Exploitation and Remedial Duties

The concept of exploitation and potentially exploitative real-world practices are the subject of increasing philosophical attention. However, while philosophers have extensively debated what exploitation is and what makes it wrong, they have said surprisingly little about what might be required to remediate it. By asking how the consequences of exploitation should be addressed, this article seeks to contribute to ﬁ lling this gap. We raise two questions. First, what are the victims of exploitation owed by way of remediation? Second, who ought to remediate? Our answers to these questions are connected by the idea that exploitation cannot be fully remediated by redistributing the exploiter ’ s gain in order to repair or compensate the victim ’ s loss. This is because exploitation causes not only distributive but also relational harm. Therefore, redistributive measures are necessary but not suf ﬁ cient for adequate remediation. Moreover, this relational focus highlights the fact that exploitative real-world practices commonly involve agents other than the exploiter who stand to bene ﬁ t from the exploitation. Inso-far as these third parties are implicated in the distributive and relational harms caused by exploitation, there is, we argue, good reason to assign part of the burden of remediation to them.


Introduction
During the last few decades, exploitation has emerged as a central topic in debates among applied ethicists about commercial surrogacy, paid organ donation, clinical trials recruiting underprivileged groups, the use of 'sweatshop' labour, and other controversial practices. 1 Paralleling and partly overlapping with these debates, moral and political philosophers have been paying increasing attention to exploitation as an ethical concept, discussing what exploitation is, how it differs from other forms of wrongdoing, and what makes it wrongif indeed it is wrong at all. 2 This article seeks to contribute to the philosophical understanding of exploitation in ways that also make it more relevant for analysing real-world practices.
Specifically, we will explore an issue that has been largely overlooked so far: namely, the duty to remediate or rectify exploitation. Whereas questions concerning the prevention and abolition of exploitative transactions and relationships have received some philosophical attention, 3 philosophers have had surprisingly little to say about what should be done to set things straight once exploitation has occurred. Our aim is to contribute to filling this gap.
There are two main questions in this area. The first concerns the content of remedial duties. After exploitation has happened, what should be done to address its consequences? What is owed to those who were exploited? We shall suggest that the answer to this question is not straightforward because exploitation differs from other wrongs in ways that make standard approaches to remediation untenable.
The second question concerns the allocation of remedial duties. Who should bear the burden of providing remediation to those who were exploited? Whose task is it to set things straight? The answer to this question is not simple either because exploitative practices in the real world commonly involve individuals and groups other than the exploiter who stand to benefit from the exploitation.
In the course of this article, we will propose replies to both questions, one by one. After a brief discussion of the wrongness of exploitation in Section 2, Section 3 will reply to the first question, Sections 4 and 5 to the second. The basic idea connecting these replies is that remediation of exploitation cannot plausibly be seen as solely a matter of redistributing the exploiter's gain in order to repair or compensate the victim's loss. While recognising the necessity of such redistribution, we argue that it is not sufficient. This is because the victims of exploitation suffer not only distributive loss but also what we refer to as relational harm, a harm which requires distinct forms of repair. Moreover, if other parties collaborate with the exploiterwhich, we argue, is often the casethen they too are implicated in these harms, and there is good reason for assigning part of the burden of remediation to them.

Wrongful Exploitation
To set the stage, consider a fairly typical putatively exploitative arrangement: Sweatshop: A poor rural area in a developing country is struck by a series of droughts which destroy the crops on which large parts of the population depends for their livelihood. Lacking training for other jobs and resources needed to migrate many agricultural workers are left unemployed. As they cannot expect support from the government or the global community, the risk of starvation looms large. An apparel manufacturer sets up a factory in the area. The wages on offer just about suffice to secure the bare necessities of workers and their families, whereas the work is hard, the working days long, and the conditions unsafe and uncomfortable. Local workers are nonetheless eager to be hired. They are fully aware of the wages and working conditions but prefer working in the factory to unemploymentthe only other option available to them. The factory generates significant profits due to the low production costs.
While cases of this sort are widely thought to raise moral concern, the reasons for that judgment are not immediately evident. To begin with, the workers appear to benefit from employment the factory. It is hard to fault their judgment that an opportunity to provide for themselves and their families, even if just barely and on the conditions on offer, constitutes an improvement in their dire circumstances. Moreover, it does not seem that the workers fail to consent to the employment. They fully understand what working in the factory entails and are not coerced, pressured, or deceived by their employer.
However, even though the workers benefit and consent, many would hold that they are nonetheless treated wrongly. More precisely, many would claim that they are wrongfully exploited. 4 This claim has significant intuitive appeal given the following features of the case. The apparel manufacturer takes advantage of the desperate situation of the workers in order to strike a highly profitable deal. They could have provided higher wages and better conditions while still profiting from the arrangement. But in order to boost their gain, they offer the workers the smallest level of benefit they will accept in exchange working in the factory.
