I am delighted to have the opportunity to respond to Benjamin Eidelson’s excellent and provocative article, Patterned Inequality, Compounding Injustice, and Algorithmic Prediction,1 and I am deeply grateful for his incisive engagement with my work. At issue is an aspect of the normative underpinning of anti-discrimination law and, especially, why policies and practices that produce a disparate negative impact on disadvantaged groups are morally troubling. As Eidelson explains, this question is particularly relevant to the use of algorithmic tools because such tools, predictably, produce this effect. For Eidelson, the reason that disparate impact on the basis of race, sex, and other legally protected traits is problematic is because the disparate impact continues and reinforces patterned inequality, which itself has serious consequences for the well-being of individuals and for society.
Eidelson and I agree that patterned inequality is morally problematic for the reasons he discusses. However, I also contend that each of us has an important moral reason to avoid compounding prior injustice. The heart of our disagreement thus centers on two questions. First, does the etiology of the disadvantage matter? For Eidelson, it does not; in his view, the reason for the patterned inequality has no moral significance. What matters is that it exists and causes harm.2 For me, by contrast, the causal etiology is relevant. I emphasize the backstory and ask whether an action or policy will compound prior injustice. If so, we have an important additional reason to avoid entrenching that injustice. The second question to which we provide different answers is this: Does the actor who compounds injustice, or reinforces patterned inequality, wrong the particular individual or group affected? For Eidelson, she does not. Rather, the bad consequences of perpetuating patterned inequality provide important moral reasons to disrupt those patterns and so justify anti-discrimination laws that prohibit both disparate treatment and disparate impact. But the individual or institution that entrenches, rather than unsettles, these patterns does not wrong the individuals affected by the policy she adopts. By contrast, I contend that individuals and institutions have personal obligations to affected individuals to avoid compounding injustice. At issue in this disagreement is the nature of the responsibility of each actor when faced with the consequences of an unjust world.
In what follows, I draw out the differences between our views on these two questions and respond to Eidelson’s insightful critiques.
I. THE CAUSAL STORY AND WHY IT MATTERS
According to Eidelson, the reason that patterned inequality exists is of no moral significance. Although he acknowledges that the patterns of racialized inequality in the United States do result from injustice, he stresses that “the moral case for reducing racially patterned inequality does not depend on that backstory; it just depends on the pattern of present facts that this history brought about.”3 In order to motivate that view, he asks the reader to imagine a society in which the same racialized patterns of inequality exist as in the current United States, “the same social meanings, the same patterns of material advantage and disadvantage, and so on—had somehow come about innocently.”4 In such a society, he contends, we would have the same reasons to disrupt these patterns as we do in a society such as ours where such patterns are the result of injustice.
This is an argument that rests on an intuition that he presents to the reader and that he hopes the reader will share. I am not sure that I do. I agree that we would have reason to reduce and dismantle the patterned inequality, as it clearly has negative effects. But would we have the same reason, and would the reasons be as pressing, as is the case when the patterned inequality is caused by prior injustice? Frankly, it is hard to know what to think about such a hypothetical because it is hard to imagine how such a world could come about. In saying this, I am not objecting to the strangeness of philosophical thought experiments generally. In most cases, I find them useful in teasing apart intuitions and in helping us to discover which features of a situation or legal case are doing the moral work. Rather, the problem is with this particular thought experiment. I take it as a sort of article of faith that there are no inherent differences among groups of people (e.g., races, cultural groups, sexes) that explain the observed group differences with regard to factors that are of legitimate interest to employers, lenders, universities, and others.5 If this faith is well placed, then my working supposition is that such observed differences are the result of injustice. That being the case, it is simply hard to grasp how there could come to be such patterned inequality, with all the same features that exist in the caste-like world we inhabit, without a causal backstory rooted in injustice. Thus, when I imagine such an unalike world, I lose my moral bearings and simply do not know what to think.
