 1998).For X and Y to relate as equals, they must (1) treat one another as equals; and (2) regard one another as equals (Cohen, 2013: 197; Lippert-Rasmussen,  2018a: 71). 11This theory is clearly different from luck egalitarianism, for example, individuals' distributive positions may reflect only their relative exercise of responsibility but they may still not relate as equals because one of the individuals regard the other as a moral inferior due to the latter's race.
We may base the mitigating discrimination argument in favour of affirmative action on this relational egalitarian theory of justice.Let us start by noting that discrimination, at least direct discrimination and some cases of indirect discrimination (Lippert-Rasmussen, 2018b), 12 is incompatible with relating as equals in the same way that racism is incompatible with relating as equals (Anderson, 1999: 312;Lippert-Rasmussen, 2018a: 73;Scanlon, 2018: 26).After all, racism may be the reason why a person suffers from discrimination.As Anderson (1999: 312) explains, relational egalitarianism 'repudiates distinctions of moral worth based on birth or social identity -on family membership, inherited social status, race, ethnicity, gender, or genes.There are no natural slaves, plebeians, or aristocrats' (see also Lippert-Rasmussen, 2018a: 73).This means that relational egalitarian justice requires removing these forms of discrimination.However, there is a further question as to when X and Y must enjoy non-discrimination in order to relate as equals.To see what I have in mind, we may, following Lippert-Rasmussen (2019), distinguish two specifications of relational egalitarianism: Whole lives relational egalitarianism: Justice requires that, from the perspective of their lives as a whole, people relate socially to one another as equals.
Time-relative relational egalitarianism: Justice requires that, at any given moment, people relate socially to one another as equals (Lippert-Rasmussen, 2019: 154).
To see that these views differ, suppose that X exploits 13 Y for 30 years after which Y exploits X for 30 years (cf.Lippert-Rasmussen, 2019: 155).In this case, X and Y relate as equals from the perspective of whole lives relational egalitarianism, but not according to time-relative relational egalitarianism since at no point in time in their lives do they relate as equals.This also matters to non-discrimination as we may take a whole lives view or a time-relative view when it comes to that question.According to a whole lives view, X and Y may relate as equals if X for 20 years discriminates against Y if this is followed by 20 years in which Y discriminates against X.This does not satisfy the timerelative view since at no point in time in their lives do they relate as equals because one of them discriminates against the other.Bidadanure argues that: on this view [relational egalitarianism], what is problematic in our examples [examples structurally similar to the one above] 14 is precisely that these societies may not be communities of relational equals at any point.Phases of domination, marginalization, or segregation cannot be thought to cancel out diachronically (Bidadanure, 2016: 246).
Thus, according to her, we should prefer the time-relative understanding of relational egalitarianism.I do not have the space to settle here whether we should prefer the whole lives view or the time-relative view, but I will follow her and assume that the time-relative view is preferable to the whole lives view.This means that justice requires that, say, X does not discriminate against Y at any point in their lives.If affirmative action is the best means of removing, or at least mitigating, such forms of discrimination in some situations, justice requires affirmative action.We may call this the relational egalitarian mitigating discrimination argument in favour of affirmative action.
In some ways at least, the relational egalitarian mitigating discrimination argument is broader than the distributive egalitarian mitigating discrimination argument, that is, it justifies affirmative action in cases where the latter does not.Suppose that men in a society regard women as moral inferiors but in a way that does not lead to distributions reflecting anything other than individuals' relative exercise of responsibility (men do not treat women differently than they would have done if they had not regarded women as moral inferiors).The reason men regard women as inferiors is that they do not believe women are capable of doing the kind of work they, as men, are doing, primarily because they do not experience them doing this kind of work.In this case, the distributive egalitarian mitigating discrimination argument does not apply.If we suppose that affirmative action in the form of hiring women for positions usually occupied by men is the best means of making men stop regarding women as their inferiors, affirmative action is justified in this case assuming the relational egalitarian mitigating discrimination argument but not assuming the distributive version.
The upshot is that a prominent argument in favour of affirmative action, the mitigating discrimination argument, may be based on relational egalitarianism as opposed to (or in addition to) a distributive understanding of justice.In this sense, both distributive and relational egalitarians qua distributive and relational egalitarians may justify affirmative action as a way of mitigating discrimination.

The Equality of Opportunity Argument in Favour of Affirmative Action
Another prominent argument in favour of affirmative action is the equality of opportunity argument.Harris and Narayan put forward this argument: Our position is that affirmative action is best understood as an attempt to promote equality of opportunity in a social context marked by pervasive inequalities, one in which many institutional criteria and practices work to impede a fair assessment of the capabilities of those who are working-class, women, or people of color.Thus, affirmative action is an attempt to equalize opportunity for people who continue to face institutional obstacles to equal consideration and equal treatment (Harris and Narayan, 2014: 452).
