1 Introduction

Paradigmatically, ‘dirty hands’ scenarios confront holders of public office (Walzer 1973; Williams 1981; Nagel 2012). A public officer is placed in the invidious position where the course of action that, putatively, they should take for the sake of the greater public good also requires committing a grave moral wrong. If the public officer accords with the former imperative, so the thesis goes, then they act in a praiseworthy fashion, but at the same time will be left with a ‘moral remainder’ – with ‘dirt’ on their hands. Some of those who accept this seemingly paradoxical possibility, have argued that a further problem arises as a function of contemporary democratic governance (Archard 2013; de Wijze 2018; Thalos 2018; Hollis 1982; Levy 2007; Thompson 1987; Nick 2019). At least within a democracy, public officers can be rightly said to act in this way on our behalf, and with our support. Thus, they argue, the dirt of public officers’ hands is also shared with the public. This has been termed the problem of ‘democratic dirty hands’ (Thompson 1987). Even more troubling, so the argument goes, since dirty hands are necessary for public officers in any stable political order, then such democratic dirty hands are necessary for any stable democracy. There is no option to live as a democratic citizen, at least over the long run, without being implicated in the grave moral wrongs of dirty hands. Our dirt is the unavoidable cost of democratic citizenship.

In this paper, I offer an argument against this disconcerting conclusion.Footnote 1 My central claim is that proponents of ‘democratic dirty hands’ have missed the import of another feature of contemporary governance: public institutions. Public institutions, as organisational agents, intermediate the relationship between public officer and public; and in so doing, the dirt necessary for stability may be ‘laundered’: the public may still gain the benefit of a public officer’s hands, but remain clean of their dirt.

First, I introduce the case of La Comisión Internacional contra la Impunidad en Guatemala (‘CICIG’), as a scenario for sustained illustration throughout the paper. Second, I define ‘dirty hands’ for instant purposes; and sketch how the case of CICIG may be understood to fit within its confines. Third, I outline the case for ‘democratic dirty hands’, and make a distinction between two possible mechanisms of dirt transmission often conflated in the literature: agency and complicity. Fourth, I turn to the import of public institutions, or more accurately, public ‘organisational’ institutions, arguing that in contemporary governance they typically stand between the public and public officers: the public stands as principal to such institutions as agents, and such institutions in turn stand as principals to public officers as agents. Fifth, I leverage this fact to show how, even if institutions themselves are granted a sufficiently wide scope of authorisation to commit ‘dirty’ acts on behalf of the public, the mandate that they in turn typically grant their own public officers is too narrow to include dirty hands activities. This does not exclude such public officers acting with dirty hands. But being unauthorised it does prevent the transmission of their dirt to the institution, and thereby onto the public. The dirt is, we might say, ‘laundered’. Sixth, I note that such successful laundering requires excluding institutional ratification and negligence. Finally, I argue that where dirty hands are uncharacteristic of an institution, then the public’s support of that institution cannot constitute complicity in them. In this way, whilst by no means conceptually impossible, democratic dirty hands is, at least to a large extent, an avoidable feature of a well-functioning democracy marked by institutions, if not always public officers, of integrity.

2 The Case of CICIG

In 2017, Iván Velásquez Gómez was the Commissioner of La Comisión Internacional contra la Impunidad en Guatemala (‘CICIG’ – The International Commission Against Impunity in Guatemala). CICIG was established in 2007 by agreement between the Guatemalan government and the United Nations (Agreement between the United Nations and the State of Guatemala on the establishment of an International Commission against Impunity in Guatemala, UN-Guatemala, 04 September 2007, UNTS v. 2472, I-44,373, the ‘Agreement’). It was a unique ‘hybrid’ international criminal justice mechanism (Zamora 2019; Roht-Arriaza 2008) established to ‘support, strengthen and assist’ the state institutions of Guatemala to identify, dismantle and prosecute members of ‘illegal security groups and clandestine security organisations’ (Agreement, Art. 1(a)). The latter were defined as groups committing illegal acts that compromised the ‘full enjoyment and exercise of civil and political rights’ and were ‘linked directly or indirectly to agents of the State or have the capacity to generate impunity for their illegal actions’ (Agreement, Art. 1(d)). Operationally, this entailed a focus upon combatting corruption (Zamora 2019; Call and Hallock 2020). In pursuit of such purposes, it had powers to investigate these illegal groups, promote their prosecutions, and offer advice and policy recommendations to other State bodies (Agreement, Art. 3(a), (b), (c), (l)). Indeed, it was empowered to ‘[t]ake all such measures it may deem necessary for the discharge of its mandate, subject to and in accordance with the provisions of the Guatemalan Constitution’ (Agreement, Art. 3(k)). Many of the staff of CICIG were recruited internationally. Ivan Velásquez Gomez, himself, was a member of the Colombian Supreme Court and appointed by the Secretary-General of the United Nations (Agreement, Art. 5(1)(a)). Funding and oversight were also provided by the United Nations (Agreement, Art 5(1)(a), (c), Art 7(1)). Thus, CICIG was able to maintain the level of independence that cognate institutions in other developing countries lacked (Zamora 2019). Indeed, such independence became the defining feature of CICIG both for the public and commentators. ‘In a nutshell, [CICIG] aimed to tackle the problem of weak prosecutorial independence, knowing that the failure to conduct criminal investigations was the gateway to impunity’ (Michel 2021: 58). In turn CICIG reinforced this perception in its public communications: ‘Is CICIG independent? Yes. From the political, organizational and financial spheres’ (CICIG 2018).