Be that as it may, Eidelson’s argument for the superfluity of the causal etiology of patterned inequality does not rest on this hypothetical alone. Indeed, he uses it merely to make initially plausible the idea that it is the patterned inequality itself that matters rather than how it came about. The heart of his argument rests on (1) showing how detrimental patterned inequality is and (2) showing that the pull that the reader may feel for the importance of the causal etiology can be liquidated, if you will, by what he terms prioritarian regard combined with the seeming expression of endorsement for the prior injustice. By prioritarian regard, Eidelson refers to the view that we should prefer actions that do not make things worse for people who are already badly off, especially when that occurs through no fault of their own.6 As I agree that patterned inequality is harmful and that its harmfulness provides a reason for us to work to dismantle it, I have no objection to the first part of this argument. I thus focus my attention on the second part of his argument. In my view, the combination of prioritarian regard and the expressive endorsement of the injustice that is often present does not fully explain the intuitive pull of the felt need to avoid compounding prior injustice.
A. The Explanatory Power of Prioritarian Regard
Eidelson rightly points out that the examples I use to demonstrate the force of the intuition that we have a duty not to compound injustice are over-determined. In my examples, the victim of injustice is also a person who is not well-off, at least as compared to some others. Therefore, part of the intuition on which I draw might be explained by the precept that in deciding how to act, we should prioritize the least well-off. If so, Eidelson points out, perhaps prioritarian regard can liquidate the force of these examples, at least to some degree.
I am not sure that prioritarian regard can do that. Take the example of the battered woman who is charged higher rates than otherwise similar purchasers of life insurance because she is more likely than they are to collect on the policy during its term. It is true, as Eidelson points out, that the “burdened class here is defined by an undeserved injury, so choosing to make the lots of those people even worse … is a problematic choice from the point of view of prioritarian regard.”7 In other words, the fact that you might think the abuse victim shouldn’t be charged higher rates than otherwise similar insurance purchasers is not due to the fact that she is a victim of injustice, but instead is due to the fact that she is among the least well-off.
But is the abuse victim different from others who pay higher-than-average rates in this respect? The life insurer charges rates that are actuarially accurate, meaning that they accurately reflect the risk each category of insurance purchaser poses. It is true that some prospective purchasers who are charged higher-than-average rates will not be appropriate subjects of prioritarian regard: the aged, who have had the good fortune of long life, and smokers, drinkers, and others whose lifestyle choices make them poor risks and so arguably deserve to pay more. Still, many others—people who are ill, for example—are no different from the abuse victim in terms of whether they should be the subject of prioritarian concern. Yet, many people think charging higher life insurance rates to battered women is different and worse than charging higher life insurance rates to those who are sick. If this is correct, prioritarian regard cannot account for this difference.
B. Confusing Intuitions About Expressive Acquiescence
Eidelson argues that “[o]nce we factor out both the effect of the employer’s conduct on the maintenance of patterned inequality and the prioritarian regard” all that is left is a “residual concern” that “in making use of unjust social conditions, one expresses a kind of comfort with or acquiescence in them.”8 The expressive acquiescence is morally relevant but the problem, as he sees it, is that this expression occurs only in some of the cases that I claim to be instances of compounding injustice. In his view, in cases where prioritarian regard is inapt, and there is no expressive acquiescence in injustice, there is no wrong left to explain. As a result, he concludes, there is ultimately nothing to the idea of compounding injustice.
Building on the work of Patrick Shin, Eidelson thinks that some cases that I term compounding injustice are better understood as ones in which the actor exploits the prior injustice in a way that amounts to expressing a positive orientation to that injustice. If so, expressing such positive orientation does (or can) wrong the victim of that injustice. But not all instances of compounding injustice (according to my account) do involve such problematic exploitation.
I have two replies to his argument here. First, I think expressive acquiescence in injustice can occur even in cases in which the actor does not exploit the injustice. Second, even when the action carries no such expressive acquiescence, I think there is still a wrong to be explained and that the idea of compounding injustice provides a compelling account.
