THE DILEMMA FOR STRUCTURAL INJUSTICE

To begin making the central puzzle more concrete, consider a widely discussed example from Young (Reference Young and Nussbaum2011, 43–4): the case of Sandy, a single mother and commuter in a typical American city who ends up facing homelessness when her current apartment building is sold to a developer, setting off a chain of events in which various structural constraints produced by the actions of countless others—from the price and availability of housing in her city to the quality of schools in different neighborhoods—prevent her from finding a new residence, leaving her on the verge of homelessness. As Young sets up the case, no one treats Sandy wrongly during this ordeal. Nonetheless, Sandy finds herself facing a common and, Young argues, unjust situation. It is unjust that anyone should face the prospect of homelessness, especially in an affluent society. Sandy experiences a structural injustice, and this is a distinct phenomenon from agential wrongdoing.

Many have found Young’s description of Sandy’s situation compelling. Others have different reactions. One is that Sandy’s situation is surely the product of at least one (collective) agent’s wrongdoing: her state. Her state should be doing more to ensure that Sandy and others like her are not vulnerable to homelessness; it bears responsibility for her situation. Another is that Sandy’s situation is not an injustice, however unfortunate it may be, given that it is not the result of anyone having wronged her in any way. Of course, she is entitled to some help to get back on her feet—but that is a matter of human decency or charity, not a matter of her having suffered an injustice.

These reactions correspond to what I call the redundancy challenge and the normative challenge, respectively. The redundancy challenge says the concept of structural injustice is unnecessary: everything we want to say with the concept of structural injustice can and should be said in terms of individual and collective wrongdoings. The normative challenge says the concept of structural injustice is implausible: to the extent that it is essentially about something other than agential wrongdoing, it lacks the normative credentials to be an injustice. Despite their important insights, existing accounts of structural injustice are vulnerable to at least one of these challenges.

Young’s account is vulnerable to the normative challenge. She maintains that structural injustice is distinct from individual and institutional wrongdoing, but she does not offer an alternative explanation of why it is unjust and not merely tragic. Situations like Sandy’s, she suggests, are obviously unjust. But this is not a satisfactory answer to the normative challenge, as David Estlund (Reference Estlund2024) has recently argued. For any situation that seems obviously unjust, he asks, suppose that the exact same situation had been produced by an unpreventable natural disaster like a hurricane. Such occurrences are commonly considered tragedies, not injustices. The relevant difference between the two is a moral, specifically a deontic, fact: unpreventable natural disasters are not a matter of anyone having treated others wrongly, indeed not a matter of human action at all.Footnote ¹ What plausibly distinguishes an unjust state of affairs from a merely tragic one is its basis in the wrongful conduct of agents, conduct that licenses moral grievances like anger or indignation on the part of the victims. According to Young’s description, Sandy’s situation is not unjust in this sense. It is not a matter of anyone having treated her wrongly. Of course, Sandy’s situation differs from a hurricane insofar as it is the product of human action, specifically the actions of many agents interacting in the housing market. But the point of the hurricane analogy is precisely that there is no discernible morally significant difference between these two scenarios.

Recent accounts of structural injustice appear less vulnerable to the normative challenge (Mantouvalou Reference Mantouvalou2023; McKeown Reference McKeown2024; Nuti Reference Nuti2019; Powers and Faden Reference Powers and Faden2019). Departing from Young, they argue that structural injustice is, at least in large part, the product of wrongdoings by the state and other powerful collective or corporate agents. In McKeown’s (Reference McKeown2024, 181 and 44) words, “powerful corporate agents like states and corporations can be held morally responsible for structural injustice, either by omission (avoidable structural injustice) or action (deliberate structural injustice)” because “powerful agents have the capacity to change the situation, [but] they actively maintain it. So the consequences are not unintentional, they are intentional.” Sandy would not be facing homelessness if her state took action, such as improving tenants’ rights and implementing rent control. By failing to implement such measures now and in the past, her state bears responsibility for her situation. Once the role of these powerful collective agents is acknowledged, the claim that structural injustice is distinct from wrongdoing looks less plausible and, in any case, irrelevant to understanding injustices here and now. Importantly, these accounts do not completely part ways with Young’s account. They maintain that individuals are not culpable for their part in structural injustice, because the conduct of ordinary individuals is not “where the action is, both in terms of understanding structural stability or change and in attributing responsibility for structural injustice” (McKeown Reference McKeown2024, 36–7). Structural injustice is primarily the product of wrongful institutional conduct. Call this the Institutional View of structural injustice.

The Institutional View seems to offer a straightforward answer to the normative challenge: structural injustice is unjust because it is traceable to the wrongful conduct of powerful collective agents. Sandy’s situation is not a tragedy; it is a deliberate or avoidable injustice perpetrated by her state, real estate corporations, and other powerful institutions.

A closer examination, however, reveals some problems with this answer. The trouble lies in the Institutional View’s ambiguous stance on whether structural injustice is simply a product of collective agents’ wrongdoing or whether it is in fact the product of a much broader and complicated set of human actions involving, crucially, ordinary individuals. Is structural injustice ultimately the product of collective agents’ wrongdoings, or is it only largely the product of such wrongdoings (and if so, how large)?

Take patriarchy as an example. Estlund (Reference Estlund2024, 353) writes, “The many instances of individual behavior that together produce a patriarchal social structure…are, in large part even if not entirely, acts that are uncoordinated across strangers.” Here he challenges, rightly, the Institutional View’s claim that the conduct of ordinary individuals is not “where the action is” for structural injustice. Social structures like patriarchy are not just what powerful institutions do. Sharon Krause (Reference Krause2013, 194) recounts a memorable example of how social processes like gender socialization among individuals play an important role in patriarchy’s reproduction: “When I was a girl my mother was constantly enjoining me to ‘take smaller steps’ so as to be more ‘ladylike’. Apparently, my natural stride was too long, perhaps too forceful, to be feminine….[H]er exhortations, intended as loving guidance, did have the effect of socializing me to norms that subtly perpetuate inequalities between men and women.” Pointing to group agents thus offers only a partial answer to the normative challenge. To establish the deontic wrongness of structural injustice, the actions of many other agents must be normatively accounted for.

