Never Merely as a Means: Rethinking the Role and Relevance of Consent

Abstract

For several decades, Kant scholars, inspired by the Groundwork false-promising example, have constructed consent-based criteria for using another merely as a means. Unfortunately, these consent-based accounts produce assessments that are both counter-intuitive and un-Kantian in relatively simple cases. This article investigates why these consent-based accounts fail and offers an alternative. The Groundwork false-promising example has encouraged a problematically narrow understanding of the conditions for using another merely as a means in virtue of the fact that the example involves a consent-sensitive duty. This article demonstrates that the scope of the prohibition on using another merely as a means includes both consent-sensitive and consent-insensitive duties and offers a duty-based account that reflects this.

1. Introduction

Immanuel Kant’s formula of humanity commands: ‘So act that you use humanity, whether in your person or in the person of any other, always at the same time as an end, never merely as a means’ (G, 4: 429). ¹ This formulation of the categorical imperative contains two broad obligations: a positive obligation to always use humanity as an end and a negative obligation to never use humanity merely as a means, both of which appear to be without exception. Notably, the formula of humanity does not prohibit using persons as means; rather, it prohibits using persons merely as means. The distinction between using someone as a means and using her merely as a means is extremely important insofar as it demarcates a type of wrongful conduct. Using another merely as a means is not the only way to behave in a morally reprehensible manner, but it is an important class of wrongful actions, namely, wrongful instrumentalization. The challenge for Kantians (and anyone else who wishes to condemn using another merely as a means) is to identify some property or set of properties that accurately describes the distinguishing features of using someone merely as a means. In the absence of tenable criteria, we cannot apply the Kantian principle in a meaningful way.

Over the past several decades, Kant scholars, inspired by the false-promising example in the Groundwork, have constructed criteria for using another merely as a means that are grounded in the concept of consent – be it possible consent, actual consent or rational consent. Possible consent accounts ask whether it is possible for an agent to give consent to our use of her, whereas actual consent accounts ask whether the agent did give her consent to our use. ² Both accounts understand consent as a normative power that agents can exercise at their discretion. (A normative power is an agential ability to bring about a normative change in the world, such as creating a permission, often through something as simple as a speech act. ³ ) Rational consent accounts, in contrast, ask whether it would be rational for an agent to consent to another’s use. ⁴ Unfortunately, these consent-based accounts produce assessments that are both counter-intuitive and un-Kantian in relatively simple cases.

This article investigates why these consent-based accounts fail and offers an alternative. I contend that the Groundwork false-promising example has encouraged a problematically narrow understanding of the conditions for using another merely as a means, and this is because the false-promising example involves a consent-sensitive duty. A consent-sensitive duty is an obligation that agents can be released from with another’s consent. Many instances of using another merely as a means are cases where an agent has transgressed a consent-sensitive duty but, I will argue, not all. In this article, I demonstrate that the scope of the formula of humanity’s prohibition on using another merely as a means includes both consent-sensitive and consent-insensitive duties, and that this is Kant’s considered view as evidenced by his later work. The presence or absence of your consent is relevant to determining whether I use you merely as a means, if and only if acting without your consent would violate a consent-sensitive duty I owe you. However, if using another merely as a means does not require violating a consent-sensitive duty, we cannot rely exclusively on consent-based criteria to determine whether we use another merely as a means.

In the following section, I review Onora O’Neill’s possible consent account (PCA) of using another merely as a means, as well as Pauline Kleingeld’s more recent actual consent account (ACA). ⁵ Multiple authors have demonstrated that O’Neill’s PCA produces false positives, which is to say that the account yields the conclusion that an agent uses another merely as a means in seemingly innocent cases. Despite her claims to the contrary, I demonstrate that Kleingeld’s ACA also produces false positives in the same cases that are problematic for O’Neill’s PCA. I diagnose the problem with these accounts as a failure to appreciate that the moral relevancy of consent depends on context. In the context of a consent-sensitive duty, another’s consent (or its absence or impossibility) is highly morally relevant; in the context of a consent-insensitive duty it is not. ⁶

In section 3, I use a pair of cases to demonstrate that the scope of the formula of humanity’s prohibition on using another merely as a means includes consent-insensitive duties, such that we can use another merely as a means even when she has given her consent. These cases involve consent to lethal harm and consent to disrespect. I employ textual evidence from Kant’s later work to argue that the position I defend is Kant’s considered view. At the end of section 3, I introduce the following duty-based account of when one uses another merely as a means: X uses Y merely as a means if and only if (1) X uses Y as a means and (2) X’s use of Y violates either a consent-sensitive duty that X owes to Y or a consent-insensitive duty that X owes to Y.

