In a series of excellent, recent papers, Deborah Hellman expounds the intuitively appealing idea that we have a duty not to compound injustice. Roughly, one compounds injustice when facts that obtain as a result of prior injustice form part of one’s reason for imposing further disadvantages on the victims of this prior injustice. This article identifies several complexities and problems motivating various amendments to Hellman’s formulation of the duty not to compound injustice. Critically, it argues that the intuitions she and others have sought to explain in terms of the duty not to compound injustice are better explained by the duty not to cause additional harm to people who are already unjustly worse off – perhaps in conjunction with a duty not to implicate ourselves in injustice.
This is a preview of subscription content, access via your institution.
Buy single article
Instant access to the full article PDF.
Price excludes VAT (USA)
Tax calculation will be finalised during checkout.
Deborah Hellman, “Indirect Discrimination and the Duty to Avoid Compounding Injustice”, in H. Collins and T. Khaitan (eds.), Foundations of Indirect Discrimination Law (Oxford: Hart Publishing, 2018), pp. 105–122; “Measuring Algorithmic Fairness”, Virginia Law Review 106(4) (2020): pp. 811–866; “Sex, Causation, and Algorithms” Washington University Law Review).
See also Rima Basu, “Radical moral encroachment: The moral stakes of racist beliefs”, Philosophical Issues 29 (2019): pp. 9–23, p. 17; Renee Jorgensen Bolinger, “The rational impermissibility of accepting (some) racial generalizations”, Synthese 197 (2020): pp. 2415–2431, p. 2417, p. 2424; Dan W. Brock, “Cost-Effectiveness and Disability Discrimination”, Economics and Philosophy 25 (2009): pp. 27–47, p. 35
See also Hellman, “Indirect Discrimination”, p. 121; Griggs v. Duke Power Co., 401 U.S. 424 (1971).
Julia Angwin et al., “Machine Bias: There’s Software Used Across the Country to Predict Future Criminals and It’s Biased against Blacks”, ProPublica (2016); Bernard Harcourt, “Risk as a Proxy for Race: The Dangers of Risk Assessment”, Federal Sentencing Reporter 27 (2015): pp. 237–243; Sonja Starr, “Evidence-Based Sentencing and the Scientific Rationalization of Discrimination”, Stanford Law Review 66 (2014): pp. 803–872.
Alexandra Chouldechova, “Fair Prediction with Disparate Impact: A Study of Bias in Recidivism Prediction Instruments”, Big Data 5 (2) (2017): pp. 153–163, p. 153, pp. 157–158; Hellman, “Measuring Algorithmic Fairness”, pp. 820–828; Northpointe Inc., “COMPAS Risk Scales: Accuracy Equity and Predictive Parity” (Northpointe, 2016), p. 8.
John Lightbourne, “Damned Lies and Criminal Sentencing Using Evidence-Based Tools”, Duke Law and Technical Review 15 (2017): pp. 327–343, p. 339; John Monahan, “A Jurisprudence of Risk Assessment”, Virginia Law Review 92 (2006): pp. 391–435, p. 416; Starr, “Evidence-Based Sentencing, p. 813. A few theorists even argue that it is unfair to women not to consider their lower base-rate recidivism risk (Margareth Etienne, “Sentencing Women”, Journal of Gender, Race & Justice 14(1) (2010): pp. 73–84, p. 82).
The duty not to compound injustice also explains why other algorithm-driven ways of treating people differentially, e.g., why it is wrong for employers to use big data on prior job experience for candidate selection purposes thereby further disadvantaging women with less job experience (Jon Kleinberg et al., “Discrimination in the Age of Algorithms”, Journal of Legal Analysis 10 (2019): pp. 113–174, pp. 139–140), and why it is wrong for financial institutions to use algorithms to assess who qualifies for a prime loan on the housing market thereby compounding injustice against African Americans (James Allen, “The Color of Algorithms: An Analysis and Proposed Research Agenda for Deterring Algorithmic Redlining,” Fordham Urban Law Journal 46(2) (2019): pp. 219–270, pp. 237–239; Matthew A. Bruckner, “The Promise and Perils of Algorithmic Lenders’ Use of Big Data”, Chicago-Kent Law Review 93(1) (2018): pp. 3–60, pp. 28–29; A.C. Ghent et al., “Differences in Subprime Loan Pricing across Races and Neighborhoods”, Regional Science and Urban Economics 48 (2014): pp. 199–215).