Defenders of the Institutional View could respond in two ways. One response is to move further away from Young’s account and say that ordinary individuals, not just states and corporations, are guilty of wrongdoing when their actions contribute to structural injustice.Footnote ² This would provide a full answer to the normative challenge. But this move is not available to the Institutional View given its central commitment to the uniquely wrongful role of powerful institutions for problems of structural injustice. More importantly, to say that individuals are also guilty for contributing to structural injustice and simply leaving the matter there would be to miss one of the important points of Krause’s example. Ordinary individuals do contribute to structural injustices like patriarchy, but the nature of their contribution is not always something that licenses guilt or blame. Krause (Reference Krause2013, 195) writes, “The truth is that we participate all the time in practices that subordinate us even as they perpetuate the domination of others, and much of the time these acts of subordination and domination are neither recognized as such nor intended, much less controlled.” Socializing one’s child into certain gender norms is now seen by many as wrongful: as Krause’s own reflections demonstrate, people have come to understand that this practice perpetuates gender inequality and stifles individual freedom. This understanding may have innocently eluded Krause’s mother who, Krause observes, was well intentioned and in many ways a victim of patriarchy herself. The normative credentials of structural injustice should not simply rest on the claim that all contributors are always acting culpably.

A different response that preserves the Institutional View’s core commitment is this: when individuals’ actions contribute to structural injustice, it is the state that bears responsibility for those actions. It is part of the state’s power, the Institutionalist can argue, to regulate individual conduct that leads to injustices like Sandy’s situation. Inadequate public regulation enables private actors to act in ways that create, exacerbate, and exploit certain forms of structural vulnerability (Mantouvalou Reference Mantouvalou2023). The state should intervene to ensure that uncoordinated acts across strangers do not lead to harm. These are not uncontroversial claims. Some may worry about the compatibility of such interventions with liberal limits on state power over individual choices, especially related to injustices like patriarchy; others may worry about their effectiveness, especially given those limits but even without them. For now, let us set these worries aside—after all, these worries may be overblown, and theorists of structural injustice need not be wedded to liberalism—and grant that the state has legitimate and effective power to regulate individual conduct in ways that would address structural injustice. This would provide a complete yet resolutely institutionalist answer to the normative challenge. The normative credentials of structural injustice are found in the wrongdoing of powerful institutions, both for their own conduct and that of individuals.

But now the redundancy challenge rears its head. The Institutional View, so understood, threatens to make the concept of structural injustice theoretically redundant, or at least significantly deflates what appeared interesting and distinctive about it. Much work in political theory has already argued for the view that social injustices are rooted in powerful institutions like the state and their profound and pervasive influence on people’s lives, what John Rawls (Reference Rawls1999) calls the “basic structure of society.” If structural injustice should be understood as a deliberate or avoidable wrong committed by powerful institutions, is it not just a species of institutional wrongdoing after all? What is gained, theoretically or politically, by calling Sandy’s situation a structural injustice rather than an institutional failure if the problem simply lies in her state’s unjust housing policies? The same can be asked about other issues like global poverty, which McKeown (Reference McKeown2024, 102) claims is an avoidable structural injustice because “international institutions and powerful states have the capacity to change the structures that underlie global poverty”—a familiar argument in the literature on global (distributive) justice (see, e.g., Pogge Reference Pogge2002). The Institutional View has the resources to fully answer the normative challenge, but in doing so, it collapses structural injustice into the familiar question of institutional justice (Browne Reference Browne2025, 31; De Bernardi Reference De Bernardi2024).

Some may argue that this is exactly the right way to think about structural injustice and what is “structural” about it. Social structure, in their view, is fundamentally about institutional design, how “the rules of the game in a society” (North Reference North1990, 3) set by institutions like the state channel human interactions in a certain direction. But there are plausible views in social and political theory that maintain that the idea of social structure encompasses much more than institutional rules. Indeed, for Young, that is precisely the point of conceptualizing structural injustice. Drawing on insights from social theorists such as Pierre Bourdieu (Reference Bourdieu and Nice1977) and Anthony Giddens (Reference Giddens1984), she insists that the concept identifies a phenomenon that is generally overlooked by theories of justice that analyze social structure by examining the state and other formal institutions. Taking a truly structural perspective on injustice requires “looking at the whole society” and the social processes driven by the actions of all agents within it, not just looking at a small institutional subset of that society that is assumed to be “more fundamental” (or more powerful) than the rest (Young Reference Young and Nussbaum2011, 70).Footnote ³ Examples of such processes include the normalization of a dominant racial or ethnic group’s cultural values and habits, the reproduction of the gendered division of labor in households, and, as Krause reminds us, the transmission of gendered bodily comportment to children. Undoubtedly, these processes are significantly influenced by what institutions do. But it would be a serious mistake to assume that these social processes can simply be placed under institutional control or that the individual behaviors driving these processes can be read off from the institutional framework of rules in which they occur, absent further facts about culture, norms, bodies, etc.

To my eye, this broader understanding of social structure is crucial for a nonredundant conceptualization of structural injustice. It is what makes the concept novel and interesting, theoretically and politically. The trouble with the Institutional View is that it gives up this important insight in its understandable quest to hold powerful institutions morally responsible for structural injustice. It claims, or must claim in response to the normative challenge, that certain institutions are the more powerful, fundamental part of society—“where the action is”—to maintain that structural injustice is the result of wrongful institutional conduct while ignoring the actions of ordinary individuals and exempting them from similar moral evaluation. The theoretical cost of this move is to make the concept of structural injustice redundant. Any so-called structural injustice could be fully redescribed in terms of institutional wrongdoing. At best, the concept of structural injustice would serve as a stepping stone for filling some overlooked gap in the theory of institutional justice, but once that theory is filled out, it would say everything the concept was meant to capture.

Together, these challenges pose a dilemma for the concept of structural injustice. The more that structural injustice is untethered from agential wrongdoing, the less it looks like a genuine injustice, rather than a tragedy that is morally indistinct from a natural disaster. Yet the more that structural injustice is grounded in institutional or some other wrongdoing, the less novel and interesting the concept turns out to be. On one horn lies the charge that structural injustice is normatively implausible; on the other lies the charge that it is theoretically redundant. Either Sandy’s situation is a product of her state or some other agent wronging her, but then it no longer needs to be called a “structural” injustice; or her situation is not a matter of anyone wronging her, but then the label of structural “injustice” looks unwarranted. Put this way, the theorist of structural injustice seems to face an impossible task: somehow, structural injustice must be both distinct from and yet nothing other than the wrongdoings of agents. But how could that be?

