PEA Soup is pleased to introduce October’s Ethics discussion, featuring David O’Brien (University of Wisconsin-Madison) paper “Can Luck Egalitarians Handle Self-Sacrifice?” with a précis by Kasper Lippert-Rasmussen (Aarhus University).

The discussion is now open.

_______________________________________________________________________________

Précis: Can Luck Egalitarians Handle Self-Sacrifice?

By David O’Brien

In his brilliant article in Ethics, David presents a challenge to luck egalitarianism concerning morally required self-sacrifice. The force of this challenge has so far been underappreciated, but thanks to David’s article, it is likely to receive the attention it deserves. Here, I reconstruct what I take to be the central structure of his challenge. In the interest of presenting it as clearly as possible, and given space constraints, I leave aside many of David’s interesting subsidiary arguments. I conclude with some reflections on how we should understand luck egalitarianism.

According to luck egalitarianism in its canonical form,

“it is unfair if some are off than others through no fault or choice of their own” (10).

In a wide range of cases, this claim has intuitive appeal. It seems unfair that some people are worse off than others simply because they were unlucky enough to be raised in poverty, while others are better off because they were raised in affluence. As Rawls might have put it, it is unfair that some are worse off than others simply because of their bad luck in the “social lottery.”

However, the canonical luck egalitarian claim becomes less intuitive when we consider cases in which some people are worse off than others because they choose to sacrifice their own well-being for the benefit of others, as they are morally required to do. Consider the following case:

Rescue I. Aodh and Brigid, passing by a river, see Conor in the river about to suffer significant harm. Either passerby could, as they each know, rescue Conor from significant harm—Aodh at only a minuscule cost to himself, Brigid at modest cost to herself. Assume that Aodh is morally required to perform the rescue but unconditionally refuses to do so. Assume further that Brigid is then morally required to perform the rescue and that, being morally motivated, she does so. She thereby produces an inequality between herself and Aodh; having incurred a modest cost in the course of the rescue, she is now somewhat worse off than Aodh is (12-13).

Brigid is worse off than Aodh. Yet it seems that she is not worse off through no “fault or choice of [her] own.” After all, she is worse off because she freely chose to rescue Conor, and she could easily—and perhaps at no cost to herself—have refrained from doing so. Hence, luck egalitarianism, as stated above, does not condemn the resulting inequality as unfair. Intuitively, however, it is unfair that Brigid is now worse off than Aodh.

Luck egalitarianism, as David formulates it above, offers only a sufficient condition for unfair inequality. Its proponents might therefore respond: “Sure, it is unfair that Brigid is now worse off than Aodh, but it is unfair for reasons unrelated to luck egalitarianism. It is unfair that Aodh places Brigid under a moral duty that she would not have been placed under, had Aodh done what he was morally obligated to do.” However, luck egalitarians generally take their theory to provide a complete account of distributive unfairness. Arguably, therefore, the canonical formulation is best understood as offering a necessary as well as a sufficient condition:

It is unfair if, and only if, some are worse off than others through no fault or choice of their own.

If luck egalitarianism is understood in this way, its proponents face the following trilemma:

  1. It is unfair that Brigid is worse off than Aodh.
  2. Brigid is not worse off than Aodh through no choice or fault of her own.
  3. It is unfair if, and only if, some are worse off than others through no fault or choice of their own.

As David effectively argues, this trilemma presents a serious problem for luck egalitarians. Since they cannot reject (3)—at least not without abandoning luck egalitarianism as formulated here—they must reject either (1) or (2).

At first glance, the most appealing option is to reject (2). Specifically, luck egalitarians might adopt a moral-requirement-sensitive understanding of what it means to be worse off through a choice or fault of one’s own, according to which:

If X is worse off than Y because X chose to act in a way that X was morally required to act, then X is, in the relevant sense, worse off than Y through no choice or fault of X’s own.

Unfortunately, this proposal generates a circularity problem. We want to know whether an agent is required to perform a particular action involving them sacrificing their own well-being for the benefit of others. What one is morally required to do depends, at least in part, on whether one’s action increases the extent to which some people are worse off than others through no fault or choice of their own. Indeed, luck egalitarians seem committed to treating this as one determinant of moral requirements. But, according to the proposed specification of what it means to be worse off through no choice or fault of one’s own, determining whether an action has this property requires us first to establish whether the action is morally required. That is precisely the question we started with: we wanted to know whether the action is morally required, and to answer it, we need to know whether it increases the extent to which some are worse off than others through no fault or choice of their own. Thus, avoiding the trilemma by adopting a moral-requirement-sensitive understanding of what it means to be worse off through a choice or fault of one’s own leads to a vicious circle.

Could luck egalitarians instead reject (1)? They could, but David argues that doing so creates a problem of objectionable arbitrariness. Suppose we adopt a prudentialist account of reasonable avoidability, according to which:

If X is worse off than Y because X chose to act in a way that was prudentially suboptimal for X, then X is, in the relevant sense, worse off than Y through a choice or fault of X’s own.

Given this account, we can reject (1). After all, rescuing Conor was prudentially suboptimal for Brigid, yet she chose to rescue him, presumably because she was motivated by a sense of moral obligation.

Unfortunately, luck egalitarianism based on this prudentialist account of reasonable avoidability faces the problem of objectionable arbitrariness. To see why, consider the following case:

Rescue IV. With near certainty, Iseult is about to suffer a significant harm (say, a trapdoor beneath her is nearly certain to swing open, causing her to fall into a river). Hugh has the options of either pressing a certain button or not doing so. If Hugh presses his button and the trapdoor beneath Iseult swings open (which is very likely), a safety net would be deployed that would prevent her from falling and thereby rescue her from the significant harm. But if Hugh presses the button and the trapdoor beneath Iseult stays shut (which is very unlikely), Hugh himself would suffer a mild shock and thereby incur a modest cost. Iseult faces a corresponding set of options concerning a trapdoor that is below Hugh…  Suppose now that you know those facts, suppose further that you know that Hugh and Iseult have each resolved to attempt their respective rescues, and suppose still further that you know that, as it happens, the trapdoor below Iseult will swing open but the trapdoor below Hugh will not… Finally, suppose that… [b]y switching around the buttons’ wiring, you can make it the case that Iseult, when she presses her button, will in fact have attempted to rescue herself (not Hugh) and that Hugh, in pressing his button, will in fact have attempted to rescue himself (not Iseult). (22-23).

According to luck egalitarianism based on a prudentialist account of reasonable avoidability, your decision whether to rewire the buttons makes a moral difference to the unfairness of the resulting inequality between Hugh and Iseult. If you do not rewire them, the inequality will reflect their reasonably avoidable, prudentially suboptimal choices and will therefore not be unfair. If you do rewire them, it will not reflect such choices, and the resulting inequality will be unfair. Yet it seems objectionably arbitrary that the unfairness of the inequality between Hugh and Iseult should depend on whether you decide to rewire the buttons, unbeknownst to them.

In sum, neither rejecting (2) nor rejecting (1) offers an easy way out of the trilemma. Of course, there is a third option: rejecting luck egalitarianism altogether. But David does not think this is the conclusion we should draw from his article—or at least not yet. Rather, he takes the immediate lesson of the trilemma to be that luck egalitarians need to reflect on the basis of luck egalitarianism.

In closing, I want to offer a reflection that may interest luck egalitarians who, as I think they should, regard David’s article as an invitation to reconsider their basic commitments and, in doing so, to articulate luck egalitarianism more adequately.

Perhaps luck egalitarians should adopt a pluralist view of fairness. One kind of fairness concerns how people are differentially advantaged by facts that lie beyond the reach of their agency. This is the kind of unfairness on which luck egalitarians have traditionally focused. Another kind concerns how people are differentially advantaged by facts that lie within the reach of their agency—for example, whether they face choices constrained by moral requirements. On this view, two people whose available options are equivalent from the standpoint of prudence may nevertheless face choice sets that are not equally good if one person’s choices are constrained by binding norms other than prudence, such as moral norms, while the other person’s choices are not. The former may be directed towards making prudentially suboptimal choices, whereas the latter is not.

Setting aside complications about responsibility for being in the choice situation one is in, inequalities arising because some people face a choice set that is prudentially equivalent to that of others, but are morally required not to select the prudentially best option, are unfair. They are unfair because they arise from people facing unequally good choice situations. These situations are unequally good because, although we care about our own well-being, we also care—and should care—about acting morally. In short, we should embrace a goodness-of-choice-set formulation of luck egalitarianism:

It is unfair if some are worse off than others as a result of having faced a worse choice set than those who are better off.

This view seems to avoid the circularity problem. We can assess the prudential equivalence of choice sets independently of whether we know which choices agents are morally required to make. Moreover, whether particular choices, such as self-sacrificing ones, are morally required or impermissible does not depend on whether the resulting distributive outcome arises from an agent’s own choice or fault. That concept plays no role in the goodness-of-choice-set formulation of luck egalitarianism.

Specifically, in Rescue I, the view implies that Brigid’s being worse off than Aodh is unfair because she faced a worse choice situation: doing what she was morally required to do imposed a greater cost on her than it would have imposed on Aodh.[1]

Furthermore, at least in a range of cases, we can assess the extent to which two agents are differentially morally constrained without presupposing any view about whether the resulting inequality is unfair and, consequently, whether it counts against the self-sacrificing action being morally required. This is true in cases where, regardless of whether the inequality resulting from the unequally good choice situations is unfair, that the reason to avoid the inequality is outweighed by moral considerations unrelated to fairness. For example, the inequality might be small, while the benefit produced by self-sacrifice is substantial. In such cases, David’s “circle of dependences” (p. 17) is broken.

Admittedly, this leaves cases in which the resulting inequality is, as it were, morally on a par with the benefit gained through self-sacrifice, as may be the case in David’s Rescue II (p. 15). Perhaps something like the circularity problem arises in such cases. But perhaps, given our ambivalence about what an agent is required to do in these circumstances, this is not a defect of the goodness-of-choice-set formulation. After all, the formulation need not deliver determinate verdicts in cases where the underlying moral considerations themselves fail to determine what an agent ought to do. David seems to assume that if luck egalitarianism generates a circularity problem in even one possible scenario, this constitutes a significant problem for the theory. Whether that assumption is justified is itself worth considering.

The goodness-of-choice-set formulation also avoids the problem of objectionable arbitrariness. Since it states only a sufficient condition for unfairness, it is consistent with the possibility that different forms of brute option luck are unfair (cf. pp. 25–27). It is therefore also consistent with the conclusion that rewiring the buttons in Rescue IV makes no difference to whether the resulting distributive outcome is unfair.