By 2017, CICIG was an unprecedented success (Open Society Justice 2016; Michel 2021; WOLA 2015; International Crisis Group 2016; Zamora 2019: 537-9; Trejo and Nieto-Matiz 2022). It facilitated filing of more than 120 cases in the justice system, implicating more than 1540 people, with approximately 200 government officers among those facing charges, and over 70 illicit networks damaged or dismantled (WOLA 2019; CICIG 2019; Call and Hallock 2020: 3). Most spectacularly, in 2015 it had uncovered an intricate corruption network in which the President of Guatemala, Otto Pérez Molina, as well as his Vice-President and other high-level officers were implicated in a scheme to defraud the state of millions of dollars of import duty revenue. This led to charges against the President and Vice-President, their resignations, a widespread citizen’s movement demanding structrual change – a ‘Guatemalan Spring’ – and, ultimately, the election of a new leader, Jimmy Morales, running on an explicitly anti-corruption platform (Call and Hallock 2020: 28–32, 36). By 2017, 70% of the population had confidence in CICIG. Guatemala’s homicide rate had almost halved since 2009, and the reported impunity rate for violent crimes had dropped from 98 to 87% (International Crisis Group 2018). CICIG is thought to have been a dominant factor in causing this phenomenon (Trejo and Nieto-Matiz 2022).

CICIG, however, had one vulnerability to the vagaries of the Guatemalan political system. Its mandate had to be renewed every two years by the President of Guatemala via an exchange of letters with the United Nations Secretary-General (Agreement, Art. 14). In the recent election, Morales had been elected promising to extend CICIG’s mandate in 2019 to 2021, and thus beyond the end of his own term as President, finishing in 2020. However, in January 2017, entirely incidental to another investigation, CICIG came across evidence that the President’s son and brother had misappropriated funds totalling approximately $14,000 USD. It was clear that the President would make every effort to stop such an investigation if it went ahead, including dissolving CICIG itself.

Velásquez, therefore, was caught in the following dilemma. On the one hand, he led a Commission with the explicit aim of stopping the impunity that arises for politicians, and their family and associates, from just this kind of access and proximity to power. Furthermore, acting independently of the political sphere in doing so had become its defining commitment to the public. On the other hand, the misappropriation of $14,000 USD was trivial compared to the focus of CICIG’s operations including large-scale multi-million-dollar corruption schemes, and extra-judicial murder, let alone the further beneficial effects that such operations were having for the quality of Guatemalan governance. Yet such consequences were at real risk if Velásquez commenced the investigation.

Let us begin by asking: is this a ‘dirty hands’ scenario?

3 Defining Dirty Hands

We might readily accept that ‘dirty hands’ scenarios always involve ‘the choice between upholding a moral principle and avoiding a looming disaster’ (Walzer 1973). But a more formal set of defining necessary and sufficient conditions has so far eluded theorists. Stephen De Wijze, however, usefully sets out six features that, with some qualification below, suffice to pin the concept for the purposes of this paper. On De Wijze’s view, a ‘dirty hands scenario’ is: (1) an unavoidable genuine moral conflict for an agent; (2) where a course of action requires a justified violation of persons or cherished values; (3) undertaking that course of action would give rise to a moral ‘remainder’ or ‘residue’ for the agent; (4) the primary motivation of the actor to commit the justified moral violation is constituted by strong moral considerations; (5) the moral violation is aimed at bringing about the lesser evil; and, (6) such a scenario arises more frequently, and often with greater urgency and import, in the political domain (de Wijze 2018: 132).

Applying these conditions, we can cast the case of Velásquez and CICIG as a dirty hands scenario. (1) There is an unavoidable genuine moral conflict, and further it has the kind of characteristic ‘principle versus consequences’ flavour that Walzer describes: on the one hand, there is keeping CICIG’s commitment to the apolitical investigation of corruption, on other hand, there is the very ongoing existence of CICIG as an institution, its far more significant anti-corruption and impunity operations, and the positive governance consequences of those operations. (2) CICIG’s commitment to apolitical investigation is, in the relevant sense, ‘cherished’ or at least ‘strong’ (Nick 2021: 196) and fundamental to its identity, and a violation of such a commitment would be tantamount to a kind of betrayal (de Wijze 2007: 12). (3) It seems plausible that if Velásquez and/or CICIG were to violate such a commitment, by shelving the investigation in order to manage the politics around CICIG’s mandate renewal, then some kind of moral ‘residue’ or ‘remainder’ would properly arise, whether that be indeed blameworthiness as Michael Walzer argues, or at least some other ‘uncancelled moral disagreeableness’ (Williams 1981: 63) (more on this below). (4) We may assume, however, that the primary motivation for such a violation would be the beneficial consequences of CICIG’s continuance, and indeed (5) the violation itself would be a ‘lesser evil’, at least on the scales of consequentialism. Finally, (6) the scenario arises with great import and indeed urgency, in what might be considered the ‘political’ domain in a broad sense (more on this, too, below).

Given this fit, however, let us also note three further points. First, ‘politicians’ have typically been cast as the central subjects of dirty hands scenarios, and indeed some have argued that such scenarios arise (or indeed only arise) because of the demands of public office (Hampshire 1989; Nagel 2012; Thompson 1987; Gowans 1994; Williams 1981; Bellamy 2010). However, it is little noted that (a) not all politicians hold public office, since one typically has to politick in order to gain office (Machiavelli 1984), and (b) not all, or indeed most, public office-holders are politicians, at least in the paradigmatic sense that they are openly partisan actors within the system engaged in open competition for control of the levers of public power and authority, with the aim of pushing their own ideological view (Philp 2007: 1). Velásquez, and indeed all the constituent members of CICIG, are non-political public officers in this sense.

Second, dirty hands scenarios are typically cast as conflicts between public imperatives on the one hand, and personal convictions (or principles of private morality) on the other (Thalos 2018: 176; Nagel 2012; Waldron 2018; Hampshire 1978; cf. Kim 2016). For example, in classic ‘torture scenarios’, one is caught between the public imperative of ‘saving the world’, and the personal conviction that torturing any human being is wrong. In Weberian form, the real challenge is about how one can live with oneself qua private person, having performed the public act (Weber 1994). However, this does not well-characterise Velásquez’s scenario at all. The relevant moral conflict arises between public imperatives (cf. Waldron 2018: 230; Hollis 1982: 393). On the one hand, there is a duty, attached to his office, of being apolitical in his decision-making. On the other hand, there are the aims of pursuing and preventing impunity from crimes in Guatemala in general, and sustaining the institution of CICIG in order to do that. These latter aims are also attached to Velásquez’s office. Indeed, the scenario remains one of dirty hands even if, personally, and indeed as a foreigner, Velásquez has no personal loyalty to Guatemala, but remains properly concerned with being a professional, that is, executing his role well and with ‘public integrity’ (Kirby 2020, 2021).