A number of philosophical accounts provide theoretical support for the view that transactions can be wrongfully exploitative despite mutual consent and mutual benefit. Consistent with their general approach, advocates of what we may call the liberal account might argue that the workers in Sweatshop are exploited because they receive an unfairly small benefit compared to the burdens they bear and compared to their employer's gain. 5 Proponents of a broadly Kantian approach might argue that this is a case of wrongful exploitation because the apparel manufacturer fails to treat the workers as persons by using their vulnerability as an opportunity to advance its own ends 6 or by failing to show proper regard for their basic needs. 7 Finally, proponents of a Marxist approach might locate the wrongful exploitation in the fact that the company violates the workers right to self-ownership 8 or engages in 'domination for self-enrichment', i.e. uses its power over the workers to subordinate and extract gain from them. 9 Whichever account one prefers, one can plausibly claim that the workers in Sweatshop are wrongfully exploitedeven though they benefit and consent.
In this article, we will largely steer clear of the debate about what precisely makes exploitation wrong. It suffices for our purposes to note that most theories of exploitation subscribe to a bipartite analysis, according to which two conditions are necessary and jointly sufficient for wrongful exploitation. 10 The first conditionthe 'Vulnerability Clause' 11is that the exploiter and exploitee are strongly asymmetrically related. The exploitee is vulnerable, needy, or in a weak bargaining position with no access to reasonable 12 or non-prohibitively burdensome 13 alternatives to transacting with the exploiter. The second conditionthe 'Advantage Clause' 14is that the exploiter uses this very asymmetry to extract a gain which is excessive or otherwise inappropriate. 15 In Sweatshop, for example, the manufacturer can secure a high profit margin only because the workers have no decent alternatives to working in the factory. Different theories of the wrongness of exploitation largely agree about this much. 16 Their differences concern what precisely makes the extraction of gain from the vulnerable morally objectionable.
We will argue that the harm involved in exploitation should be understood by reference to these two conditions and their specific combination. It is because the exploiter not only gains excessively but gains excessively taking advantage of another's vulnerability that she inflicts both a distributional loss and a relational harm on him, two harms which both call for remedial action. Our account as such is not committed to any specific theory of the wrongness of exploitation, but we will occasionally highlight points where its details or practical implications may depend on one's theoretical commitments.

Compensation and Remediation
Suppose we have established that a person or group, such as the workers in Sweatshop, has been wrongfully exploited. Then what? Since the charge of exploitation is morally serious, we presumably expect it to have some practical bite. In particular, it is plausible to think that the victims are owed some kind of remediation. But what does it mean to adequately remediate the exploitation?
At first sight, the answer might seem simple. Exploitation by definition involves the extraction of benefit which means that exploiters always stand to gain from the transaction. Correspondingly, there is a sense in which the exploitee always loses, and the loss is incurred specifically because of the exploitative exchange. Intuitively, it seems natural to assume that remediating wrongful exploitation is a matter of undoing this distributive pattern, i.e. of using the exploiter's gain to repair or compensate for the exploitee's loss.
Although such redistribution is surely part of remediating exploitation, we will highlight two difficulties with an exclusively redistributive approach. First, while it is true that victims of exploitation suffer a loss requiring compensation, redistribution itself is complicated by the fact that the nature of this loss is different than in other cases of wrong-doing. This difference has crucial implications for what counts as adequate compensation. Second, in addition to this loss, victims of exploitation suffer relational harm by being involved in an unjust relationship. This latter harm cannot be remediated by simply redistributing the gains resulting from the exploitative transaction.

Distributive Loss and Compensation
To start with the first issue, in many cases repairing wrongful harm is paradigmatically a question of restoring the status quo ante, i.e. returning the victim to the state he was in before the harm was inflicted. Consider theft: if I have stolen your bike, my primary duty is to give it back. If I have lost or sold the bike, I owe you another one or its monetary equivalent. Either way, you were harmed relative to a pre-interaction baseline, and remediation is determined with reference to that baseline.
This baseline, however, cannot be adequate for all cases of wrongful harm. If an employer refuses to hire the most qualified candidate for a job because she is a woman, it has wrongfully harmed her and owes her recompense. But she has not been made worse off than before the discrimination occurred when she also did not have the job. Rather, she is worse off than she would have been had she not been discriminated against and would have been given the job. Here the harm and the compensation are determined relative to a counterfactual baseline. 17 Neither baseline works, however, in the case of compensating the victims of exploitation. This is because these victims, when exploitation is mutually beneficial, are not harmed in either of the two senses just considered. 18 The exploitation has made them better, not worse off than before and better, not worse off than they would otherwise have been. In Sweatshop, for instance, employment in the factory represents the workers' only chance of providing for themselves and their families. They did not have that chance before being hired and would not have the chance had they not been hired. It would be counterintuitive to propose remediating the exploitation by realising the workers' pre-employment situation or the situation that would now prevail had they not been employed. Either way they would be deprived of what they see as their best available option.