Perhaps embracing the goodness-of-choice-set formulation requires us to reject my modified canonical formulation of luck egalitarianism, which states both a necessary and a sufficient condition of unfairness. However, one might argue that, like David’s formulation, which states only a sufficient condition, the modified canonical formulation is merely an attempt to capture the underlying motivation of luck egalitarianism. David’s challenge may show that the choice-set formulation captures that motivation more successfully than the canonical formulation does. In short, rejecting (3) in the trilemma and replacing it with the goodness-of-choice-set formulation need not amount, in the relevant sense, to abandoning luck egalitarianism’s core commitment. Rather, it might amount to articulating it more precisely in the light of David’s forceful challenge and the spirit in which he puts it forward.

[1] An additional complication is that Aodh is responsible for her being in a choice situation she would not have been in, had Aodh done his duty. This relates to my pluralism about fairness point above.

33 Replies to “Ethics Discussion: David O’Brien – Can Luck Egalitarians Handle Self-Sacrifice?”

  1. Kasper, I’m so grateful for the (very) generous précis! I’m also delighted by your suggested alternative way of avoiding the circularity problem, which I think is both ingenious in its own right and also raises some very interesting issues. In the hopes that it helps in kicking off discussion, I’ll just note here a couple of initial thoughts that I’ve had about it in the past few days.

    1. Kasper suggests an alternative way of formulating luck egalitarianism that avoids the circularity problem that I discuss in my paper. His suggestion, as I’m thinking of it, has got two key moves. The first move is a new dimension of egalitarian assessment: not assessing people’s *choices* but rather assessing people’s *choice sets* (i.e., the sets of options *from which* people make their respective choices). The second move is the *kind* of choice-set assessment that’s done: assessing what Kasper calls the “goodness of choice sets”. A salient way that one person’s choice set can be worse (in Kasper’s sense) than another’s is when some options in the first set are morally forbidden but none in the second set are. I’ll call this property of a choice set its ‘m-goodness’ (just to avoid ambiguity about the sense of ‘goodness’ that’s at issue here). When you put those two moves together, you get what Kasper calls ‘goodness-of-choice-set luck egalitarianism’ (which I’ll call GCS luck egalitarianism for short). GCS luck egalitarianism entails that A’s being worse off than B can be unfair when (and because) A faced an m-worse choice set than B faced. [i]

    As Kasper notes, applying GCS luck egalitarianism to a variant of my Rescue I case helps to illustrate how GCS luck egalitarianism avoids the circularity problem that I discuss in my paper. In that variant case, we can imagine, one person A faces a choice of whether to engage in a near-costless rescue of some third party, and A does so, whereas another person B never faces such a choice. GCS luck egalitarianism can then deliver the result that the resulting inequality between A and B is unfair, because A faced an m-worse choice set than B did. The moral-requirement-sensitive luck egalitarianism (MRS luck egalitarianism for short) that I discuss in the paper can deliver that result too. But it’s the *way* that GCS luck egalitarianism recovers that result that lets it avoid the circularity problem that I discuss in my paper. The circularity problem that I discuss arises for moral theories that incorporate MRS luck egalitarianism, on which (roughly) what A ought to do depends (in part) on whether there’s unfair inequality or not, which in turn depends (in part) on a property *of A’s choice* (namely, whether it’s what she ought to do or not). And so: a cycle of dependences. A moral theory that incorporates GCS luck egalitarianism, by contrast, makes what A ought to do dependent (in part) on whether there’s unfair inequality or not, but makes *that* depend (in part) only on a property *of A’s choice set* (namely, how m-good it is). And so: no cycle of dependences. (Per Kasper: The idea that there’s a link between the moral status of “*particular* choices, such as self-sacrificing ones” and the inequalities that they may produce “plays no role in the goodness-of-choice-set formulation of luck egalitarianism”.) Lovely!

    The considerations above help to illustrate what’s so clever—and, to my mind, delightful—about Kasper’s suggestion. GCS luck egalitarianism doesn’t have the particular feature of MRS luck egalitarianism that generates the circularity problem that I discuss, and yet it retains the broad ‘spirit’, so to speak, of moral-requirement-sensitivity. As I say in the paper, I myself am broadly sympathetic *to* the judgments that lead to MRS luck egalitarianism. And so I think there’s a lot for luck egalitarians (or, at least, luck egalitarians like me) to like in Kasper’s suggestion.

    Now let me point to a couple of interesting issues that I think Kasper’s suggestion raises.

    2. As Kasper notes, it looks like a ‘revenge’ version of the circularity problem may arise, after all, when one attempts to apply GCS luck egalitarianism to cases like my Rescue II. Indeed, I think it isn’t a great surprise that this should be so. After all, as noted above, GCS luck egalitarianism preserves at least the ‘spirit’ of moral-requirement-sensitivity. And so one might then expect that circularity concerns might show up, in *some* form, when one attempts to apply GCS luck egalitarianism to cases like my Rescue II. The concern, in brief and setting aside some details, is that in such cases when one attempts to determine *whether* the rescuer faces an m-worse choice set, one needs to know the moral status of the options *in* the rescuer’s choice set—but, in cases like Rescue II, it seems harder to deny that the inequality produced by the rescue, if unfair, *would* bear importantly on the moral status of one or more of the options in the rescuer’s choice set. And so circularity threatens again. So I share Kasper’s suspicion that GCS luck egalitarianism may be vulnerable to a kind of ‘revenge’ version of the circularity problem.
    Nonetheless, as Kasper further adverts to, there might still be something to like about GCS luck egalitarianism, even in the key range of cases like Rescue II. Here’s one way, I think, of seeing why. Because what it’s assessing is (the m-goodness of) *choice sets*, not particular choices made within those sets, GCS luck egalitarianism arguably has some extra wiggle room in *what* assessment it makes *of* the rescuer’s choice set in cases like Rescue II. For example, it might be able assign m-goodness-values to choice sets in such a way that the rescuer’s choice set in Rescue II is (e.g.,) *on an m-par with*, or only *roughly as m-good as*, others’ choice sets. By introducing the right kind of indeterminacy or vagueness into its choice-set assessments, in other words, one might try to get GCS luck egalitarianism’s entailments about Rescue II to match our intuitive sense of “ambivalence” (as Kasper puts it) about such cases. Because GCS luck egalitarianism enjoys this extra degree of freedom (as compared with MRS luck egalitarianism), then, there might still be reason for optimism that GCS luck egalitarianism can handle even a ‘revenge’ circularity problem. Of course, the GCS luck egalitarian would still have to identify a plausible basis *for* assigning the right kinds of m-goodness-values (e.g., in the way I suggested above), in order to handle the ‘revenge’ circularity problem on a well-motivated basis that isn’t objectionably ad hoc. So, while I do think there’s a potentially nice strategy available to the GCS luck egalitarian here, I also think there’s still a question mark over whether it can be made to work.

    3. That leads me to my second thought. Suppose now that GCS luck egalitarianism can—e.g., via the development of it that I suggested above—be made to avoid the circularity problem. That then leads us to one of main things I was hoping my paper might do – namely, focus attention on questions about the *basis* of luck egalitarianism. Here’s how I’d be inclined to see the resulting situation. We’d then have at least two candidate ways of addressing the circularity problem while remaining faithful to the broad ‘spirit’ of moral-requirement-sensitivity. One would be MRS luck egalitarianism, combined with one of the explanations of the basis of luck egalitarianism that I discuss in the final section of my paper (e.g., via an appeal to an underlying account of fairness). The other is GCS luck egalitarianism, combined with some explanation of why fairness pertains in the first instance to choice sets, and only then to the choices made within those sets, in the ways Kasper describes above. To make further progress in assessing which is the more theoretically choiceworthy way of addressing the circularity problem, it seems to me, what would be nice would be to have an (ideally, independently-motivated) account of fairness that we could then consult to ask *which* of those two ways of specifying luck egalitarianism is more faithful to that underlying account of fairness. Or, put in the terms that I use in the paper, what we could ask is which of those two versions of luck egalitarianism has a more adequate answer to the grounds question about it—i.e., the question of why, or in virtue of what, the kind of inequality with which it’s concerned is a kind of inequality that’s morally significant. Now I’m not sure which of MRS luck egalitarianism or GCS luck egalitarianism comes out more favorably in that comparison. But that’s exactly that kind of question that I think is quite an exciting new way of making progress in our understanding of luck egalitarianism. And that’s another reason that I love Kasper’s suggestion here. It ultimately points, I’m inclined to think, toward that fruitful set of research questions for future work on egalitarianism.

    The above, of course, are just some initial thoughts—looking forward to continuing to discuss these and/or other issues with Kasper and/or others here.

    —–

    [i] It may be worth noting that, although GCS luck egalitarianism assesses choice sets in the first instance, nonetheless it *isn’t*, as I understand it, equivalent to the kind of opportunity-focused, or choice-set-focused, version of luck egalitarianism that Kasper (rightly, in my view!) criticized in a lovely and instructive exchange with Richard Arneson a few years ago (K. Lippert-Rasmussen (1999), ‘Arneson on equality of opportunity for welfare’, Journal of Political Philosophy 7(4) 478-487; R. Arneson (1999), ‘Equality of opportunity for welfare defended and recanted’, Journal of Political Philosophy 7(4): 488-497.) That’s because, as I emphasize in the paragraph to which this note is appended, the kind of assessment of choice-set quality that Kasper’s proposing, in the present context, is quite a different one. It *might* be interesting to ask, though, whether some of the considerations canvassed against choice-set-focused versions of luck egalitarianism, in that earlier exchange, might carry over to the present context.

  2. Given Kasper has ably summarized David’s main arguments, I will jump straight into my critical commentary. In brief, I offer three tentative grounds for supposing that moral luck egalitarianism can avoid the circularity problem that is the main contribution of David’s important article.

    First, in the central case, Rescue II, David considers the possibility “that rescuing Eamon is morally required” (p. 16). He concludes that “the rescue thus produces only a modest gain in aggregate welfare but a significant increase in unfair inequality. But then the rescue is not morally required, contradicting our initial assumption.” But as the case is described, the weightings could be such that the rescue is morally required. Suppose, for instance, that our pluralist theory says a unit of welfare has 10 moral value, and a unit of unfair inequality has -1 moral value. Rescue “would leave Deirdre at welfare 75, although she is more deserving at level 0, and leave Eamon at welfare 100, although he is less deserving at level –1” – i.e. there is 1750 moral value from the welfare minus 25 moral value from the unfair inequality, for a total of 1725 moral value. How does this compare to not rescuing? It “would not produce an unfair inequality—it would leave Eamon at welfare 70 and Deirdre at welfare 100, but since Eamon at level –3 would be less deserving than Deirdre at level –2, the inequality would not be unfair” – i.e., it gives 1700 moral value from the welfare with no deduction for unfair inequality as there is none. So with these weightings, and assuming consequentialism, rescuing Eamon is straightforwardly morally required, as it yields the highest moral value. Of course, the case’s numbers could be modified to avoid this outcome, but the weightings could then be adjusted in turn to restore it. The pressure of the example and revisions of it is thus to give very strong weighting to welfare as opposed to unfair inequality, though the pressure does not seem such that the weighting has to be lexical, so the theory remains the kind of pluralist egalitarianism that David targets. Moral luck egalitarians may not like being forced towards specific weightings but the example doesn’t seem to defeat their view.