In section 4, I consider cases where the absence (or impossibility) of another’s consent fails to establish that they are used merely as a means. These cases involve consent-insensitive duties to render aid to persons with impaired agency and to execute role-based responsibilities as agents of the state. In section 5, I complete my defence of the duty-based account of using another merely as a means. I maintain that the moral wrong of using another merely as a means is normatively complex in the manner that wrongful coercion is normatively complex, namely, these wrongful behaviours cannot be identified without reference to other normative content (e.g. particular rights and obligations). Consent-based accounts that provide purely descriptive necessary and sufficient conditions for using another merely as a means yield problematic assessments. The duty-based account avoids producing counter-intuitive and un-Kantian assessments precisely because it does not attempt to explain using another merely as a means in purely descriptive terms.

It will be evident, however, that the duty-based account does not purport to function in a manner comparable to consent-based accounts. When we utilize the duty-based account, we must know what at least some of our duties are prior to drawing the conclusion that some form of conduct would use another merely as a means. This might give the troubling impression that we cannot derive duties from Kant’s formula of humanity despite its status as a supreme principle of morality. Thankfully, I do not believe that accepting the duty-based account necessitates this conclusion. This article draws the more modest conclusion that we cannot derive duties from Kant’s formula of humanity in the manner that some commentators have proposed. I argue that we cannot treat the two parts of the formula of humanity (always treat humanity as an end, never use humanity merely as a means) as if they were independent. The conclusion of this article is that the negative prescription – never use humanity merely as a means – is logically dependent on the positive prescription – always treat humanity as an end. ⁷ In section 3, I provide examples of how we can derive duties from the formula of humanity command to always treat humanity as an end and then use these duties to produce conclusions about when we use persons merely as means. The alternative account I offer is compatible with deriving duties from the formula of humanity, but not from what some commentators have called the Mere Means Principle. ⁸

While the conclusion that we need prior normative content to demarcate conduct that uses another merely as a means may strike some as unappealing, I hope to show that the advantages of this approach are numerous. In addition to avoiding false positives, the duty-based account is more consistent with fundamental features of Kant’s ethics than consent-based accounts. Moreover, the duty-based account allows us to see how the two parts of the formula of humanity are connected, and demonstrates that the notion using merely as a means is a thick Kantian concept that cannot be easily co-opted by those who value consent practices without connecting them to concepts like dignity. The duty-based account preserves the idea that using merely as a means is a particular kind of failure to treat as an end, namely, wrongful instrumentalization, while prompting us to ask deeply important questions about what acknowledging the dignity of humanity requires of us in different contexts.

2. Previous attempts to demarcate using another merely as a means via consent

In this section, I consider two attempts to employ the concept of consent to demarcate conduct that uses another merely as a means. These attempts are unsuccessful insofar as they produce highly implausible, as well as un-Kantian, assessments in common cases. At the conclusion of the section, I diagnose what I take to be a common problem with these consent-based accounts: they incorrectly assume another’s consent (or its absence or impossibility) is morally relevant in all contexts.

2.1 O’Neill’s possible consent account

In the Groundwork, Kant presents false promising from a motive of self-interest as an example of using another merely as a means. At this point in the Groundwork, Kant has already presented the false promising example twice. On both occasions, he demonstrates how a maxim of false promising from self-interest cannot be willed as a universal law (G, 4: 402; 4: 422). In the latter part of section II, Kant returns to this example in order to demonstrate that the humanity formulation of the categorical imperative would also condemn this behaviour. According to Kant, ‘he who has it in mind to make a false promise to others sees at once that he wants to make use of another human being merely as a means’ (4: 429). He proceeds to explain the wrongness of false promising in a single sentence: ‘For, he whom I want to use for my purposes by such a promise cannot possibly agree to my way of behaving toward him, and so himself contain the end of this action’ (4: 429–30). This sentence is the source of inspiration for many consent-based accounts of using another merely as a means, including Onora O’Neill’s possible consent account (PCA). ⁹ According to O’Neill, ‘Not to treat others as mere means introduces a minimal, but indispensable, requirement for coordinating action in a world shared by autonomous beings, namely that nobody act in ways others cannot possibly consent to, so in principle precluding their autonomous action’ (O’Neill 1985: 263). O’Neill contends that consent is not possible when there is no genuine possibility for dissent, and ‘those closely involved in or affected by a proposal have no genuine possibility of dissent unless they can avert or modify the action by withholding consent and collaboration’ (p. 259). ¹⁰ O’Neill identifies deception and coercion as behaviours that preclude consent. According to O’Neill, because victims of deception and coercion are denied a choice between consent and dissent – they cannot avert or modify the action by withholding their consent or collaboration – those who engage in these behaviours use others merely as means (pp. 262–3).