EMERGENCE, COMPLEXITY, AND SPONTANEOUS ORDER

This section begins developing an account of structural injustice that unravels this puzzle, building on Young’s important but underdeveloped thought that structural injustice is a product of complex social processes driven, in large part, by ordinary individuals and their actions. A major step toward this account is taken in Sally Haslanger’s (Reference Haslanger2022; Reference Haslanger2023) recent work in social ontology, which integrates ideas from social theory and philosophy into an original theory of society as a complex dynamic system structured by the social practices of individuals. One notable passage reads: “Societies are not planned and implemented by anyone. They are self-organizing systems. Although legislatures, economists, planning boards, and policy makers undertake to construct and intervene in social systems, and what they do matters, they are not in control. The society, so to speak, has a life of its own, and much of what happens is not designed or intended by anyone” (Haslanger Reference Haslanger2023, 9). This is a strikingly different picture of society from the Institutional View. Here, the state (and other powerful institutions) is but one agent among many in the system, and as comparatively powerful as it may be, it is not “in control” of the society it aims to manage. Why not?

Consider the issue of racial residential segregation, “the linchpin” (Massey Reference Massey2020) of racial inequality in the United States as a problem of structural injustice. It is now widely recognized that ending legally enforced or de jure segregation has not addressed de facto segregation: a recent study finds that most metropolitan areas were more racially segregated in 2019 than they were in 1990 (Menendian, Gambhir, and Gailes Reference Menendian, Gambhir and Gailes2021). Part of the explanation for this is a history of culpable discriminatory practices by political, financial, and real estate institutions, such as redlining, that have only recently been outlawed. But it would be a mistake to attribute the persistence of residential segregation to institutional actors alone. Residential segregation is also driven by social processes that the state does not control, most notably the choices of many individuals navigating the housing market. Thus, Elizabeth Anderson (Reference Anderson2010b, 189) closes her book-length argument for racial integration as an imperative of justice with the sobering observation that “the project of integration cannot be left to state initiatives alone. Most of the work of integration inevitably rests with the spontaneous actions of citizens in civil society.”Footnote ⁴ Notice again the gap between this picture and the Institutional View. Contrary to the Institutionalist View’s claim that powerful institutions like the state are “where the action is,” here state action cannot succeed unless ordinary individuals take the necessary actions and develop the necessary attitudes and habits for a truly, not just formally, integrated society (Anderson Reference Anderson2010b, 116).

It is tempting to conclude from these considerations that ordinary individuals must also be culpable for their part in structural injustice. Regarding residential segregation, it may be argued that responding to the racial diversification of one’s neighborhood by moving to white enclaves or “white flight” is a culpable behavior for which white individuals bear some responsibility for racial inequality (Kruse Reference Kruse2007). What ordinary individuals do within the remit of their everyday lives is socially consequential, and they should be held responsible for their actions.

But the issue is not so straightforward, and this is where the gap between structural injustice and agential wrongdoing begins to appear. Racial hostility is not required for segregated neighborhoods to result from individual residential choices. As Thomas Schelling (Reference Schelling1969; Reference Schelling1971) famously demonstrated, segregated neighborhoods can form even when agents navigate a housing market with a preference that at least roughly one out of three neighbors belong to their own racial group, a result that intensifies when the preference is raised to one-half.Footnote ⁵ In other words, individuals who are perfectly content to be the racial minority in their neighborhood by a ratio of 2-to-1 end up forming segregated neighborhoods. Racial residential segregation, in principle, need not involve culpable actions like redlining and white flight; surprisingly, it can be the product of many agents acting on what intuitively appear to be innocuous in-group preferences about their neighborhood. The housing market, Schelling (Reference Schelling1969, 488) concludes, is “a complex system with collective results that bear no close relation to individual intent.” And as Joseph Heath (Reference Heath2022, 299) observes, if collective results and individual intentions come apart, then “it becomes impossible to say anything about the relative levels of discrimination or tolerance merely from observing the spatial distribution of individuals of different races.” We cannot infer culpable attitudes or actions toward other racial groups from the occurrence of segregation alone.

Schelling’s model illustrates what in my view essentially distinguishes structural injustice from familiar forms of individual or institutional wrongdoing. In a complex system, the actions of many agents create emergent properties that are not simply reducible to the properties of the agents that create them. These properties are what license Haslanger’s claim that “society has, so to speak, a life of its own,” thus presenting a distinct object of study—and critique—from the individuals and institutions that comprise it.

The basic idea of emergence, often associated with the phrase “the whole is greater than the sum of its parts,” is an old one. Émile Durkheim ([1895] Reference Durkheim and Lukes2013) invoked it to justify social science as a distinctive field of study: society exhibits empirically investigable properties that are not reducible to properties of individuals even though society contains nothing but individuals. But emergence is not specific to sociology. Challenging the purported hierarchy in the natural sciences from (fundamental) physics to (more applied) biology, P. W. Anderson (Reference Anderson1972, 393) writes: “The behavior of large and complex aggregates of elementary particles, it turns out, is not to be understood in terms of a simple extrapolation of the properties of a few particles. Instead, at each level of complexity entirely new properties appear.” The concept of structural injustice is premised on a similar claim: the interactions of human agents in a complex society create emergent properties, properties that are not simply reducible to the properties of individuals themselves.

While familiar, the basic idea of emergence requires elaboration, for it can be given more and less plausible interpretations. Recent years have seen a proliferation of philosophical accounts of emergence that respond to longstanding worries about its metaphysical and scientific plausibility. For my purposes, it is important to distinguish between two varieties of emergence: ontological emergence and epistemic emergence.Footnote ⁶ Each offers a different gloss of the claim that an emergent property of a system is not simply reducible to the properties of its constituent parts.

In ontological emergence, an emergent property is metaphysically distinct from its constituent lower-level properties because there are reasons to hold that the property exists in its own right. Such reasons include causal novelty (the emergent property can cause some effect, and this ability is different from even the aggregated causal powers of the lower-level properties) or non-deducibility (the emergent property is not deducible from the lower-level properties even if one knew everything about them). So understood, emergence is heavily criticized for creating more mysteries than it solves: how can something be generated entirely by lower-level properties, and yet have novel causal powers or be non-deducible from those properties? (Kim Reference Kim1999). Some deny on these grounds that there are any ontologically emergent properties, pointing to successful reductionist explanations of properties that formerly were considered emergent in the history of science (Graham Reference Graham2025). Others, such as David Chalmers (Reference Chalmers, Clayton and Davies2008), argue that there is exactly one plausible instance of ontological emergence: consciousness, as a property not deducible even from complete knowledge of the physical properties of the brain. These criticisms apply to the issue of residential segregation. It seems implausible to say that residential segregation is an ontologically emergent property. Indeed, Schelling’s model shows precisely the deducibility of segregated neighborhoods from the preferences of agents making choices under certain conditions, though the relationship between the individuals and the emergent property is surprising and far from straightforward.

