Beyond right and wrong: on the conditionality of dirty hands

Abstract

Dirty Hands theorists disagree about how agents should resolve a high-cost moral dilemma, but their disagreement is partly because they tend to discuss widely different cases of a broad and heterogeneous phenomenon. Moralists are typically concerned with problems that often involve an agent who is under coercion and is asked to engage in an activity that will cause severe and certain harm to individuals. Realists, on the other hand, base their observations on cases where political parties negotiate to form coalitions or policy platforms; these compromises may affect the political integrity and representative credibility of the agent, but less so their moral integrity as measured by universal moral standards. Yet, both types of Dirty Hand scenarios concern the same phenomenon: an urgency to make a morally costly compromise. As a result, we propose to evaluate Dirty Hands problems by placing them on a dual continuum based on two conditions: their projected outcomes, and their external circumstances. We propose that the position of a moral problem on this continuum affects the extent to which a compromise is or is not excusable. Finally, we consider the implications of our findings for the Dirty Hands debate and for the study of political ethics more broadly.

Introduction

In October 2023, following a series of coordinated attacks by Palestinian militant groups from the Gaza Strip onto bordering areas in south Israel, the Israeli armed forces engaged in a wide-scale counter-attack on Gaza. Seen that Palestinean fighters were spread out over a densely populated area and hiding in underground tunnels, Israel started to bomb areas in northern Gaza where they knowingly caused multiple civilian casualties. From an ethical point of view, the politicians authorising these strikes had dirt in their hands: even if these airstrikes are accepted as an act of ‘self-defence’ and collateral damages seen as justified, they still involve a moral transgression because killing innocent civilians is per se an immoral act.

When faced with a pressing moral conundrum, politicians are often forced to sacrifice a universally shared, deontologically considered moral principle in order to better serve the immediate public interest. This is commonly described as the Dirty Hands (DH) problem. The problem is paradoxical because the morally right action is also morally wrong. Although there is general agreement that such a problem is particularly likely to occur in politics (Hall and Sabl 2022, pp. 7–9), there is little consensus on when and how often it occurs.¹ First suggested by Walzer (1973), the ‘ticking bomb’ scenario is often invoked as the standard example of how politicians may confront a DH dilemma: a politician is tempted to authorise the torture of a suspected terrorist to prevent a bomb attack against civilians. Even if ordering to torture the terrorist suspect is the right thing to do under the circumstances, torture still remains wrong. Hence, even after resolving the dilemma, the politician is left with a moral residue, which is why Walzer and other Moralists² (Yemini 2014; de Wijze 2018; Nick 2022) think the agent should acknowledge their guilt. In the case of the Israeli government, this would require a public acknowledgement of guilt for the loss of lives of innocent civilians in Gaza.

Admittedly, and fortunately, ‘ticking bomb’ scenarios are rare. Yet, some claim that the DH problem is not limited to such rare and exigent circumstances; in other cases, the wrongdoing can be less severe than authorising torture and more akin to making concessions. For scholars who draw from the Realist tradition, DH dilemmas are comparable, reducible, or perhaps identical to the dilemmas involved in making political compromises (Bellamy 2010; Tillyris 2015, 2016, 2017). Like DH problems, political compromises often entail violating some deeply held principles for the sake of the public interest. And because political compromises are ubiquitous, a pervasive feature of political life, they argue, so is the DH problem. In addition, because political morality differs from private morality, politicians should not be held accountable for acts that would otherwise be reprehensible. In the case of the Israeli government, they would not have to publicly acknowledge guilt for the loss of lives of innocent civilians in Gaza.

In this paper, we show that the disagreement between Moralist and Realist DH theorists is misguided, and offer a roadmap for a more accurate comparison between moral problems faced by political agents. In the following “Moralists v. Realists: a misguided debate” section, we explain what the problem with the current state of the art is: the two sides base their reflections on incommensurate premises, which makes the debate counterproductive. In the main part of the paper (“How Moralists frame the DH problem” and “How Realists frame the DH problem” sections), we compare DH problems typically discussed by Moralists and those typically discussed by Realists, and find that they involve a different set of contextual, agential, and other considerations. Based on these reflections, we present in “Evaluating DH problems: a dual continuum framework” section a dual continuum on which DH problems can be placed, which takes into account the different ‘origins’ of the problem (e.g. if the agent is subject to immoral coercion or not) and the different ‘outcomes’ produced by making a DH concession (e.g. how severe the harm is). We suggest that the place of a moral problem on this scale affects the degree to which a compromise is excusable or not. Lastly, we reflect on the implications of our findings for the future of the DH debate and for studies of political ethics more broadly.