Yet there is a sense in which even mutually beneficial exploitation does inflict a harm or loss on the victim. In the bipartite analysis of exploitation, the Advantage Clause entails that by extracting something to which they are not entitled under the circumstances, exploiters deny their victims something to which the latter are entitled. The exploiter gains excessively (or inappropriately), and while the victims are better off than before the transaction and better off than they would otherwise be, they are worse off than they ought to be. In other words, they are harmed relative to a normative baseline, understood as a counterfactual state of affairs in which the parties have transacted but neither has gained more or less than they are entitled to. This is the sense in which the workers are harmed in Sweatshop: they are denied a set of benefitsa higher wage and less burdensome working conditionsthat they would enjoy if the manufacturer had not secured an excessive gain from the exchange, a gain to which the manufacturer is not entitled. 19 If this analysis is correct and assuming (plausibly) that remediation in general is at least partially a matter of repairing wrongful harm, remediating exploitation requires realising that counterfactual state of affairs. To visualise: in Figure 1, the bottom solid line represents the exploitee's pre-transaction level of welfare (or some other relevant good). The dashed line in the middle represents the exploitee's welfare level in the actual, exploitative transaction. The dotted line at the top represents the exploitee's welfare level in the counterfactual, non-exploitative transaction. The exploitee's loss consists in the shortfall between the dotted line and the dashed line. Remediation requires eliminating this shortfall.
As an illustration, consider Antidote 20 : Hiker B is bitten by a venomous snake and is about to die. Fortunately, hiker A happens to be nearby and has the antidote. However, A charges $10,000 for the antidote whereas the ordinary retail price is $10. B accepts the offer and survives. While B derives a significant benefit (his life), he is nonetheless exploited. Valdman rightly suggests that a necessary part of the explanation as to why B is exploited is that he is charged too much. It follows that there is a sum (possibly $10) such that if A had charged this sum he would not have exploited B. If this is correct, then, on our account remediation requires A to return $10,000 minus this sum to B (i.e. to return his excessive gain, the gain that exceeds that to which he is entitled). Analogously, if one holds that the workers in Sweatshop were wrongfully exploited because they were denied a set of benefits that their employer would have been able to provide had the employer not profited excessively from hiring them, then one should also advocate remediating the workers by providing them with these benefits.
Notice that this account does not presuppose any particular theorye.g. Kantian, Marxist, or liberalof what makes exploitation wrong. It is compatible with all theories that hold that mutually beneficial transactions can be wrongfully exploitative. Any such theory will recognise a distinction between exploitative and non-exploitative levels of benefit (e.g. in Sweatshop, between receiving subsistence wage in exchange for hard but much-needed work, on the one hand, and receiving a higher wage for less onerous labour, on the other). That said, it is possible that some such theory is needed for determining the precise location of the baseline for remediation. We might set the baseline in one place if (on a liberal theory) we think compensation is a matter of bringing about a fair distribution of benefits and burdens between the parties, in another place if (on a Kantian theory) we are concerned with satisfying the need on the part of victim that the exploiter wrongfully neglected, and so on. But it is also possible that different theories would converge on the location of the baseline, though for different reasonsor, perhaps, that its exact location is indeterminate on some or all theories.

The Relational Injustice of Exploitation
This brings us to the second difficulty. While compensation of the sort discussed so far is necessary for adequately remediating exploitation, it cannot be sufficient. This is because exploitation, even when it is mutually beneficial and consensual, causes not just distributional loss, but also what we will refer to as relational harm. Therefore, exploitation cannot be fully remediated without redressing this distinct harm.
Our argument is based again on the bipartite analysis of exploitation. Recall that the Vulnerability Clause entails not just that the exploitee is vulnerable and the exploiter is not, but that the exploitee is vulnerable to the exploiter because the former has no reasonable or non-prohibitively burdensome alternative to transacting with the latter. In other words, there is a background asymmetry between the parties that makes the exploitative extraction of gain possible.