Samuel Kerstein, Japa Pallikkathayil and Derek Parfit have independently noted that the PCA of using another merely as a means generates counter-intuitive results in a number of relatively simple cases. An unconscious jogger that you happen to come upon in the park cannot consent to your performing CPR on him. According to Kerstein, ‘it seems wildly implausible to contend that your attempt to save him was morally impermissible’ (Kerstein 2009: 174). ¹¹ Equally implausible is the verdict that an arresting officer treats a suspect merely as a means insofar as the suspect cannot avert or modify the officer’s way of treating them by withholding consent or collaboration. Pallikkathayil has demonstrated that this conclusion is not merely counter-intuitive; it is ‘at odds with Kant’s political philosophy, where he acknowledges that the use of force is sometimes permitted and perhaps even sometimes obligatory’ (Pallikkathayil 2010: 118). She reports that Kant acknowledges at least four cases where the use of force against another is at least permissible:

I may use force to defend myself and the objects in my possession in the state of nature if the establishment of a civil condition is not possible … I may use force to compel others to leave the state of nature and enter the civil condition … agents of the state may use force to enforce its laws, as when the police interfere in the commission of crimes … [and] agents of the state may use force to punish criminals. (Pallikkathayil 2010: 119)

Kant’s view is that the use of coercion or force against another is sometimes permissible, if not morally necessary. I take the counter-examples described above to be sufficient to establish that O’Neill’s PCA is unacceptable.

2.2 Kleingeld’s actual consent account

Pauline Kleingeld has recently argued that ‘Kant considers the actual consent of someone who is used as a means to be a necessary but not sufficient condition for the agent to avoid using this person merely as a means’ (Kleingeld 2020a: 397). She defends the following actual consent account (ACA):

An agent uses another person merely as a means if and only if (1) the agent uses another person as a means in the service of realizing her ends (2) without, as a matter of moral principle, making this use conditional on the other’s consent; where (3) by ‘consent’ is meant the other’s genuine actual consent to being used, in a particular manner, as a means to the agent’s end. (Kleingeld 2020a: 398) ¹²

Kleingeld’s ACA would appear to be vulnerable to many of the same counter-examples that are problematic for O’Neill’s PCA. An unconscious jogger cannot give actual consent to another performing CPR on her, and thus an agent who performs CPR on the unconscious jogger cannot make the jogger’s actual consent a condition of her performing CPR. Kleingeld agrees that this conclusion is implausible and attempts to circumvent it by maintaining that the first condition of her ACA is not satisfied in the case of the unconscious jogger. According to Kleingeld,

[W]hen Kant’s prohibition is read literally, as concerning cases of using others, then the case of the unconscious jogger causes no problems. Suppose you are a good Kantian and you are acting on your maxim of helping others in need, a maxim you have adopted from duty. In this scenario your end is to save the jogger’s life, and the means by which you realize this end is your giving her CPR. You are not using the jogger as a means, and hence by implication you are not using her merely as a means. (Kleingeld 2020a: 399)

Kleingeld’s treatment of the unconscious jogger case is unconvincing for several reasons. First, Kleingeld appears to merely assert that the unconscious jogger is not used as a means rather than derive this assessment from a general account of when one uses another. Later in the article, she offers the following: ‘an agent uses another person as a means if and only if, in the agent’s practical reasoning underlying the action, the other serves to realize or promote the agent’s ends’ (Kleingeld 2020a: 400). Unfortunately, even with this account of when one agent uses another as a means, it is not obvious why the agent who performs CPR fails to satisfy this condition. Kleingeld herself asserts that ‘[Kant] argues … that in a human person this [rational] capacity is inseparably connected with the body and that using part of an entity constitutes using that entity … In other words, using a person’s body or using a person’s rational capacities counts as using a person’ (Kleingeld 2020b: 212). Given that the unconscious jogger’s presence is clearly necessary for the realization of the beneficent agent’s end (saving the jogger’s life), it is difficult to see how, in the agent’s practical reasoning, the jogger does not serve to realize the agent’s end.