That brings us to epistemic emergence. Following Mark Bedau (Reference Bedau1997), a property is epistemically emergent insofar as it is deducible from its lower-level properties only by simulation: “set up the system, let it roll, and see what happens.”Footnote ⁷ Simulation is required because although the emergent property is fully determined by its lower-level properties, how those lower-level properties interact to produce the emergent property is too complex a process to be predictable from knowledge of the lower-level properties. In Schelling’s model, agents following the simple decision rule “Move into a house so long as at least roughly 1/3 of the neighbors are members of my racial group” end up forming segregated neighborhoods, a result that only becomes apparent to us by simulating its implementation. In this epistemic sense, this result is not simply reducible to the actions or intentions of the agents, even if it is their actions and intentions that produced it. The process by which residential segregation is generated by agents’ choices is not explicable by the linear cause-and-effect reasoning that typically suffices in everyday life.

My contention is that the kinds of outcomes that plausibly count as structural injustice are epistemically emergent. The concept of structural injustice invites us to look at society as a kind of complex system where multiple overlapping social processes, driven by many agents engaged in economic, cultural, linguistic, and other interactions, produce some vulnerability or deprivation as their emergent outcome. Some social processes, such as economic interactions in a market, are themselves a kind of complex system that produces emergent outcomes—a fact that puts considerable stress on domestically oriented theories of economic justice as well as conceptions of “society” as a closed, self-contained nation-state in an era of globalization. When they have harmful effects on human lives, such outcomes can be described as structural injustice. But in its broadest scope, structural injustice describes the emergent outcomes of multiple overlapping social processes in society: how culture, the economy, politics, housing, education, and so forth, all interact with one another to produce some injustice. Such outcomes include racial disparities in infant mortality (Wallace et al. Reference Wallace, Crear-Perry, Richardson, Tarver and Theall2017), economic inequalities perpetuated by people’s accents (Donnelly, Baratta, and Gamsu Reference Donnelly, Baratta and Gamsu2019), and colonial atrocities like sexual slavery that end up disproportionately impacting poor women (Lu Reference Lu2017, 128–38). These outcomes are, by nature, not linearly related to or straightforwardly predictable from the actions or intentions of the agents who bring them about. Indeed, the outcome may depart dramatically from what any agent intended or foresaw or could have foreseen.

So understood, structural injustice presents a distinct problem from familiar kinds of wrongful individual and institutional actions like exiting a neighborhood out of racial prejudice or depriving a racial minority of equal political status. Such wrongdoings are routinely part of the processes that generate structural injustice, but they will not fully explain the emergent outcome. Indeed, focusing on such wrongdoings would obscure the fact that the outcome can arise without them, as when individuals with integrationist preferences still produce residential segregation. This is true even if wrongdoing is stretched beyond the intentional infliction of harm to the negligent failure to avoid foreseeably causing harm to others, since social interactions may produce emergent outcomes that were unforeseeable to the agents who produced it. In a complex system, it is possible for social interactions to produce vulnerability or deprivation without any agents having acted wrongly. It is this possibility that is foregrounded in Young and Haslanger’s accounts of structural injustice.

Notably, Rawls also recognizes this possibility in his discussion of “background justice.” Social processes driven by the actions of ordinary individuals can have inimical results even within a just society: “the conditions necessary for background justice can be undermined, even though nobody acts unfairly or is aware of how the overall results of many separate exchanges affects the opportunities of others” (Rawls Reference Rawls1993, 266). The next line is crucial for understanding Rawls’ point: “There are no feasible rules that it is practicable to require economic agents to follow in their day-to-day transactions that can prevent these undesirable consequences.” In principle, he suggests here, there might be individual-level rules of economic interaction that would fully preserve background justice. But such rules are not relevant in practice due to the epistemically emergent nature of economic outcomes in a market economy: “the rules governing agreements and individual transactions cannot be too complex, or require too much information to be correctly applied” (Rawls Reference Rawls1993, 267). A society’s institutions may conform with Rawlsian principles of justice, but the justice they aim to uphold can erode over time from individual actions that are permitted in those institutions.

In short, structural injustice is, at heart, a problem of complexity. This conceptualization of structural injustice bears a strong resemblance to F.A. Hayek’s notion of “spontaneous order,” at least in its ontological and epistemic claims (Hayek Reference Hayek1945; Reference Hayek1967; Reference Hayek1973; Reference Hayek1976; Reference Hayek1979). Hayek himself would deny that emergent social outcomes count as an injustice, as I will discuss in the next section. While neither Young nor Haslanger reference Hayek’s work, their understanding of structural injustice similarly describes “the result of human actions but not the execution of any human design” (Ferguson Reference Ferguson1767, 187). Hayek emphasized the unintended benefits of a market economy as opposed to a planned economy, but complex social processes do not only exist in the market, nor do these processes only produce unintended benefits.Footnote ⁸ Playing on “spontaneous order,” structural injustice is an emergent disorder in a complex society.Footnote ⁹ Call this the Complexity View.

The Complexity View avoids the redundancy challenge: structural injustice does not simply collapse into agential wrongdoing, whether by individuals or institutions. The concept names a distinctive social phenomenon. Before turning to the normative challenge, let me address a potential worry about this view, namely that it is premised on a controversial social ontology involving mysterious “emergent” properties. To the extent that structural injustice can be conceptualized without it, one might argue, the better.