    Second, couldn’t the moral luck egalitarian just say that in Rescue II-style cases rescuing and not rescuing are equally good? After all, if neither option can coherently be said to be morally required, it seems quite plausible that they are no better than each other. This is quite a normal situation, applying for instance where two outcomes are the same in terms of welfare and unfair inequality, or perhaps more analogously, where there is radical uncertainty.

    Finally, the moral luck egalitarian could maintain that it is impossible for one agent to change the expected desert of another agent. I defend this position, which I call Fixed Expectations, in a forthcoming paper in Oxford Studies in Political Philosophy, vol. 12. There I suggest it as a response to the objection that desertists and luck egalitarians will be motivated to morally corrupt badly off individuals, in order to reduce the badness of their disadvantage. I wonder if it might also work to defuse Rescue II, which implicitly assumes that Deirdre can change Eamon’s expected desert. (There is a column labelled for Eamon’s “deservingness,” but it must be expectations about Eamon’s deservingness that are relevant to the choice whether to rescue.) Indeed, it is quite counterintuitive that Deirdre can choose to raise or lower Eamon’s expected desert, as though it tracks her will rather than his.

    I’d love to hear David’s thoughts!

  3. Thanks, David and Kasper, for a great paper and a great precis, respectively!

    My question (to David) is rather simple and, perhaps, betrays a misunderstanding on my part. I wonder whether the luck egalitarian view that Thaysen and Albertsen (T&A) advance in their 2017 paper “When Bad Things Happen to Good People” can give us the right verdict in cases of self-sacrifice without involving any circularity.

    T&A propose to distinguish between disadvantage creation and disadvantage distribution. They write: “An agent is responsible for creating a (dis)advantage if, and only if, she is responsible for behaving in such a way that somebody was (dis)advantaged. An agent is responsible for distributing a (dis)advantage if, and only if, she is responsible for behaving in such a way that X, rather than Y, was (dis)advantaged” (p. 94).

    In light of this distinction, they propose to “revise”* the core luck egalitarian principle in the following way: “A distribution is just if, and only if, people’s comparable positions reflect nothing but their comparable responsibility for creating advantages and disadvantages” (p. 94).

    This principle can be applied to Rescue I in a straightforward way: Brigid isn’t responsible for the creation of disadvantage; she just redistributes it from Connor to herself, so inequality between her and Aodh is unjust. So, like the Temkin-proposed revision, it gives us an intuitively correct verdict in Rescue I. In Rescue II, I think there are different versions of the T&A-style view that can give different verdicts in this case, but I can’t see how any of them would lead to a contradiction in the way Temkin’s view does because the view, at least on the face of it, is not moralized – the statement that I quoted above doesn’t invoke any moral terms, either directly or by implication, that would create the circularity issue. It does, of course, refer to responsibility, but I take it that whether one’s responsible for X doesn’t depend on the deontic status of X.

    You mention the T&A view, along with a fairly close one by Jesse Spafford, in footnote 27, where you say that it doesn’t avoid the circularity problem, but it isn’t clear to me where circularity arises here, so it would be interesting to hear more about this

    *I put “revise” in quotation marks because, although that’s the term they use, I think it is more appropriate to characterise it as a specification, not a revision.

  4. Carl, thank you so much for these marvelous comments and very careful reading of my paper! Perhaps, as I did with Kasper’s précis, I’ll just again give a few initial thoughts here, and then I’d love to follow up again if you have further thoughts. In reverse order:

    3. Your forthcoming paper sounds fabulous! I definitely agree that, if you’re right about the limits of how one agent’s actions can affect another’s (moral) deservingness, that would strike at the heart of what’s assumed to be possible in my Rescue II case. And I also share the thought that it doesn’t seem good to have a theory of moral desert on which B’s deservingness can change as a mere result of what A wills. When I was originally cooking up the case, a concern along these lines had briefly—and I’m now realizing, perhaps too briefly!—occurred to me. I’d hoped to avoid that kind of concern by building the case such that Deirdre’s rescue would have the effect of changing Eamon’s moral deservingness by way of making it the case that Eamon doesn’t form a malevolent intention that *he* would otherwise have formed. So I’m wondering if there’s ways of spelling out the mechanism here that might avoid the concern—for example, following a suggestion that I think I might briefly mention in a footnote somewhere, if Deirdre’s ‘rescue’ consists of a course of education that shapes Eamon’s dispositions, character, and so on (and *thereby* affects what intentions he expectably forms). Would that version of the case still run afoul of what you say in the forthcoming piece, though?

    2. I think you’re right that this is a way out of the circularity problem! (Anything’s better than a contradiction!) But I think that, by itself, this way out of the problem would leave the pluralist egalitarian consequentialist (who endorses MRS luck egalitarianism as their account of unfair inequality) with even *more* explaining to do. Bracketing the point about the relative weights given to each moral factor for now (more on that below), let’s assume the kind of pluralist egalitarian consequentialism that I use to illustrate the circularity problem—i.e., one that recognizes aggregate welfare and unfair inequality as moral factors and which takes “each of those moral factors to be of independent and more than trivial degrees of moral significance … giving neither lexical priority over the other”. Then if there’s a nominal amount of one factor at stake in the choice in question but a notable amount of the other factor, it cries out for explanation *how* these factors could be in the kind of equipoise needed for both options to be permissible. If both options are permissible, *something* must be coming in to make it the case that the two factors are equipoise—but what? (And if there really were no explanation other than ‘Otherwise one would have a contradiction’, I was thinking that that was an uncomfortable position for the MRS luck egalitarian to be in, vis-à-vis her dialectical opponents.)

    1. Yes, I do think you’re right that, once one fixes the magnitudes of aggregate welfare and inequality that are at stake in any given version of Rescue II, one can then construct a version of pluralist egalitarian consequentialism (that incorporates MRS luck egalitarianism) that won’t face a circularity problem in the case, just by specifying appropriately the relative weights that it assigns to those two moral factors. (As your example nicely shows.) But I’d been thinking that the point I needed was the inverse of this. Once the pluralist egalitarian consequentialist (who embraces MRS luck egalitarianism as their account of unfair inequality and satisfies the other condition quoted in my #2 above) tells me what their relative weighting of aggregate welfare and unfair inequality is, I can then go and cook up a case like Rescue II, with appropriate magnitudes of aggregate welfare and inequality at stake in it, that’ll generate the circularity problem. And, I’d been thinking, if there were even one such case (for every such version of pluralist egalitarian consequentialism), that would be very bad news for the view. (But perhaps your point here connects with something I didn’t address in Kasper’s précis. Kasper wrote, correctly, that “David seems to assume that if luck egalitarianism generates a circularity problem in even one possible scenario, this constitutes a significant problem for the theory. Whether that assumption is justified is itself worth considering.” So maybe that is also an assumption that you implicitly wanted to put into question in your point here?)

    Apologies that these are somewhat sketchy initial thoughts—just done with a long day of teaching! But would love to hear if they prompt further questions or thoughts, from you or others.

  5. Thanks to David for his thought-provoking paper and to Kasper for kicking off the discussion with his lucid précis.

    I would like to ask David what he thinks about two possible moves that might avoid the circularity problem.

    To see what the first one would consist in, note that David holds that the circularity problem for luck egalitarians arises as a result of their subscribing to the following claims, which create “a circle of dependences” (p. 17):

    1) Whether it is morally required to rescue depends on whether the rescue would be best;
    2) Whether the rescue would be best depends on whether the inequality produced by the rescue would be unfair;
    3) Whether the inequality produced by the rescue would be unfair depends, in turn, on whether it is morally required to rescue.

    But do luck egalitarians subscribe to 3)? They could, instead, subscribe to:

    3*) Whether the inequality produced by the rescue would be unfair (i.e., whether it would be fair to let the costs of rescue fall only on those who choose to undertake the rescue) depends on whether, *if* there is a requirement to rescue, this requirement binds everyone or only the rescuer. To settle whether 3) holds, we need to know whether the rescue is in fact morally required, and circularity threatens. But to settle whether 3*) holds, we only need to know whether, if the rescue were morally required, the requirement is one that would bind the rescuer, specifically (as would be the case if the rescuer were responsible for causing the person that needs rescue to be in that position); or whether it would bind everyone (as would be the case if there were a general duty to rescue).

    A second possible way of avoiding the circularity problem takes off from David’s own suggestion (at p. 30), which – as I see it – involves rejecting 2): the rescue’s producing unfair inequality does not affect whether the rescue would be best; rescuing would be best, and morally required, even if the inequality the rescue produces were unfair.

    With regard to taking this route out of the circularity problem, David asks, “…how could it be true that an unfair inequality of significant magnitude does not outweigh the other considerations at stake…? (p.30)”; and he goes on to suggest that since the answer to this question cannot consist in “positing a constraint to rescue others from significant harm that is absolute with respect to any kind of good”….the luck egalitarian will need to “supply an account of the basis of the moral factor in question – namely, reducing unfair inequality -in a way that vindicates the claim that the constraint to rescue is absolute with respect to it” (p. 32).

    My suggestion, in response to David’s question, is that it we should note that *failing* to rescue produces an unfair inequality: the person whose rescue would not be undertaken would end up at a disadvantage, both in absolute and in relative terms, that, at least in the range of cases in which she is not responsible for needing the rescue, would be unequivocally unfair. If this is correct, it seems that the luck egalitarian has at hand a fairly straightforward story as to why the rescue would be morally required even if it produced unfair inequalities: failure to undertake the rescue would also produce unfair inequalities and presumably – in the cases of rescue which are plausible candidates for morally required rescues – greater unfair inequalities than those which the rescue would cause.