Crucially, this background asymmetry can in itself be innocuous in many cases. For instance, in Antidote there is nothing morally objectionable per se about the fact that hiker B happens to desperately need a medication which hiker A happens to possess. What turns this into a case of wrong-doing is the fact that A takes advantage of B's desperate need to extract an exorbitant price. Cases of this sort suggest that it is not necessary for wrongful exploitation to occur that the exploitee is the victim of unjust background conditions. 21 In fact, we think a stronger claim is warranted: wrongful exploitation can occur not only in cases where the background asymmetry is nobody's fault, but also in cases where the exploitee himself is to blame for it. Suppose in Antidote that the area is notoriously snake infested and that B, despite numerous warnings and despite the antidote being widely and cheaply available, had nonetheless set out on the hike without it. Unsurprisingly, B is bitten and needs the antidote, which (again) A proposes to supply for $10,000, a proposal which (again) B accepts. It seems clear that A wrongfully exploits B in this iteration of the case too, even though the vulnerability that A takes advantage of arises from B's own reckless conduct. 22 Granted, in many real-world cases the background asymmetry which makes exploitation possible is morally suspect or straightforwardly unjust. For instance, the exploitation of indentured labourers is made possible by the labourers' morally objectionable dependence on their exploiters. Similarly, it might be argued (not implausibly, we think) that the unemployment and poverty that prompt the workers to seek employment in the factory in Sweatshop are serious injustices. 23 In any case, and this is the crucial point, whether the background asymmetry is in itself morally objectionable or not, the exploitative transaction creates a relationship which is objectionably asymmetrical or unequal, thereby harming the exploitee. By deliberately taking advantage of the background asymmetry (as posited by the Vulnerability Clause) to extract an excessive or otherwise inappropriate gain from the exploitee (as posited by the Advantage Clause), the exploiter positively affirms this asymmetry. She does not just accept or neglect the exploitee's vulnerabilitywhich, depending on the case, it may be permissible for her to dobut treats it as a lever for profit-maximisation. For notice that the terms of the transaction are entirely dictated by her. The exploitee, on the other hand, though strictly speaking free to refuse the transaction, is denied any opportunity to influence its terms and is thus pressed to give up something to which he is entitled. Such a stance towards other people's vulnerability is, we submit, incompatible with granting them a minimally acceptable (let alone equal) standing in one's relationship to them.
In other words, such an arrangement is defective as judged by relational ideals that specify the shape that any just or adequate social relationship must take in regard to the relative rank, status, and power of the parties and the acceptable level of solidarity between them. 24 While these ideals have been most extensively elaborated and defended in the literature on relational justice, 25 they have a broader appeal since proponents of other theories of justice also have reason to appreciate the significance of relational goods. 26 We cannot here fully spell out or defend any particular conception these ideals. However, we will show how important themes in the relational justice literature help clarify the nature and significance of the relational harm involved in exploitation.
First, it is commonly recognised that the demands of relational justice are not independent of distributive concerns. The distribution of (non-relational) goods may impact negatively on people's standing, and significant distributive inequalities may even be incompatible with relational ideals. 27 This observation helps explain how the relational harm involved in exploitation is connected to the distributive loss explored above. The exploiter not only arrogates exclusive power over the exploitee, thus failing to grant her adequate standing. The point of doing this is to appropriate something to which the exploitee is entitled and the exploiter is not. Thus, the transaction distributes goods in a way that inflicts a loss on the exploitee (relative to a normative baseline) and that reflects the exploiter's failure to treat the exploitee as a person of equal (or sufficient) standing. More precisely, the relational harm is caused not by the distribution per se nor by the exploiter's taking advantage of the exploitee's vulnerability per se, but rather by the taking advantage of the vulnerability for the specific purpose of realising the impermissible distribution. In addition, the resulting distribution will, as it were, concretely manifest the relational injustice. As a result of the exploitation, the parties will be asymmetrically related in respect to the holding of a specific set of goods, and asymmetrically related in this way because of the wrong-doing on the part of one of them. This will further undermine the exploitee's standing in the relationship. While this may be a comparatively minor worry in cases of a one-off exploitative interaction (e.g. Antidote), it is significant in cases where there is a long-term exploitative relationship (e.g. Sweatshop).
Second, while our discussion of relational injustice has focused on harms to the exploitee, several authors have noted that the negative impact of unjust relationships is at least to some extent reciprocal. That is, the harm of such relationship tends to be suffered not only by the weaker party, but also to some degree by those at the longer end of the stick due to, among other things, the erosion of trust and solidarity. 28 This observation brings out a seldom noticed feature of exploitation, namely that exploitative transactions can be mutually bad in relational terms, despite mutual distributive gain (relative to no transaction).
Third, many believe that relational goods are non-substitutable, meaning that no amount of (say) material resources can adequately offset massive asymmetries of power, rank, or status. 29 If this is indeed true, our insistence on the relational harm of exploitation and the need to remediate it becomes particularly important. However, our argument does not strictly speaking presuppose non-substitutability, but only that wrongful exploitation involves negative relational cost, which may or may not be outweighed by the distributive gain.