Moralists v. Realists: a misguided debate

On the face of it, the divide between Moralists and Realists is principled: the former require individual politicians with dirty hands to show moral regret, while the latter grant them a pass in the name of the art of politics. In our mind, however, the divide is at least partly caused, and certainly exacerbated, by definitional confusion. It arises from the fact that the two camps collapse different types of moral problems political agents face into one unified reference point. Put differently, they treat cases that feature substantial differences as if they are the same. Moralists tend to draw on examples such as the ticking bomb (Walzer 1973) or the sinking boat scenario³ (Stocker 1992), whereas Realists are almost exclusively inspired by cases of political parties negotiating over coalition-formation or policymaking (Bellamy 2010, 2012; Tillyris 2017).⁴In our mind, these examples feature crucial differences, which we illustrate below, and should therefore involve a different set of reflections. The use of different types of scenarios is not just a matter of framing and illustration but informs the selection, omission and prioritisation of different aspects that affect moral decision-making. There cannot be an identical moral calculus for very distinct cases, not unless their differences are duly incorporated into that calculus.

One negative consequence of this asymmetry between perceived DH problems is the tendency to over-generalise. Because little distinction is made between types of moral problems and little attention is paid to whether they share key characteristics, both camps assume the existence of broad and universal standards that presumably apply in all cases where politicians are called upon to make morally costly decisions. For example, Moralists argue that when politicians get their hands dirty, they must always express guilt, in order to deflect immoral and opportunistic behaviour (Walzer 1973; de Wijze 2018; Nick 2022). Conversely, Realists take it for granted that politics is, by definition, always a dirty business that requires moral compromise, for which politicians should not express guilt (Bellamy 2010; Tillyris 2015). But perhaps expressing guilt is a reasonable demand when a politician engages in borderline legal behaviour for the sake of the common good but seems less urgent during national budget negotiations between elected parties. In our view, blanket statements commonly made by DH theorists are inevitable unless a distinction is made between political compromises involving ideological shifts, which are indeed a daily phenomenon in politics, and deeper moral compromises that violate basic universal norms. Applying the conclusions of one set of examples to the other is not so straightforward. The (perceived) DH problem is all too often decontextualised and treated in the abstract, losing much of the nuance and comparability that a more concrete treatment would allow. The gap that emerges between the two camps then does not reflect a genuine disagreement but a collective misperception.

A second problem that follows from the above is that the dichotomies established by this misdirected discussion are fictional. In the absence of an analysis of the conditions that structure a moral-political dilemma, the result is the creation of conceptual antinomies with a weak basis. For example, DH theorists disagree as to whether the DH problem is frequent or not, with Realists describing it as an everyday phenomenon, while Moralists tend to see it as a matter that arises rarely and usually only in crisis situations.⁵ Does the question of frequency not depend on the precise nature of the problem in question? Few would disagree that government coalition formation is a frequent phenomenon, whereas being blackmailed by terrorists is not so frequent (and will hopefully remain so).

The question of frequency fails to capture the full spectrum of DH problems, but it also introduces a deeper disagreement about a possible division between normal and exceptional politics. For those who accept this dichotomy, events that qualify as ‘exceptional’ have a higher degree of permissibility compared to ‘normal’ ones. In our view, however, it is not productive to think about the nature of politics in these black-and-white terms. A useful way forward for the DH debate, on the other hand, is to think about specific contextual conditions that are crucial for moral decision-making.

To be sure, the lack of definitional clarity has accompanied the DH debate from its very beginning. In his pioneering analysis, Michael Walzer borrows the term ‘Dirty Hands’ from a 1944 play by Jean-Paul Sartre of the same title (Le Mains Sales). The play, however, contains a series of moral dilemmas and transgressions committed by political agents without ever clarifying the qualitative differences between them. For the sake of illustration, let us briefly describe parts of the play.

In a fictional Eastern European country at the end of WWII, Hoederer, a Communist Party leader, plans to collaborate with non-socialist parties, including fascist and liberal ones, in order to help his own party take power. Hoederer claims that he has ‘dirty hands up to the elbows’ and has ‘plunged them in filth and blood’. He asks whether it is possible to ‘govern innocently’ (Sartre 1989, p. 218). One could be forgiven for assuming here that Hoederer’s compromise is at the heart of the play’s moral reflection. But then, a fellow Party member named Hugo is ordered by the party to assassinate Hoederer in order to preserve the ideological purity of the Party. Although he has moral qualms about murder, Hugo carries out his mission. The play goes on to show how, despite being abandoned by the party, Hugo reveals his initially hidden political motives for the crime and accepts his punishment (Sartre 1989, p. 241). Killing someone for political reasons then adds to the moral problematisation of the play.

With this dramatic sequence of events, Sartre’s play underlines two types of dirty-hands scenarios: one involving collaboration with hated enemies, and one in which committing a politically motivated murder is the dirty act. Both acts serve political ends, strategic ones in the first case, and ideological ones in the second. But the former involves a situation of political-ideological compromise, while the latter involves the criminal act of murder. Even if a political compromise may render someone passively complicit in the crimes committed by one’s future political associates (as Hoederer fears), it falls far beneath the active commitment of a crime. From the very beginning, the conflation of these two moral problems has endowed the debate with confusion.