Similarly, Sher argues that: the key to an adequate justification of reverse discrimination [affirmative action] [is] to see that practice, not as the redressing of past privations, but rather as a way of neutralizing the present competitive disadvantage caused by those past privations and thus as a way of restoring equal access to those goods which society distributes competitively (Sher, 2002: 61). 15  According to the equality of opportunity argument, justice 16 requires equality of opportunity, and in some cases, affirmative action is the best way of securing equality of opportunity or at least less inequality of opportunity.Thus, justice requires affirmative action in such cases (cf.Lippert-Rasmussen, 2020: 72).I understand equality of opportunity such that: X and Y enjoy substantive equality of opportunity (vis-á-vis one another) with regard to a certain position, P, if, and only if, when X and Y have the same native talent required for the position and the same ambition (i.e., they commit the same level of efforts to achieve the relevant position), they enjoy equal chances of getting it (Lippert-Rasmussen, 2020: 78).
This argument is closely related to the mitigating discrimination argument, primarily because discrimination often increases inequality of opportunity.They are different, however.In a society in which everyone suffers from discrimination, there may be equality of opportunity.In that case, the equality of opportunity argument in favour of affirmative action would not apply but the mitigating discrimination argument would.
Why is equality of opportunity a requirement of justice?According to the luck egalitarian view of justice presented above, inequality of opportunity entails that distributive positions not only reflect individuals' relative exercises of responsibility.In a society with inequality of opportunity, X and Y may work equally hard, be equally talented and end up unequally well off, say, because Y, and not X, suffers from racism when applying for jobs.Inequality of opportunity is thus unjust according to luck egalitarianism.Thus, the distributive understanding of the equality of opportunity argument justifies affirmative action when and because it is the best means of securing equality of opportunity or at least less inequality of opportunity.The argument is usually put forward and defended in this distributive version (see, for example, the quote from Sher above).
As was the case with regard to the mitigating discrimination argument, I now want to show that we may base the equality of opportunity argument in favour of affirmative action on relational egalitarianism instead of luck egalitarianism.We may start with the claim that inequality of opportunity is incompatible with relating as equals.I take it to be uncontroversial that a necessary requirement of equal relations is that people have equality of opportunity. 17Suppose that men, for no good reason, have greater opportunity than women in society in terms of getting prestigious jobs.We would not say that there are equal relations between men and women in this society. 18Indeed, a woman in this society might reasonably see such inequality of opportunity as an expression of the fact that men's interests are considered more important than women's interests (cf.Scheffler, 2015: 25).
This means that relating as equals requires that X and Y have substantive equality of opportunity at any given moment in their lives.If we add the premise that affirmative action in some cases is the best means of securing equality of opportunity, we may conclude that relational egalitarian justice requires affirmative action in such cases.This we may call the relational egalitarian equality of opportunity argument in favour of affirmative action.This argument is different from the equality of opportunity argument in favour of affirmative action with which we started this section since it is based on a relational as opposed to a distributive conception of justice.
The relational egalitarian equality of opportunity argument in favour of affirmative action is different from the relational egalitarian mitigating discrimination argument in favour of affirmative action.If we boost discrimination against a privileged minority, say, a small group of extremely wealthy people, we may reduce inequality of opportunity (Lippert-Rasmussen, 2020: 75).The relational egalitarian equality of opportunity argument may justify affirmative action in such a case but the relational egalitarian mitigating discrimination argument would not. 19 The scope of the distributive and relational understandings of the equality of opportunity argument does not differ.That is to say, the cases that violate equality of opportunity in the distributive understanding of justice and justify affirmative action also violate equality of opportunity in the relational egalitarian understanding of justice and likewise justify affirmative action.This is to say that the distributive-based and relational egalitarian-based equality of opportunity arguments in favour of affirmative action are extensionally equivalent, but the ultimate reason for why affirmative action is required differ on the two arguments: on the former, it is required to secure just distributions whereas on the latter, it is required to secure equal relations.
We have now seen that another prominent argument in favour of affirmative action, the equality of opportunity argument, can be given a relational egalitarian interpretation and not only a distributive justice interpretation.In this sense, both distributive and relational egalitarians qua distributive and relational egalitarians may justify affirmative action as a way of securing equality of opportunity or less inequality of opportunity.Before turning to discuss the site(s) of affirmative action when affirmative action arguments are based on relational egalitarianism, let us turn to a third and final argument in favour of affirmative action.