  6. Serena, I really couldn’t be more grateful to you for reading and engaging with my paper, and for suggesting these two brilliant possible moves in response to the circularity problem that I raise. On your two suggestions, in reverse order:

    2. I think your point here—viz., that “*failing* to rescue produces an unfair inequality: the person whose rescue would not be undertaken would end up at a disadvantage … that, at least in the range of cases in which she is not responsible for needing the rescue, would be unequivocally unfair”—is *certainly* true in many, or even *all*, real-world analogs of my Rescue II case. And so I agree that, when attention is restricted to those analogs of Rescue II, this does offer a nice (and, as you say, a nicely straightforward) explanation of why a rescue producing an unfair inequality (even one of significant magnitude) needn’t affect whether the rescue is morally required. In the passages you quoted, though, I’m trying to think through an explanation that would work in Rescue II itself, and I’d been thinking that the issues in *that* case make it harder to deploy an explanation of the kind you suggest here. As I’d tried to build it, that’s a case in which, if the rescue *weren’t* undertaken, the inequality that one would thereby be left with, although large in magnitude, comes out *not* to be unfair. (It comes out not to be unfair, at least by the lights of moral-requirement-sensitive (MRS) luck egalitarianism, because of the funny feature of the case that Carl Knight, in another response in the discussion here, was querying—namely, the rescue affecting not only the rescuee’s welfare levels but also his level of moral deservingness.) Nonetheless, it’s *also* true, in that Rescue II case, that the rescuee is not *responsible for needing* the rescue. And so suppose one thought that, even in the Rescue II case, that fact *alone* (viz., not being responsible for needing the rescue) were sufficient to render the resulting inequality (as you put it) “unequivocally unfair” (notwithstanding the level of moral deservingness of the potential rescuee). Then I think the version of luck egalitarianism that one would have in mind would be quite different from MRS luck egalitarianism. Whether it offers a more or less plausible way of avoiding the circularity problem than MRS luck egalitarianism would depend, I think, on that key issue that seems to separate it from MRS luck egalitarianism—namely, whether fairness requires rescuing even the very morally undeserving, in circumstances where they aren’t responsible for needing to be rescued. I think my own tentative judgments might tend against that, but that’s something that I’d like to think some more about.

    1. Actually, in light of the notes above, I’m now wondering if there might be an interesting connection between your second suggestion above and the first of your two suggestions—viz., the thought that the unfairness vs not-unfairness of a given rescue might turn on whether there was a *general* duty to rescue in play vs a *special* duty to rescue obligating only the potential rescuer. A unifying thread that runs through the two suggestions, I’m now thinking, is the thought that questions about fairness in such cases turn more on situation-specific facts about the relations between, and the histories of, the parties involved than it does on less situation-specific facts like their levels of moral deservingness. In any event, I *do* think that that thought would represent an interesting alternative to MRS luck egalitarianism, and I’d want to think through some more cases to figure out how plausible an alternative I think it represents to MRS luck egalitarianism. My initial suspicion, though, would be that even this alternative version of luck egalitarianism would face circularity problems. Here, briefly, is why I suspect that. If I’m understanding your condition 3* correctly, what it is says that: (i) if an inequality is produced by a rescuer who was morally required to rescue in virtue of a *special* duty of rescue obligating only her, it’s *not* unfair to leave that inequality in place, whereas (ii) if an inequality is produced by a rescuer who was morally required to rescue in virtue of a *general* duty of rescue (one that needn’t have obligated only her), then that inequality is at least eligible to be unfair. It’s in that second kind of case (ii) that I’m thinking circularity concerns might still threaten. What I’m thinking is that one might then build a case in which *whether* that general duty to rescue generates a moral requirement to rescue would crucially depend, in part, on whether the inequality produced by the rescue is unfair or not. (This is related to some of the issues I discuss, at one point, about avoiding the circularity problem via an appeal to moderate nonconsequentialist constraints.)

    I should add though—stepping back from those details—that if the broad alternative to MRS luck egalitarianism that you suggest can be made to work in such a way that it avoids the circularity problem, then I think I’d repeat the substance of what I said in the last part of my response to Kasper’s précis above. The position then, as I see it, would be that we’d have *two* ways of avoiding circularity: on the one hand, MRS luck egalitarianism (combined with one of the underlying accounts of fairness sketched in the final section of my paper) and, on the other hand, the version of luck egalitarianism you suggest here (combined with another underlying account of fairness). To make further progress, I would think, one would then turn to questions about *basis* of luck egalitarianism—i.e., asking whether each version of luck egalitarianism can be shown to appropriately match up with its underlying account of fairness, and asking which of the two underlying accounts of *fairness* is the more plausible one (hopefully, by checking each one against some independently-motivated constraints on a plausible account of fairness). That, I think, represents a potentially fruitful way to make more progress in understanding luck egalitarianism (and is, ultimately, why I think the circularity problem is an interesting one!)

  7. Dmitry, thank you for your comment (which I’d missed earlier; apologies)! Yes, I do think the issues that are raised by Thaysen & Albertsen’s paper are distinctive, for the reasons that you note here. There’d been a longer discussion of their and others’ views in a long footnote on an earlier draft of the paper, which ultimately ended up in a very compressed footnote in the final paper. So, expanded a bit, here’s the thought I’d had in mind, when I tersely said (in the paper) that “similar remarks apply to” their view as those that apply to my discussion of “‘moralized’ reasonable-avoidability” versions of luck egalitarianism. Following Jesse Spafford, I’m inclined to think that one concern about Thaysen & Albertsen’s view as formulated in their (wonderful!) 2017 paper is what Spafford labels “neglect of distributive considerations”. (This is a point Spafford makes in his (equally wonderful!) 2021 paper ‘Luck egalitarianism without moral tyranny’ (Philosophical Studies 179(2): 469-493, 492; the case that nicely illustrates his point is at p. 491.) Along similar lines, I’m inclined to think that to be revised to treat distributive considerations *plausibly*, Thaysen & Albertsen’s view would have to make the unfairness of an inequality resulting from self-sacrificial choice depend on whether it produces a *reasonable* redistribution of already-existing disadvantage, where a redistribution of disadvantage is reasonable when it is at least as favored by the balance of moral reasons as the alternative. That is to say, the point I’d had in mind (building on that critical point of Spafford’s) is that Thaysen & Albertsen’s view would really have to have a moralized element in order to be plausible and, once it incorporated such a ‘moralized’ element, it would be subject to similar remarks as ‘moralized’ reasonable-avoidability versions of luck egalitarianism. (Thaysen & Albertsen might disagree!)

  8. Hi All. This is a great paper and a great discussion!

    I haven’t thought about this issue much before. What I say below may not work, but I’d be interested in anyone’s thoughts on it.

    Intuitively, I am attracted to MRS luck egalitarianism. It does seem unfair if, through no fault or choice of my own, I end up with a demanding obligation that others don’t have.

    So, like others, my inclination is to find a way out of the circularity.

    My thought is that instead of weighing the welfare effects and the equality effects of a putative obligation simultaneously, we judge them at two different stages. And instead of ‘assuming’ it’s an obligation (and then finding it isn’t, which is a contradiction), we ask what the world would look like if it WERE to be an obligation. And if we find that would produce too much unfairness, it isn’t.

    I think the key here is to have a presumption of non-obligation to help. David’s circle gets going by presuming (or imaging) it IS an obligation, and then finding it’s not. Or presuming (or imagining) it is NOT an obligation, and then finding it is. My suggestion is we assume it isn’t an obligation unless it passes two tests ( in which we consider how things would be if it WERE an obligation, without presuming it is). There is a contradiction between assuming x is an obligation and then finding its not. There is no contradiction between a defeasible presumption it isn’t an obligation unless it can pass a test, and then finding it passes the test (in which we consider how the world would look if it was an obligation) and so finding it is.

    That way, you can put a *possible obligation* to the test. And if it isn’t an obligation, this doesn’t create a contradiction, and you don’t then enter a loop where the non-obligation becomes an obligation and then a non-obligation….

    The process I am thinking of would go like this.

    A needs help. Is B required to help?

    1. The two way comparison: What would be the costs to B compared to the benefits to A? If the benefits are sufficiently weighty compared to the cost, there is a *possible obligation*?
    2. the equality test: if this *possible obligation* really were an obligation, what effect would that have on fairness? If it would make things too unfair, it’s not an obligation. If it would not make things too unfair, it is an obligation.
    3. If it is an obligation, acting will produce unfairness, but unfairness that is ‘worth it’. If it isn’t an obligation, acting will produce no unfairness.

    Because of the presumption of non-obligation, putative obligations that have gone through this process cannot re-enter it unless something significant changes. However, unlike some of the things ways out David considers, it doesn’t require lexical priority of one consideration over the other.

    What is wrong with this? I don’t think this is quite the route David considers when he moves beyond consequentialism.

  9. Thank you to David for writing this excellent paper, and thank you to Kasper for the excellent summary.

    I found it deeply inspiring and learned a great deal from it. I have read the article repeatedly, and in the course of reading it I developed two rough ideas:

    1. Whether non-MRS luck egalitarianism might face a problem that’s similar to the circularity problem? I call this the “reflexivity problem”.

    Suppose that A and B make identical choices, each with a prudential value of 10. Owing to a difference in brute luck, however, A actually receives 10 units of welfare, whereas B receives only 6.
    A therefore has an obligation under LE to compensate B by transferring 2 units of welfare, so that the distribution becomes (8, 8).

    Case 1: A chooses to compensate B.
    By doing so, A voluntarily makes an additional choice with a prudential value of 8. Consequently, the prudential value of A’s choice is now lower than that of B’s choice, whose prudential value remains unchanged at 10.
    A’s distributive share should therefore be lower than B’s, in order to reflect the difference between the prudential values of their choices.
    If we accept proportionality, A’s prudential value is now four-fifths of B’s. Accordingly, A’s distributive share should also be four-fifths of B’s, yielding the following distribution:
    This conflicts with LE’s initial distributive target of (8, 8).

    Case 2: A refuses to compensate B.
    In this case, the distribution remains (10, 6), which likewise conflicts with LE’s initial distributive target of (8, 8).

    Yet both cases violate LE’s fundamental motivation: to neutralize the differential effects of luck on individuals’ benefits.
    After A and B make their initial choices, the resulting distribution is (10, 6), even though their choices have equal prudential value. LE therefore initially requires the distribution to be adjusted to (8, 8).
    However, if A refuses to compensate B, the distribution remains (10, 6). If A chooses to compensate B, the distribution should instead be adjusted to (64/9, 80/9). Both outcomes conflict with the initial target of (8, 8).
    (Since this example involves no effect of option luck, as the expected value of A’s choice equals its actual outcome, brute-luck egalitarianism and all-luck egalitarianism will reach the same judgement in this case.)

    I conjecture that the source of the contradiction is this: When two individuals make choices of equal prudential value, prudential LE requires the individual who is better off owing to luck to fulfil an egalitarian moral obligation: compensating the individual who is worse off owing to luck, thereby satisfying the requirement of distributive equality.
    However, the very act of fulfilling this compensatory obligation changes the prudential value of the agent’s choice. This generates a new distributive requirement that conflicts with the requirement initially imposed by LE.

    2.Whether there might be another version of luck egalitarianism that avoids the circularity problem, by appealing to the distinction between creating/distributing disadvantage?