Third, as implied above, there is disagreement within the literature as to the exact nature of the ‘moral residue’ – the dirt – left by dirty hands action, that is, what ought to be our moral appraisal of such an action, or at least the appropriate ‘reactive attitude’. Whilst I myself am inclined to follow Walzer in seeing such ‘dirt’ as blameworthiness that at the same time comes with praiseworthiness, giving dirty hands its paradoxical quality (Walzer 1973), the argument in this paper is consistent with most other conceptions, whether that be agent-regret (Roadevin 2019: 135), shame (Stocker 1990: 9), lingering sense of wrongness (Thompson 1987: 13), guilt (Levy 2007: 49–50); distaste (Thalos 2018; Waldron 2018: 226), or ‘tragic-remorse’ qua a recognition of genuine wrong-doing, but also the bravery and good done (De Wijze 2005; de Wijze 2013). What is important for the argument of this paper, however, is that whatever one’s conception of the ‘dirt’ of dirty hands, mere consequential liability for the wrongful action does not suffice. To be precise, an individual is ‘consequentially liable’ for a wrong, if they bear obligations to compensate or otherwise ensure reparative actions for the wrong.Footnote 2 Paradigmatically, we expect such consequential liability to follow from culpability, that is, where one is to blame for the wrong, then one has reparative duties with respect to that wrong. However, the two kinds of responsibility, often come apart: a professional insurer may be liable to compensate for the negligence of a doctor but not to blame for their malpractice; a parent might be liable to clean up after their child but not to blame for their mess; indeed, I might even be liable in tort to compensate for property damage performed whilst mentally incapacitated, but not open to blame under criminal law.Footnote 3 Further, where blame typically grounds the propriety of punishment, condemnation and/or disesteem, mere consequential liability without culpability precludes the propriety of punishment, condemnation and/or disesteem. So my point, in the instant case, is that even if one does not think that the ‘dirt’ of dirty hands is blameworthiness per se, it must be something that – like blame – is more than mere consequential liability. Another way put: dirty hands presupposes the attribution of the relevant ‘dirty’ action to the actor as something that they are in some way morally accountable for – whether that be blame or some other evaluative and/or affective response. Dirty hands cannot arise merely via the attribution of reparative duties with respect to that dirty action. We might say: dirty hands requires the propriety of morally owning the action, not merely owning the consequences.

4 Two Mechanisms of Democratic Dirty Hands: Agency and Complicity

Let us assume, therefore, that Velásquez faces a dirty hands scenario so defined. This sets us in a position to address our central question: under what conditions does a dirty hands scenario, such as Velasquez’s, also constitute a case of democratic dirty hands? ‘Democratic dirty hands’ putatively arises where, because of some feature of democracy, the public shares the dirt of their public officers (Thompson 1987; Bellamy 2010: 427; Hollis 1982: 389, 396; Levy 2007: 48) – or indeed, as some argue, it is fully transferred to them (Thalos 2018). As a result, we too are ‘polluted’ and ‘sullied’ by, ‘responsible’ for and must own the dirty hands actions of our public agents. This is important, not merely for the sake of simple honesty about our moral position within a democracy, but also for evaluating our own standing as a public with respect to the ‘dirty’ public officers. If we too are dirty, then prima facie we lack standing to punish, or perhaps even blame, such public officers for their actions (Thompson 1987: 18; Roadevin 2019: 134). To do so would be, to say the least, hypocritical (Todd 2019; cf. Bellamy 2010: 426-7; Hollis 1982: 396), if not tantamount to suppressing our own guilt (Levy 2007: 40).

Despite the growing chorus of support for the possibility, actuality and, indeed, prevalence of, such ‘democratic dirty hands’ there is, however, ambiguity with respect to the precise feature of democracy that supposedly grounds it. Indeed, there has been a tendency to conflate two possible mechanisms: agency and complicity. It is important to distinguish these two mechanisms since, as I shall explore, one might obtain without the other.

First, most proponents of democratic dirty hands claim that the feature that grounds the transfer of dirt between public officer and their public is the existence of a kind of agency relationship, that is, the public officer stands qua ‘agent’ to the public qua ‘principal’ (Hollis 1982: 396; Levy 2007: 48; Thompson 1987: 18; Archard 2013: 781; Thalos 2018: 177; see also, Beerbohm 2018: 5; Roadevin 2019: 133-4; de Wijze 2018: 134). Whilst these concepts are used in a range of cognate ways across disciplines, in the immediate context they are taken to denote a relationship known primarily to law where one actor – the ‘agent’ – is authorised to act on behalf of (or ‘in the name of’) another – the ‘principal’ (Restatement (Third) of Agency (2007) § 1). In this capacity, they may be said in Thomas Hobbes’ terminology to ‘personate’ the principal, such that their acts may be attributed to the principal as if they had performed them herself (Hobbes 1996: XVI.1; Miller 2018: 37; see also Waldron 2018: 230, n 31). The relevant act of ‘authorisation’ of this putative agency relationship is, according to these theorists, regular elections (and/or other forms of participation) by which public agents and/or their policies are ‘authorized’ (Beerbohm 2018: 5), ‘endorsed’ (de Wijze 2018: 130), or given our ‘consent’ (Thompson 1987: 18).