To solve this confusion, in the rest of the paper, we compare the way Moralists and Realists have framed the DH problem, and the different constitutive elements they have emphasised. All these elements are later brought together to inform our proposal for an analytical framework that can help to distinguish and evaluate different types of DH problems.

How Moralists frame the DH problem

Within the DH debate, Moralists can be divided into two types: pure and mixed. Pure Moralists, including absolutists (or deontologists) and utilitarians (both act- and rule-utilitarians), do not experience the DH dilemma: for them, there is only one right course of action. For act-utilitarians, for instance, it would be the action that serves the greatest utility for the greatest number (Nielsen 2000). Nonetheless, very few people are pure Moralists in such a sense. Most people harbour mixed moral intuitions—they are moved by both deontological and consequentialist considerations—and as a result would experience the DH dilemma (Nagel 1972, pp. 124–125; Yemini 2014, p. 168). Even people in whom one consideration is stronger than the other will encounter the DH problem: although they may not experience guilt, they must recognise that in order to do a right, they have committed a wrong in the eyes of others (Yemini 2014, p. 177). Walzer and other Moralist DH theorists who advocate acknowledging wrongdoing are therefore mixed Moralists. For them, the DH dilemma exists, and it manifests in cases where it is impossible to satisfy the public interest without sacrificing a deontological moral concern. It is only to these mixed Moralists that we refer when we use the term Moralists in this paper, leaving pure Moralists entirely outside the scope of our analysis.

Moralists have used various examples to discuss the DH problem, the most classic being the ticking-bomb scenario. For Yemini, it captures the essence of the ‘dirty hands’ problem, ‘an acute conflict between neutral, consequential reasons and deontological restrictions’ (Yemini 2014, p. 171): preventing the death of many innocent civilians is a consequence anyone would condone, whereas torturing the captured rebel is a violation of a deontological restriction. Yet, not all DH theorists would agree on this definition. The absence of an agreed-upon description of the DH problem has created an endless source of debate, with various examples being added to illustrate its complexity. The issue, however, is that discussing various cases without underlining their similarities and differences over-broadens the category and creates an inevitable confusion, which, as we said, carries on to the disagreement with Realists.

Walzer, like Sartre, convolutes the ticking bomb scenario itself by adding to the puzzle that the politician who is called upon to authorise torture has also made a political commitment against torture. This additional claim leaves him vulnerable against his Realist critics, who correctly argue that politicians cannot always expect to fulfil their campaign promises entirely.⁶ In essence, a DH situation need not always involve a betrayal of one’s expressed political or even moral position. A DH situation may involve a conflict between two incompatible moral duties that would apply to everyone (de Wijze 2007, p. 9). Consider the example of a DH situation from Nick involving a politician facing immoral coercion by a criminal group threatening to destroy places of cultural worship unless the government releases its imprisoned members immediately (2019, p. 85). There is no indication in this example that the politician under scrutiny has made any electoral promise or is morally committed to safeguarding the places of cultural worship or holding the criminal group accountable for their actions. The politician, however, still ends up facing a moral conflict in which performing the duty to uphold the rule of law will negate the responsibility to protect cultural heritage and vice versa.

Besides the ‘ticking bomb’ scenario, in his original 1973 article, Walzer invites us to consider the following ‘corrupt ward boss’ scenario: a political candidate must reach an agreement with a dishonest ward boss for a school construction project to secure crucial votes despite his (sic) moral commitment to winning the election fair and square. Yet, he is tempted to do it, for it is worth winning, even though Walzer does not say what makes it worth winning (1973, p. 165). He argues that because the politician reaches the deal reluctantly, with compunctions, we know him to be a moral politician; indeed, we want him to take the deal precisely ‘because he has scruples about it’ (Ibid., p. 166). However, unlike the public good of saving innocent civilians from a bomb explosion, in this case, the extent to which winning the elections is a public good is contestable. As Yemini points out, Walzer is mistaken when he suggests that the difference between the ‘corrupt ward boss’ scenario and the ‘ticking bomb’ scenario is a matter of degree, whereas there is a difference in substance between the two: the former is not an example of a DH situation because the ‘objective utility’ for everyone of our political agent winning the elections is questionable, it does not have a neutral value that everyone would have a reason to promote (2014, p. 172). We would not go as far as to disqualify this scenario as a DH case but nevertheless agree that there is a crucial distinction to be made: what is missing here is a clear indication that this particular election is correctly deemed to be an exigent occasion. In what follows, we explain in more detail our understanding of ‘exigency’ and the value we afford to it as a constitutive element of a DH problem. But, before that, we focus on what we consider the most important criterion that makes the paradigm DH cases stand out and shape the Moralist prescription: the type and level of harm induced on others by dirtying one’s hands.