Second, it is not clear that we should accept Kleingeld’s account of when one agent uses another as a means. Perhaps most troubling is the fact that Kleingeld’s account of using another as a means does not appear to be capable of picking out what Kant considers to be a paradigmatic case of using another merely as a means – assaults on the property of others (G, 4: 430). Consider a thief who opens an unlocked car door and steals a shopping bag left on the passenger’s seat. It is unlikely and certainly not necessary that the owner of the shopping bag has entered into the thief’s practical reasoning, nor that the owner serves to realize or promote any of the thief’s ends. Nonetheless, Kant maintains that an agent who violates another’s property rights ‘intends to make use of the person of others merely as means’ (G, 4: 430). Thus, even if Kleingeld could explain why the agent who performs CPR on the unconscious jogger fails to satisfy her condition for using another as a means, she would avoid the counter-intuitive result in this case at the cost of having to exclude (contra Kant) many (if not most) property crimes as instances of using another merely as a means. ^{13, 14}

In his 2013 book, How to Treat Persons, Kerstein articulates the following account of when one agent uses another as a means: ‘an agent uses another (or, equivalently, uses or treats another as a means) if and only if she intentionally does something to or with (some aspect of) the other in order to realize her end, and she intends the presence or participation of (some aspect of) the other to contribute to the end’s realization’ (Kerstein 2013: 58). ¹⁵ Kerstein’s account of using another can identify cases of theft as using another as a means, even if the owner does not enter into the thief’s practical reasoning. ¹⁶ This makes Kerstein’s account of using another as a means preferable to Kleingeld’s at least as an interpretation of Kant. However, the agent who performs CPR on an unconscious jogger also satisfies Kerstein’s condition for using another, as does an arresting officer. If we employ Kerstein’s account of when we use another as a means, Kleingeld’s ACA will generate many of the same false positives as O’Neill’s PCA. In the absence of a compelling explanation for why Kleingeld’s account of using another as a means is preferable to Kerstein’s, as well as an explanation for why the unconscious jogger does not figure into the beneficent agent’s practical reasoning, it is fair to conclude that these cases are as problematic for Kleingeld’s ACA as they are for O’Neill’s PCA.

The counter-examples above demonstrate that the criteria specified in O’Neill’s PCA and Kleingeld’s ACA are problematic when applied in contexts that do not involve consent-sensitive duties. The problem common to both accounts appears to be that they attempt to read from Kant’s analysis of the false-promising case a criterion for using another merely as a means that is relevant in all contexts. Working from the example of false-promising, O’Neill identifies the impossibility of consent as a sufficient condition for using another merely as a means, whereas Kleingeld identifies the absence of actual consent as sufficient. Were we to consider only cases involving consent-sensitive duties, I doubt either account would produce a single false positive. In the remainder of the article, I will argue that the relationship between consent and using others merely as a means is more complicated than the accounts above acknowledge. Specifically, I will argue that the relationship between another’s consent and the impermissibility of our use of her depends on whether actual, normatively transformative consent is at least necessary, if not sufficient, to render our conduct permissible. Sometimes another’s consent is necessary to avoid using her merely as a means; in other contexts, consent is not required. This is (partially) explained by the fact that we owe others both consent-sensitive and consent-insensitive duties.

3. Consent-sensitive and consent-insensitive duties

Legal theorist Victor Tadros has written extensively on the topic of consent. In his 2016 book, Wrongs and Crimes, he employs a distinction between consent-sensitive and consent-insensitive duties. While this is not a distinction that Kant explicitly makes in his work, Kant’s ethical theory contains both types of duties, as well as the resources for parsing this distinction in an intelligible manner. I suggest that labelling duties in this way can elucidate the work that consent can and cannot do in Kant’s ethics. In this section, I will argue that the scope of the formula of humanity, including the prohibition against using persons merely as means, includes both consent-sensitive and consent-insensitive duties, and this is why the presence or absence of consent (or the possibility or impossibility of consent) is not always a reliable indicator of permissible or impermissible conduct. The absence or impossibility of consent is a reliable indicator of wrongful conduct only in the context of a consent-sensitive duty.

Tadros provides the following account of a consent-sensitive duty:

Typically, when X owes a consent-sensitive duty to Y not to v:

1. 1) X owes a duty to Y not to v;

2. 2) Y can release X from this duty by consenting to X ving.