Consider first what is expressed in cases in which there is no exploitation. For example, suppose an employer selects applicants on the basis of academic achievement, and further, suppose that racial differences in academic achievement are the result of prior injustice. For Eidelson, the fact that in a world without injustice, the employer would still find it useful to select employees on this basis means that this is not a case of exploitation. As a result, he would say that this employer does not express acquiescence in the prior injustice by using educational attainment as a screening method. Eidelson does not deny that the employer might express something, and indeed, that what the employer expresses in this situation could be morally troubling. As he explains, the employer expresses “not an attitude toward past injustices, but a troubling willingness to sustain a society characterized by patterned inequality and its devastating effects.”9 For Eidelson, these two attitudes are importantly different. But are they? My doubts about this are similar to my uncertain reaction to his thought experiment regarding a society with patterned inequality and no prior injustice. If the existence of patterned inequality is necessarily tied to prior injustice, it will be difficult to pry apart intuitions about the real world, as compared to Eidelson’s hypothetical one. For the same reason, it will be difficult to separate expressing acquiescence in prior injustice from expressing acquiescence in the continuation of patterned inequality. For that reason, I think it is more likely than Eidelson acknowledges that what the employer (or other relevant actor) expresses in these situations is an acquiescence in prior injustice. This is one reason I think his deflation via exploitation argument is unsuccessful.
If we put this argument aside and assume that Eidelson is correct about the expressive claim, he then concludes that there is nothing left to the idea of compounding injustice. The intuition I mine based on the cases I offer are, in his view, completely liquidated by prioritarian regard and expressive acquiescence in injustice. Where neither applies, there is nothing of moral concern that remains. To make this point, Eidelson relies on a twist on my early release case in which the state uses a personality test to determine whom to release early.10 Eidelson posits that victims of child abuse are more likely to have a personality trait that is correlated with recidivism and so are less likely to be eligible for early release. On my account, the state’s use of this personality test will compound injustice, which is a conclusion that Eidelson finds problematic.
If there is no expression of acquiescence in the injustice (a point I disputed earlier but will accept for the sake of argument here), how could the abuse victim be wronged by the state? To Eidelson, the fact that the personality test would be a useful tool, even in a world with no child abuse, denudes its use of expressive problems and so there is nothing wrong with using it. In my view, the combination of amplification and mixing—where an actor takes the fact of the injustice or its direct effects as his reason for action and thereby entrenches that injustice—is wrong. The disagreement centers on the idea of “mixing” and what is required for an actor to involve himself in the prior injustice. Eidelson challenges my account by pressing that “it is difficult to see how his incorporating what are in fact ‘effects of the injustice’ in his decision—without incorporating them under any description relating to the injustice—would ‘implicate him in the [prior] wrongdoing.’”11 For Eidelson, the actor must “orient his or her agency to the prior wrong” to even plausibly mix with it. For me, this active orientation is not necessary. For me, negligence or passivity in the face of injustice suffices for moral responsibility. This may be simply a point at which our intuitions lead in different directions. If so, that is helpful to clarify, and this clarification allows the reader to mine her own intuitions on these points.12
II. PERSONAL WRONGS AND UNCERTAINTY ABOUT CAUSATION
The second dimension of difference between Eidelson’s view and my own relates to the responsibility of the actor—whether an individual or an institution—who adopts a policy that continues the patterned inequality (Eidelson’s perspective) or compounds the past injustice (my perspective). Does the actor wrong the person negatively affected by her action? For Eidelson, there is a personal wrong only in cases in which the action expresses something morally troubling. When this problematic expression is absent, the actor does not wrong the victim of injustice, and what the actor owes in such a context “is precisely what it owes to others who are equally badly off, bear equal responsibility for their circumstances, and pose equal risks to others.”13
In response, I want to make an Eidelson-esque move. The sort of case we are considering is one in which there is a screening tool that would be employed in a world without injustice: educational attainment for employment, the psychological test for early release in the hypothetical proposed by Eidelson. In such cases, Eidelson believes that the employer and the state do not wrong the victim of injustice, even where the harm of that prior injustice is augmented or entrenched. Here I want to suggest that the reason these cases feel less compelling—and thus lead Eidelson to claim the victim of prior injustice is not wronged by the act in question—can be traced to uncertainty about whether the particular individual at issue has or lacks the relevant trait due to prior injustice. If only victims of injustice had weak educations or bad scores on the personality test, then Eidelson couldn’t posit that the screening tools would be used in a world without injustice. So, it must be the case that some employees and inmates have these traits absent injustice. As a result, with regard to any particular person with a weak education or a bad score on the personality test, we (and the state or employer) cannot be sure whether that person is in fact a victim of injustice. This uncertainly may be infecting our intuitions about whether a person who is a victim of injustice has been wronged.