The Diversity Argument in Favour of Affirmative Action
We have now examined the mitigating discrimination and the equality of opportunity arguments in favour of affirmative action.In this section, I turn to an argument that has been highly popular in both legal and moral contexts, to wit, the diversity argument in favour of affirmative action.It was expressed by Justice Powell who argued with regard to the attainment of a diverse student body that: this clearly is a constitutionally permissible [438 U.S. 265, 312] goal for an institution of higher education . . .Ethnic diversity, however, is only one element in a range of factors a university properly may consider in attaining the goal of a heterogeneous student body.Although a university must have wide discretion in making the sensitive judgments as to who should be admitted, constitutional limitations protecting individual rights may not be disregarded (Regents of the University of California v. Bakke 438 U.S. 265,1978; see also Grutter v. Bolinger 539 U. S. 306, 2003).
Investigating the diversity argument, Lippert-Rasmussen presents the argument as follows: 1. Affirmative action promotes diversity better than non-affirmative action policies.2. Promoting diversity is good overall.3. Promoting diversity does not violate any deontological constraints.4. If 1-3, then affirmative action is morally justified all things considered.5. Thus, affirmative action is morally justified all things considered (Lippert-Rasmussen, 2020: 124-125). 20  What is important for present purposes is premise 2, specifically with regard to why promoting diversity is good overall.Usually, proponents of diversity arguments point to how diversity has epistemic benefits. 21Consider Appiah who says: Why is diversity a good?Because diversity of social identities makes education and research better.Exposure to people of the major social kinds in your society on terms of equality and in the college setting is part of an education for life in a society of diverse identities.Universities are also research institutions and in large areas of the social sciences and the humanities interest is differentially distributed by social identity, in both senses of the term: social identity affects what you find interesting and it affects what stakes you have in the answers to questions (Appiah, 2011: 278).
Similarly, Bok and Bowen (2002: 179) argue that 'race almost always affects an individual's life experiences and perspectives, and thus the person's capacity to contribute to the kinds of learning through diversity that occur on campuses' (see also Anderson, 2010: 141;LeFevre, 2003;Lippert-Rasmussen, 2020: 124).
What I want to show now is how a diversity argument in favour of affirmative action may be based on relational egalitarianism.This argument does not claim that promoting diversity is good due to epistemic benefits. 22Instead, diversity is good when and because it prevents power from being concentrated in some group (cf.Bergmann, 2002: 147), and preventing power from being concentrated is good because people cannot relate as equals if power is concentrated in some group.If we start with the former claim -that is, diversity is good when and because it prevents power from being concentrated in some group or people -suppose that only White men go to prestigious universities which educate the future elite of society.Bringing in non-White men or women to these universities through affirmative action may prevent power from being concentrated in one group in society in the future.For another example, bringing in (more) non-White men and women on the boards of powerful companies may prevent power from being concentrated in one group in society, not only in the future, but also in the present.
That equal power is necessary to relate as equals has been argued by relational egalitarians.'An egalitarian relationship requires', Viehoff argues: rough equality of power over the interactions that make up the relationship.Relational equality is, in turn, threatened where the parties to the relationship have significantly different power over how they interact with and relate to one another.We might call this the threat of subjection, and refer to equal power as a matter of nonsubjection.Where the parties to a relationship lack equal power, the relationship lacks the distinctive value associated with relating to one another as equals . . .nonsubjection requires not only that the parties in fact have roughly equal power, but also that the parties be committed to having equal power (Viehoff, 2014: 355; see also Kolodny, 2014: 295;cf. Moreau, 2020: 51-52).
As the last sentence points out, it does not suffice for an equal relation that one party has more power but never exercises this power.In that case, the powerholder could exercise his power to get the non-powerholder to do what he wants if he were to change his mind.This is classically explained in the example of the slave-master relation in which the slave still suffers from domination even if his master is benevolent since the master does have the power to arbitrarily interfere with the slave (see, for example, Pettit, 1997Pettit, , 2012)).Let us call this argument the relational egalitarian diversity argument in favour of affirmative action. 23It states that promoting diversity is good when and because it prevents power from being concentrated in some group because unequal power is incompatible with relating as equals and concludes that affirmative action is pro tanto justified in cases where affirmative action better promotes this form of diversity than non-affirmative action policies.
One may object that such affirmative action may at most equalize power between groups, but the remarks from Viehoff suggest that relational egalitarians care about equality of power between individuals, not between groups.So in what sense is this diversity argument actually relational egalitarian? 24I have two arguments for why it is a relational egalitarian argument.First, relational egalitarians care about how groups relate to each other.For instance, Anderson says: [t]he relational theory of inequality [relational egalitarianism] locates the causes of economic, political, and symbolic group inequalities in the relations (processes of interaction) between the groups, rather than in the internal characteristics of their members or in cultural differences that exist independently of group interaction (Anderson, 2010: 16).