At law, such agency relationships are a particular species of fiduciary relationship (Restatement (Third) of Agency (2007) § 1.01). Whilst in all fiduciary relationships an actor – the ‘fiduciary’ – is empowered to act on behalf of another – a ‘beneficiary’ – in the ‘personation’ sense described, the agency relationship is distinguished from other fiduciary relationships by the principal’s control over their agent (DeMott 2019: 321). This control is manifest in the need for the principal’s consent to the relationship; and their ability to instruct the agent. The agent’s defining duty is to carry out these instructions. They set the scope of the agent’s authorisation to act on behalf of their principal (Penner 2020: 129). In turn, and most important for instant purposes, these elements of the principal’s (a) consent, (b) control and (c) instructions ground the possibility of the personated principal being attributed not merely the agent’s rightful but also wrongful actions.Footnote 4 Furthermore, if such actions are performed within the scope of an agent’s authorisation,Footnote 5 then the principal not merely bears the consequential responsibility of their agent’s actions but also shares culpability. The principal is treated as if they themselves committed the wrong, and hence is liable to punitive damages (Restatement (Second) of Torts (1977) § 909, see also (Parlee 1984; Sturley 2010), or, indeed, if relevant, in many jurisdictions, criminal sanction (Pieth and Ivory 2011). The rationale is that culpability is appropriate because the principal’s consent, control and instructions are sufficient to establish that they have willed the act, or at least voluntarily accepted that it might done, in their name. This is termed ‘direct liability’ (Restatement (Third) of Agency, § 7.04; Dal Pont 2020: 22.3–22.6).

It is this feature of the agency relationship that proponents of democratic dirty hands ultimately need to avail themselves of. It is such direct liability that can transmit not merely consequential liability, but also dirt – qua culpability (or whatever one considers to constitute the moral remainder) – of public officers – qua agents – to the public – qua principal.

Of course, at this point, some might question whether any democratic relationship between public and public officers, that actually or could even possibly exist, would fulfil anything analogous to the conditions of consent, control and instruction required for the principal-agent relationship so defined. We do not actually consent to the relationship as individuals, we cannot as individuals or even as a group deliver anything like the kind of binding instructions to our public officers, (elections are, arguably, better understood as opportunities to change the identity of our public fiduciaries, on the basis of their tentative policy preferences, rather than a set of binding instructions on policy), and our formal mechanisms of control between elections are largely limited to administrative and constitutional law rather than the kind of operational control typical of agents (cf. Roadevin 2019: 134; Thalos 2018: 182). However, I put this style of objection to one side in this paper. I concede, arguendo, that sufficient consent, control and instructions may exist to establish that the public may be treated as a ‘principal’ and that individual ‘public officers’ as agents. My argument, instead, will just be that, in general, institutions intermediate any such relationship: the public are, if anything, a principal to public institutions; and public institutions are, almost invariably, principals to public officers.

But before turning to this argument, even if one dismisses the very existence of agency relations in democratic contexts, the threat of ‘democratic dirty hands’ does not entirely disappear. This is because, whilst the agency mechanism is undoubtedly the primary mechanism of dirt ‘transmission’ suggested by its discussants, there is an alternative possible mechanism sometimes alluded to: complicity (Beerbohm 2012: 226; de Wijze 2018; Archard 2013; Roadevin 2019: 134).

Complicity is where one actor – the ‘accessory’ – shares the culpability for the wrongful actions of another – the ‘principal’. In one classic legal formula, complicity rises where one acts to ‘aid, abet, counsel or procure’ the relevant act of another (Accessories and Abbettors Act 1861 (UK), s 8). For our purposes, however, we might readily draw upon the growing normative literature, where complicity is construed more broadly as some form of participation in the wrong of another; a form of ‘support’, ‘encouragement’ or ‘enhancement’, for which one is in some relevant way responsible (Pauer-Studer 2018). One’s complicity, and hence culpability, may vary in degree with the fulsomeness of one’s support (see de Wijze 2018: 138). However, most importantly, complicity requires something more than mere participation. Typically, this is taken to be ‘intentional participation’ (Kutz 2000: 138) or ‘knowing contribution’ (Lepora and Goodin 2013: 81 − 3).

Unlike the relationship of principal to agent, complicity does not presuppose that the accessory authorises the principal to act on her behalf, or ‘in her name’, let alone that she controls or instructs them. Indeed, this should be apparent by the fact that the so-called ‘principal’ in the case of complicity is the one committing the act, rather than being the one attributed the act of another as in the agency relationship. The contrast is further made by the paradigmatic case of both direct liability and complicity. At law the paradigmatic case of direct liability in an agency relationship is where an organisation (e.g. an employer) is directly liable for the actions of their organisational member (e.g. employee). By contrast, a paradigmatic case of complicity is where an organisational member (e.g. employee) is held to be complicit in the collective activities of the organisation (e.g. employer) (Pauer-Studer 2018; Kutz 2000; Lepora and Goodin 2013). The lines of transmission of culpability are running in opposite directions, even if the employer is termed ‘principal’ in both contexts.

So, being distinct from the agency relationship, it is therefore open to argue that the public are complicit in the dirty hands of public officers, even if they do not consent, control or indeed instruct them to do so. This alternative argument for democratic dirty hands is less developed in the literature, in part because it often seems to be confused with the mechanism of authorisation (e.g. Archard 2013: 786; Beerbohm 2018). But the argument might be sketched as follows.

On the one hand, it is clear that, at least within a democracy, citizens act in ways that voluntarily – that is, not compelled by binding directions or coercion – support their own governance structures, and the public officers within them. Most obviously, at least where voting is non-compulsory, we voluntarily participate in processes for choosing who some of those public officers will be through elections. We might also offer support to such actors via voice, resources, time and money given to their parties, causes or policies. We might engage with such actors as state contractors, participants in deliberative processes, serving as an employee, or offering expertise, information and advice. We might offer ‘robust’ support, in the form of an expressed willingness, if necessary, to lobby, protest or even fight in defence of such structures and public officers. In this way, one might think we generally ‘participate’, directly or indirectly, in facilitating the ongoing ability of public officers to act with dirty hands, even if this does not always pass the stricter test of ‘but-for’ causation (see, (Kutz 2000). On the other hand, we know (or should know) that so supported public officers will act with dirty hands. Indeed, if we accept the Machiavellian realist argument, then we must appreciate that it will be necessary to preserve order itself (Machiavelli 1984: 45 − 6; Philp 2007). As such, so the argument goes, we can be taken to accept such acts as a consequence of the support we offer such public officers. We might not intend their acts, but we intend to play a role in placing them in a position where we can expect that they will so act. By virtue of this fact, therefore, it is thought we must share any dirt arising.