Outcomes: severe and certain harm

Surely, for a DH situation to arise, there must be evidence of ‘dirty’ conduct. Walzer himself does not further problematise the nature of moral transgressions that occur in DH situations. Coady and O’Neill (1990, p. 263) distinguish among different immoral acts on the basis of their ‘awfulness’ while warning that the acts might be justifiable only partially or combinedly. By contrast, de Wijze has suggested that ‘[g]enuine DH scenarios lie on a continuum ranging from those involving relatively minor moral infractions to cases where there is the commission of terrible moral crimes’ even though those involving serious moral crimes are fortunately rare and ‘are often offered as paradigm examples of DH….’ (2018, p. 132).

Let us examine in detail the kind of moral infractions that paradigmatic DH cases (those typically used by Moralists) involve. The most typical example is torture, illustrated by the ‘ticking bomb’ case. Since, as a practice, torture is prohibited by numerous human rights conventions, the question of whether it is permissible in a ‘ticking bomb’ scenario has engendered an intense scholarly debate. Luban (2014, p. 44) cites the UN Convention Against Torture and the US Bill of Rights and other acts as proof of how established the prohibition of torture as an international moral norm is. Meisels refers to the International Convention on Civil and Political Rights that imposes an absolute ban on torture but nevertheless argues that torture may be excusable, even if never justifiable, in genuine DH scenarios (2008, p. 222). Likewise, Archard (2013, p. 778) contends that torture remains wrong even when it brings a highly desirable outcome. Alongside torture, another controversial outcome often discussed by Moralists is the targeting of non-combatants in war (as our opening example illustrates). This is also seen as a DH scenario, possible to excuse when it saves more lives in the long run, for example by bringing about a quick victory (Walzer 1973; Yemini 2014; Kramer 2018).

The reason why torture and the bombing of civilians are difficult to excuse, even if they serve undeniably good purposes, is that they are serious violations of human rights. Human rights are widely shared and consolidated moral principles that emanate from our common desert for human dignity. They are de facto universal because they are enshrined in international covenants signed and ratified by the overwhelming majority of countries in the world today. It is the violation of such universal moral principles that characterises these extreme DH cases typically discussed by Moralists.

In terms of the harm caused by authorising torture or the killing of civilians, these cases stand out because they cause grave and certain harm to human beings. The gravity of the harm is linked to the violation of a human right, as opposed to other, less fundamental rights. The harm is certain, i.e. its effects are direct and immediate, compared to the harm that is only eventual and expected, at a distance from the actual transgression (Elm-Schulin 2023, pp. 155–161). So, even if it is uncertain that a compromise will yield the intended results, as others have successfully argued (Hall and Sabl 2022, p. 6), what is certain is that it will result in immediate harm. The gravity and certainty of harm to individuals is what increases the cost of the moral transgressions typically discussed by Moralists. Their examples are characteristic of how harmful the outcomes of DH problems are. Moreover, they typically involve a high-stakes context and other constraints to which we turn our attention below.

Exigent circumstances: physical and immoral coercion

The ‘ticking bomb’ scenario and other real and hypothetical examples of what are often called ‘extreme’ DH scenarios involve an element of exigency, also described as emergency, necessity or urgency. This element of exigency is not necessarily related to time. The problem is not that a decision is urgent, i.e. that it cannot wait for the time that would be required, for example, to call a cabinet or parliamentary vote on the matter, or seek a judicial opinion or permission to act from a court, or to commission an intelligence or other scientific analysis of the viability of the options at hand. While these time constraints may still be part of the picture, they are not the defining element of what makes the circumstance exigent. Nor is urgency linked to the abnormality of the situation. As Margalit rightly points out, situations such as war or terrorist attacks may be extreme in relation to peace, but they are both realistic and expected (Margalit 2009, p. 84).

Rather, it is the presence of a coercive element in the decision-making situation that creates the exigency. As de Wijze has argued, dirty hands differ from ordinary moral conflicts in that ‘the agent is immorally coerced to further an evil project’. They arise when another person deliberately creates circumstances that lead an agent to betray a person, a group, a value or a moral principle and cooperate in their immoral project in order to avert a greater evil (de Wijze 2007, pp. 15–16). This coercive element may involve the use or threat of physical force, such as when an army has invaded another country’s territory, a hostage situation is unfolding, or a ticking bomb is about to explode. In addition to life-or-death situations, there may be other types of coercion that are equally immoral, such as blackmail by threatening to divulge highly classified or damaging information, manipulating one’s public or private image, or fabricating or planting various types of incriminating evidence. In essence, coercion is created by someone acting immorally, often using force or the threat of force, to compel another agent to act involuntarily.

In fact, it is often an uncompromising stance that precipitates the exigency. An uncompromising stance implies an unwillingness to negotiate and make concessions. In the ‘ticking bomb’ scenario, it is the terrorists who maintain an uncompromising stance (assuming that, like most of them, the terrorist attack is an end in itself and not a strategic means of extracting concessions). They refuse to negotiate or seek any alternative peaceful means of achieving their goals, and instead seek to impose an evil project unilaterally. This suggests that they have ruled out any compromise with the government. Consequently, in order to avoid a catastrophe, the politician in charge is forced to take a drastic decision—to torture the terrorist suspect—and thus commit a serious moral transgression. What is crucial here is that the act is carried out under conditions of immoral coercion.