Similarly, Voigt, in describing relational egalitarianism, says: [w]hile the exact requirements of relational equality have not yet been spelt out in much detail, the kinds of phenomena that relational approaches single out as problematic typically include hierarchies of certain kinds, oppression and stigmatization of particular individuals or groups (Voigt, 2018: 438-439; see also Schemmel, 2011a: 124;Young, 1990). 25  Second, how groups relate to each other may affect how individuals relate to each other.If women as a group stand as inferior to men as a group, individual women may internalize this and start to view themselves, as individuals, as inferior in relation to individual men (cf.Hindriks, 2017: 422). 26Højlund (2022: 4) illustrates this -and argues why it is objectionable according to relational egalitarianism -through the case of servility, where 'servility is a failure to recognize oneself as an equal moral agent with an equal standing to have one's views and claims taken seriously'.As she says, servility manifests itself in many ways.It may manifest itself in how women are more reluctant than men to run for political office (Højlund, 2022: 1).If the public view in society constantly reminds women as a group that they are inferior to men, it is not surprising that this eventually also influences how individual women regard and treat themselves in relation to individual men.Thus, that the diversity argument, in the first instance, focuses on group standing does not mean that it is not a relational egalitarian argument.
The scope of the relational egalitarian diversity argument in favour of affirmative action is different from the scope of the diversity argument focusing on epistemic benefits.The latter may justify affirmative action where the former does not.We may gain epistemic benefits by affirmative action in cases that do not affect the power distribution in society.In such cases, the argument focusing on epistemic benefits, but not the argument focusing on power, may justify affirmative action.The relational egalitarian diversity argument may also justify affirmative action where the diversity argument focusing on epistemic benefits does not, for example, when the student population has a sufficiently high level of diversity to gain epistemic benefits but where this does not hinder power from being concentrated in some group.Thus, the two arguments are not extensionally equivalent. 27 Note also that the diversity argument based on relational egalitarianism differs from Anderson's (2010) integrationist argument in favour of affirmative action -an argument that has received attention in recent years.Although the two arguments resemble each other, they are different.Writing in an American context, Anderson (2010: 148) contends that 'Americans live in a profoundly racially segregated society.De facto racial segregation unjustly impedes socioeconomic opportunities for disadvantaged racial groups, causes racial stigmatization and discrimination, and is inconsistent with a fully democratic society'. 28Affirmative action, in the integrationist argument, is justified as a way of securing non-segregation and non-stigmatization in society.
The integrationist argument is different from the diversity argument based on relational egalitarianism in that the latter may justify affirmative action where the former does not.Consider the case of women.In most Western societies, women are not segregated from the main institutions of society: 'they are not geographically concentrated in particular areas of cities, are not absent in mainstream social institutions and associations, and workplaces are typically mixed in terms of gender' (Lippert-Rasmussen, 2020: 146).In this way, women differ from other groups, such as African Americans (e.g.Shelby, 2007).When women are not segregated from the main institutions of society, the integrationist argument does not justify affirmative action in favour of women.However, the diversity argument based on relational egalitarianism may still justify affirmative action for women in this case because even if there is non-segregation between, say, men and women, there may still be unequal power between them.
In another sense, the integrationist argument is broader than the diversity argument based on relational egalitarianism.In a society in which different groups are almost totally segregated from each other, power may be equal between the groups, in which case, the diversity argument based on relational egalitarianism does not justify affirmative action, but the integrationist argument may justify affirmative action to secure less segregation in society.Thus, this shows that the integrationist argument in favour of affirmative action is different from the relational egalitarian diversity argument in favour of affirmative action.Note, finally, that an upshot of the arguments in sections 'The Mitigating Discrimination Argument in Favour of Affirmative Action', 'The Equality of Opportunity Argument in Favour of Affirmative Action' and 'The Diversity Argument in Favour of Affirmative Action' is that since the mitigating discrimination, equality of opportunity and diversity arguments may be based on a relational egalitarian understanding of justice, relational egalitarians who are in favour of affirmative action do not have to rely on Anderson's integrationist argument.This is an advantage inasmuch as we may want to pursue affirmative action in cases in which a group is disadvantaged with regard to power but where it does not suffer from segregation, for example, in the case of women.

The Site(s) of Affirmative Action
I have now discussed three of the most prominent arguments in favour of affirmative action, to wit, the mitigating discrimination argument, the equality of opportunity argument and the diversity argument.I argued that we may base these arguments on a relational egalitarian understanding of justice, as opposed to a distributive understanding of justice.These arguments in favour of affirmative action based on relational egalitarianism differ from the distribution-based arguments in favour of affirmative action in terms of when they justify affirmative action.Moreover, I argued that the diversity argument based on relational egalitarianism differs from Anderson's integrationist argument in favour of affirmative action.