I argue in § 7 that the existence of public institutions can disrupt this secondary complicity mechanism of dirt transition too. But first, let us focus upon the alleged primary agency mechanism § 4–6.

5 Institutional Agency, Institutional Dirty Hands?

Theorists of democratic dirty hands often cast themselves as updating for contemporary governance a doctrine developed by Niccolò Machiavelli for 16th Century Italian city-state princes: democracy itself being the salient innovation (Thompson 1987: 11; Shugarman 2000: 231-2). However, contemporary governance also distinguishes itself by the centrality of public institutions. What exactly do I mean by an ‘institution’? I am using ‘institution’ in the ‘organisational’ sense. Institutions that qualify as ‘organisations’ are more than mere structures of actions and norms that sustain themselves over time (Hodgson 2007: 98). They are structures that define a collective (or group): determining membership, and roles for such members within the collective. In turn, such members are then ordered by internal rules, authority structures and decision-making procedures (Isaacs 2011; French 1979, 19841996); ‘rules’ for short. Such rules enable the determination of policies that each member of the collective is obliged to accept as joint aims of the collective, and thus as something that, ceteris paribus, each member has reason – a reason of office – to facilitate the collective achieving (French 1995: 31; Arnold 2006: 289). On this view, many public structures of actions and norms qualify as institutions in the ‘organisational’ sense: the Department of Health, the House of Lords, the Central Bank, the Cabinet, or indeed, in the present case CICIG.

Now, I shall make an assumption that some may take as controversial, but rests upon a long, growing, and arguably now dominant line of theory: organisational institutions can be moral agents, in the sense that they are able to act intentionally on the basis of moral reasons (Isaacs 2011; French 1979, 19841996; List and Pettit 2011; Byerly and Byerly 2016; Lovett and Riedener 2021; Collins 2019; Pasternak 2017). Different theorists aim to substantiate this claim in different ways, but a standard view might go as follows. When policies are made in accordance with the rules of the collective, they may be considered the ‘intentions’ of the collective. When those rules authorise members actions to facilitate the realisation of such policies by members – whether individually or jointly – then those acts may be considered the intentional ‘acts’ of the collective. When decision-making process constituted by such rules permit conceptions of what such intentional acts of the collective morally should aim to do as the basis of the decision, then the collective may be considered able to act intentionally on the basis of moral reasons. So able, on this kind of view, the collective can be said to have moral agency,Footnote 6 and thus be morally responsible for its (in)actions. And as an agent, the collective qua organisational institution will find its moral position framed by collective commitments, as well as duties, permissions, and other deontic incidents (cf. Lovett and Riedener 2021; Collins 2019; Pasternak 2017).

Returning to the domain of contemporary governance then, in contrast with Machiavelli’s time, it is more accurate to say that we are governed by institutions rather than by individuals. Laws are made by Parliament, Congress or the Supreme Court. They are administered by the police force; the Department for Social Services; or Home Office. And promises, contracts and representations, are made by such institutions to stakeholders, whether that be individual citizens, contractors, other government bodies, or the public at large. Public officers, as individuals, and including politicians, as a rule act as members of such public institutions. They act on their behalf; and those public institutions, in turn, act on behalf of the public, or indeed, other superordinate institutions that do (Miller 2018: 43 − 4).

To illustrate via our case study. We can assume that CICIG is one of these institutions qua an organisational moral agent. This is consistent with, although not determined by, the fact that the Agreement grants CICIG personality at law (Agreement, Art 4.). It also fits the language in which I have, inevitably, sometimes fallen into above. It is first and foremost CICIG as an organisation, not Velásquez, that has the commitment, owed to the Guatemalan people, to be independent. By virtue of such a commitment CICIG, as an organisation, is duty bound to pursue the investigation of President’s son and brother. And it is also first and foremost CICIG as an organisation, not Velásquez, that, by virtue of its fundamental governing purpose, has an underlying imperative to pursue the elimination of impunity in Guatemala more generally. Indeed, and this is the fundamental point, it is CICIG that first and foremost (at least relative to Velásquez) acts on behalf of the Guatemalan public.Footnote 7 Velásquez, in turn acts on behalf of CICIG acting on behalf of the Guatemalan public. He ‘represents’ CICIG (Agreement, Art 5(1)(a)). To use the Hobbesian language of ‘personation’ introduced above, Velásquez personates CICIG which, in turn, personates the Guatemalan public.

Is this chain of personation relationships also a chain of agency relationships? For instant purposes, subject to the possible objections noted above (§ 3), let us assume so. The Guatemalan public stands as principal to CICIG as agent, and CICIG stands as principal to Velásquez as agent. Velásquez is authorised to act on behalf of CICIG at its consent, within its control, and subject to its instructions. This may sound somewhat odd given that Velásquez is the leader of CICIG. But distance remains between an institution as principal and its leader as agent. The institution’s decision-making mechanisms, and its rules, do not collapse into his whim, instead he is subject to them, even though he may play an outsized role in the process of making them. Velásquez is subject to oversight mechanisms (although note Open Society Justice 2016: 8), the opinion of other Commission members, the founding document of the body, internal management policies, binding commitments made by other agents of CICIG, acting on its behalf, and so on. At least at law, such a relation of corporate leader to incorporated body is a, if not the, paradigmatic principal-agent relationship. Acting as agent for the institution as principal, Velásquez still owes it to the public to act independently and to pursue the elimination of human rights abuses, corruption and impunity in Guatemala in general. However, he only owes these obligations to the public as CICIG’s agent, acting on its behalf.Footnote 8

If this is correct, however, then it follows that not merely Velásquez, but also CICIG itself faces a dirty hands scenario: it must either act on principle – its foundational commitment to apolitical action – and risk its ongoing and vital contribution to Guatemalan governance; or compromise principle in order to best promote that contribution. Further, assuming the public stands as principal to CICIG as agent, and CICIG as principal to Velásquez as agent, then, prima facie, the principle of direct liability – one might think – would apply all the way down the chain. Has our argument so far done nothing then but confirm current ‘democratic dirty hands’ theory by way of more complicated means? Not if we now return to the principles of direct liability.