Eidelson characterized our disagreement as intramural, in that we both agree that algorithmic selection processes involve problems that go beyond mischaracterization.14 In addition, it may seem merely theoretical. When and how would it matter whether discrimination law should prohibit certain actions because they will further patterned inequality or instead because they will compound prior injustice? In many instances, it will not. Each of our accounts can explain and justify much anti-discrimination law. For example, both of our accounts provide a justification for disparate impact liability. But, at other times, the different justifications we each offer may matter.15
The recent debate about how to interpret Section 2 of the Voting Rights Act presents an example. When a facially neutral law affects voting, a question arises regarding whether the law at issue denies a person protected by the statute an equal opportunity to vote.16 This question was at issue in the recent case of Brnovich v. Democratic National Committee.17 In this case, the Supreme Court considered two aspects of Arizona law, one which limited who could collect completed early ballots and the other requiring the rejection of election-day ballots cast in the wrong precinct.18 Lower courts applying Section 2 in similar contexts have addressed questions like this one using an analysis that resembles the approach I endorse. These courts have asked whether the laws disproportionately burden minority voters (the disparate impact prong), and whether this disparate impact results from the interaction between the law at issue and prior racial discrimination (the compounding injustice prong—to use my terminology).19 Such an approach differs from traditional disparate impact analysis under other statutes,20 and my account provides a justification for this alternative.
The fact that a law yields a disparate racial impact because of its interaction with prior race discrimination matters precisely because the law will compound injustice. As Dale Ho puts what amounts to the heart of my argument: What Section 2 requires is that states “refrain from adopting restrictions on voting that reiterate or amplify existing patterns of racial discrimination or exclusion.”21 For Eidelson, by contrast, this amplification of prior exclusion would not be relevant. Rather, he would reject the emphasis on the etiology of the disparate impact and instead focus only on whether such disparate impact is justified. For me, by contrast, the approach that lower courts have adopted in these vote denial cases is on to something important. Moreover, this emphasis on the interaction between current law and past practices is not anomalous. In fact, it implicitly animates aspects of our equal protection doctrine as well.22
For now, this debate about Section 2 will remain academic as Justice Alito, writing for the Court in Brnovich, rejected both of our views. He explicitly rejected the singular focus on disparate impact23 and ignored the argument that the interaction between a current law and conditions caused by past injustice is relevant.24 That said, the debate between Eidelson’s view and my own could again become relevant. If, and when, it does, courts will be called upon to assess whether it is patterned inequality alone that matters or instead whether the compounding of injustice is also morally, and legally, relevant.
To close, let me reiterate how much I have learned from the challenges to my view so clearly and forcefully articulated by Eidelson. In this exchange, and in our conversations, Eidelson’s arguments have made me reexamine by views and revealed to me where they are underdeveloped and in need of elaboration and refinement. For that, I am deeply grateful.
Benjamin Eidelson, Patterned Inequality, Compounding Injustice, and Algorithmic Prediction, 1 Am. J. L. & Equal. 252.
For Eidelson, “the very existence of patterned inequalities gives rise to grave social ills and undermines human flourishing in ways that other inequalities do not.” Id. at 257.
Id. at 263.