    The revised principle:A distribution is just if, and only if, people’s comparable positions reflect nothing but their comparable responsibility for creating advantages and disadvantages(Jens Damgaard Thaysen and Andreas Albertsen 2016).Dmitry Sereda has already presented this issue rather fully above, so I will only add a brief point.

    Returning to Rescue II.Because Deirdre carries out a costly rescue, she does not create disadvantage; she only distributes disadvantage.Therefore,her responsibility for creating disadvantage is always 0 (Eamon’s initial desert is -1, but because he does not create disadvantage, his initial responsibility is 0).

    ① If rescuing Eamon is morally required

    If she rescues, Deirdre has responsibility 0 and welfare 75, and Eamon has responsibility 0 and welfare 100. The inequality caused by the rescue is unfair.

    If she does not rescue, Deirdre has responsibility 0 and welfare 100, and Eamon has responsibility -2 and welfare 70. No unfairness is caused.

    The rescue brings about a moderate increase in total welfare, but it significantly increases unfairness. Therefore, it is not morally required, which contradicts the initial assumption.

    ② If not rescuing Eamon is morally required

    If she rescues, Deirdre has responsibility 0 and welfare 75, and Eamon has responsibility 0 and welfare 100, causing unfairness.

    If she does not rescue, Deirdre has responsibility 0 and welfare 100, and Eamon has responsibility -2 and welfare 70. No unfairness is caused.

    Not rescuing leads to a moderate decrease in total welfare, but it significantly reduces unfairness.
    Therefore, it is morally required, which conforms to the initial assumption.

    Therefore, it appears that the revised principle will not face the circularity problem.

    I would very much like to hear David’s views on this.

  10. Many thanks, David, for your replies!

    Just a couple of quick related thoughts, one in response to one of your replies, one about the similarity between your circularity problem and some of Rawls’ remarks about moral desert.

    In response to my suggestion that we think about the moral requirement to rescue as itself an expression of the commitment of luck egalitarianism, you note that, since whether one is morally deserving and whether one’s disadvantage would be unchosen can break apart, the view of luck egalitarianism which my suggestion presupposes “would be quite different from MRS luck egalitarianism”.

    The view I have in mind would not appeal to moral desert, it is true, and instead appeal to whether or not people are agent-responsible for their disadvantage. It would be different from views like Temkin’s, Arneson’s and, I take it, Eyal’s, but I think it would be close to Cohen’s – and my own judgement is that it would be more plausible than one that appeals to moral desert.

    One reason why it would be is given by Rawls, some of whose remarks on moral desert point to a circularity problem that I think is close to yours:

    “The essential point is that the concept of moral worth does not provide a first principle of distributive justice. This is because it cannot be introduced until after the principles of justice and natural duty and obligation have been acknowledged. Once these principles are on hand, moral worth can be defined as having a sense of justice..Thus the concept of moral worth is secondary to those of right and justice, and it plays no role in the substantive definition of distributive shares. The case is analgous to the relation between the substantive rules of property and the law of robbery and theft. These offenses and the demerits they entail presuppose the institution of property which is established for prior and independent social ends. For a society to organize itself with the aim of rewarding moral desert as a first principle would be like having the institution of property in order to punish thieves. (TOJ section 48)”

  11. Excellent paper and summary. Sorry I haven’t had time to read all the comments. Just a quick thought on Rescue I David.

    Rescue I. Aodh and Brigid, passing by a river, see Conor in the river about to suffer significant harm. Either passerby could, as they each know, rescue Conor from significant harm—Aodh at only a minuscule cost to himself, Brigid at modest cost to herself. Assume that Aodh is morally required to perform the rescue but unconditionally refuses to do so. Assume further that Brigid is then morally required to perform the rescue and that, being morally motivated, she does so. She thereby produces an inequality between herself and Aodh; having incurred a modest cost in the course of the rescue, she is now somewhat worse off than Aodh is (12-13).

    Why can’t luck egalitarians just analyse the case like this. Inequalities that arise due to our complying with our duties don’t count as chosen, because we don’t have any moral alternative. So the inequality between Brigid is unjustly worse off than Aodh if Brigid is required to rescue Conor. However, Brigid is nevertheless required to rescue Conor, because if she doesn’t, there will be an even worse unjust inequality between Brigid and Conor. She is required to rescue to prevent this even worse inequality occurring.

    There is no circularity here because the inequality that explains why Brigid is required to rescue Conor (that between Brigid and Conor) is not the same as the inequality that explains why Brigid is nevertheless a victim of injustice (that between Brigid and Aodh).

    You might say that this won’t work for

    Rescue II. Eamon has previously committed a minor moral wrong
    and will shortly form the intention to act wrongly again. 11 Unlike
    his previous minor moral transgression, however, this time Eamon
    would commit a serious moral wrong and also seriously injure him-
    self in the course of doing so. Deirdre can rescue Eamon (by pre-
    venting him from forming the intention in question, and thereby
    preventing him from committing the serious moral wrong), but
    she would thereby incur a moderate cost to herself. Alternatively,
    she can fail to rescue Eamon and thereby avoid incurring any cost
    to herself. Other things are fixed.

    But can’t we analyse this case as follows. Deirdre may be required to act because the inequality between her and Eamon will otherwise be too large, even though his wrongdoing will be chosen. She is then worse off than Eamon, but she is required to bear this inequality to avoid an even greater inequality that will arise if she doesn’t act. However, that inequality is still unjust, because it was possible to achieve equality where everyone is better off by Eamon not being disposed to act wrongly in the future. The fact that he won’t do what he could do is sufficient to generate a requirement. The fact that he could do so is sufficient to make the resulting inequality unjust. Again, there’s no circularity here, because we are appealing to the role of inequality in two different counterfactuals that are both relevant to justice – one based on what Eamon will do, and another based on what he could do.

  12. Anyway, thank you again, it was fun engaging with your paper and perhaps we can continue the conversation offline!

  13. Thank you all for this very interesting discussion. We definitely should not count discharging duties as a form of gambling whose resulting disadvantages are fair. Women who are disadvantaged because they care for dependants whom men should also care for would then be described as cases of option luck, deserving no correction. Thus, I agree that it is very important to settle these matters if luck egalitarianism is to be put into practice.

    That said, I approach the issue from a somewhat different angle, since I have always understood luck egalitarianism as a proposal in political philosophy, with the attention to institutions, Rawlsian publicity, legal simplicity and the metrics for interpersonal comparison that this often entails. For example, sufficiency-constrained luck egalitarianism forms the skeleton of many plausible public policies, such as taxing risky habits to pay for safety nets. On this view, individuals who squander their share of resources are not restored to equality but granted a minimum that both limits how low they can fall and maintains the plausibility of holding them liable for their subsequent decisions. I did not think of luck egalitarianism in terms of going through my day constantly recalculating my duties in view of my changing comparative position.

    Consider a scheme I learned about in Cuba. It was decided that people whose workplace was beyond walking distance would receive a bike. If a policy of Constant Redistribution were adopted, those who wrecked or sold their bikes would receive new ones. Soon there would be no bikes left, even for essential services. And those who had looked after their bikes would complain when theirs were suddenly taken back for redistribution, making them lose the beneficial activities they had built around their bikes, effectively punishing them for having acted correctly.

    So, to accommodate incentives, the legitimate complaints of prudent cyclists, stability and socially beneficial outcomes, a replacement bike could be claimed only if the first had been lost unavoidably or while discharging social duties, such as civil defense activities, which were socially needed but voluntary: it did not matter who did them, provided they got done. The concerns that led to the rejection of Constant Redistribution require luck egalitarian distribution to take this form. If those who lost their bikes rescuing flood victims could not get a replacement, the system would fail to provide the values for whose sake Constant Redistribution was rejected.

    By avoiding both Constant Redistribution and Constant Redefinition of Duties (that is, deciding case by case, in light of the resulting distribution, whether a given act was a duty discharge), the scheme had a clear grounding, avoided circularity, and was simple enough for people to understand and endorse.

    I agree that luck egalitarian views, Dworkin’s included, require further specification in various respects (I list some in “Responsibility and Gender”, 2023). But I also think that adopting an institutional perspective often avoids problems we might otherwise find intractable, such as that of people being mistaken about their true duties. So I wonder if the problems that have emerged would be more manageable if luck egalitarian views remained within the institutional setup for which many of them were conceived. What counts as a duty discharge can then be fixed by public rules justified on independent grounds. And where the considerations are closely balanced, as in the cases Carl’s second point concerns, institutions can settle the matter, rather than leaving imperfectly informed individuals to recalculate unilaterally.

  14. Thanks for replies, David!

    3. My Fixed Expectations claim is quite sweeping and precludes changing people’s desert via education. The basic rationale is that the education one receives as a child is a matter of brute luck and therefore cannot raise or lower expected desert. It would seem unfair if you could be raised to have predictably lesser desert and therefore a predictably lesser distributive claim. Note that I do allow desert (as opposed to expected desert) to be changed by intervention: if I put a knife in front of a usually good but momentarily enraged person, I do not change their expected desert but I do change their actual desert; it either goes up compared to the no-knife scenario as they successfully resist the knife, or it goes down because they stab their enemy, these two effects equalling out (in terms of desert weighted by probability) such that expected desert is unchanged. Still, even if Fixed Expectations addresses circularity cases, it is quite a departure from the usual assumptions of moral luck egalitarians, and to some may seem just as counterintuitive as other escape routes such as prudential luck egalitarianism.

    2. Yes, it’s a fair point about whether making both options permissible would be acceptable given there may be significant differences in the magnitudes of the welfare and unfair inequality. I’ll reflect on that.

    1. I’m happy to accept that even one case can be a big problem for a theory. Whether there’s such a case here feels like it depends on whether we’re looking at the moment of the counterexample, the moment of the re-weighting to defeat the counterexample, the moment of the reworked counterexample, and so on! But that can’t really be right. I suppose if we assume infinite iterations, the luck egalitarian will be chased back to lexical weighting, and won’t be a pluralist in your sense anymore – i.e. the objection would go through. I wonder if there’s a way to block that. More to think about!

  15. What a great paper! Just a few comments, some of which are rather minor.

    1. A minor point: according to what David calls the canonical slogan form of luck egalitarianism (LE), ‘it is unfair if some are worse off than others through no fault or choice of their own.’ LE is compatible with the view that all inequalities are, in the relevant respect, due to luck. Call this latter view the ‘all luck view’. Combining LE and the all luck view implies that all inequalities are unfair. It may be suggested that this isn’t really a version of LE but rather of outcome egalitarianism, but I think this would be a mistake. Rather, it would be a way of deriving outcome egalitarianism from the conjunction of a more basic egalitarian principle, LE, and a claim about luck. Anyway, this conjunction would seem to avoid both the circularity problem and the arbitrariness problem. As I said, this is a minor point, because most luck egalitarians endorse luck egalitarianism at least in part because it allows them to hold people responsible for their choices but not for (brute) luck. So this point is really only about the scope of the challenge to LE.