What I will do in this section is to show that basing these arguments in favour of affirmative action on relational egalitarianism has an interesting implication when it comes to the site(s) of affirmative action, that is, with regard to where we should pursue affirmative action.In legal and philosophical contexts, affirmative action is usually discussed in relation to admissions into universities and in hiring on the job market (Anderson, 2010: 135).All else equal, pursuing affirmative action in the political domain is preferable to pursuing affirmative action in these sites according to the relational egalitarian-based arguments in favour of affirmative action, as I will now argue.
When the arguments in favour of affirmative action are based on a relational egalitarian understanding of justice, the ultimate purpose of affirmative action is to secure that people relate as equals, as opposed to inferiors and superiors, in the community (this is also to say that if there were not, and would not be, unequal relations, these arguments would not justify affirmative action).Importantly, the political domain is particularly important for securing equal relations, as Kolodny (2014) has forcefully argued (compare Anderson, 1999: 313).According to him, the political domain is particularly important for three reasons.
The first reason is that political decisions cannot usually be escaped at will.To be able to escape a decision at will makes a difference to what would otherwise be an unequal relation.If a slave can exit a slave contract, the relation between the slave and the master is not unequal in the way it is if the slave cannot escape the contract at will.Indeed, 'the freer one is to exit what would otherwise be a relation of social inferiority, the less it seems a relation of social inferiority in the first place' (Kolodny, 2014: 305).One cannot typically escape political decisions at will in the same way that one can escape nonpolitical decisions. 29 The second reason is that political decisions have final de facto authority.In motivating this, Kolodny asks us to: suppose that lord and servant set terms at the start of each year, somehow with genuinely equal influence, over how the lord is to boss the servant around . . .In such a case, the fact that they have equal influence over decisions higher up, as it were, the chain of command, which set the terms for how other, lower-order decisions are to be made, plays a role in avoiding, or moderating, the social inferiority that unequal influence over those decisions would otherwise entail (Kolodny, 2014: 305).
Thus, equality at the higher-order decisions eliminates, or at least moderates, the inequality between the lord and the servant with regard to lower-order decisions.Political decisions are important because they have final de facto authority, that is, they are, as it were, the highest-order decisions.This means, first, that if there is inequality in influence over political decisions, this cannot be moderated by equality in influence at a higher level.Second, equality of influence with regard to higher-order decisions, when those decisions have authority over lower-order decisions, moderates the threat to equal relations posed by unequal influence over lower-order decisions, or as Kolodny (2014: 306) aptly puts it, 'the threat to social equality that hierarchy would otherwise pose, one might say, is moderated by the fact that whatever hierarchy there may be is ultimately regulated or authorized from a standpoint of equality' (Kolodny, 2014: 306).
The third reason is that political decisions involve force.Force is particularly important to relational (in)equality 'because, as a contingent matter, the power to use force is the "final" power . . .[in the sense that it is] the power that usually determines the distribution of other powers' (Kolodny, 2014: 307).That I am smarter than you does not do me much good if you can subject me to force.'One cannot', Kolodny (2014: 307) explains, 'reliably have superior powers of other kinds over others where they have superior powers to subject one to force'.Thus, asymmetries in the capacity to use force is important for relational equality and political decisions, unlike non-political decisions, usually involve (the threat of) force.
Thus, political decisions are particularly important for relational (in)equality because (1) they cannot usually be escaped at will; (2) they have final de facto authority; and (3) they involve force.That political decisions are particularly important for relational (in) equality is important in relation to the relational egalitarian-based arguments in favour of affirmative action.Remember that these arguments -the mitigating discrimination argument, the equality of opportunity argument and the diversity argument -in their relational egalitarian understanding justify affirmative action as a way of securing equal, or less unequal, relations.We now know that the political domain is a particularly important site to secure equal, or less unequal, relations.In an ideal situation, equal opportunity for influencing political decisions may be the best means of securing equal relations (Anderson, 1999: 318;Kolodny, 2014: 291). 30However, in non-ideal situations, where some are inferiors in relation to others -which are the situations in which we usually discuss the justifiability of affirmative action -these reasons become reasons to implement affirmative action by giving unequal opportunity to influence political decisions, that is, by granting better opportunities to influence political decisions to those who are inferiors in the society in question to move them closer to equal relations with the superiors (compare Kolodny, 2014: 309).In this way, the political domain may be an efficient 'relations booster' because of the importance of political decisions for how we relate to each other.