I explore this contention in Deborah Hellman, The Epistemic Commitments of Nondiscrimination, in Oxford Studiesinthe Philosophyof Law (John Gardner et al. eds., forthcoming 2021).
Eidelson, supra note 1, at 268.
Id. at 268–69.
Id. at 271.
Id. at 274.
Id. at 272–73.
Id. at 273.
I am very grateful for the exchange with Eidelson on this point. His challenges have helped me to get a better grasp on my view. However, I remain somewhat unsatisfied with the way I have captured my position and encourage readers to engage in the Rawlsian process of reflective equilibrium Eidelson suggests both to determine whose intuitions you share and what principles support them. Id. at 274, n.83.
Id. at 273.
Id. at 256.
Eidelson mentions the case of Dothard v. Rawlinson, 433 U.S. 321 (1977), as one such example. Id. at 266, n.60. As Eidelson notes, the compounding injustice account would not cover this case, as the reason women are shorter than men on average is not the result of injustice. As I agree that disrupting patterned inequality is also important, I am happy to endorse the outcome of this case on other grounds. That said, I think it presents a less morally pressing case than others where injustice is compounded. The fact that my account treats a case like Dothard as peripheral therefore vindicates its moral appeal in my view.
Section 2(b) of the Voting Rights Act specifically provides that:
A violation of subsection (a) is established if, based on a totality of the circumstances, it is shown that the political processes leading to the nomination or election in the State or political subdivision are not equally open to participation by members of a class of citizens protected by subsection (a) in that its members have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice.
Nos. 19-1257 & 19-1258, slip op. (U.S. July 1, 2021).
Ariz. Rev. Stat. Ann. § 16-1005(H) (Supp. 2020). For a description of Arizona’s out-of-precinct policy, see Democratic Nat’l Comm. v. Hobbs, 948 F.3d 989, 999–1000 (9th Cir. 2020).
See, e.g., Ohio St. Conf. of the NAACP v. Husted, 768 F.3d 524, 544 (6th Cir. 2014); see also Dale E. Ho, Building an Umbrella in a Rainstorm: The New Vote Denial Litigation Since Shelby County, 127 Yale L. J. 799, 821 (2018) (surveying the post–Shelby County vote denial cases and explaining that in Ohio, Wisconsin, and North Carolina plaintiffs made claims that tied the disparate impact to prior injustice and discrimination).
Nicholas O. Stephanopoulos, Disparate Impact, Unified Law, 128 Yale L. J. 1566, 1570 (2019) (arguing for a unified approach to disparate impact across various statutes and, in particular, for resisting a different approach under Section 2 and emphasizing that “courts have focused on the explanations for racial disparities in voting, especially the extent to which they are intertwined with past and present discrimination” and lamenting that development).
Ho, supra note 19, at 822.
See, e.g., Deborah Hellman, Sex, Causation, and Algorithms: How Equal Protection Prohibits Compounding Prior Injustice, 98 Wash. U. L. Rev. 481 (2020).
See Brnovich v. Democratic Nat’l Comm., Nos. 19-1257 & 19-1258, slip op. at 20 (U.S. July 1, 2021) (noting that “the disparate impact model employed in Title VII and Fair Housing Act cases” is not useful).
The rejection of the compounding injustice approach can be seen in the Court’s distancing itself from Thornburg v. Gingles, 478 U.S. 30, 47 (1986) (a vote-dilution case), which stressed that “[t]he essence of a §2 claim is that a certain electoral law, practice, or structure interacts with social and historical conditions to cause an inequality in the opportunities” of minority voters to vote. Seeid.
Deborah Hellman is the David Lurton Massee, Jr., Professor of Law and the Director of the Center for Law & Philosophy at the University of Virginia School of Law. Discrimination law and its moral foundations are a main focus of her work. She is the author of When Is Discrimination Wrong? (2008) and coeditor of The Philosophical Foundations of Discrimination Law (2013). In addition, she has written articles for both law journals and philosophy journals on topics related to discrimination and to the fairness of algorithmic decisionmaking.