Before turning to the issue of remediating the relational harm of exploitation, we shall suggests that recognising this harm might help respond to the so-called non-worseness claim, a powerful challenge to any account holding that mutually beneficial and consensual exploitation can be impermissible. According to this claim: 'it cannot be morally worse for A to interact with B than not to interact with B if: (1) the interaction is better for B than non-interaction, (2) B consents to the interaction, (3) such interaction has no negative effects on others'. 30 We warn here against adjudicating (1) without considering the relational harms that may be caused by interaction, harms which will be absent from non-interaction. Once these harms are factored in, it is entirely possible that non-interaction is better for B despite the distributive gain of interaction.
We do not, of course, deny that the distributional gain to B (compared to no transaction) could outweigh the relational harm such that the interaction is on balance better for B than non-interaction. In such cases, the non-worseness claim remains a challenge to the view that interaction is seriously wrong. There is probably no general criterion for adjudicating when gain in one dimension outweighs harm in the other and thus for demarcating the range of cases where the non-worseness claim applies. However, adding relational harm to the equation explains (a) how non-interaction can be better than exploitative interaction with mutual distributive gain, and (b) how even when the interaction is on balance better due to the distributive gain, it can nevertheless be harmful in relational terms. 31

Remediating the Relational Injustice of Exploitation
We have argued that exploitation inflicts a relational harm by severely undermining the exploitee's standing in her relationship to her exploiter. Exploitation therefore cannot be fully remediated through redistribution of the gains from the exploitative transaction. In addition, the exploiter must acknowledge and seek to repair the relational harm. Consider Antidote again. Suppose that A, once back from his forest walk with B's money on his bank account, realises that his treatment of B was wrong and immediately transfers the $10,000 (minus perhaps whatever amount he is entitled to keep) back to B. We find it intuitively plausible that that this does not suffice to fully repair the wrong suffered by B.
There are several complementary ways to restore the damaged status of the exploitee. First, the exploiter can be expected to apologise to B and seek B's forgiveness. We understand here an apology as a request to repair the damage to the relationship between the wrongdoer and the victim caused by the former's wrongful action. Correspondingly, the victim's forgiveness is interpreted as the willingness to grant that request. 32 In this regard, Antidote represents something of a limiting case since A and B are strangers who may have no relationship beyond their one-off interaction. However, even in this case, if A genuinely wishes to make up for wronging B, then A should not only return the money but also express his recognition of B's status by apologising and seeking B's forgiveness.
More often than not, however, exploitation takes places within ongoing relationships. In such cases, second, remediation requires the exploiter to express his willingness to renegotiate the terms of the interaction, including by granting the exploitee influence over these terms. By doing so, the exploiter accedes to empowering the exploitee so that the shape of their relationship will be at least minimally acceptable as regards their relative rank, status, and power. This will also show at least a modicum of solidarity with the exploitee's plight. As a result, successful remediation will also typically lead to at least a reduction of objectionable economic and social disparities between the parties. Some might agree that exploitation causes relational (in addition to distributive) harm, but object that this does not reveal anything specific about exploitation. After all, other kinds of moral wrong may also bring about relational harm, which any attempt at remediation must address. In response, while agreeing that other wrongs too may cause relation harm, we want to emphasise the specific nature of relational injustice in exploitation. Unlike other wrongs, exploitation necessarily presupposes an asymmetry and necessarily distributes benefits and burdens in a way that reflects this asymmetry, thus affirming the unequal relationship that made the extraction possible. While other wrongs may certainly cause relational harm, we argue that the precise nature of the harm and the mechanism behind it are specific to exploitation and that this specific form of objectionable relational harm is what remedial action should target.

Exploitation with Third-Party Beneficiaries
In this section, we will highlight another way in which the relational dimension of exploitation matters. Here we consider the issue of allocation: who ought to remediate, and why? The difficulty here is that real-world exploitative transactions and relationships tend to be more complex than the cases discussed so far, because they commonly involve third parties who stand to benefit from the exploitation. 33 For example, in Sweatshop the picture might look like this: The apparel manufacturer is a supplier to a large multinational retail company which strongly expects it to deliver garments on schedule at an agreed price. The multinational facilitates the manufacturer's operations by providing technical, managerial, and legal support. It is aware of the wages and working conditions in the factory but chooses not to inquire too deeply into or address these. Moreover, the manufacturer's operations are facilitated by the country's government, which provides stimulus for the domestic textile industry through tax breaks and other incentives. Also, while the wages and working conditions in the factory fail to comply with labour laws in the country, the government does not consistently enforce these. Nor does it do much to reduce unemployment in poor rural areas or mitigate its effects on workers.
Finally, partly because of the low production costs of the manufacturer, individual consumers in developed countries are able to buy clothes more cheaply than they otherwise could.