That might be true. But it is not a point in favor of the Institutional View. A crucial assumption of that view is that the state and multinational corporations are themselves (collective) agents distinct from the (individual) agents that make them up. Following Christian List and Philip Pettit’s (Reference List and Pettit2011) influential account, the Institutional View claims that collectives or groups are agents when and because they possess the constitutive capacities of agency due to their internal organization and procedures. But group agency itself, it turns out, is an emergent property in the epistemic sense. List and Pettit (Reference List and Pettit2011, 76) write, “While the agency achieved by a group supervenes on the contributions of its members—while it is not ontologically autonomous—it is autonomous in another, related sense. The agency of the group relates in such a complex way to the agency of individuals that we have little chance of tracking the dispositions of the group agent, and of interacting with it as an agent to contest or interrogate, persuade or coerce, if we conceptualize its doings at the individual level.” Although they frame the issues in terms of supervenience and autonomy, the substantive point is that group agency is an epistemically emergent property of a group of individuals organized in a certain way. The Institutional View, then, rests on the acceptance of one particularly striking case of epistemic emergence: group agency. The idea of emergent properties is thus no more mysterious than the idea of treating the state and other institutions as agents in their own right.

CULPABILITY, CONTRIBUTION, AND PASSIVE INJUSTICE

But can the Complexity View now defend the deontic wrongness of structural injustice? One might worry that this view has made the normative challenge unanswerable. If structural injustice is essentially distinct from agential wrongdoing, how could deontic wrongness ever reach it? The worry is heightened by the fact that several political theorists attuned to social complexity, including notably Hayek, maintain that emergent social outcomes are neither just nor unjust. For the reasons Estlund lays out, these moral concepts properly describe only how agents conduct themselves, not to the undesigned and unforeseen outcomes of those interactions. Is structural injustice, then, merely bad and not wrong?

I contend that this worry overlooks sources of deontic wrongness that apply to undesigned and unforeseen yet socially produced outcomes, as various critics of Hayek have argued (Friedman Reference Friedman2013; Levy Reference Levy2019; Lukes Reference Lukes1997; Lister Reference Lister2013; Shklar Reference Shklar1990). An important feature of emergence in the context of social structures is that, ultimately, it is all a product of human agency. As agents, our knowledge of the emergent properties of our actions and our ability to intervene in society in response to them are not static and immutable facts but things that can change and develop over time. The fact that society contains epistemically emergent properties does not make impossible for agents to investigate and intervene in such phenomena as they become known—though it does impose significant limits on how agents can and should engage in such social reform, as the next section will explain. History contains many examples of agents collectively learning to navigate the emergent problems created by their social practices, such as the multiple revisions that economic theory and policy has undergone in the last 100 years in response to various economic crises (Widmaier Reference Widmaier2016). Giddens (Reference Giddens1990, 38) calls this process of social learning reflexivity: “social practices are constantly examined and reformed in the light of incoming information about those very practices, thus constitutively altering their character.”Footnote ¹⁰ These reforms, he notes, often produce new emergent problems that we could not have foreseen in implementing these new practices, thereby generating new information and starting the cycle anew.Footnote ¹¹

Agency, understood not as the capacity to prospectively predict and avoid the emergent properties of human action but as the capacity to reflexively examine and respond to them as they arise, is the key to answering the normative challenge. Changes to what we know and can do about an emergent social problem also changes our moral relationship to that problem. In a complex society, we should expect such changes to be a frequent occurrence. As Seyla Benhabib (Reference Benhabib2004, 104) observes in the context of burgeoning problems of global justice, “we are constantly being challenged in understanding the moral implications of our actions by discovering unintended consequences…Such knowledge creates moral responsibility.” I argue that there are two plausible and compatible, though independent, explanations of how deontic wrongness is created this way. Hayek overlooks these sources of deontic wrongness. Young also seems to overlook them, leading her to draw an immutable and thus untenable distinction between structural injustice and agential wrongdoing.Footnote ¹²

One explanation is what I call the argument from contribution, which builds on some proposals by critics of Young’s distinction between structural injustice and individual wrongdoing (Atenasio Reference Atenasio2019; Barry and Ferracioli Reference Barry and Ferracioli2013; Goodin and Barry Reference Goodin and Barry2021; Reiman Reference Reiman2012; Sangiovanni Reference Sangiovanni2018). In their view, actions that contribute to producing structural injustice is a form of wrongdoing. This follows from the general principle that one ought not cause, or in collective action cases be part of the cause of, harm to others. There is a difference, however, between acting wrongly and acting culpably. Wrongful actions can be excused and not culpable when performed under certain conditions, including, notably, out of (reasonable) ignorance of their consequences. Those actions are no longer excused and become culpable, however, once this ignorance is dispelled (or is now itself culpable). Combining this with the earlier discussion of epistemic emergence and reflexivity gives us the following picture. Agents are not culpable for producing an emergent outcome, at first. They are excused by their ignorance. But the moral situation begins to change once the harmful outcome has been produced and people are suffering. Claiming ignorance of the harm produced by their social practices becomes increasingly an untenable excuse. The critical mistake in Young’s account is failing to distinguish culpability from wrongdoing: structural injustice that involves no culpable wrongdoing is possible in a complex society, but it is not a permanent condition for agents with the capacity to examine and re-evaluate their social practices in response to emergent problems.

That agents are engaged in wrongdoing when they unwittingly contribute to structural injustice may be disputed. Some philosophers maintain it is not wrongful to act in ways that cause harm if one could not have known about the consequences (see, e.g., Kiesewetter Reference Kiesewetter2016); if that is right, then actions that unwittingly contribute to structural injustice are not wrongful, undermining this explanation of its wrongness.Footnote ¹³ But surely there are instances in moral life where it is possible to do something wrong in a non-culpable way out of reasonable ignorance. To offer one prominent philosophical example (Rosen Reference Rosen2004): if I unwittingly serve my friend a cup of tea with a spoonful of arsenic instead of sugar because the box of “sugar” in my pantry in fact contains arsenic for reasons I could not possibly have known about, I have done something wrong to my (now dead) friend, even though I am not culpable for it. Structural injustice, I suggest, can happen in a similar way. Agents may be ignorant of the harmful emergent properties that result from their actions via complex social processes, and that ignorance excuses them for these actions that wrong others.Footnote ¹⁴

A different objection to the argument from contribution is that no individual’s contribution to structural injustice is causally significant. One person moving homes makes a tiny, perhaps even superfluous, contribution to the pattern of residential distribution that ultimately results from many people’s actions in the housing market. It may be argued that such contributions to a collectively produced harm are not wrongful, and further, that if no individual contributor to structural injustice is acting wrongly, then the outcome they collectively produce cannot be wrongful either. This objection raises an important question about collectively produced harms: if the actions of many individuals collectively produce a harm, yet no individual’s action considered on its own makes a difference to that harm, then what, if anything, makes those actions morally wrong? Wrongdoing typically refers to actions that have a direct and decisive causal connection to harm, but as Young (Reference Young and Nussbaum2011, 107–10; see also McKeown Reference McKeown2018, 494–6) points out, that kind of connection does not obtain in structural injustice.