Daniel Butt, “On Benefiting from Injustice”, Canadian Journal of Philosophy 37 (2007): pp. 129–152; Alexandra Couto, “The Beneficiary Pays Principle and Strict Liability: Exploring the Normative Significance of Causal Relations”, Philosophical Studies 175 (2018): pp. 2169–2189; Göran Duus-Otterström, “Benefiting from Injustice and the Common Source Problem”, Ethical Theory and Moral Practice 20 (2017): pp. 1067–1081; Robert E. Goodin, “Disgorging the fruits of historical wrongdoing”, American Political Science Review 107 (2013): pp. 478–491; Robert E. Goodin and Christian Barry, “Benefiting from the Wrongdoing of Others”, Journal of Applied Philosophy 31(4) (2014): pp. 363–376; Bashar Haydar and Gerhard Øverland, “The Normative Implications of Benefiting from Injustice”, Journal of Applied Philosophy 31 (2014): pp. 349–362; Edward Page, “Give It up for Climate Change: A Defence of the Beneficiary Pays Principle”, International Theory 4 (2012): pp. 300–330; Tom Parr, “The Moral Taintedness of Benefiting from Injustice”, Ethical Theory and Moral Practice 19 (2016): pp. 985–997; Judith Jarvis Thomson, “Preferential Hiring”, Philosophy & Public Affairs 2 (1973): pp. 364–384; for critiques, see Robert Huseby, “Should the Beneficiaries Pay?” Politics, Philosophy & Economics 14(2) (2015): pp. 209–225; Kasper Lippert-Rasmussen, “The Benefits of Injustice and Its Correction”, Journal of Political Philosophy (2021), online first: https://doi.org/10.1111/jopp.12263; “Affirmative Action, Historical Injustice, and the Concept of Beneficiaries”, Journal of Political Philosophy 25(1) (2017): pp. 72–90.
Hellman, “Indirect Discrimination”, p. 113.
Explaining the implication component, Hellman writes, “the actor must take the prior injustice or its effects as her reason for action” (“Indirect Discrimination”, p. 113. It is natural to assume that by “reason for action”, Hellman has the agent’s motivating reason for action in mind, i.e., the reasons the agent take to favour and justify her action and that guide her in acting (whether consciously or not). There are other notions of reasons, but I am not sure they would sustain the charge of implication.
On this definition, the amplification element is defined in terms of actual effects. It could be defined with expected effects. My arguments below apply to probabilistic versions of the amplification element too, but I do not have the space to discuss these issues further.
Hellman, “Indirect Discrimination”, p. 108.
Griggs v. Duke Power Co. mentioned in Section I is a case of this kind.
Hellman, “Indirect Discrimination”, p. 116.
Hellman, “Indirect Discrimination”.
Hellman, “Indirect Discrimination”, pp. 110–111.
This problem is similar to the one about closeness in the doctrine of double effect (Philippa Foot, “The Problem of Abortion and the Doctrine of Double Effect”, in P. Foot, Virtues and Vices (Berkeley: University of California Press, 1978), pp. 19–32, p. 21; Victor Tadros, “Wrongful Intentions Without Closeness”, Philosophy & Public Affairs 43(1) (2015): pp. 52–74; Gideon Yaffe, “Criminal Attempts”, The Yale Law Journal 124 (2014): pp. 92–156, pp. 106–115.
It is controversial whether this sort of counterfactual offers the right sort of test of what one intends (Frances Kamm, Intricate Ethics (Oxford: Oxford University Press, 2007), pp. 21–22).
A further question here is whether the effect needs to be negative. Sometimes injustices have positive effects for the wronged party (see note 34).
Hellman, “Indirect Discrimination”, p. 113.
It might be replied that while G1 wrongs G2 by compounding injustice against G2, G1 does so non-culpably and, thus, that Hellman can non-counterintuitively insist that Failed Attempt I does involve a violation – albeit a non-culpable one – of the duty not to compound injustice. In my view, this reply does not work even though I accept that one can wrong another person, e.g., by violating a right of that person, non-culpably, e.g., because one has no way of knowing that one is violating that person’s rights. Suppose we tweak Failed Attempt I such that G1 reasonably believes that the affirmative action scheme will benefit G2’s members and intends to bring about these benefits, but that the prior injustice to G2 in no way determines G1’s action and, thus, involves no violation of a duty not to compound injustice. I do not believe this variant of Failed Attempt I differs from Failed Attempt I in terms of wrongness in the way the proposed reply implies, i.e., that these two cases differ with respect to whether they involve the wrong of compounding injustice. (On my view, neither does.)