Not all Moralists agree that exigency always involves immoral coercion. Nick (2021) argues that immoral coercion—which she calls the ‘external immorality condition’—is not present in all cases of DH. According to both Nick and de Wijze, a DH problem is supposed to elicit a certain type of emotional response from the agent. de Wijze (2004) then distinguishes between ‘agent regret’ and ‘tragic remorse’. Because DH involves immoral coercion, de Wijze says it will always lead to tragic remorse (caused by active and intentional participation in wrongdoing) rather than agent regret (caused by unintentional wrongdoing). Nick disagrees, arguing that it is not helpful to use this index to distinguish DH cases from other moral conflicts because situations involving unintentional wrongdoing can also evoke a sense of tragic remorse. So, she does not think immoral coercion is a necessary determinant of the DH problem. For her, if immoral coercion is not necessary to evoke tragic remorse, then we do not need it as a determinant of the DH problem.⁷

However, we do not agree that the DH problem should be tied to a particular emotional response when there is a striking variety of mental dispositions among individuals. Therefore, we do not consider moral regret to be a determinant of the DH dilemma. Including moral regret as a determinant would mean that cases in which politicians do not experience tragic remorse would have to be excluded from DH scenarios. However, as Yemini (2014) points out, this is problematic because of the difference between the external perspective of an observer and the internal perspective of a situated agent. Because of this difference, the politician who responds to a DH situation may not experience moral regret and yet remain a moral politician. A better criterion for judgement, Yemini argues, and we agree, is to determine ‘whether such person is capable of understanding why in the eyes of other people—or society as a whole—she may still be considered to have done a moral wrong’ (2014, p. 177). The point is to recognise that a moral wrong has been committed, even if the agent feels justified and unrepentant in committing it.

Still, there are DH cases that do not seem to involve immoral coercion: those in which politicians respond to natural and man-made emergencies with no malicious intent. For example, in a situation where two teams of hikers are trapped in a forest fire, a politician may have to weigh in on deciding whether to send all the firefighters to save one team with certainty, or whether to divide the forces and try to save both teams but with less certainty. The coercion exerted on the agent may not be immoral because the genesis of the problem does not involve human intentionality, but it still involves physical force caused by natural elements. Thus, for Moralists, DH problems always seem to involve an element of—immoral or physical—coercion.

In a nutshell, most, if not all, Moralist descriptions of DH scenarios involve serious harm and some element of coercion. Having compared different views, we can now better understand the nature of the harm that a paradigmatic DH scenario involves for Moralists. We also have a clearer picture of what exigency entails: a situation characterised not by time pressure or scale of devastation but by a physical, i.e. violent or natural, threat that compels a decision-maker to act. In the next section, we turn our attention to how the DH problem is understood, defined and described by Realist DH theorists.

How Realists frame the DH problem

In recent years, the most substantial criticism of the Moralist DH thesis—the idea that DH situations are rare and that politicians should plead guilty to dirtying their hands—has come from the Realist school of political theory. By Realists, we do not refer to the realpolitik school of thought, where politics is seen as motivated only by self-interest and power. Realists do recognise that morality has a place in politics, but they see it as a sui generis morality, not as an extension of a universal moral code projecting private morality onto politics. Instead, they view political decision-making as guided by its own role-specific moral standards (Bellamy 2010; Tillyris 2015). It is a morality determined by the politician’s responsibility of securing a public good, a responsibility not shared by private individuals.

Being a good politician thus entails ‘making difficult choices between rival moral claims’ in order to perform political duties such as meeting the conflicting needs of citizens, dealing with an adversary, or forming a coalition to implement a favoured policy, and none of these can be achieved by adhering to any pre-political universal moral norm (Bellamy 2010, pp. 419–420). In our mind, the political (as opposed to universal) moral code that Realists advocate may involve contradictory but equally legitimate role-specific moral obligations: to serve (a) the general public interest, (b) the interests of their constituency, (c) their party lines, or (d) their ideologies. Each of these goals, which are often at odds, encapsulates an eligible interpretation of the public good from the perspective of the acting agent.

Even so, Realists argue that serving the public good—on whatever interpretation—may involve reprehensible acts from a private morality viewpoint. Politicians would not need to exhibit guilt, however, because these acts lie outside the distinct political moral code that applies to them. The DH dilemma arises for Realists not because politicians face circumstances where they are tempted to violate universal moral codes as Moralists think, but because they face circumstances where practical reason does not dictate which is the right course of action. Hence, although both (mixed) Moralists and Realists agree about the existence of the DH dilemma, they think of it as embedded in different moral codes: universal morality in the first case and political morality in the second.