    2. Like Carl, I was thinking that, in Rescue II, the judgement that rescuing Eamon is morally required could be derived from the weighting function. To this David responds, if I understand him correctly, that for any weighting function, an example analogous to Rescue II can be devised. This may well be true, but given the complexity that such weighting functions can have, this may be a bit more difficult to show. Or perhaps all I am saying here is that in my own case, I would need to think more about it.

    3. A couple of points about the discussion of Rescue II, assuming that not rescuing Eamon is morally required. (a) What if the measure of unfair inequality were sensitive not only to who is most (un)deserving, but (also) to the difference in their level of deservingness? Since the difference between their levels of deservingness is greater if Deirdre doesn’t rescue Eamon than if she does, might that not impact the level of unfair inequality (say, such that the difference between their welfare levels is less objectionable if Eamon isn’t saved)? (b) Small point, but it is assumed that aggregate welfare is a relevant concern in this comparison (and so might speak in favour of rescuing Eamon). Later, this concern is motivated with a concern about levelling down. However, all that is required to avoid a commitment to levelling down is that egalitarianism should satisfy the Pareto principle. And in Rescue II, the two outcomes are Pareto incomparable.

  16. Xu Bowen, thank you for sharing these questions here, as well as in your longer email to me that goes into more detail on each of your two questions. I’m delighted to have such a careful and thoughtful reader! I’ll reply later, separately, to your longer email with some more thoughts on some of the more detailed points that you make here. But for the purposes of this discussion, let me just say a couple of bigger-picture things. In reverse order:

    2. Yes, I think your second question here—i.e., the question about a version of luck egalitarianism according to which creating vs merely redistributing disadvantages marks a line between those inequalities that are not unfair and those that are—is closely related to the question that Dmitry Sereda raised earlier in the discussion. Your working through of the details of how that version of luck egalitarianism would apply to my Rescue II case looks correct to me. So the main thing I would want to say, as I noted in the earlier reply, is just that I think there are independent reasons to think that this version of luck egalitarianism would be plausible only if it were revised to track whether someone is redistributing disadvantage in a *reasonable* way (for the reasons I summarize in that earlier reply). And then because, in Rescue II, the moral deservingness of the potential rescuee (Eamon) *differs* depending on which option is chosen (thereby affecting whether the rescue would *be* an instance of redistributing disadvantage in a reasonable way), the ‘moralized’ version of this view will still encounter the circularity problem. That’s still just a quick sketch, but hopefully that helps to fill in a little more of my thought than my earlier reply did.

    1. I’m still thinking through the case that illustrates your first thought, but if I’m understanding it correctly, it’s quite a fascinating point! What it might show is that there’s a *broader* class of problems, of which the circularity problem is just *one*, that might affect a *broader* range of versions of luck egalitarianism beyond the MRS luck egalitarianism that I discuss in the paper. I’ll just note for now that one thing I’d like to think about further—this is what I’d like to follow up with you about by email—is how similar or different, ultimately, the circularity problem is to what you call the reflexivity problem. As you note, one thing that seems to be driving your reflexivity problem is choices that different people make at different times, and how luck egalitarianism seems to have to change its assessment of the former choices in light of the latter choices. That may mark an interesting difference from the circularity problem that I discuss, but I’m not 100% sure yet. I’ll look forward to further discussing this with you over email!

  17. Serena, just a quick note on your reply to say thank you so much—and, yes, I’d love to follow up with you further about this, and will drop you an email! The set of deeper reasons that you suggest here—and that that passage from Rawls also nicely suggests—why one’s account of moral deservingness (or moral worth) can’t be ‘prior’ to one’s account fairness (or justice) are questions that I would really like to, and should, think much more about. (The issues here also now strike me as not unrelated to the issues that Paula Casal raises below—i.e., the thought that this kind of ‘technical’ problem might be avoided once luck egalitarianism is housed in, or seen in the first instance as, as an account of justice.)

  18. Patrick T: Oh wow, I love this! (I’m now realizing that one consequence of writing a paper that proposes a problem is that one invites far more ingenious potential solutions than one could have thought of oneself!) I’m still thinking this proposal through, so could I ask you to help me with a small first question about it? A key moving piece here, if I’ve got it right, is the background ‘non-obligation by default’ principle—it’s that background principle that breaks the symmetry that afflicts (what I’m now inclined to describe as!) ‘naïve’ MRS luck egalitarianism, thereby breaking the path toward the circularity problem and contradiction, etc. My initial thought then was that one might just as well have assumed a background ‘obligation by default’ principle, and one would then get a mirror-image view that would get the opposite result in my Rescue II. Is that right? Or is there a reason to like ‘non-obligation by default’ better?

  19. Hi all,

    Thanks for a wonderful paper and a very interesting discussion so far! I apologize if I’m repeating things people have said already, I only quickly skimmed the comments, though I hope to read them in more detail later.

    A couple quick thoughts. First, I think characterizing the problem as a cycle of dependence understates it. Cycles of dependence come in better and worse kinds, and this is the bad kind, the kind that generates an outright inconsistency: in Rescue II, if the act is required then it isn’t required, and if it isn’t required then it is. David already grants this in his reply to Carl (“anything’s better than a contradiction!”), but I think the point is a bit stronger than the equipoise worry he raises there. Carl’s option, saying that in Rescue II-style cases neither act is required, isn’t merely in need of an explanation; it refutes itself the same way the two original horns do. If neither act is a wrong, Deirdre’s desert is 0 whichever she chooses. Then rescuing puts the more deserving person worse off (unfair) and not rescuing doesn’t, so not rescuing is best, so it is required after all. The only stable way for the theory to “go quiet” is for the indeterminacy to run all the way around the loop: no fact about whether rescue is required, hence no fact about Deirdre’s desert, hence no fact about whether the inequality is unfair. That’s coherent, but it’s a commitment to metaphysical indeterminacy about what Deirdre ought to do, not just to the two options being equally good. (Maybe that’s what David was thinking, anyway?).

    I think (and I want to emphasize that this is really not my area) that whether a cycle generates an inconsistency (rather than mere “ungrounded” bootstrapping, like the truth-teller, which many philosophers find easier to live with) depends on the signs of the explanatory arrows and on there being a jump somewhere in the loop. Basically, what generates inconsistency is a negative feedback loop in which some step is discontinuous (here: bivalent requirement, desert dropping from 0 to −2) and the jump is big enough to leap over any consistent assignment. I’ll come back to the issue of continuity below, since I think it might offer a way out.

    Second, I think the paper may pitch the problem too narrowly by attaching it to luck egalitarianism per se. (David’s notes 15 and 17 gesture at this, but he leaves the generality open without committing.) Staying in the distributive neighborhood for now, the liar structure arises for non-egalitarian comparative desert views, too. Those views say that inequality is bad whenever the desert-sensitive LE view says it’s unfair, then add on top of that that some equality is bad, too, and nothing about Rescue II depends on the egalitarian default. What it depends on is comparison: in a comparison, the wrongdoer’s being better off counts as a cost of the right act (the complier is now the deserving one left worse off than a wrongdoer), which is what flips the sign. Interestingly, I think non-comparative desert views, by contrast, give rise to the truth-teller structure but not the liar: on an absolute (Kagan-style) view the wrongdoer’s being well off is a bad added to the wrong act, so requirement raises the value of complying rather than lowering it, and you get two consistent fixed points instead of none.

    More generally, once you see that any view on which explanatory factors have this particular structure gives rise to this problem, it seems clear that there are going to be other examples in ethics, including ones that have nothing to do with requirement. Just off the top of my head: (1) heroic self-sacrifice (of the right kind) non-instrumentally makes one’s life better; (2) whether an act is a self-sacrifice depends on whether it makes your life on the whole worse. If the heroic bonus exceeds the gross cost of what’s given up, the view is inconsistent: if it’s a sacrifice, it’s heroic, so you come out ahead, so it isn’t; if it isn’t, there’s no bonus, so it is. You can restore consistency by constraining the intrinsic value of a sacrifice to always be less than the cost of the stuff sacrificed (gross, not net, or the constraint is itself circular, which in effect means the bonus must scale as a fraction of what’s given up). That’s a moderate bound like the one David rejects for RII in fn 46, and I think the contrast is instructive: the bound looks principled for sacrifice, since the value of a sacrifice is plausibly derivative from what’s given up and so can’t exceed it, whereas David rightly rejects the analogue for RII, both because it can be revenged by retuning the numbers and because the cap looks undermotivated.

    This is where I want to come back to the issue of continuity, because I think moving from a bivalent variable (required or not) to a continuous one might offer a way to have the virtues of the moderate approach without the problematically unprincipled character. Suppose we let the desert cost of a wrong track something continuous, like how seriously wrong a failure to rescue would be, where seriousness is all-things-considered: how decisively the balance of moral reasons, including the fairness consideration itself, favored rescue. If the dependence is continuous then a negative loop has to have a unique consistent assignment of values (think of it like the equilibrium a negative feedback loop reaches in a dynamic system, like a thermostat reducing the power to the air conditioner as the temperature gets closer to its target). In RII, welfare considerations create a strong reason to rescue, and the negative feedback from unfairness of undeserved inequality attenuates the seriousness of the wrong, but that feedback is self-limiting. The fairness reason feeds on the seriousness of the wrong (of failing to rescue), so it consumes its own fuel. The reason this has the virtues of the moderate approach without the stipulated cap is that there is no cap. The fairness reason can’t win not because of a restriction placed on it, but because its ground (Deirdre’s undeserved disadvantage) is proportional to the seriousness of her wrong, which shrinks as the reason approaches winning. There’s no free parameter to retune, so no revenge case. In RII the upshot is that Deirdre is required to rescue, but her failing to do so would be a less serious wrong than the welfare stakes alone suggest — which strikes me as the right answer, and a familiar kind of mitigation rather than a paradox.

    So, I guess my question, David, is: What do you think about going continuous about the bearing of moral demands on desert as a way out?

  20. Paula, thank you for this! I think you’re just 100% right that the problem I’m focusing on here arises when one just takes a rather abstractly defined normative-ethical view (like an act-focused pluralist consequentialist moral theory that incorporates MRS luck egalitarianism as its account of unfair inequality) and then mechanically applies it to a rather abstractly defined case (like my Rescue II case). And I also think you’re just 100% right that, rather than fussing around with the details of that rather abstractly defined view (and the details of what might ground it), one might do better to step back and ask (as I take you, in effect, to be doing) if that was *ever* the most plausible way to think about the view in question at all. Maybe this sort of ‘technical’ problem only arises if one plucks luck egalitarianism out of its natural habitat—as you say, as a view in *political* philosophy in the first instance, with all the extra constraints that implies, and that the cases in your post nicely illustrate—and is a problem can be straightforwardly sidestepped by just putting the view back *in* its natural habitat!