If we were merely to implement affirmative action in university admissions or in hiring for jobs for those who are social inferiors, with the purpose of realizing equal, or less unequal, relations, there may still remain inequality at the higher-order decisions in the political domain.In this case, the equality at the lower level would be moderated from a standpoint of inequality; after all, the affirmative action policies at the lower-order level could be removed by a decision at the higher-order level. 31This is a problem from the point of view of relational egalitarianism.This problem would not arise (to the same extent) if we were to pursue affirmative action in the political domain.In that case, there may still be inequality with regard to lower-order decisions but that would be moderated by an inequality that favours the inferiors with regard to higher-order decisions. 32In this way, implementing affirmative action in the political domain would be better at securing equal, or less unequal, relations than implementing affirmative action in admissions to universities or in hiring for jobs, all else equal.This follows from the importance of the political domain for relational (in)equality.This means that, all else equal, we have more reason to pursue affirmative action in the political domain than in the sites usually discussed in philosophical and legal contextsthat is, in university admissions and in hiring for jobs -when we justify affirmative action based on relational egalitarianism. 33This is a surprising conclusion in the sense that I take it that most people prima facie find affirmative action more justifiable in university admissions or in hiring for jobs than in the political domain.

Concluding Remarks
This article has shown how three prominent arguments in favour of affirmative actionthe mitigating discrimination argument, the equality of opportunity argument and the diversity argument -may be based on a relational egalitarian understanding of justice, as opposed to a distributive understanding of justice.Affirmative action is usually discussed in relation to university admissions and in hiring for jobs.Basing the arguments in favour of affirmative action on relational egalitarianism has an interesting implication for the site(s) of affirmative action: it implies, all else equal, that we have more reason to pursue affirmative action in the political domain than in university admissions and in hiring for jobs.This is the case because the political domain is a particularly important constituent of relating as equals.In non-ideal circumstances, the political domain may thus be a particularly important site in which to improve the standing of members of a disadvantaged group vis-á-vis members of other groups.
I would like to end this article by briefly reflecting on how to pursue relational egalitarian-based affirmative action in the political domain.For reasons of space, I will not argue for which institutional solution relational egalitarians should choose.But I will, in a preliminary way, point to some considerations that relational egalitarians should take into account when they are to settle the institutional matter.We may distinguish two solutions.The first is what we may refer to as the differential voting weights solution. 34This solution grants a larger voting weight at the voting stage to individuals from the targeted group than to individuals from non-targeted groups.This means that they will be granted the opportunity to have a larger influence in the elections.In this way, this solution deviates (temporarily) from the common 'one person, one vote' rule.A benefit of this solution is that the inequality in opportunity in favour of members of targeted group at the higherorder level improves their standing vis-á-vis other people in the community.Whatever inequality there may be at the lower-order level to the detriment of the targeted group is moderated by an inequality at the higher-order level in favour of the targeted group (cf.Kolodny, 2014: 306).However, one worry about this solution is that some relational egalitarians -such as Anderson (1999: 313, 318) -are strict in arguing that relational egalitarianism requires that we do not deviate from a 'one person, one vote' scheme.This principle is as treasured as any. 35 To escape this worry, such relational egalitarians may instead pursue affirmative action in the political realm by using a quota which reserves x amount of seats in the legislative assembly to the targeted group.Let us call this the quota solution.This solution expectedly targets the most resourceful people from the targeted group (cf.Goldman, 1976: 191), that is, those who have the resources necessary to run for office.And they may then benefit other members of the targeted group through providing cultural and social capital (Anderson, 2010: 150) and by serving as role models (Allen, 2002;Bidadanure, 2015;compare Scanlon, 2018: 48). 36But this also points to a potential worry with this solution, namely, that it benefits the most resourceful people in the targeted group.This is a common argument against affirmative action. 37As Mulligan (2018a: 207) presents this criticism, 'because it is a racial, not a socioeconomic, instrument, it provides unjust benefits to blacks born into social advantage [compared, inter alia, to other black people and to white people born into poverty]'.The differential voting weights solution avoids this criticism, at least to a larger extent, since it targets every member of the targeted group directly.
This points to some of the considerations that relational egalitarians will have to take into account in deciding how to pursue affirmative action in the political domain.There may be further considerations, and there might also be further ways of implementing affirmative action in the political domain.What is clear, however, is that if we base arguments in favour of affirmative action on relational egalitarianism, we have more reason, all else equal, to pursue affirmative action in the political domain than in more traditional sites such as in university admissions and in the job market.