On these grounds, Realists argue that the Moralist DH thesis not only misrepresents and underestimates the complexity of political life but also advocates a moralist brand of politics by offering a false standard of political excellence and an illusion of social harmony (Bellamy 2010; Tillyris 2017). The DH problem is not a rare episode, they say, but a pervasive feature of everyday life. The DH problem is thus not necessarily limited to the paradigmatic ‘extreme’ cases, as the Moralists would have us believe. To show that politicians frequently get their hands dirty, Realists draw our attention to the ubiquity of political compromises, which they argue share the core features of the DH problem. Politicians get their hands dirty when they make a political compromise deemed to be in the public interest by betraying aspects of their stated moral position.

Tillyris (2017, p. 485) therefore finds problematic the implicit assumption of Moralist theorists that only rare, exigent circumstances challenge a politician’s moral integrity. He argues that political compromises also feature the DH problem, where politicians betray their values and pre-election commitments. Faced with the prospect of never realising their political goals, politicians opt for a ‘lesser evil’—partial realisation of their political objectives—instead of abandoning them altogether (Tillyris 2017, pp. 490–491). As examples, he refers to the compromises both Democrats and Republicans had to make to pass the 1986 US Tax Reform Act and the compromises the British Conservatives and the Liberal Democrats had to make to form the 2010 coalition government (Ibid).

Similarly, reflecting on the 2010 coalition formation in the UK, Richard Bellamy shows how the limits of transparency in politics force politicians to get their hands dirty by violating their campaign promises. For instance, feasibility constraints often force governments to experiment with trade-offs between different sorts of health services, where the interests of a particular group are prioritised over the others so that they can build a ‘good enough’ healthcare service, even though they promise a universally best healthcare service capable of equally meeting the needs of everyone (Bellamy 2010, p. 423). Politicians cannot be entirely candid in their policy promises. They often resort to such ‘smoke and mirrors’ practices because, even when a majority of citizens support a policy, they support it ‘for a variety of different and possibly mutually inconsistent reasons’ (Ibid.). In his example, Bellamy considers that the politician still acts out of concern for the general interest, which is why this scenario resembles a DH scenario.

Given the ubiquity of the DH problem, Bellamy (2010, p. 427) invokes Machiavelli to argue that not being good also implies keeping up the façade of a virtuous politician; they need to hide their dirty hands with the clean gloves offered by liberal idealism. Tillyris (2016, pp. 173–174) goes further by arguing that not only should they wear clean gloves, but the moment they enter politics, politicians also need to master the ‘task of hypocritical concealment’—a reality that the static Moralist DH thesis fails to capture because only a dynamic account of the DH thesis can recognise the importance of hypocrisy in politics.

What is clear from the above discussion is that for Realists, the moral dirt comes from politicians compromising their intended goals and violating their pre-electoral commitments in order to serve the public good. As we have seen, the ‘ticking bomb’ scenario also involves violating a pre-electoral commitment. However, as a paradigmatic case, the ‘ticking bomb’ scenario also includes additional features not found in a political compromise. Conversely, political compromises have certain features not always found in extreme or paradigmatic DH cases. For instance, a political compromise features mutual concessions (Rostboll and Scavenius 2019, p. 4). This is not the case in paradigmatic DH scenarios, where an uncompromising stance from an adversary forces politicians to take drastic actions to avert a catastrophe. Additionally, the harm committed by reaching a political compromise does not necessarily reach the level, certainty and directness of harm resulting from torture or murder.

Unlike political compromises, paradigmatic DH situations involve immoral coercion where an individual or a group unilaterally attempts to bring about a terrible outcome or threatens such an outcome unless their demands are met and refuses to express their demands through democratic channels. Not all Realist DH theorists accept immoral coercion as a determinant of a DH problem. Hall (2022) for example contends that immoral coercion is not needed for dirty hands to arise. He holds that ‘one dirties one’s hands when one makes a good-faith choice between plural and conflicting values in a way that generates “residual moral claims” from other agents’ (Hall 2022, p. 4). Any morally charged decision that leaves one party with a justifiable grievance qualifies as dirty hands.⁸ However, room for manoeuvre and making alternative choices is simply much larger when there is no coercive threat involved. In a political compromise, one for instance has the option to compensate the grieving party by means of a complementary policy or decision to their advantage. There are different options for balancing the scale. Conversely, when immoral coercion is involved, the scales are already tipped at your disadvantage. Causing serious harm is not an option among many, but one that appears commensurate to the actions that caused the situation in the first place. The grounds for suspending moral conduct, its probability and its justifiability become eligible only because someone has extended a credible threat that calls for an appropriate response.

That said, the DH problem as we understand it differs from the concept of ‘rotten compromises’ introduced by Margalit. For him, a rotten compromise is ‘an agreement that establishes or maintains an inhuman political order’, whose victims are often not a party to the agreement (Margalit 2009, p. 41). For example, a compromise that abets crimes against humanity is rotten and should never be allowed (Ibid., p. 46). These compromises may or may not be the result of coercive threats. Their unacceptability is premised on the outcome of the agreement—a dehumanising regime—and not on the process. In our view, however, not only the evil nature of the act but also the element of coercive threat at the time of a decision renders it a ‘dirty’ decision. It not only results in immoral conduct, but it also originates from it.