    A couple of thoughts on that: (1) If that *is* ultimately the upshot of the paper, it’d be a different upshot than *I’d* been thinking, while writing it, that the paper might have. But it wouldn’t, I think, be an uninteresting upshot! It would be an interesting reason to think that luck egalitarianism *can really only* plausibly be taken as a view in political philosophy. (2) I myself would be downcast to discover that was true! I’ve come around to thinking that questions about whether luck egalitarianism is eligible to be part of an account of political justice are very complicated (for some of the reasons that Gina Schouten discusses in her excellent recent book The Anatomy of Justice (Oxford University Press, 2024), although Gina ultimately thinks there *is* a role, albeit a somewhat limited one, that luck egalitarianism can play in such an account). And so I’d been motivated to think that, whether or not luck-egalitarian unfairness is the special kind of bad that rises to the level of political injustice, it was at least worth investigating whether it was possible to defend the thought that the world is, in one way, *worse* if it contains luck-egalitarian unfairness. It was thoughts along those lines that I took to motivate the more complicated project that I try to engage in, in the paper, of seeing if there *is* a way to work out luck egalitarianism as a simple normative-ethical view. (3) One thing I’m wondering—and this is now a bit more speculative—is whether a *structurally* similar kind of circularity concern might arise even for the strictly political-philosophical version of luck egalitarianism. I do take your point that given the principled reasons to avoid constant redefinition of duties, avoid constant redistribution, have principles of fairness apply to institutional arrangements in the first instance, and so on, that it would be harder to develop a case that conclusively shows circularity to be a potential possibility for such a version of luck egalitarianism. But, as I mention (very) briefly in a footnote in the paper, offhand I’m inclined to think that a choice with *something* like the structure of my Rescue II case could in principle by faced by, e.g., decision-makers who are designing educational institutions (given that such institutions expectably affect both future people’s welfare and how expectably deserving—institutionally and otherwise—those people will be). And so I’m wondering if my circularity problem might be at least of some *indirect* interest, to those inclined toward strictly political-philosophical versions of luck egalitarianism, insofar as there’s something in the vicinity of it that they, too, will need to build their theories to avoid.

  21. ANNOUNCEMENT

    Hi All. I’ve been in touch with David. For some technical reasons yet to be identified a whole load of replies by him have not appeared on the blog.

    If you’ve posted something, please hang tight and hopefully those replies will appear soon.

    In the interim, thanks so much for contributing to a great discussion of a really fascinating paper.

  22. Patrick: Really grateful to you for posting that interim announcement. Just trying a couple of tests here to see if I can fix this glitch.

  23. Okay, I’m going to take advantage of this apparently working again to try to get a few replies out!

    Serena, thank you so much for your further note—and, yes, I’d love to follow up with you about this, and will drop you an email! The reasons that you suggest here—and that the passage you quote from Rawls nicely suggests—why one’s account of moral deservingness (or moral worth) can’t be ‘prior’ to one’s account of fairness (or justice) are questions that I would really like to, and should, think much more about. (The issues here also now strike me as not unrelated to the issues that Paula Casal raises below—i.e., the thought that this kind of ‘technical’ problem might be avoided once luck egalitarianism is housed in, or seen in the first instance as, as an account of justice. Some more on them in my further reply below.)

  24. Xu Bowen, thank you for sharing these questions here, as well as in your longer email to me that goes into more detail on each of your two questions. I’m delighted to have such a careful and thoughtful reader! I’ll reply later, separately, to your longer email with some more thoughts on some of the more detailed points that you make here. But for the purposes of this discussion, let me just say a couple of bigger-picture things. In reverse order:

    2. Yes, I think your second question here—i.e., the question about a version of luck egalitarianism according to which creating vs merely redistributing disadvantages marks a line between those inequalities that are not unfair and those that are unfair—is closely related to the question that Dmitry Sereda raised earlier in the discussion. Your working through of the details of how that version of luck egalitarianism would apply to my Rescue II case looks correct to me. So the main thing I would want to say, as I noted in the earlier reply, is just that I think there are independent reasons to think that this version of luck egalitarianism would be plausible only if it were revised to track whether someone is redistributing disadvantage in a *reasonable* way (for the reasons I summarize in that earlier reply). And then because, in Rescue II, the moral deservingness of the potential rescuee (Eamon) *differs* depending on which option is chosen (thereby affecting whether the rescue would *be* an instance of redistributing disadvantage in a reasonable way), the ‘moralized’ version of this view will still encounter the circularity problem. That’s still just a quick sketch, but hopefully that helps to fill in a little more of my thought than my earlier reply did.

    1. I’m still thinking through the case you sketch that illustrates your first thought, but if I’m understanding it correctly, it’s quite a fascinating point! What it might show is that there’s a *broader* class of problems, of which the circularity problem is just *one*, that might affect a *broader* range of versions of luck egalitarianism beyond the MRS luck egalitarianism that I discuss in the paper. I’ll just note for now that one thing I’d like to think about further—this is what I’d like to follow up with you about by email—is how similar or different, ultimately, the circularity problem is to what you call the reflexivity problem. As you note, one thing that seems to be driving your reflexivity problem is choices that different people make at different times, and how luck egalitarianism seems to have to change its assessment of the former choices in light of the latter choices. That may mark an interesting difference from the circularity problem that I discuss, but I’m not 100% sure yet. I’ll look forward to further discussing this with you over email!

  25. Patrick T: Oh wow, I love this! (I’m now realizing that one consequence of writing a paper that proposes a *problem* is that one invites far more ingenious potential *solutions* to it than one could have thought of oneself!) I’m still thinking this proposal through, so maybe I could ask you to help me with a small first question about it? A key moving piece here, if I’ve got it right, is the background ‘non-obligation by default’ principle—it’s that background principle that breaks the symmetry that afflicts (what I’m now inclined to describe as!) ‘naïve’ MRS luck egalitarianism, thereby breaking the path toward the circularity problem and contradiction, etc. My initial thought then was that one might just as well have assumed a background ‘obligation by default’ principle, and one would then get a mirror-image view that would get the opposite result in my Rescue II. Is that right? Or is there a reason to like ‘non-obligation by default’ better?

    (Probably you can see that the impetus behind this question—just in case the discussion closes here before we get a chance to talk more about it here—is the worry that perhaps this way out would leave luck egalitarians without good grounds to choose between the ‘no obligation by default’ and the ‘obligation by default’ versions of the view, at least unless one appended some underlying account of why fairness favors one over the other [i.e., without getting in to the basis-of-luck-egalitarianism issues that I think the MRS luck egalitarian, too, has to get in to].)

  26. Paula, thank you *so much* for this! I think you’re just 100% right that the problem I’m focusing on here arises when one just takes a rather abstractly defined normative-ethical view (like an act-focused pluralist-consequentialist moral theory that incorporates MRS luck egalitarianism as its account of unfair inequality) and mechanically applies it to a rather abstractly characterized case (like my Rescue II case). And I also think you’re just 100% right that, rather than fussing around with the details of that rather abstractly defined normative-ethical view (and the details of what might ground it), one might do better to step back and ask (as I take you to be doing) if that was *ever* the most plausible way to think about the view in question at all. Maybe this sort of ‘technical’ problem only arises if one plucks the view out of its natural habitat—as you say, as a view in *political* philosophy in the first instance, with all the extra constraints that views in that domain may bring with them, and which the cases in your post nicely illustrate—and maybe such ‘technical’ problems can then be straightforwardly *sidestepped* by just putting the view back *in* its natural habitat!

    A couple of thoughts on that: (1) If that *is* ultimately the upshot of the paper, it’d be a different upshot than *I’d* been thinking, while writing it, that the paper might have. But it wouldn’t, I think, be an uninteresting upshot! It would be an interesting reason to think that luck egalitarianism *can really only* plausibly be taken as a view in political philosophy. (2) I myself would be downcast to discover that were true! I’ve come around to thinking that questions about whether luck egalitarianism is eligible to be part of an account of political justice are very complicated (for some of the reasons that Gina Schouten discusses in her excellent recent book The Anatomy of Justice (Oxford University Press, 2024), although Gina ultimately thinks there *is* a role, albeit a somewhat limited one, that luck egalitarianism can play in such an account). And so I’d been motivated to think that, whether or not luck-egalitarian unfairness is the special kind of bad that raises to the level of political injustice, one could at least defend the thought that the world is, in one way, *worse* if it contains luck-egalitarian unfairness. And I took that kind of thought to motivate the more complicated project that I try to engage in, in the paper, of seeing if there *is* a way to work out luck egalitarianism as a simple normative-ethical view. (3) One thing I’m wondering—and this is a bit more speculative—is whether a *structurally* similar kind of circularity concern might arise even for the strictly political-philosophy version of luck egalitarianism. I do take your point that given the principled reasons to avoid constant redefinition of duties, avoid constant redistribution, have principles of fairness apply to institutional arrangements in the first instance, and so on, that it would be harder to develop a case that conclusively shows circularity to be a potential possibility for such a version of luck egalitarianism. But, as I mention (very) briefly in a footnote, I’m inclined to think that a choice with *something* like the structure of my Rescue II case could in principle by faced by, e.g., decision-makers who are designing educational institutions (given that such institutions, e.g., expectably affect both future people’s welfare and how expectably deserving—institutionally and otherwise—those people will be). And so I’m wondering if my circularity problem might be at least of some *indirect* interest, to those inclined toward strictly political-philosophical versions of luck egalitarianism, insofar as there’s something in the vicinity of it that they, too, will need to build their theories to avoid.

  27. Victor, thank you!

    1. I co-sign the way you’re thinking about Rescue I. The circularity problem doesn’t arise in *that* case, and it doesn’t arise for the reasons you say. If it arises anywhere, it’s only in cases in which an inequality produced by a rescue *if unfair* would ‘tip the balance’ as to whether or not the rescue is morally required (and only for versions of luck egalitarianism on which the unfairness of the inequality, in turn, would depend on whether the rescue were morally required). Rescue I isn’t like that, because as *that* case is built, the inequality even if unfair wouldn’t be enough to ‘tip the balance’ and change which option is morally required. So one certainly shouldn’t think that what I call moral-requirement-sensitive (MRS) luck egalitarianism is vulnerable everywhere to circularity problems. They arise, if they arise at all, only in a quite peculiar kind of case.