3. 3) If X vs without Y’s consent, X wrongs Y; but not if Y consents. (Tadros 2016: 204)

Consent-sensitive duties are numerous and common. We owe a duty to others not to interfere with their property; however, they can selectively release us from this duty by consenting to a loan or a transfer. Similarly, though we have a stringent obligation to refrain from interfering with the persons of others, agents may selectively release other agents from this obligation by giving their consent to a massage, an examination, a boxing match or sex. The presence or absence of consent is enormously important to the moral and legal assessment of many forms of conduct. Taking another’s property without consent is theft. Surgery without consent is battery. And sex without consent is rape. When agents give their consent, they exercise their normative power to transform otherwise impermissible (and sometimes also illegal) conduct into morally permissible (as well as legal) conduct. This normative transformation occurs only if the consent is valid. An agent’s consent is valid only if it is produced under the right circumstances. The literature on what constitutes valid consent is extensive and beyond the scope of this article. ¹⁷ For my purposes, when I use the phrase valid consent, it should be understood to mean consent given voluntarily, by a decisionally competent agent, with understanding appropriate for the context. ¹⁸

Kant’s false-promising example involves a consent-sensitive duty. Specifically, it involves the transfer of resources from one agent to another which is permissible (juridically and ethically) only with the original owner’s valid consent. In the Groundwork example, the conduct is wrongful because the false promiser acquires the victim’s resources without the victim’s valid consent. The victim consents to a loan; she does not consent to how the false-promiser uses her. Many commentators who endorse consent-based accounts appear to have assumed that Kant intended this single example to provide us with a universal rule for determining when we use another merely as a means. Kleingeld, for instance, maintains that Kant’s analysis of the false promising case ‘clearly suggests that the prohibition against using others ‘merely as means’ should be understood in terms of a consent requirement’ (Kleingeld 2020a: 391). ¹⁹ I suspect commentators have read too much into Kant’s description of the false promise. The lying promise illustration in the Groundwork is precisely what Kant says it is, a single example of one agent using another merely as a means. If Kant never intended his explanation of the wrong-making features of the false promise to be universal to all instances of using another merely as a means, and I do not think he did, then this common misinterpretation, at the heart of various consent-based accounts, would explain why these accounts produce problematic assessments.

It seems clear that violating a consent-sensitive duty is sufficient to establish that one has used another merely as a means. In some respects, these duties are the easy cases. The rights violations that Kant refers to in the Groundwork – assaults on the freedom and property of others – are (for the most part), like the false-promising case, examples of transgressing consent-sensitive duties. Below I argue that the scope of the formula of humanity’s prohibition on using others merely as means extends to consent-insensitive duties as well. I consider two different consent-insensitive duties. Each duty is such that were an agent to violate the duty while using another as a means she would be guilty of using the other merely as a means regardless of whether the other gave valid consent.

3.1 Consent to harm

The volenti maxim – volenti non fit injuria – maintains that no wrong is done to one who has consented (Bergelson 2009: 165). If the volenti maxim were true, it would follow that you do me no wrong in inflicting lethal harm provided you had my valid consent to do so. The volenti maxim expresses an extreme view that few endorse without qualification. If, contra the volenti maxim, an agent can be wronged even when she has given her valid consent, there must be consent-insensitive duties, duties we cannot release each other from by an exercise of our normative power of consent. Consent-insensitive duties would thereby demarcate the limits of this normative power.

Whereas consent-sensitive duties have a consent condition – it is wrong to do x unless you have consent from Y – consent-insensitive duties have no consent condition. Negative consent-insensitive duties prohibit forms of conduct regardless of whether the conduct has been consented to. In a recent paper, Peter Schaber identifies the duty to refrain from exploitation as a consent-insensitive duty. ‘If one exploits another person one commits a wrong by wronging the victim, even if the victim has consented to it’ (Schaber 2020: 86). According to Schaber, it is wrong to exploit others even with their consent because this form of conduct has wrong-making properties that are not consent-sensitive. ²⁰ Unfortunately, Schaber provides no further mechanism for identifying consent-insensitive duties, and without such a mechanism he is vulnerable to the charge of being arbitrary.

Tadros offers us a little more. He suggests that the limit of what our consent can normatively transform is indicated by what we morally owe to ourselves. Tadros is particularly interested in the limits of our ability to transform otherwise wrongful harming acts into permissible ones by way of consent. He proposes: ‘Whether consent to harm is morally [transformative] … depends in part on whether the person consenting violates a self-regarding duty in giving it. And that depends on whether she would violate a self-regarding duty in harming herself for the sake of the goal that harming her aims at’ (Tadros 2016: 265). According to Tadros’ proposal, if it would be wrong for you to do X to yourself (as a means to some end of yours), then it would be wrong for me to do X to you (as a means to some end of mine), even if you have given me your valid consent to do X.