Some might think that this is so because G1 acts disrespectfully towards – perhaps even demeans – G2, and that G1 thereby wrongs G2. An account along those lines would seem congenial to Hellman’s (When Is Discrimination Wrong? (Cambridge: Harvard University Press, 2008) respect-based account of what makes discrimination wrong.
I am not sure that, ultimately, we should embrace a duty not to implicate ourselves in injustice. (For a restrictive view of, roughly, such a duty applying only to cases of exploiting the fact of injustice, see Benjamin Eidelson, “Patterned Inequality, Compounding Injustice, and Algorithmic Prediction”, American Journal of Law and Equality 1(1) (2021): pp. 252–276, pp. 270–275). My point here is simply comparative: I am saying that this duty (and a duty not to cause additional harms to the unjustly worse off – see below) fits the relevant range of moral intuitions better than the duty not to compound injustice.
The strictness and the laxity problems reappear in this connection. However, since they are common to the duty not to compound injustice and the duty not to implicate ourselves in injustice, they cannot help us to determine which of the two duties we have most reason to accept, and thus they can be ignored here.
To see the need for the qualification “or in some other morally appropriate way” consider a case where we take a prior injustice as our reason for educating the wrongdoer about the wrong that they committed, but without mitigating or eliminating the injustice. (Suppose these two responses are not on the table.) Without the qualification this would count, counterintuitively, as a case of wrongful implication. However, assuming, plausibly, that moral education is a morally appropriate response to wrongdoing, and provided we include the qualification, it is not a case of that. Admittedly, it is not entirely clear what would count as morally appropriate ways of addressing injustice, and therefore this qualification leaves the duty not to implicate ourselves in injustice somewhat fuzzy. However, for present purposes – to construct a challenge to the duty not to compound injustice – this is a relatively minor complication.
Hellman, “Indirect Discrimination”, p. 109; see also Stephen G. Post, “The Echo of Nuremberg: Nazi Data and Ethics”, Journal of Medical Ethics 17 (1991): pp. 42–44, p. 43; Arthur Schafer, “On Using Nazi Data: The Case Against”, Dialogue 25(3) (1986): pp. 413–419, p. 418.
If one accepts the possibility of posthumous dignitary harm (e.g., Joel Feinberg, Harm to Others (New York: Oxford University Press, 1984), pp. 86–87; T.M. Wilkinson, Ethics and the Acquisition of Organs (New York: Oxford University Press, 2011), pp. 29–42), one might disagree with this point.
Hellman, When Is Discrimination Wrong?, esp. pp. 29–37.
Failed Attempt II is an attempt to compound injustice, and, generally, failed attempts to treat people disadvantageously (etc.) are no less demeaning – and, thus, on respect-based accounts of the wrongness no less wrong – than successful attempts.
Some of the cases I discuss in Sections IV and V rely on this distinction. Note also that some occurrences of the phrase “compounding injustice” refer to duties not to cause additional harms to the unjustly worse off rather than to Hellman’s duty not to compound injustice (e.g., Allen, “The Color of Algorithms”, p. 234) in relation to segregation and algorithmic discrimination and H. Shue, “The Unavoidability of Justice”, in A. Hurrell and B. Kingsbury (eds.), The International Politics of the Environment (Oxford, England: Oxford University Press, 1992), pp. 373–397).
Hellman, “Indirect Discrimination”, p. 107.
Hellman, “Indirect Discrimination”, p. 10.
We need to know more about how to distinguish different domains from one another, and about why transferring harm from one sphere to another sphere matters morally. Here I simply rely on some familiar classifications that appear plausible.
In several places, Hellman writes about amplifying the “injustice” rather than amplifying the “harm” – e.g., “In order to satisfy this element, the actor must not only harm the victim, she must also amplify the original injustice” (Hellman, “Indirect Discrimination”, p. 109). In simple cases, when an injustice causes harm, and one amplifies that harm, one also amplifies the injustice. But perhaps there can be harmless injustices, and thus ways of amplifying such injustices without amplifying any harm. Consider an unjust paternalistic act benefitting the recipient. Suppose I take the benefit (or this injustice) as my reason for treating the recipient differently in a different domain, perhaps thereby benefiting this person even further in an unjust paternalistic way. Perhaps by Hellman’s lights this should be seen as a case of compounding injustice too, even if it involves only benefits and no harms. If so, her definition needs to be slightly amended. I will be ignoring this issue and focusing on the simple cases.
Its policy of charging female victims of domestic violence higher premiums could, for instance, be seen as demeaning (see Section III), but this charge would be a different one from the one of compounding injustice.