when it comes to nursing but he is given the job because he is a man (say, the hospital wants more diversity).I will not settle this issue since I will be focused on affirmative action for groups who are disadvantaged in the global sense.6.For instance, giving better paternity leave conditions than maternity leave conditions to make it more favourable for women compared to men to pursue their careers.7.In this article, I will not justify the latter assumption.I assume, alongside defenders of affirmative action, that there are some instances in which affirmative action is the best way of eliminating unjust discrimination.The same assumption -about affirmative action being the best means in some circumstances -applies to the equality of opportunity and diversity arguments discussed in sections 'The Equality of Opportunity Argument in Favour of Affirmative Action' and 'The Diversity Argument in Favour of Affirmative Action'.8.Many opponents of affirmative action argue that affirmative action is unjust because it is discrimination, that is, worse treatment of some compared to others based on race, sex and so on (e.g.Cohen in Cohen and Sterba, 2003;Hook, 2002;Pojman, 2014).Note, however, that Lippert-Rasmussen's definition of discrimination is non-moralized, that is, discrimination is not unjust per se.This is why the mitigating discrimination argument says that justice requires that we eliminate unjust discrimination.As previously explained, this is not a paper in which I defend affirmative action -I show what follows if we accept relational egalitarian-based arguments in favour of affirmative action -so I do not have to show that affirmative action in some cases is not unjust discrimination (but this, or at least that there would be less unjust discrimination overall if we pursued affirmative action, is what a defender of affirmative action would have to prove to avoid this charge from opponents).I thank an anonymous reviewer for asking me to clarify this.9.For reasons of space, I will use luck egalitarianism as the account of distributive justice throughout the paper, as this is arguably the most prominent theory of distributive justice.However, there are others, for example, a sufficientarian view (see, for example, Axelsen and Nielsen, 2015;Crisp, 2003;Frankfurt, 1997).Thus, when I speak to the differences and similarities between distributive and relational egalitarian-based arguments in favour of affirmative action, I am referring to differences and similarities between luck egalitarian-based affirmative action and relational egalitarian-based affirmative action.The arguments may not necessarily carry over to other theories of distributive justice, for example, a sufficientarian view may in general require less affirmative action than an egalitarian distributive view, assuming that realizing the former ideal is less demanding than realizing the latter (people may all have enough even though they do not have the same).For objections to affirmative action on grounds of distributive justice, see Mulligan (2018a).10.For an argument in favour of affirmative action in health based on luck egalitarianism, see Segall (2013: ch.9).11.The section 'The Site(s) of Affirmative Action' adds more to this description of relational egalitarianism.12.As I said above, I assume the following understanding of discrimination: X discriminates against Y relative to Z if, and only if, (1) X treats Y worse than Z; (2) this is (a) due to the fact that X believes that Y is a member of a certain socially salient group and believes that Z is not, or (b) explained in a certain way by the fact that Y is a member of a certain socially salient group and Z is not (Lippert-Rasmussen, 2020: 55).
Cases that satisfy 1 and 2a are cases of direct discrimination whereas cases that satisfy 1 and 2b are cases of indirect discrimination.For cases of indirect discrimination which violate relational equality, see Moreau (2020: ch.2).13.By exploitation, I mean 'an agent, A, taking unfair advantage of another agent, B', where it may be unfair in either (or both) a procedural sense or (and) a substantial sense (Zwolinski and Wertheimer, 2017).On exploitation violating relational equality, see Anderson (1999: 312).14.One of which is the following example: imagine a feudal society with two castes that swap position every 20 years.The first caste dominates the second for 20 years, then the second dominates the first for the subsequent 20 years, and so on.At the end of their lives, the two castes will have exerted equal amounts of control over each other (Bidadanure, 2016: 241 16.Remember that I by justice mean luck egalitarian justice (see footnote 9).17.Note also that there may be inequality of opportunity due to discrimination (compare the mitigating discrimination argument).18.This is in line with prominent relational egalitarians.As Lippert-Rasmussen says: [W]e find both in Anderson and Scheffler suggestions to the effect that equality of opportunity is not just desirable as an instrument for promoting suitably egalitarian relations, but that it is either independently desirable or desirable as a constitutive part of the relational ideal (Lippert-Rasmussen, 2018a: 211).
19.However, in many cases, these justifications will have the same implications.This is not surprising given that discrimination usually increases inequality of opportunity (Lippert-Rasmussen, 2020: 75).20.Since I base the diversity argument on relational egalitarianism, the argument will be different from Lippert-Rasmussen's version of the diversity argument in that it will be expressed in terms of what justice requires and not merely in terms of when affirmative action is morally justified.I thank an anonymous reviewer for asking me to clarify this.21.By focusing on epistemic benefits, the diversity argument may not necessarily be based on a distributive understanding of justice (although it could be).22.Perhaps a relational egalitarian argument could also point to epistemic benefits, for example, with a more diverse campus community, one gets a better appreciation of people's differences and become better equipped to deal with different people in egalitarian fashion.If so, there could be two different relational egalitarian-based diversity arguments in favour of affirmative action, one focusing on epistemic benefits, the other focusing on power.In this section, I will develop and focus on the power argument.I thank [Callum MacRae] for discussion on this.23.One may question whether this is not merely a particular version of the equality of opportunity argument which is focused on equality of opportunity for power.The equality of opportunity argument and the diversity argument in favour of affirmative action are clearly different.There may be substantive equality of opportunity, assuming the understanding of substantive equality of opportunity presented in 'The Equality of Opportunity Argument in Favour of Affirmative Action' section, in a society even though there is unequal power, for example, because those with more power than others never exercise their power.24.A related worry is that the example I give above -of pursuing affirmative action to secure a more diverse future elite of society -is a form of affirmative action that could at most equalize power between groups; it would not address the issue that there is a future elite of society.It would just make that elite more diverse.And an opponent of affirmative action may thus argue that this diversification is in fact entrenching the very inequalities that egalitarians should be trying to dismantle.Remember, however, that affirmative action is a tool that is needed under non-ideal circumstances (under conditions of background injustice).