6 The Institutional Laundry

At law, a principal is not directly liable for all possible acts performed by their agent, or even all performed in their name. In the main, a principal is only directly liable where the agent acts within the scope of their authorisation, that is, their ‘actual authority’ (Third Restatement, § 7.04). The scope of an agent’s authorisation is set by their instructions (or ‘manifestations’: Third Restatement, § 2.02, see (DeMott 2019)). Instructions themselves may be explicit, involving imperatives, permissions or purposive statements: Do X! You may do X. Our purpose is to do X. They can be particular commands to particular agents, or general policies applying to a set of agents. Instructions can also be implicit, constituted by the norms established by practices and processes, particularly within an institutional context. Either way, the scope of such instruction will also be taken to include what might be considered reasonably necessary or incidental to what is instructed. But this presumed scope may, in turn, be limited by other instructions. These limits, themselves, may be express, constituted by prohibitions articulated by directions or policies made by the principal: do X, Y or X, but not if it leads to G. Or those limits may be implied from the overall purposes of the relationship, or a defeasible presumption that the principal does not authorise a breach of general law. Subject to some qualifications (see § 6), where an agent’s action lies outside of the scope of their authorisation, then the principal is not directly liable for any wrong so committed. The principal may still be held consequentially (or ‘vicariously’) liable where the wrong arises in the agent’s wider scope of ‘employment’ by the principal.Footnote 9 But the principal cannot be held culpable (Morgan 2012: 617; Dal Pont 2020: 22.3). This reflects the fact that, having not authorised the agent’s action, they cannot be said to have willed, intended or otherwise accepted it might be performed on their behalf.Footnote 10

Theorists of democratic dirty hands are aware that a moral principle, akin to these legal principles, must also constrain the scope of any public agent’s authorisation to commit dirty hands on behalf of the public. Absent the intermediation of any institution as discussed above, however, they tend to see this scope as incredibly broad. The basic picture is that citizens elect individual public officers (‘politicians’), who are thereby authorised to act on our behalf as our agents. Our ‘instructions’ to such agents are of two kinds: purposive and constraining. On the one hand, there is a fundamental purposive direction to pursue the overall public interest, or the public’s ‘will’: (cf. Levy 2007: 48; Thompson 1987: 22). On the other hand, there are some basic constraints to the pursuit of that purpose, such as, the existing law, constitutional provisions, human rights, fundamental principles, and/or the bare constraints of legitimacy (Thompson 1987: 22; Levy 2007: 48; Thalos 2018: 179). We then maintain control over these public officers via elections, other forms of accountability, and the ability to reverse and/or disapprove of policies thereby.Footnote 11 Furthermore, it is assumed that it is well known that it is often reasonably necessary and incidental to the pursuit of the purpose(s) of public office that one must sometimes act with dirty hands, and hence dirty hands falls within its implied scope. As Neil Levy puts it in short, ‘[o]ur politicians are our agents, authorized by us to carry out policies for our benefit. Moreover, it is common knowledge that these actions are often grubby, and sometimes downright dirty. We ask them to get their hands dirty on our behalf’ (Levy 2007: 48).

We can now say, however, that the problem with this picture is the absence of institutions. If the public can be construed as granting such a wide scope of authorisation within an agency relationship, then they grant it to their public institutions not to the public officers within them. Even politicians, who are elected directly by us, are only elected to serve in and as members of such institutions. It is these institutions that, in turn, determine their own scope of authorisation to such public officers to act on their behalf by their policies, practices and processes. And the scope of that authorisation to public officers can be far narrower than what institutions might themselves have been granted by the public.

To illustrate. Let us return to CICIG. The scope of CICIG’s authorisation, we might assume, is broad. It has its purposive instructions: to ‘support, strengthen and assist’ the State institutions of Guatemala in identifying, dismantling and prosecuting members of illicit security groups that tend to enjoy functional immunity for their criminal activities (Agreement, Art. 1(a)), including (Agreement, Art. 1). It also has a broad sweep of powers that, bar some minimal constraints, permit it to ‘[t]ake all such measures it may deem necessary for the discharge of its mandate, subject to and in accordance with the provisions of the Guatemalan Constitution’ (Agreement, Art. 3(k)). CICIG, as an organisation, within the scope of this authorisation, has created for itself a clear policy, manifest in its public proclamations, internal messaging and practice, that it will maintain a kind of independence from the political sphere that requires investigating President Jimmy Morales given the evidence. But CICIG could always change, repeal or simply decide to breach its own policy, even if just internally, even secretly. Acting with dirty hands, in this case, we might assume does lie within the scope of CICIG’s authorisation from the public. However, and this is the all-important point, without such a change in policy, it does not lie ex ante within the scope of Velásquez’s authorisation from CICIG to so act himself. Whilst Velásquez is the leader of CICIG, he is not CICIG. He must initiate a change in CICIG’s policy before he can act on behalf of CICIG within the scope of that new policy. So this suggests not just one but two possible ways in which Velásquez could act with dirty hands.

The first option is that Velásquez decides that CICIG should cease the investigation. He then convenes the senior leadership team. They discuss options, and collectively form a decision, as policy, not to move forward, and do so specifically because they understand the two horns of the dirty hands dilemma. They do not proclaim the decision to the world – that would defeat the point. But it is still a decision, even if secretly made, validated by the organisation’s own decision-making processes, deciding to pursue the greater good as policy, knowing it is a breach of the organisation’s external commitment to apolitical action. In line with this policy, Velásquez formally declines to pursue the investigation, acting within the new scope of his authorisation from CICIG. In this case, Velásquez’s wrong would transmit to CICIG and CICIG itself would act with dirty hands; and indeed, so the argument for democratic dirty hands goes, so would the dirty transmit on to Guatemalan public.