Thus extended, our case (henceforth Sweatshop II) involves not just an exploitative manufacturer and exploited workers but also a multinational, a government, and consumers in developed countries. These third parties are not exploitersat least not in the sense that the manufacturer is. But they are connected to the exploitation in ways that seem potentially morally significant. We will suggest that it is useful to divide third parties involved in exploitation into three principal categories: innocent beneficiaries, knowing contributors, and collaborators. 34 In the next section, we argue that this taxonomy has important implications regarding remedial duties in response to exploitation. So, consider the involvement of the consumers first. The most straightforward view is that they are innocent beneficiaries: they stand to gain in consequence of the exploitative conditions in the factory, but they are not involved in any significant way in the exploitation itself. This view could be challenged, however. It might be argued that low production costs and related exploitative hiring practices are predictable responses to consumer demand for cheap clothes, and, moreover, that consumers are (or should be) aware that their market behaviour has such effects. On this alternative view, the consumers knowingly contribute to (i.e. are complicit in 35 ) the exploitationeven if only marginally or on an aggregate level.
While the consumers' involvement is thus somewhat ambiguous, the involvement of the government appears to fall squarely into this latter category. The government benefits from the exploitation because the profitable factory generates tax revenues and contributes to economic growth. In addition, the government can rightly be said to make the exploitation possible, both by failing to consistently enforce labour laws and by omitting to address unemployment thus helping to maintain a pool of exploitable workers. Moreover, it seems plausible that it is (or should be) aware of playing this role. The government thus qualifies as a knowing contributor to the exploitation.
The multinational also benefits from the exploitation because the factory's low production costs are reflected in the price it pays for the garments. And it contributes to the exploitation. But its contribution differs from that of the government. The multinational's involvement is such that it can be said to collaborate with the manufacturer in regard to the latter's exploitative conduct. In other words, the manufacturer and the multinational exploit the workers together.
The distinction we make between knowingly contributing and collaborating is not uncontroversial. However, it can be defended by drawing on joint action theory. We will use Bratman's theory for simplicity's sake here, 36 but the basic distinction could be established using rival theories of joint agency as well. 37 Bratman argues that people act jointly when three conditions are fulfilled: (a) the participating individuals' intentions are shared, (b) they are mutually responsive to one another, and (c) it is common knowledge 38 among the participating individuals that (a) and (b) obtain. In short, when these conditions are met and people are indeed acting together, then participants must necessarily be mutually responsive to one another, and furthermore they must be so intentionally and this is 'all out in the open'.
When the joint action is exploitative, this means that the third party is aware of the exploitation and works hand in hand with the exploiter towards realising the exploitative transaction or maintaining the exploitative relationship. Plans may be elaborated together, contracts may be signed, contributions mutually adjusted, and so on. We believe that this is true of the relationship between the manufacturer and the multinational in Sweatshop II. The multinational not only enjoys the fruits of the workers' cheap labour such as lower production costs, but also helps to plan and maintain the labour and production practices in the factory, and can thus be said to share the intention with the manufacturer to exploit the workers. By contrast, such intentional overlap and mutual responsiveness seem absent from the manufacturer's relationship to the consumers or the government.
That said, this distinction may admittedly play out differently in different real-world cases of sweatshop labour. In some cases, government institutions may be sufficiently intimately involved with exploitative manufacturers to qualify as collaborators rather than contributors. Conversely, multinationals' supply chains may sometimes be too complex for them to interactlet alone collaboratewith exploitative suppliers, and so they may be more accurately classified as contributors. And perhaps in some cases neither government institutions nor multinationals are involved in either way. However, while this observation calls for caution in applying the framework we sketch here, it does not undermine the framework itself.
In sum, joint action theory can help us distinguish between the class of third-party beneficiaries of exploitation who cooperate with the exploiter, on the one hand, and the class of third-party beneficiaries who do not, on the other. Emphatically, the point of making this distinction is not to straightforwardly rank third parties in terms of moral culpability: cooperating with an exploiter is not necessarily worse or more blameworthy than contributing to or benefiting from her conduct. Rather, the point is that when the third party acts together with the exploiter, their remedial obligations will be different. We now turn to this issue.

Allocating Remedial Duties
So different agents may be involved in different ways in exploitative real-world practices. Which among these agents ought to shoulder the burden of remediating the victims? For example, who in Sweatshop II should ensure that the exploited workers are provided with adequate wages, working conditions, and opportunities to influence their situation?

Normative Grounds for Distributing Remedial Duties
David Miller's framework for allocating remedial duties is a useful point of departure for tackling this question. Miller defines remedial duties as 'special [duties] to put [some] bad situation right', i.e. as duties 'towards the deprived or suffering party that [are] not shared equally among all agents'. 39 Miller argues that such duties can be assigned to agents on four distinct grounds: (a) moral responsibility, (b) causal responsibility, (c) community, and (d) capacity. Thus, I may be required to 'put some bad situation right' because I am morally responsible for its occurrence, because I (innocently) played a role in causing it, because I have some special relationship to those who suffer from it, or because I am especially capable of providing alleviation.