But individual contributions may well be wrongful even when these actions, considered on their own, make no difference to the outcome. Julia Nefsky (Reference Nefsky2017; Reference Nefsky2019) argues that agents have a moral reason to refrain from actions that help to bring about a harm, such that (1) one’s action is or would be part of what causes the harm and (2) it remains an open possibility that the harm might not happen if enough agents refrained from such actions. Building on her account, Robert Goodin and Christian Barry (Reference Goodin and Barry2021, 348) argue that from a moral point of view we should “care about all those who helped bring about that [structural] injustice, even if none were causally decisive.” To borrow their toy example, if ten people dump toxic waste into a lake but only seven were enough to make it uninhabitable, it cannot be that none of the ten are responsible for polluting the lake because each can say their own contribution made no difference to the outcome. Surely all ten have done something wrong: each has helped to pollute the lake. The same can be said of structural injustice as a kind of collectively produced harm: none of the actions may be causally significant when viewed in isolation, but each has helped to produce an unjust situation. We do not have to say that each contribution is causally significant to explain what is wrongful about structural injustice.Footnote ¹⁵

The other explanation of structural injustice’s normative credentials is what I call the argument from passive injustice, building on Judith Shklar’s (Reference Shklar1990) influential discussion. Shklar contends that it is a mistake to think that whether something is an injustice depends only on its causal story, such that harmful outcomes that did not originate in wrongful conduct or “active injustice,” including natural disasters, must therefore be mere misfortunes. Equally important is the question of whether agents organized in a political community have the capacity to prevent or mitigate that harmful outcome. Where the capacity exists, it is “passive injustice” for the community to simply let the outcome happen, to be indifferent toward the vulnerable.Footnote ¹⁶ Passive injustice applies to natural and social outcomes alike. Indeed, the fact that humans have repeatedly developed new ways of altering their natural environment should lead us to question the very distinction between what is natural and what is social, and with it, what is inevitable and what is contingent. As she observes, “[s]ome misfortunes of the past…are now injustices, such as infant mortality and famines” (Shklar Reference Shklar1990, 5). Advancements in technology and infrastructure, for example, in medical care and agricultural production, move these problems from the realm of natural disaster to the realm of social injustice. The point extends to Estlund’s analogy to hurricanes: perhaps one day climate science and disaster mitigation will advance to the point where the occurrence of any hurricane damage to human lives is seen entirely as a social failure, not a natural disaster.

Shklar’s notion of passive injustice leads to a different explanation of structural injustice’s normative credentials. In a complex society, the emergent outcome of agents’ interactions can be harm, vulnerability, and deprivation even though those interactions did not involve wrongdoing or active injustice. But once this outcome is produced and agents become aware of these emergent properties, they have the capacity to respond, or fail to respond, in ways that turn this emergent harm into an injustice. As agents, they need not passively resign themselves to living with these problems forever; they can at least try to prevent these harms from reoccurring and mitigate their effects on people’s lives. On this argument, structural injustice could, in principle, arise without any wrongdoing, but structural injustice does not simply go on without wrongdoing given agents’ capacity to evaluate and reform their society in response to emergent problems.

This capacity will often be exercised, at least in part, through institutions; in this respect, the Institutional View is right to insist that institutions can act in response to Sandy’s situation.Footnote ¹⁷ But the Institutional View conflates the fact that institutions like the state play an important role in remedying structural injustice with the claim that structural injustice is essentially about what institutions—not ordinary individuals—do. That claim is false: the defining feature of structural injustice is its origin in complex social processes where the actions of many agents, including ordinary individuals, create emergent properties that are harmful. In other words, the Institutional View confuses (a part of) the solution with the nature of the problem itself.

These two arguments, one pointing to contribution and the other pointing to passive injustice, answer the normative challenge. To support my case, let me address two further objections raised by Estlund that bear on these arguments. Addressing them illuminates some important differences between the two explanations and their implications.

The first objection is that duties to prevent bad outcomes cannot ground claims of injustice. Estlund (Reference Estlund2024, 336) writes, “[T]here might yet be grievance toward people who violate their duties of superintendence: requirements to help prevent or remedy the conditions in question. Still, the conditions themselves, while allegedly unjust, would not themselves warrant any such grievance. We wouldn’t yet know how it is an injustice that needs remedying or preventing.” Even merely bad outcomes can generate duties to prevent or remedy them, and so appealing to those duties cannot explain why an outcome is unjust. Failing to aid the victims of an unpreventable hurricane may be wrongful, but that does not make the hurricane itself an injustice. This comes out, Estlund (Reference Estlund2024, 343) observes, in the reactions of victims: “They will properly react [to structural injustice], in action as well as attitude, very differently from their reaction to being unpreventably harmed by a hurricane.”

The argument from contribution has a reply: the difference between structural injustice and a hurricane is that one is a socially produced outcome of human action and the other is not.Footnote ¹⁸ Where there is structural injustice, there are agents who have failed their duties to not harm others, not merely those who have failed duties to remedy or prevent harm, however it was brought about. This failure may have been excused in the past due to agents’ ignorance of the emergent properties of their actions. But that excuse is now unavailable, and they are warranted targets of grievance by victims of structural injustice. By contrast, a hurricane itself is not, nor could ever be, a warranted target of grievance since it was not produced by (even excused) wrongdoing.

The argument from passive injustice has a different response: the difference between structural injustice and a hurricane is that one is a socially preventable outcome and the other is not, or at least not entirely preventable given the current state of our ability to manage or even reliably forecast atmospheric phenomena. Where there is structural injustice, there are agents guilty of passive injustice, of tacitly condoning a socially preventable harm. This failure may have been excused in the past due to agents’ ignorance or inability. But that excuse is now unavailable, and they are warranted targets of grievance by victims of structural injustice. By contrast, a hurricane itself is not a warranted target of such grievance since it is not socially preventable.