Alternatively, Velásquez still decides that CICIG should not move forward with the investigation, but instead of triggering the internal decision-making processes of CICIG to reframe policy, Velásquez leaves the policy of the organisation as it is: explicitly excluding such political behaviour from its activities, and thus the authorisation of any agent acting on its behalf, including himself. Velásquez uses alternative means to kill off the investigation. He creates internal management roadblocks. He acts as a decisional bottleneck. He calls for more background research. He mispresents, dissembles and delays. Perhaps, in the end, he simply decides not to pursue the investigation, without giving (truthful) reasons to his peers. The immediate result is the same as in the first option: the investigation does not go ahead. However, Velásquez acts outside of his own authorisation in effecting this result. In this case, therefore, CICIG is at most consequentially liable for his actions (subject to the qualifications in the next section § 6). It cannot be said to have authorised them. Having not authorised the act constituting dirty hands, CICIG cannot be said to have dirty hands, and thus neither can the Guatemalan public. The dirt has not been passed on, even though the beneficial consequences to the organisation, and to the Guatemalan public may.Footnote 12 The dirt is, we might say, being laundered.

7 Successful Laundering: Avoiding Ratification and Negligence

As a rule, therefore, a principal – like CICIG – may avoid direct liability for the wrongs of its agents – like Velásquez – when those wrongs are performed beyond the scope of authorisation. However, two further conditions must be fulfilled in order for such a principal to remain clean.

First, a principal only maintains the required distance between themselves and their agent’s unauthorised act, if they subsequently disavow the act, if that act becomes known to the principal. A failure to do so leads de facto or de jure to recognition that it accorded with the previously inchoate, or at least unarticulated, intention of the principal. The act may be retrospectively authorised, or ‘ratified’ in the terminology of agency law (Restatement, § 4.01, § 7.04). At an institutional level, we might assume this requires the institution to take proportionate action against the agent: actions that both express disapproval, or indeed condemnation, and also demonstrate the intention to prevent such disobedience occurring again within the institution. This has dramatic consequences for the institutional laundering of dirty hands. It means that, in order for the institution to remain clean, it must take action against the member who acted with dirty hands. If we continue with the second option detailed above, if Velásquez’s actions are ultimately discovered by others within CICIG, then CICIG’s decision-making mechanisms must initiate a process of disciplining Velásquez: most obviously by terminating his appointment, and publicly condemning his action. Whilst seemingly harsh – hanging the proverbial laundry out to dry – this is consistent with what most advocates take to be the ‘moral remainder’ of dirty hands (contra Levy 2007). It grounds the propriety of blame and punishment (Walzer 1973), or at least the recognition of genuine wrong-doing and the need for punishment, even if also combined with recognition of the good done (de Wijze 2013).

Second, even if the agent’s dirty act is unauthorised and unratified, the principal may still be sullied by the dirt of an adjacent wrong: negligence (Giliker 2018: 512). Such negligence would arise where the principal has not discharged an appropriate standard of care to limit the likelihood of such actions being performed with the powers it has granted its agent. Instead, they are negligent in ‘selecting, training, retaining, supervising, or otherwise controlling the agent’ (Restatement, § 7.05). In an institutional context, like that of CICIG, we might expect this standard of care to be typically discharged by appropriate processes, practices and policies, establishing clear instructions, incentives and cultural norms, combined with monitoring and accountability (Giliker 2018: 5120-3). The overall effect of such measures need not be to make unauthorised wrongful actions by an institution’s members impossible. Indeed, discharging the relevant standard of care will be consistent with non-specific knowledge that a limited number of wrongful actions are even inevitable over time (Beever 2009: 102). The institution as principal is merely required to have taken reasonable steps to manage that risk (even if, ‘reasonable’ here is, I admit, hiding another debate on its own contours (Beever 2009: 102-6)). In this way, the institutional principal can avoid any dirt associated with the unauthorised, unratified act of its agent by meeting this standard of care. However, having such kinds of processes, policies, and practices in place, it can thereby also go one better, and be said to maintain its integrity despite the dirty hands act.

8 Integrity and Complicity

A number of recent theorists have sought to define what it means for public institutions to have, themselves, a kind of ‘integrity’ analogous to that of individuals, that is, a kind of collective virtue (Byerly and Byerly 2016: 43), associated with consistency and coherency, moral permissibility, praiseworthiness, and trustworthiness (Brock 2014: 5–6; Breakey et al. 2015; Wueste 2005: 21; Buchanan and Keohane 2006: 422-3; Gordon 2021: 863). In previous work, I have argued that ‘public institutional integrity’ is ‘the robust disposition of a public institution to pursue its purpose efficiently, within the constraints of legitimacy, consistent with its commitments’ (Kirby 2021: 1627; see also Kirby 2022). Such a ‘robust disposition’ is typically instantiated where the institution has reliable mechanisms to realise the other constituent traits of institutional integrity, in particular the very kinds of policies, processes and practices previously mentioned to discharge its standard of care. These policies, processes and practices are what will make the institutional member’s wrongful action uncharacteristic of the institution as a whole. Indeed, via such policies, processes and practices, the character of the institution as a whole is geared or ‘disposed’ towards preventing the wrongs of its members as best it can. We might say that, by the act, at worst, the institution’s integrity is marked, but not undermined.

Does the integrity of CICIG persist if Velásquez continues as leader having committed such dirty hands unauthorised, and if it remains unknown to others? I concede this raises a tension. On the one hand, the integrity of institution does remain if the action remains uncharacteristic. On the other hand, this implies the existence of processes, policies and practices within the institution geared not merely to preventing such wrongs by its members ex ante, but also revealing them and sanctioning them ex post. Hence, the institution will either maintain its integrity if it subsequently reveals such a wrong on the basis of such processes, policies and practices, and then acts to condemn and distance itself from the act, or if Velásquez hides the wrong so well as that it cannot possibly be seen as a failure of the institution as a whole to find it. Either way, the ‘success’ of Velásquez’s dirty hands would require at least a degree of secrecy until CICIG’s mandate was renewed, but one might imagine a subsequent revelation by which CICIG not merely preserves, but reinforces its integrity in the eyes of the public by sanctioning, on principle, the very person who would have saved it.