We will briefly make three points about this framework and its application to the issue of remediating exploitation. First, whereas Miller discusses remedial duties in general, our scope is more limited. We are not concerned with any special duty that somebody may have to improve the situation of exploited people, but with duties arising specifically from exploitation. This is important because in many cases exploiters and third-party beneficiaries can have duties towards the exploitee which are independent of the exploitative transaction. For instance, in Sweatshop II one might plausibly hold that the government, and perhaps also the multinational, the manufacturer, and/ or consumers in affluent countries should contribute to reducing the poverty and vulnerability of the workers 40quite independently of each party's connection to the exploitation of them. However, we focus on duties that agents have over and above any such preceding duties, i.e. on duties directly connected to exploitation.
Second, it should be emphasised that the general practice of attributing remedial duties is based on a concern with alleviating suffering and deprivation. As Miller notes, we all have a strong interest in receiving aid should we end up in such a condition, and any credible philosophical account of allocating duties should foreground this interest. From this perspective, remedial duties must not be too widely dispersed for then no agent is likely to act. But it is at least as important that some agent can always be tasked with providing remediation. So far, so good. However, Miller also suggests that, from the victims' perspective, it is relatively unimportant who is assigned this task as long as remediation is forthcoming. We disagree with this point. To begin with, at least as far as exploitation is concerned, it misses the relational aspect of remediation. As noted, remediating exploitation does not just require redistributing goods but also repairing the morally defective relationship brought about through the exploitative exchange. Whereas the first of these requirements can be satisfied by anybody who possesses the relevant goods, the other calls for the exploiter specifically to engage in remediation. Moreover, and more generally, the concern with alleviating the victims' plight generates strong secondary reasons to ensure that our justifications for selecting remediators are potentially acceptable to the remediators themselves as well as to third parties who are in a position to sanction them should they fail to assume that role. Otherwise little remediation will happen. This suggests that attributions of remedial duties must reflect widely recognised normative considerations that are to some degree independent of the victims' interest in remediation.
Third, and relatedly, while Miller holds that the four principles are not hierarchically ordered, some are clearly more compelling than others in most cases. As Miller recognises, ordinary moral practices and legal norms typically accord some priority to moral responsibility as a basis for remedial duties. That is, the finding that some harm has been caused by the blameworthy conduct of a clearly identifiable agent is often by default considered a prima facie reason for allocating the remedial duty to that agent. By contrast, causation, community, and capacity are more controversial as grounds for remediation. Furthermore, even when these grounds are recognised in principle, they often fail to pick out salient duty-bearers. There may simply be too many agents around who have the capacity to remediate, who have contributed causally to the harm, and who have communal ties to the victim.

Joint Remediation
With these points in mind, we will now outline an approach for allocating the burden of remediating exploitation in complex real-world cases. The key idea is that when exploitation involves joint action the task of remediating the victims is also a joint one. The point is not simply that the remedial duty is assigned to all contributors to the joint exploitation, but also that discharging the duty requires joint action. Since the parties exploited together, they should now seek to address the consequences of the exploitation together. More precisely, since each of them acted wrongly, each must play a part in the joint remedial action seeking to ensure that the other does so as well, and adjust their respective performance to that of the other as needed to make the joint activity successful.
To make this more concrete, consider Sweatshop II again. Here the multinational and the manufacturer should together ensure that the exploited workers are granted adequate wages and working conditions. Each of them should make this their plan and, making use of their established relationship, seek to ensure that the other does so too. In view of this joint plan each of them should then form subplans for their respective individual contributions. For instance, the manufacturer might have to ensure that concrete safety measures are taken, employment contracts amended, and so on, whereas the multinational might have to provide funding and advice needed for these improvements. Furthermore, when implementing these subplans, each needs to continually monitor what the other does and if necessary tailor their own activities accordingly.
Importantly, this involves mechanisms of assistance and enforcement. If one of the parties is unable or unwilling to play its part, the other is required to provide support, apply pressure, or (depending on the situation) even 'pick up the slack' to some extent. So, for instance, if the manufacturer is disinclined to improve working conditions in the factory, the multinational may have to leverage its bargaining power to ensure that it does. And if the multinational provides insufficient funding for the requisite improvements, the manufacturer may have to turn to other partners for financial support.