Of course, Estlund’s worry is precisely that the mere failure to prevent something bad does not transform that thing into an injustice. But that is also precisely Shklar’s point: there are situations where (passively) failing to prevent something bad is morally tantamount to (actively) bringing it about. Estlund (Reference Estlund2024, 349) himself suggests that some of the deontic wrongness of structural injustice might lie in the distinction between acts versus omissions, specifically the category of culpable omissions. Passive injustice is a kind of wrongful omission. This claim can be strengthened by filling in the fuzzy idea of a “bad or harmful outcome” with a more specific account of the kinds of outcomes that agents arguably have an obligation to socially prevent when it is possible to do so. Not all socially preventable harms constitute an injustice, or as Jonathan Wolff (Reference Wolff, Browne and McKeown2024) puts it, there may be structural harms that are not structural injustices. Individuals may have claims to the prevention of hunger and homelessness but not headaches. This has a broader and unsurprising implication, namely that the Complexity View needs an accompanying normative theory to explain what kinds of social outcomes should count as unjust, what sorts of socially produced harms we ought to try to prevent.Footnote ¹⁹ What the Complexity View tells us is that those outcomes, whatever they are, count as structural injustices insofar as they are the emergent outcomes of complex processes driven by human actions.

Estlund’s (Reference Estlund2024, 349–50) second objection is that these arguments, while valid, still do not establish warranted grievance attitudes for structural injustice in all relevant cases since agents may not be culpably failing their duties. Everyone may already be doing what is reasonable to demand of them vis-à-vis their duties to avoid contributing to harmful social processes and to prevent or mitigate certain bad outcomes. People cannot reasonably be expected to do more to prevent injustice or to refrain from contributing, but structural injustice continues. In this vein, Young (Reference Young and Nussbaum2011, 131–3) suggests that agents typically face structural constraints that exculpate their contribution to structural injustices. In those cases, Estlund worries, since no agent is culpably failing their moral duties, there can be no warranted grievance about structural injustice. In this respect, structural injustice looks more akin to a natural disaster than a deontic wrong.

The argument from contribution can address this worry as follows. It is possible to have fully excused wrongdoing: an agent can do something wrong but in a completely blameless or non-culpable way, whether due to ignorance or constrained choices. It is true that in such cases, grievance attitudes like anger or resentment would be unwarranted toward structural injustice.Footnote ²⁰ However, the fact that some wrongdoing is fully excused does not diminish its wrongness. Killing your friend with poisoned tea is a wrong, but you may not be culpable for it. The same goes, it may be argued, for these cases of structural injustice: no one may be to blame for these outcomes in a world where no one is culpably failing their duties, but the outcomes are nonetheless unjust.

This reply is not available for the argument from passive injustice. That argument implies that if agents have done enough to avoid passive injustice, the remaining harm is merely bad, not wrong. Shklar (Reference Shklar1990, 70) remarks about the Great Famine in Ireland, “Had the victim’s voice been treated as the privileged and primary one and had the most activist of the policies that were reasonably proposed been followed, there would still have been plenty of hunger, but no injustice.” But I would argue that this is not an implausible verdict.Footnote ²¹ Where agents really are doing what they should to prevent or mitigate structural injustices in their society, then the remaining suffering and vulnerability and inequality would be bad, but not wrong—for the time being. As Anderson (Reference Anderson2010a, 5) writes, “Where all agents conduct themselves justly—where they successfully comply with all reasonable demands—the state of affairs resulting from their conduct is just.” But this is hardly a fatal concession for the concept of structural injustice. Such circumstances, Shklar rightly observes, are vanishingly rare in human affairs. Passive injustice is the rule, not the exception, for how agents treat emergent forms of vulnerability and deprivation in their society.Footnote ²² That is enough to explain the grievances regarding many structural injustices today.

SOCIAL REFORM IN A COMPLEX SOCIETY

Recent theoretical challenges, I have argued, pose a dilemma for the concept of structural injustice: either it is distinct from agential wrongdoing but then lacks the normative credentials of injustice, or it is primarily a matter of institutional wrongdoing but then is theoretically redundant. The Complexity View offers a way out. Structural injustice is distinct from (culpable) wrongdoing because predicting the emergent outcomes of human actions is too epistemically difficult, such that structural injustice is not simply a product of (culpable) wrongdoing even though it is ultimately a product of human action. However, structural injustice as emergent disorder is unjust, not merely bad, because we are reflexive agents with the capacity to examine and reform our social practices in light of new information about their effects. Two forms of wrongdoing explain the wrongness of structural injustice: contributing to social processes that harm others, and failing to prevent or mitigate bad outcomes that are socially preventable or mitigable. So understood, structural injustice is a normatively plausible and theoretically valuable concept, one that identifies a distinctive social phenomenon for criticism and redress.

One important but easily overlooked implication of this view, however, is that the pursuit of social reform itself faces epistemic limitations in a complex society.Footnote ²³ I will conclude by dwelling on this implication. Goodin and Barry (Reference Goodin and Barry2021, 343) point out that the very same complexity that distinguishes structural injustice from agential wrongdoing may also threaten the justifiability of holding agents responsible for failing to remedy structural injustice. Structural injustice arises as an unintended consequence of what agents do, but the issue of unintended consequences does not magically disappear when agents introduce social reforms to try to address them. Well-meaning, intuitively attractive social reforms can generate bad unintended consequences, including exacerbating the very problem they were meant to address. Recall Sandy’s housing situation: recent empirical evidence suggests that rent control as a policy response to the housing crisis in cities can worsen the crisis in the long run by lowering the rental supply despite helping incumbent renters in the short run (Diamond, McQuade, and Qian Reference Diamond, McQuade and Qian2019; see also Sankaran Reference Sankaran2025). Here, we again see a connection to Hayek’s theory of spontaneous order and, quoting Daniel Luban’s (Reference Luban2020, 77) recent commentary, its “genuine insights about history, above all concerning the limits of rationalism and the importance of unintended consequences.”Footnote ²⁴ Faced with a complex society, how can agents be confident that their proposed social reform will effectively remedy a structural injustice, rather than make it worse or create other problems?

Without a compelling answer to this question, there is a lingering worry that the Complexity View is still trapped in the dilemma. On the one hand, if agents do not and cannot know how to effectively remedy structural injustice under conditions of complexity, then it would be unjustified to hold them culpable for failing to take remedial action. But then structural injustice would lack the normative credentials of injustice in this respect.Footnote ²⁵ On the other hand, if we simply say there is no real epistemic difficulty in predicting what some reform will actually do in a complex society, then this casts doubt on the claim that there is such a difficulty for the ex ante prediction of structural injustice in the first place. Surely agents who can do one can also do the other. But then structural injustice would be redundant after all. It would amount, as Martha Nussbaum suggests (Young Reference Young and Nussbaum2011, xxi), to a species of negligence: the wrong of failing to guard against the foreseeable harmful consequences of one’s actions, whether the foreseer is individuals when interacting with each other or institutions who could antecedently design the “rules of the game” to prevent these undesirable outcomes.