Given the importance Kant assigns to self-regarding duties, I find Tadros’ proposal to be particularly amenable to the task of identifying Kantian consent-insensitive duties. ²¹ Self-regarding duties are, by nature, consent-insensitive duties, insofar as we lack the normative power to release ourselves from these obligations. As Kant observes, if we could release ourselves, we would not be bound to the duty at all which would be a contradiction (MS, 6: 417). Tadros’ proposal is that self-regarding duties can be used to identify other-regarding consent-insensitive duties. This raises the question: how do we determine what would be morally impermissible to do to oneself? Kerstein has suggested the following strategy for identifying perfect, self-regarding duties: ‘If an agent cannot have the end he is pursuing in treating himself in some way, then the agent treats himself merely as a means. An agent cannot have this end just in case … he cannot pursue it without thwarting the pursuit of some end that he is rationally compelled to have’ (Kerstein 2008: 211). While I find nothing objectionable about Kerstein’s proposal, it is not particularly helpful. It simply replaces the question, what do we owe to ourselves with the question what ends are we rationally compelled to have?

A more promising approach looks to what Kant calls the ground of the formula of humanity. According to Kant, the ground of this supreme practical principle is that rational nature exists as an end in itself (G, 4: 428–9). ²² For Kant, rational beings exist as ends in themselves in virtue of their capacity for morality, that is, their capacity to give themselves the moral law and to act from duty alone (G, 4: 434). In the Doctrine of Virtue, Kant draws a connection between being an end in oneself and having dignity.

But a human being regarded as a person, that is, as the subject of a morally practical reason, is exalted above any price; for as a person (homo noumenon) he is not to be valued merely as a means to the ends of others or even to his own ends, but as an end in itself, that is, he possesses a dignity (an absolute inner worth) by which he extracts respect for himself from all other rational beings in the world. He can measure himself with every other being of this kind and value himself on a footing of equality with them. (MS, 6: 434–5)

Insofar as we are capable of morality, we exist as ends in ourselves, and to be an end in oneself is to possess dignity. Kant’s assertion that persons have dignity entails two things. First, to have the status of dignity is to be elevated above things, as well as skill, imagination, wit and diligence in work, all of which have mere price (G, 4: 434–5; MS, 6: 462). ²³ On the other hand, to have the status of dignity is to stand in a relationship of equality with all other persons (MS, 6: 434–5).

Given that persons have dignity in virtue of their capacity for morality, what we are obliged to respect in ourselves and others is our capacity for morality including our equal membership in the moral community. Humanity is a limiting condition on all our actions (G, 4: 431). The relevant question to ask when seeking to determine whether some form of conduct would transgress a self-regarding duty is the following. Would engaging in this conduct (i) undermine or harm the source of your dignity or (ii) repudiate your moral standing as elevated above things and equal with all other persons?

One self-regarding duty Kant identifies in the Doctrine of Virtue is the duty to refrain from suicide (MS, 6: 422–4). Given that lethal self-harm would destroy our rational capacities including our capacity for morality, the answer to the relevant question articulated above seems to be a clear yes. It would follow that there is a strict, self-regarding obligation not to commit suicide. If we employ Tadros’ proposal, it would then follow that because it is wrong for me to inflict lethal harm on myself (as a means to one of my discretionary ends), it is wrong for you to inflict lethal harm on me (as a means to one of your discretionary ends) even if you have my valid consent. The argument for the consent-insensitive duty would proceed as follows. I transgress a perfect self-regarding duty when I act in a manner that undermines or harms my source of dignity or repudiates my equal moral standing. Because killing myself would harm the source of my dignity, I have a perfect, self-regarding duty not to kill myself, and when I violate this duty, I use myself merely as a means. If I would use myself merely as a means when I kill myself to attain a particular, discretionary end, then you would use me merely as a means when you kill me to attain that end – with or without my consent. ²⁴ We derive this consent-insensitive duty from the formula of humanity command to always treat humanity as an end.

Allow me to digress for a moment in anticipation of a practical question. It would seem to follow from the argument I have just made that Kantians must oppose physician-assisted suicide and euthanasia. While I lack the space to treat this topic with the thoroughness it deserves, I feel compelled to indicate that I believe Kant’s ethics allows for more nuance than an unqualified condemnation of these practices. Kant himself suggests this in his presentation of casuistical questions following his account of the duty not to commit suicide. He describes a man who has been bitten by a rabid dog. ‘He explained, in a letter he left, that, since as far as he knew the disease was incurable, he was taking his life lest he harm others as well in his madness (the onset of which he already felt)’ (MS, 6: 423–4). Kant asks whether the man infected with rabies acted wrongly in killing himself, but supplies no answer in response. The morally relevant features in this case are the imminent loss of humanity, the degraded state the man would exist in prior to death and the potential for causing harm to others. David Velleman has argued that sometimes it is more respectful to destroy something rather than allow it to exist in a degraded state (Velleman 1999). There is an argument to be made that the case of a terminally ill patient who commits suicide when the loss of personality is imminent is morally different from Kant’s Groundwork example of an agent who commits suicide to escape suffering. ²⁵ There seems to be room in Kant’s ethics to permit some practice of physician-assisted suicide and euthanasia while also justifying serious restrictions on these practices.