The significance of the joint action perspective is particularly salient in regard to the relational side of remediation. In cases where two people together wronged a third, it is clearly incumbent on each of them to recognise and apologise for her or his contribution to the wrong-doing ('I am sorry for what I did'). But furthermore, each should be prepared to acknowledge that this contribution was part of a collective misdeed ('I am sorry for what we did'). Not even this seems quite enough, though. Rather, full remediation appears to require the wrong-doers to, when at all possible, together recognise and seek the victim's forgiveness for the wrong ('We are sorry for what we did'). We also noted earlier that remediation of the relational harm caused by exploitation requires renegotiating the terms of the ongoing interaction so that the exploitee is granted at least an acceptable standing, which includes giving her a voice. Clearly, in complex cases involving third parties, such renegotiation may also require cooperation. In Sweatshop II, for instance, the manufacturer may not be able to ensure that employment contracts are improved and that workers are empowered to influence their working conditions (e.g. through unionisation) unless the multinational helps make this possible by removing administrative obstacles and contributing funding or expertise.
By stressing joint remediation, we are not denying the importance of individual efforts. On our approach, each contributor to joint exploitation did wrong, and each is therefore required to engage in remediation. This individual duty is what grounds the requirement of joint remediation. So even when joint remediation cannot be organised, each party remains required to address the victim's plight in some way. This strikes us as highly plausible in regard to distributive as well as relational forms of remediation. For instance, suppose in Sweatshop II that the manufacturer goes out of business and that its former manager is nowhere to be found. It still seems clear that the multinational ought to compensate the workers significantly, though not necessarily shoulder the entire compensatory burden that would have been required had the manufacturer still been present. Furthermore, it still seems clear that the multinational (or its representatives) owe the workers recognition of and apology for the joint exploitation and that it ought to unilaterally initiate the dialogue to improve working conditions, even though such gestures will inevitably be somewhat incomplete when performed without the manufacturer's participation.
We believe this approach is attractive partly because it forcefully recognises the victims' interest in remediation. Assigning remedial duties to exploiters as well as their collaborators could help ensure that remediation happens in cases where the former are no longer around. Also, insisting on joint remediation where possible requires exploiters and collaborators to seek to overcome the motivational, practical, and resource-related obstacles that often would preclude individual remediation. At the same time, our approach is supported by powerful considerations independent of the victims' interests. Collaborators are singled out along with exploiters because they too engaged in the exploitation and shared the exploiters' objectionable ends. In other words, the remedial duty is attributed on the basis of moral responsibility, the most compelling of Miller's principles, and there is no need to invoke more tenuous connections (e.g. causal responsibility or community). In this way, our approach aligns forward-looking and backward-looking considerations that otherwise tend to pull in opposite directions.
Emphatically, we are not denying that third parties who did not collaborate with the exploiter may nonetheless have duties to remediate the exploitation. But these will not be duties they bear together with and have to discharge together with the exploiter. For instance, in Sweatshop II governmental institutions may well be required to facilitate the manufacturer's efforts to remediate the exploited workers, just like they facilitated the exploitation in the first place. (And again, the government plausibly has duties to address poverty and unemployment among its citizens and to enforce labour laws independently of its role in the exploitation.) However, the governmental institutions are not required to monitor and adjust to the manufacturer's performance to the extent that the multinational is. And they will not be to blame for any failure to remediate on the part of either of these parties. 41 We have argued in the last two sections that collaborators who act jointly with the exploiter are morally responsible for the exploitation and are therefore particularly salient bearers of remedial duties. This argument lends further support to the claim that remediation of exploitation cannot merely be a matter of redistributing gains from exploiters to exploitees. As argued above, exploitation inflicts a relational harm in addition to a distributive one, and adequate remediation has to reflect this fact. We can now see that in joint exploitation cases not just exploiters but also their collaborators are implicated in both these harms. Therefore, in such cases, collaborators have to participate in remediating these harms and where possible collaborate with the exploiter in doing so.

Conclusion
Despite a growing interest in exploitation among philosophers and applied ethicists, surprisingly little has been said so far about what it might mean to set things straight after exploitation has occurred. This article intends to contribute to filling this gap.
We have identified and begun addressing three main issues. First, it is unclear what remediation of mutually beneficial exploitation even means given that the victims have already benefited. We have argued that even if exploitees benefit relative to no transaction, they are deprived of something they are entitled to. Compensating this loss is a necessary part of remediation. However, second, we have also argued that compensation is not sufficient because remediation must also address the fact that exploitation leads to a relationship which is deficient as judged by ideals of relational justice. Third, it is unclear how the burden of remediation should be allocated in cases involving third parties. Here we have suggested that joint action matters. When exploitation resulted from joint activity, its remediation should also (where possible) be performed jointly, and certain remedial duties can only be adequately discharged if the exploiter and the collaborating third party act together. This is also necessary because in joint exploitation cases collaborators are part, on the wrong side, of the morally deficient relationship between exploiters and exploitees.