    2. Ah, what you say about Rescue II is lovely! On the view that you might be inclined to take about it, if I’m putting its different entailments together correctly: (a) Deirdre *is* required to rescue; (b) the resulting inequality *is* unfair, and nevertheless (c) it would also be (equally as?) unfair *not* to have had that inequality. So one can have one’s cake and eat it too, so to speak, by recognizing lots of different *kinds* of unfairnesses that are floating around in the case, and (e.g.) do one’s luck-egalitarian theory-building by choosing the grounds of the rescue being required in such a way that they differ from the grounds of the inequality that results from the rescue being unfair. Much as I said to Patrick’s suggestion, this is just *another* ingenious kind of way of thinking about the case that simply hadn’t occurred to me. Although inclined to pluralism, I hadn’t thought about *that* kind of more thoroughgoing pluralism about different kinds of (equality-related) unfairness, as I’m framing your thought to myself. (And yet, anything’s better than a contradiction, so I *don’t* think that your way of thinking about the case can be just waved away.) The resulting view would be quite a distance from the simpler MRS luck egalitarianism. MRS luck egalitarianism takes a simpler view of the sources of *unfairness* in the case. But of course, as I discuss in the paper’s final section, it too has to introduce complexity *somewhere*, in order to avoid entailing a contradiction in Rescue II. The MRS luck egalitarian opts to introduce the further complexity in its account of the source of inequality-given moral reasons and (relatedly) how they interact with the other reasons at stake in the case. And so I’m inclined to think that, ultimately, litigating whether your suggestion is a more or less theoretically choiceworthy way to avoid the circularity problem than MRS luck egalitarianism is will come down to figuring out which of the two views has had to introduce the *least-nicely-explainable* kind of complexity. That strikes me as important (and really hard) further work to do.

  28. Nils, thank you *both* for the very kind words *and* for these comments! In brief reply:

    1+2. Yes, I do agree that all-luck egalitarianism offers a potentially attractive way out of the problem. I say a little in section III of the paper, however, about some complications with this way out. Boiling down what I say there, I think (take my cue from some remarks of Carl Knight’s is his tremendous recent work on all-luck egalitarianism) that all-luck versions of luck egalitarianism can be divided into ‘prudentialist’ and ‘moral(ized)’ versions (depending on whether or not they take the relevant kind of bad luck to be that which pertains only to one’s own interests or not), and that ‘prudentialist’ all-luck egalitarians will face a version of what I call the problem of arbitrariness whereas ‘moral(ized)’ all-luck egalitarians will face a version of the circularity problem. (I only briefly explain why in the paper – one needs a risk-involving variant of my Rescue II case to show this, and I agree that, as you also say in your point #2, it would be good to take a look at concrete examples of that and other cases, if only to rule out the possibility that they would have some feature that would allow some *other* way to avoid or address circularity concerns.)

    3. On your 3(b) here: yes, I think that it would take a bit of further work here to show that a version of the circularity problem would afflict any plausible moral theory that incorporated MRS luck egalitarianism, no matter which other moral factors it incorporated. You’re right that it was a bit quick of me to say that something like aggregate welfare is needed as one of those other moral factors. I guess I was just taking it that it would be heroic for the MRS luck egalitarian to try to wave away the circularity problem on foot of her having the optimistic thought that there could be *no* other factor in the correct moral theory that could operate, vis-à-vis unfair inequality, relevantly like aggregate welfare does in my Rescue II.

    What I’m very very taken with, though, is your point 3(a) here. I think this really does point to a potential way out of the Rescue II problem for the MRS luck egalitarian. It *wouldn’t* represent a massive revision of the view that would require significant further explanation. And it wouldn’t require any fussing around with questions about how the inequality-given moral reason in the case interacts with the other moral reasons in the case. The idea here also reminds me of some remarks of Shelly Kagan’s, in his The Geometry of Desert, to the effect that the intrinsic disvalue of people not having what they morally deserve might vary, depending *on* the absolute levels that they morally deserve. (We might justifiably care more about the very morally undeserving or the very morally deserving getting what they deserve, and less so about moral mediocrities like the rest of us. I’m temporarily forgetting what Kagan calls that—in his graphical terms, a V-shaped skyline for desert, perhaps?) So I think it arguably enjoys some independent motivation too. It may be worth noting that the resulting view would depart from MRS luck egalitarianism at least as it appears in the writings of those who I take to be its canonical defenders in the literature, and that I’m not quite sure yet myself what I think about all the above. But I *do* think this idea is a starter. The thing I’d want to look into next is how essential it *is* to Rescue II that there is such a *significant* difference in the absolute levels of moral deservingness of its characters, under each option that that the rescuer has. If this isn’t an essential feature of the case, then perhaps the case could be further tweaked to re-generate a circularity problem even for this more sophisticated version of MRS luck egalitarianism that you in effect suggest here. But this would take further work to check. What a brilliant thought—thank you.

  29. What a fabulous set of comments, Jake Z; thank you!

    On your first point: In response to your question at the end of this part of your post, yes, this is roughly the thought that I’d implicitly had in mind in those earlier replies—thank you for spelling it out so clearly. What I say in the paper, which I think is sympatico with what you say here, is that the *problem* for pluralist egalitarian consequentialism (w/ MRS luck egalitarianism as one component of it) is that it entails a contradiction in Rescue II, and *what explains why* it does so is the (particular kind of) cycle of dependences that it, that moral theory, induces in that case.

    On your second point: Right, as you note here, there’s a general formal structure—one that hasn’t anything to do with distributive ethics or luck egalitarianism in particular—in the background of the problem here. As I’m glad you noted, that formal structure has an affinity with the structure of the famous liar paradox. And, as you say, one can multiply cases in normative ethics and beyond with that structure. (In fact, in some earlier work with the phenomenal Ben Schwan, we’ve explored some cases with a similar structure that implicate issues in metaethics. That’s in our ‘Wrongfulness rewarded’ (2021), Synthese 199, 6897-6916. The cases are not unlike your examples here.) Why focus on luck egalitarianism then? Might one not just want a more general solution to a general problem? My thought here was that, at least in the context of distributive ethics, some views that are candidates to be the truth about distributive ethics *aren’t* obviously going to be vulnerable to the problems associated with cases with that general formal structure. Even non-MRS luck egalitarianism isn’t, for example. And while some of luck egalitarianism’s competitors may be (e.g., as you note, views that take one fairness-related moral factor to be people having what they absolutely deserve), many others may *not* be. To the extent that some versions of sufficientarianism or prioritarianism, for example, aren’t *motivated* by an underlying concern for fairness (but, e.g., rather by an underlying concern for compassion as some of their defenders have sometimes suggested), they may not face analogous pressure to incorporate the equivalent of a moral-requirement-sensitivity condition into their view. (It’s that underlying concern about fairness, as I take it, that looks like the initial motivation for being a moral-requirement-sensitive egalitarian.) And so, I was thinking, if a view like MRS luck egalitarianism is unhappily *distinctive*, vis-à-vis its relevant competitors, in facing an instance of this general problem, there’s then motivation (insofar as we’re interested in figuring out the correct view about distributive ethics) to ask whether *it* can escape the problem, *either* by appealing to something distinctive to it and its underlying concern about fairness (as I and, inter alia, other commentators here have suggested) *or* (as you I think in effect suggest) by looking to some strategy that looks like it will help with the more *general* problem (for *all* the views, in normative ethics and elsewhere, that face it) of which the circularity problem in the paper is a particular instance.

    And then finally on your third point, the positive proposal: I think there’s a couple of different ways of developing this, as I’m understanding it. I thought initially when you said “moving from a bivalent variable (required or not) to a continuous one”, you meant revising the moral theory of which MRS luck egalitarianism would be a part to be a kind of scalar consequentialism—i.e., one that doesn’t recognize the traditional deontic categories (like permissible or impermissible). But then reading further on I thought you meant something more like, not altering the deontic categories, but ‘fixing’ how the moral factors that *determine* the deontic category of an action ‘work’ in the theory, in precisely the way needed to avoid anything with the Rescue II structure arising. (Your helpful air-conditioner analogy helped me to glom on to this idea. *That* idea reminds me a bit of, in a different context, some luck egalitarians’ observation that one can avoid all-things-considered levelling-down problems in part by observing that, in simple cases, the degree to which levelling down improves things, equality-wise, is always exactly counterbalanced by the degree to which it disimproves things, welfare-wise.) So I think, if I’m understanding it correctly, that there’s at least those two ways of developing your ‘continuous variable’ strategy for avoiding the problem. I’d probably want to think a bit more about each development of the idea; my immediate thought is that these seem like at least *somewhat* theoretically costly ways to go, given their potential implications in other cases, to save the luck egalitarian from this problem. And then, if I’m understanding it correctly, the feature here that saves the day is to have the net weight of an inequality-given moral reason turn on, not on the ‘competition’ between it and other moral factors, but rather (in part) on whether it, if sufficiently weighty, would induce circularity. That strikes me at first blush as an unusual *kind* of fact to be determining the the weight of a moral reason (in effect, having the deontic level feeding back, so to speak, to the factoral level, whereas pre-theoretically I would have wanted the former to be determined *by* the latter). Maybe that’s not ultimately unpalatable, but I think there’d be a bit of work to do, to compare the theoretical costs and benefits of this proposal as against retaining MRS luck egalitarianism in its simpler form but appending to it one of the underlying explanations of how fairness-given moral reasons work that I discuss in the final section of the paper. Phew! Really interesting stuff.

  30. Carl: Thank you for that follow-up comment! Okay, I will definitely have to read that forthcoming paper of yours. In light of what you say in your #1, I’m now coming back now to thinking more carefully about the expected vs actual desert distinction and realizing that I hadn’t been carefully enough appreciating how that bears on the ‘problem’ Rescue II case and its construction. I certainly hadn’t had it enough in mind in constructing the case, and now I’m wondering (something you raised in your first comment) about whether I think luck egalitarianism *is* most plausibly formulated to be sensitive to expected or actual desert. (The issues here are not the exactly same as, but are suddenly reminding me a bit of, the kind of thing that Nir Eyal & Anders Herlitz discuss in their marvelous* ‘Input and output in distributive theory’ article. Rescue II is formulated to target, as I’m remembering the distinction they draw, a broadly ‘output’-style version of MRS luck egalitarianism.)

    * I keep referring in my replies to other people’s work and saying that it’s marvelous &c., and I hope that isn’t sounding vacuous at this point. I really do just love a lot of the recent literature on distributive ethics!

  31. My recollection is that the discussion is due to close shortly, so just in case it does, let me say—what I hope is already implicit in many of my replies above—just an absolutely heartfelt thanks to everyone who took part in this (and to the organizers for inviting me to do it, and for setting it up and running it all so well). I wasn’t sure beforehand what it would be like to do philosophy in this format, but now I feel like I would happily do it every week if I could!

  32. Thank you all for a wonderful discussion. Special thanks to David O’Brien for the paper, to Kasper Lippert-Rasmussen for his thoughtful comments, and to everyone who joined the conversation.

Comments are closed.