Having established the difference in substance between the paradigmatic DH cases on the one hand and DH cases involving political compromises on the other, we now present an analytical framework that captures the nuances between different DH scenarios and the implications they carry for normative theorising. In the concluding section, we explain why the conceptual disambiguation between different types of DH scenarios matters for the way solutions to the DH problem are drawn.

Evaluating DH problems: a dual continuum framework

Our main argument starts from the observation that the difference between potential DH problems is not merely a matter of degree—from frequent to infrequent or from mundane to extreme (as de Wijze argues)—but primarily of substance. What is at stake differs from one situation to another, in two ways. On the one hand, there are substantial differences regarding the outcome of a DH compromise: what exactly is compromised, the extent of harm caused, but also what kind of harm is caused and to whom. On the other hand, DH cases also differ significantly with respect to their origins: the circumstances out of which a DH problem arises and the situation an agent is placed in. Our main argument is that DH problems faced by politicians are too diverse with respect to both of these dimensions to invite uniform solutions.

To illustrate, consider again two examples of the DH problem, the one typically used by Moralists, and the other by Realists. In the first case, the so-called ‘ticking-bomb’ scenario, a politician is informed that a bomb is planted in some public place and that the suspect is detained. The politician is called to authorise torture and other dehumanising treatment to extract information from the suspect that would save the lives of dozens. Doing so, however, compromises their moral integrity on a rights-based theory of morality, i.e. their duty to respect and protect others’ rights.⁹ For Realists (Bellamy 2012; Tillyris 2017), the typical example is one of political compromise. It involves the British Conservatives and the Liberal Democrats agreeing to form the 2010 coalition government. In doing so, they face the prospect of betraying their values and pre-election commitments. Not compromising at all is worse, however, as it would leave them out of government, thus completely unable to fulfil their political goals (in the present electoral term).

Both of these examples involve politicians who find themselves tempted to violate certain moral principles. In both cases, certain practical necessities urge them to do so, and both agents think that a moral transgression would serve the public interest best. Yet, there is a substantial difference as to which kind of moral principles they are tempted to violate. In the first case, by authorising torture, they will violate fundamental and universal moral standards, such as the human right not to be tortured. Any violation of human rights entails a compromise of the human dignity of other human beings and, as a consequence, of the moral integrity of the perpetrator. Protection of others’ rights is not only in accordance with rights-based moral theories but also with social morality more broadly.¹⁰ By contrast, forming a governing coalition with a political adversary does not involve an assault on anyone’s human dignity. It is a compromise on ideological beliefs and political priorities. To some extent, such compromises are expected in party-based representative democracies. In that sense, they are also more excusable. The compromise is less moral and more political in nature.

However, in between these two extremes—violation of universal moral standards v. ideological compromise—we find other possibilities in which a person’s integrity is compromised. For example, a politician lying to their constituents also loses some degree of moral integrity as honesty is also a socially shared principle. Breaking their campaign promise is another way in which one’s moral character may appear compromised, as it deducts from a politician’s trustworthiness and reliability. Above all, what is affected here is the politician’s quality as a representative, their representativeness. But because deceiving a constituency, sometimes unintentionally, is still much less harmful than failing to protect innocent lives,¹¹ the consequences of DH problems should be placed on a different scale.

A second important dimension affecting the quality and severity of a DH problem is the circumstances under which it arises. In our two examples, one agent is dealing with an uncollaborative terrorist using physical violence (a ticking bomb); the second agent is dealing with a political adversary, a professional politician whose job is to deliberate and make political deals. These circumstances are incommensurate as to the availability of options to act differently. In the first case, constraints in time and resources make it impossible to extract the information about where the bomb is hidden in other ways. At the same time, keeping one’s hands clean may result in failing to protect innocent civilians. In the second case, one has the option of putting on the table different proposals for coalition-building. In this case, keeping one’s hands clean would merely result in spending the term outside government, in opposition. The room for manoeuvre seems broader, and the conditions for negotiation are clearly less coercive.

As a result, our suggestion is to think of the DH problem as a dual continuum incorporating two axes. The first regards the origins of the dilemma, ranging from dilemmas triggered by political opportunism¹² to those caused by immoral or physical coercion. The second axis regards the outcome of a DH decision and ranges from losing credibility and representative credit to causing direct harm to one or more individuals. Moralists and Realists focus on the same continuum, yet their normative prescriptions differ because they concentrate on cases situated on opposite ends of the continuum. We argue against this practice, which at first sight only serves illustration purposes but ends up muddling the conceptual architecture of the entire debate. In doing so, we do not take sides in favour of either the Moralist or the Realist camp. Instead, we suggest a context-sensitive approach to replace the current one-size-fits-all responses (Table 1).