Regardless, why is it important that our public institutions not merely remain clean of the dirt of their agents, but also can maintain such ‘institutional integrity’ in doing so? It is important because it allows us to explain how the public may not merely avoid dirt transmitted by way of the primary mechanism of authorisation, but also the secondary mechanism of complicity.

As sketched in § 3, the putative mechanism of complicity rests on twin claims about the public’s relationship with public officers. First, the public offers forms of voluntary support for such public officers and the governance structures in which they operate. And second, the public does so knowing (or in a position to know) that such public officers will act with dirty hands. Hence, it is inferred that we accept such acts as a consequence of the support we offer (even if we cannot necessarily be said to intend them). This, putatively, constitutes a form of complicity sufficient for the transmission of at least some degree of dirt.

The intermediation of institutions of integrity between the public and individual public officers, however, disrupts this argument. This is because, it is possible for citizens to intend to support their institutions, not their individual public officers per se. Take the act of voting, for example. I can construe my vote as a form of support for a particular politician, or I can vote in order to support the public institutions in which they might be placed. I participate in a process that I may merely consider necessary to constitute a legislature, and/or an elected executive, which in turn helps to sustain the other institutions of government, and indeed government as a whole. I play my role in the practice necessary to sustain these institutions. Similarly, when I act as a state contractor, or a participant in deliberative processes, or indeed as an employee, or when offering expertise, information and advice, I can take myself as supporting such public institutions, not any specific individual or set of individuals within them. In doing so, I can take myself to be supporting an institution in part because of and in part to sustain, its integrity (Kirby 2022). In this case, the object of my support is an institution of integrity, not any particular individual within it.

Of course, when supporting an institution, even one of integrity, one ought to concede that it is possible, perhaps even predictable, that some public officers within it will act in an unauthorised manner with dirty hands. Indeed, some may even consider it desirable that such public officers so act when the occasion arises. However, this is not sufficient for complicity. This is because nothing in our support for an institution of integrity aims at promoting dirty hands behaviour within it. Indeed, we do the opposite. Since the institution of integrity, by definition, is geared towards preventing such dirty actions as much as possible through its policies, processes and practices, then our support for such an institution is support for such prevention measures. We simply intend to support an institution that is committed to preventing wrongful actions by its own members, including dirty hands. We can be said to accept the risk that dirty hands will be performed by public officers given our support for such institutions, but we cannot be said to promote such actions themselves, since we are actively supporting the institution’s measures to limit that risk so far as is reasonable. We are no more culpable for accepting such a risk than anyone who performs (or supports) an act that has a risk but meets the appropriate standard of care in managing it.

Indeed, this is the most obvious construal of the support of Guatemalan’s for CICIG in 2017. As stated above, 70% of the population were polled as having confidence in CICIG (International Crisis Group 2018). Indeed when they voted in a landslide for Jimmy Morales in 2015, they reportedly voted in large part in support of CICIG, its work and the integrity it was displaying. In particular, its integrity was manifest in the political independence necessary to investigate almost the entire Cabinet of the previous government. Velásquez may have been the head of CICIG, but leaders come and go, and are selected as a product of the very rules of CICIG as an institution. We might reasonably assume the public were supporting that institution by virtue of, and in part to promote, its integrity – unprecedented within Guatemala. Hence, implicitly, they were also supporting whatever internal policies, practices and processes existed to ensure such integrity, and prevent wrongs by its own members. And, in particular, they were supportive of those policies, practices and processes that buttressed independence from political pressure. In doing so, therefore, it is hard to see how the Guatemalan public could be complicit in Velásquez doing exactly the opposite. They might have accepted, if asked, the risk that he might do so, but nothing in their support for an institution of integrity geared to limiting that risk reflects an endorsement, indifference or even acceptance of such an act. Any argument for democratic complicity, therefore, fails.

9 Conclusion

In January, 2017, Velásquez moved forward with the investigation of the President’s son and brother, and they were arrested. The investigation later expanded to include Morales himself (Lakhani 2017). Morales’ response was swift and destructive. He mobilised an effective disinformation campaign against the Commission, and built an alliance with the media, business and political elites: ‘a pact of the corrupt’.Footnote 13 He declared in 2018, that he would not renew the Commission’s mandate and ordered the immediate transfer of functions to other government agencies, hobbling the Commission’s power to act until its mandate expired in January 2019. Velásquez was forced to flee the country, declared persona non grata. Some had only recently heralded CICIG as a uniquely stable yet successful anti-corruption institution within a post-conflict developing country (Zamora 2019), offering an exceptional chance for Guatemala to break the depressingly common norm of political cycle of corruption in such contexts. Today, in the wake of Iván Velásquez Gómez’s decision, however, CICIG is abolished, and Guatemala is reverting to that norm (Goldman 2022; Ortega 2021).

Did Velásquez make the wrong decision? Obviously, it is hard to judge, at least from the philosopher’s armchair. However, this paper has offered an argument that, even if the decision were not wrong qua a breach of a duty to act with dirty hands, it may not have been the most praiseworthy course of action. Ceteris paribus, if my argument is correct, then the praiseworthy action would have been to act with dirty hands and halt the investigation into the President’s relatives, but to have done so, without the authorisation of CICIG itself, that is, to have deliberately gone ‘rogue’. In this way, CICIG as an institution may have had chance at both surviving, and of maintaining its integrity, both because of and in spite of Velásquez’s act.

Perhaps not, of course: no such thing is certain in politics. But the very possibility, which permits generalisation across contexts and a variety of public institutions, disrupts the conclusions of those who press the inescapable force of the mechanisms of authorisation and complicity that underpin so-called ‘democratic dirty hands’. Institutions of integrity provide are both an instrument by which the public can support structures geared to preventing the commission of wrongs in public office, but also incorporated entities that can house members who, despite such just such structures, are brave enough to commit unauthorised the necessary wrong of dirty hands. There is a chance, arguably a good chance, that in this way the Guatemalan public could have kept its own hands clean by continuing to support CICIG as an institution of integrity, even though it owed its ongoing existence to the brave but culpable dirty hands of its leader, Velásquez. But that was not to be the case.