Although it would be better if there was no future elite of society, I assume that such an elite would be present under non-ideal circumstances (it would not be feasible to just remove the elite).And given this assumption, it is certainly better, from the point of view of relational egalitarianism, that the elite is more diverse such that power is not concentrated within one group of people.I thank an anonymous reviewer for asking me to discuss these questions.25.My emphasis.26.For an empirical study of how this happens in the case of people with mental illness, see Evans-Lacko et al. (2012).27.This also means that there will be some cases that will not be captured if we move to a relational egalitarian-based justification of affirmative action.However, it is an open question whether this speaks in favour of or against relational egalitarian-based affirmative action.And, as mentioned in footnote 32, the luck egalitarian-based and relational egalitarian-based arguments in favour of affirmative action are not mutually exclusive.28.For an argument as to why it is unclear that Anderson's argument, contrary to her intention, should be understood as a non-distributive argument, see Lippert-Rasmussen (2020: 150-151).29.In those cases in which one cannot escape non-political decisions at will, they do pose a threat to equal relations (Kolodny, 2014: 305).30.For an argument as to why unequal opportunity for influencing political decisions is not necessarily incompatible with relating as equals, see Bengtson (2020).31.Cf.Pettit's (2012: 50; see also Pettit, 1997) notion of domination according to which 'someone, A, will be dominated in a certain choice by another agent or agency, B, to the extent that B has a power of interfering in the choice that is not itself controlled by A'. 32.I am not, in this article, taking a stand on whether relational egalitarian-based affirmative action should complement or replace distributive-based affirmative action.But let me note that the two approaches are not mutually exclusive.For instance, it would be possible to pursue relational egalitarian-based affirmative action at the higher-order level (the political domain) and distributive-based affirmative action at the lower-order level.33.Does luck egalitarian-based arguments in favour of affirmative action also support affirmative action in the political domain?They might in some circumstances, but not, I think, in the same categorical way as in the case of relational egalitarian-based arguments.If some are worse off in the political domain than others through no fault of their own, a luck egalitarian-based argument may favour affirmative action in the political domain.Suppose it is much more difficult for a Black person to be elected for office than a White person because most voters are racists and believe Black people to be inferior to White people.In that case, a luck egalitarian-based argument may support affirmative action in the form of reserving quotas in the legislative assembly for Black people.The reason I think that the political domain is not special in the same way for luck egalitarian-based arguments for affirmative action as it is for relational egalitarianbased arguments is that the former is ultimately concerned with distributions and not relations.In terms of securing distributions that only reflect individuals' relative exercise of responsibility, distributions in the economic sphere, or educational sphere, may be as important as distributions in the political sphere.If ascetic warriors secure such luck egalitarian distributions in society (and deprive themselves of personal liberties and material comforts to compensate for their administrative role), as Kolodny (2014: 294-295) explains, there will not be a luck egalitarian complaint, but there may well be a relational egalitarian complaint.If this is true, it seems that the political domain is more important for relational egalitarianbased arguments than for luck egalitarian-based arguments.But, as argued, this does not mean that luck egalitarian-based arguments never justify affirmative action in the political domain.34.For a recent approach to differential voting, see Mulligan (2018b).35.But see Bengtson (2020) for an argument as to why differential voting weights are compatible with equal relations.36.That the quota solution does provide gains for the targeted minority has recently been confirmed in a study by Gulzar et al. (2020).They explore the quota scheme in India where seats are reserved for historically disadvantaged Scheduled Tribes and find that 'quotas deliver no worse outcomes overall and that gains for targeted minorities come at the cost of the relatively privileged rather than other historically disadvantaged groups' (Gulzar et al., 2020(Gulzar et al., : 1245)).Although they focus on distributive effects, we can expect that less distributive inequality may have a beneficial effect on social relations as well, at least when there is a large degree of distributive inequality in society (Anderson, 2010: 18-19;Lippert-Rasmussen, 2018a;Schemmel, 2011b).37.For a discussion of this argument, see, for example, Lippert-Rasmussen (2020: ch.10).