Table 1 Dual continuum: evaluating DH problems based on their origins and outcomes

To what extent a political agent could be excused (and thus may not need to exhibit remorse) depends primarily on the position of the DH problem on the above continuum. The more harm a moral transgression leads to, the less excusable it becomes. In addition, moral transgressions may become excusable when committed under extreme coercion but are less excusable under limited or no coercion. To illustrate, transgression A, which is not precipitated by a coercive event but is nevertheless causing major harm, would be impermissible and unjustified. Conversely, transgression B, precipitated by a coercive event and caused no severe harm, is in principle, justified. But when both of the two dimensions (coercion or harm) are simultaneously high or low, then we are entering more contentious terrain. In these cases, moral transgressions may not be justified but are nonetheless excusable under certain conditions.

How, for instance, should we evaluate moral transgressions (like C) that cause severe harm—and are therefore in principle not excusable—but are triggered by immoral coercion—which offers grounds for excuse? Moralists argue that these may be excused but only if the harm is acknowledged and appropriate reparations are paid to the victim(s) (de Wijze 2018; Nick 2022). The reparations are paid on behalf of both the responsible politician and citizens in whose interest the transgression was committed. This is a fair resolution. By using our dual continuum, however, it is possible to reach more nuanced views about when, why and with how much reservation a transgression is acceptable and can be forgiven depending on which exact spot on the continuum a DH dilemma occupies.

On the opposite side of the spectrum are DH problems (like D) triggered by narrow political goals. There is little to no coercion involved, and the resultant harm is often limited, uncertain and indirect. What we have is the reverse conditions compared to C: little harm—which makes the compromise excusable—yet little coercion, too—which makes the compromise less excusable. Of course, the price of the moral transgression is automatically paid by the loss of credibility and representativeness. Against this background, the Realist prescription to not plead guilty makes sense, for the price is already extracted. But mirroring the resolution offered by Moralists and building on our dual continuum framework, we think such political compromise can be excused only if the agent offers valid reasons for treating the situation as an emergency, i.e. for why they felt coerced to agree to a shady deal. Unless such reasons are publicly provided and well-grounded, others can rightly question why the agent acted the way they did; in this case, even with little to no concrete harm caused to individuals, the compromise is unacceptable.

In brief, when the conditions of ‘severe harm’ and ‘no coercion’ apply, an act is in principle, unjustifiable. Conversely, when there is ‘no severe harm’ and ‘coercion’, an act is in principle, justifiable. When these conditions are present in different combinations than the above (severe harm + coercion as in the ticking bomb scenario, or no harm + no coercion as in most cases of political coalition formation), an act is by all likelihood excusable. In this case, the higher the value of the condition that applies (e.g. the more severe the harm or the less obvious the coercion), the more and higher reparations are required. With the dual continuum in mind, DH cases can be interpreted and evaluated on their own terms, yet still drawing on a coherent and thorough standard of political ethics.

Conclusion

DH problems vary depending on their circumstances and their outcomes. Some arise when a politician is compelled to act in a manner inconsistent with human rights principles. We have called these the extreme or paradigmatic Dirty Hands (DH) problems. And then there are DH problems in which political representatives need to sell out on some of their principles or promises in order to reach a favourable agreement—these are DH problems involving compromise of ideological beliefs and political commitments. DH problems resulting from coercion and involving grave harm and those triggered by political goals and causing no direct harm are governed by different contextual criteria. Failing to distinguish between them leads to unproductive theoretical debate, where conflicting claims about what politicians ought to do and how we should hold them accountable result from incommensurate premises rather than from different moral standards.

As an antidote to this conceptual confusion, we have proposed a dual continuum that distinguishes between different DH problems based on their origins and outcomes. It also aligns with, and where necessary corrects, the different normative prescriptions put forward by Moralist and Realist DH theorists. The answer to the question of how we should hold politicians accountable will depend on the nature of the DH problem under consideration and its position on this dual continuum. An outcome involving serious harm may be excusable if it is caused under extreme coercion and only after reparations have been paid to those affected. On the other hand, DH problems involving compromises on campaign promises or policy proposals occur under little or no coercion; they too can be excused as long as they result in less serious harm, but only if the reasons for the compromise are publicly explained and accepted. The key to unlocking a DH problem lies in the particular way in which its constitutive elements are manifested and combined in each individual case. We hope our dual continuum will provide a useful tool for scholars and practitioners to better understand, assess and resolve DH problems.

In our view, there is no one-size-fits-all approach to the DH problem and to questions of political ethics more broadly. Certainly, we need pre-established standards for evaluating them to avoid ad hoc solutions. However, individual cases vary considerably in terms of which conditions apply, how strongly they apply, and how they are combined. Evaluation frameworks should take account of this diversity, be broad enough to cover a whole phenomenon, and yet provide specific advice adapted to the conditions that apply in each case.