To fully escape the dilemma, the Complexity View thus needs an accompanying account of social reform that has two features. First, this account must accept that agents face serious epistemic barriers in predicting the outcomes of their actions in a complex society, including actions undertaken in the pursuit of social reform. Second, it must show that agents can nonetheless overcome these barriers to effectively address emergent injustices in their society. Social reform, this account would explain, does not require powers of prediction that are made unavailable by complexity; agents have other means for figuring out how to remedy social problems.

We can find such an account, or at least major steps toward it, in recent pragmatist and experimentalist approaches to social reform (Anderson Reference Anderson2006; Barrett Reference Barrett2020; Barrett and Buchanan Reference Barrett, Buchanan, Sobel and Wall2023; Jaeggi Reference Jaeggi2018; Knight and Johnson Reference Knight and Johnson2011; Thrasher and Vallier Reference Thrasher and Vallier2018; Wiens Reference Wiens2012). They contain, I suggest, the final piece of our puzzle. Their key contention and insight is that deliberate and effective social reform is possible even in a complex society so long as agents can imagine and try out alternative social arrangements in response to newfound social problems, learn from their experiences of living with those arrangements, and collectively decide whether the results are sufficiently positive to make those arrangements worth keeping and emulating. Social reform, they argue, should be understood as a fallibilistic process of experimenting with different social arrangements and evaluating their results ex post—precisely because of the epistemic difficulties of ex ante prediction.Footnote ²⁶ As Jacob Barrett (Reference Barrett2020, 119) puts it, “We are better at evaluating institutional changes after they are implemented than we are at predicting what they will do, better at evaluating where we have ended up than predicting where we will go.” By exercising their capacity for experimentation and evaluation, agents can find effective social reforms to emergent problems in a complex society.

To be sure, this approach to social reform does not guarantee certainty about what to do to remedy structural injustices here and now. Agents may disagree about what reforms to try, what lessons to take from past and present experiments, and so on. Recent efforts to relax zoning laws in response to the housing crisis to encourage new construction or “upzoning” have found more success in certain cities than others, and many are unconvinced that such “supply-side” interventions offer a sustainable and transposable solution (Ellickson Reference Ellickson2021; Freemark Reference Freemark2020; Greenaway-McGrevy and Philips Reference Greenaway-McGrevy and Phillips2023; Murray and Limb Reference Murray and Limb2023; Wetzstein Reference Wetzstein2021). But this kind of uncertainty is to be expected and far from the kind of ignorance and helplessness that would undermine political demands for social reform and, with it, moral complaints of passive injustice. Passive injustice, as Shklar’s remark on the Great Famine suggests, consists in the wrongful failure to pursue the most serious reforms among those reasonably proposed in response to a problem and the failure to heed the victim’s experience of the problem—not the failure to immediately come up with the perfect solution. These wrongs certainly carry over to the context of structural injustice. Agents routinely ignore or dismiss others’ negative experiences of structural injustices within current social arrangements, are unwilling to experiment with alternative arrangements for reasons like perceived self-interest or apathy, and so on. Passive injustice encompasses all of these sorts of moral failures.

Some will object that this approach does not really provide a solution to the problem of unintended consequences so much as simply accept the problem as part of the “process” of social reform. Elizabeth Anderson (Reference Anderson2006, 13) writes in her explication of pragmatism: “Unfavorable results—failures to solve the problem for which the policy was adopted, or solving the problem but at the cost of generating worse problems—should be treated in a scientific spirit as disconfirmations of our policies.” But if we are not adept at predicting whether a reform will generate “worse problems,” some will argue, then perhaps it is better not to take the leap at all. Perhaps passivity in the face of structural injustice is not morally wrong but a reasonable response to the potential consequences of intervening in a complex society under deep uncertainty.

But this conservative position is not a better answer to the problem of unintended consequences.Footnote ²⁷ For one, there is no reason to think that the problem of unintended consequences will go away if agents refrain from trying to reform their complex society. Social reform is hardly the only sphere of human activity that generates unintended consequences, as issues like climate change serve to remind us. Nor do we have any reason to expect that reforms will always or even generally lead to worse problems. Such pessimistic expectations are unsupported, as Luban (Reference Luban2020, 78) observes, by “the raw unpredictability implied by the idea of unintended consequences on its own.” Given these facts, the possibility of bad unintended consequences is an acceptable risk for the sake of improving the situation when confronted with real and pressing problems.

My point is not that the problem of unintended consequences has no bearing on the pursuit of social reform. Rather, it is that the right response to the problem is not avoiding social reform altogether but managing its risks, for instance, by conducting experiments that limit the number of people potentially affected and that can be halted or reversed (Barrett and Buchanan Reference Barrett, Buchanan, Sobel and Wall2023, 137; Perrow Reference Perrow1999, 304–52; Scott Reference Scott1998, 345). This may sound like an overly conservative position of its own. But political thinkers of many different stripes have taken the problem of unintended consequences seriously in thinking about how to successfully reform a complex society, even—or especially—when the aim of reform in their view is ultimately to create radically different social arrangements.Footnote ²⁸ To simply ignore or dismiss the problem would be naïve.

Admittedly, these implications of the Complexity View may come across as a disappointment, especially considering the structural injustices of contemporary society. But if the preceding arguments are correct, some disappointment on this front may be unavoidable if we are to have a compelling explanation of what is distinctive and plausible about the concept of structural injustice. The very issue that distinguishes structural injustice from other wrongs, namely the complexity of contemporary society and the emergent properties of human actions within it, also sets some important, though surmountable, limits on how agents can remedy it. As is often the case, the definition of the problem delimits the contours of the solution. Filling in these contours, of course, will require further, more fine-grained theoretical and empirical work: investigating specific structural injustices, uncovering their origins in complex webs of human action, and discovering latent possibilities for social reform. Still, by probing the concept of structural injustice itself and asking how it may be defended from recent theoretical challenges, we can come to a clearer view of this prominent concern in contemporary politics.