Probably, Hellman’s primary concern is to show that certain legal norms cohere with underlying moral norms. Since the main cases she considers – cases of indirect discrimination and cases of algorithmic discrimination – are not cases where the prior harm is reduced while being transferred to a different sphere, from a practical point of view, the present possibility is irrelevant to her primary concern.
Ben Bradley, “Doing Away with Harm”, Philosophy and Phenomenological Research 85 (2012): pp. 390–412; Feinberg, Harm to Others, pp. 136–145; Nils Holtug, “The Harm Principle”, Ethical Theory and Moral Practice 5(4) (2002): pp. 357–389; Lippert-Rasmussen, “Indirect Discrimination Is Not Necessarily Unjust”, Journal of Practical Philosophy 2(2) (2014): pp. 33–57, pp. 44–48; Born Free and Equal? (Oxford: Oxford University Press, 2013), pp. 157–160; Katharina Berndt Rasmussen, “Harm and Discrimination”, Ethical Theory and Moral Practice 22 (2018): pp. 873–891.
Hellman sometimes uses a non-moralized concept of harm – e.g., “Our actions often harm others and yet are permitted” (Hellman, “Indirect Discrimination”, p. 107). Sometimes this concept is employed in its intrapersonal version and sometimes (as when she says “…the state has a strong reason not to disadvantage abuse victims relative to other prisoners when determining whom to release early” (Hellman, “Indirect Discrimination”, p. 108)) in its interpersonal one.
cp. Bradley, “Doing Away”, p. 397.
Imagine yourself in the shoes of a privileged female victim of domestic violence complaining about the injustice of your insurance premium being raised on account of your being a victim of domestic violence to an audience of unjustly disadvantaged women who pay a much higher insurance premium on account of their being the victim of much more disadvantage-inducing injustices, and whose premiums would be raised if unjustly privileged people like you did not have to pay a higher premium on account of your specific disadvantage, actuarially speaking. How would you answer them if they say in response to your complaint: “Hang on! Surely, if anyone is being charged unjustly high premium, it is us not you”?
Cp. Lippert-Rasmussen, “Indirect Discrimination”, pp. 47–48.
“G1”, “G2”, and “G3” most naturally refer to groups of individuals. However, for simplicity, I write as if they refer to individuals.
Here I express disagreement with the message in the following passage: “the account presented here requires not only that the group affected have suffered injustice but also that the policy at issue compound that injustice” (Hellman, “Indirect Discrimination”, p. 121). Also, recall the discussion of Attempts III and IV in Section III, which, unlike the pair of cases discussed here, involved a pair of cases in which both disadvantaged groups were subjected to someone else’s unjust treatment.
One might submit that this just shows that we should adopt a notion of harm different from the moralized, intrapersonal one. However, as I argued above, there are other problems with these accounts. Also, these accounts have similar implications with regard to Good Brute Luck and No Good Luck as the duty not to compound injustice based on the moralized, intrapersonal concept of harm.
Suppose G1 has exactly what it is entitled to, whereas G2 has slightly less. I can either impose a huge cost on G1 or a tiny cost on G2. Here the duty not to cause additional harms to the unjustly worse off should be understood in such a way that G1 would be “unjustly worse off” if I select the first option.
For related claims, see Shelly Kagan’s views on curved “desert mountains” (The Geometry of Desert (Oxford: Oxford University Press, 2012), pp. 227–243) and the priority principle (Derek Parfit, “Equality and Priority”, in A. Mason (ed.), Ideals of Equality (Oxford: Blackwell Publishers, 1998), pp. 1–20, p. 12) on which benefits matter more, morally speaking, the worse off the recipients are.
Recall from Section III that one can appeal to this duty in conjunction with the duty not to implicate oneself in justice. Someone who does that will typically agree with Hellman about the moral assessment of most concrete cases, where she would appeal to the duty not to compound injustice. However, the two appeals would diverge in cases of unjustly privileged victims of injustice such as the one I have discussed in this section. In any case, it is theoretically important whether there is a duty not to compound injustice.
A previous version of this paper was presented at the Arctic University of Norway-UiT, Oct 20, 2020, and at the “Algorithmic Fairness Workshop” held (online) at University of Copenhagen, Nov 12, 2020. I would like to thank Reuben Binns, Ben Eidelson, Göran Duus-Otterström, Sarah Fine, Andreas Føllesdal, Deborah Hellman, Nils Holtug, Annabelle Lever, and Michael Morreau for their helpful comments. This work was funded by the Danish National Research Foundation (DNRF144).
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Lippert-Rasmussen, K. Is there a duty not to compound injustice?. Law and Philos 42, 93–113 (2023). https://doi.org/10.1007/s10982-022-09460-y