At this point, I take myself to have demonstrated that there is at least one consent-insensitive duty such that, were we to violate this duty while using another as a means, this would constitute using another merely as a means even if we act with the other’s valid consent. While the precise contours of this duty require specifying, I suspect few will find the notion that we may not engage in consensual killing as a means to achieving our ends to be a terribly radical proposal. If there is a counter-example to the volenti maxim, inflicting lethal harm seems to be a good candidate, and I have argued that there are Kantian grounds for this position.

3.2 Consent to disrespect

In the Doctrine of Virtue, Kant divides self-regarding duties into duties we owe to ourselves as an animal being and duties we owe to ourselves as a moral being. The duty to refrain from suicide belongs to the former, whereas the duty to refrain from servility belongs to the latter. We are guilty of servility when we disavow our dignity or equal moral standing with others, for instance, when we lack proper moral self-esteem or invite or allow others to treat us with less than the full respect we are entitled to, that is, to ‘tread with impunity on [our] rights’ (MS, 6: 436). Kant explains, ‘Humanity in his person is the object of the respect which he can demand from every other human being, but which he must also not forfeit’ (MS, 6: 435; emphasis added). I read Kant here as saying that we lack the normative authority to waive our moral claim to respect, which is to say we lack the normative power to release others from obligations of respect. If this is correct, then duties of respect are consent-insensitive duties.

Later in the Doctrine of Virtue, Kant explicates our other-regarding duties ‘arising from the respect due them’ (MS, 6: 462). There he contends that ‘Every human being has a legitimate claim to respect from his fellow human beings and is in turn bound to respect every other. Humanity itself is a dignity; for a human being cannot be used merely as a means by any human being (either by others or even by himself) but must always be used at the same time as an end’ (MS, 6: 462). It is notable that Kant invokes the formula of humanity prohibition on using others merely as means in this section with no mention of consent. A simple explanation for the absence of any reference to consent is that duties of respect are consent-insensitive duties. Insofar as I lack the normative authority to renounce my claim to your respect (MS, 6: 464), my consent cannot change this obligation for you. As Kant puts it, ‘to deny them the respect owed to human beings in general, is in every case contrary to duty’ (MS, 6: 463; emphasis added).

While Kant identifies arrogance, defamation and ridicule as contrary to the duty of respect owed to others, he does not describe these practices as using another merely as a means. This may be because he takes himself to be describing vices of character. ²⁶ Nonetheless, understanding something like ridicule as using another merely as a means is fairly natural. The one engaging in ridicule is clearly using another as a means to some end, perhaps redirecting attention away from his own deficiencies. Insofar as disavowing one’s claim to equal respect (for some discretionary end) would violate a self-regarding duty and use oneself merely as a means, to deprive another of equal respect is also contrary to duty, and an instance of using another merely as a means, regardless of whether the other has given valid consent to the disrespect. Again, what I have done above is derive a consent-insensitive duty from the formula of humanity command to always treat humanity as an end and then used this consent-insensitive duty to produce a conclusion about using another merely as a means.

3.3 Consent-based accounts revisited

Can consent-based accounts register consensual killing or disrespect as instances of using another merely as a means? O’Neill’s PCA cannot, given that she understands the conditions for possible consent in non-normative terms, namely, the ability to avert or modify another’s use by withholding consent or cooperation. Provided that you make my consent a condition of your killing or disrespecting me, O’Neill’s PCA would not register the conduct as wrongful. ²⁷ What about Kleingeld’s ACA? Recall Kleingeld’s ACA stipulates that:

An agent uses another person merely as a means if and only if (1) the agent uses another person as a means in the service of realizing her ends (2) without, as a matter of moral principle, making this use conditional on the other’s consent; where (3) by ‘consent’ is meant the other’s genuine actual consent to being used, in a particular manner, as a means to the agent’s end. (Kleingeld 2020a: 398)

Cases like consensual killing or disrespect will be problematic for Kleingeld’s ACA unless she understands genuine actual consent in a manner that excludes consent to these practices. A passage in her paper suggests that she may have thought about the normative limits of valid consent. She writes:

[I]t is important to note that those who give consent ought to meet specific moral requirements governing their own agency. Kant emphasizes that you ought not to use yourself merely as a means either (G 4: 429, MM 6: 420–37), and this poses constraints on the types of use to which you are morally allowed to consent (cf. MM 6: 236). You should not consent to being used as an accomplice in a scheme of deception, for example, and you should not agree to sell yourself into slavery. Thus there is a set of normative conditions that should be met by anyone who is asked to serve as a means. Agents intending to use others as means in turn ought to take these conditions into account when asking for consent, since they must not ask others for their consent to being used in degrading or otherwise morally impermissible ways. (Kleingeld 2020a: 404–5)

Unfortunately, Kleingeld confuses the issue by mixing examples. Take the case of consenting to serve as an accomplice in a scheme of deception, for instance, by loaning the deceiver a piece of property that will aid the deception. There is certainly a sense in which you should not give your consent to this use just as you should not engage in any wrongful conduct. However, you nonetheless can give normatively transformative consent to this use. If you loan the deceiver your property, he has not committed theft and has not wronged you. Your consent was normatively transformative in this case even though, morally, you should not have consented.

The case of selling yourself into slavery is rather different. Not only is it the case that you should not consent to the sale of your person, it is also the case that you can not give normatively transformative consent to this transaction (MS, 6: 283). Even if you express consent, this expression does not transform the conduct; the one who uses you as a piece of property wrongs you even if you have given voluntary, informed and decisionally competent consent to this use.

Another example from earlier in Kleingeld’s paper is telling. She asks us to:

Imagine a scenario in which a genocidal dictator selects several individuals from the ethnic group she is annihilating, in order to subject them to dangerous medical experiments that promise to yield valuable medical insights that would benefit the rest of humanity. Now suppose that one of the selected individuals happens genuinely to consent to the treatment. (Kleingeld 2020a: 393; emphasis added)

Kleingeld maintains that her ACA can identify the dictator’s use of the individual as mere means use despite genuine consent. This is because ‘Whatever maxim [the genocidal dictator] is reasoning from, she patently fails to limit her use of others to cases in which they consent’ (Kleingeld 2020a: 405). In other words, the dictator fails to satisfy the second clause of the ACA because she does not make her use of the volunteer conditional on his consent. She would have used him even if he had not consented. The case of the genocidal dictator is a case of consent to harm; however, Kleingeld does not question whether normatively transformative consent to painful, dangerous medical experiments that promise no benefit to the subject is possible in a Kantian ethical system. Even if Kleingeld’s ACA can produce an acceptable verdict in this case, it may not produce this verdict for the right reason. ²⁸

At this point, one might question whether I have actually identified a defect in Kleingeld’s ACA. After all, even if Kleingeld’s ACA does not yield the conclusion that (at least some) consensual killing or consensual disrespect are instances of using another merely as a means, it is open to Kleingeld to respond by saying that these forms of conduct are nonetheless wrongful on Kantian grounds insofar as they fail to respect others as ends in themselves. ²⁹ This response is certainly available to Kleingeld and those sympathetic with her account. Forms of conduct can be wrongful even when they are not instances of using another merely as a means. However, it is worth noting that all instances of wronging another (including instances of using them merely as means) could be explained in terms of a failure to respect persons as ends in themselves. It is the broader category. As Thomas Hill describes it,

The humanity formula obviously contains two related prescriptions: (1) never treat humanity in any person merely as a means and (2) always treat humanity as an end. The second encompasses the first: if we always treat humanity as an end, we will never treat humanity merely as a means. Also the meaning of the first is apparently dependent on the second: to treat humanity merely as a means is just to treat it as a means without also treating it as an end. (Hill 2012: 299)

The proposal that we explain the wrongfulness of (at least some) consensual killing or consensual disrespect in terms of a failure to respect persons as ends in themselves invites the following questions: What is the point of labelling some forms of conduct as instances of using merely as a means if the wrongfulness of the conduct in these cases, as well as others, could be explained in terms of a failure to treat persons as ends in themselves? Why have the category of using merely as a means at all?

To these questions I would respond that the category of using merely as a means is important and useful insofar as it describes a particular kind of wrongful conduct, namely, wrongful instrumentalization. All cases of wrongful instrumentalization will also be failures to respect persons as ends in themselves, but not all cases of failing to respect persons as ends will be cases of wrongful instrumentalization. Kant identifies indifference to others as one such case (G, 4: 430), and Martin Sticker has recently argued that we need the category treating as a mere thing to describe ways we fail to treat others as ends in themselves through neglect, indifference and complacency (Sticker 2021). In these cases, we do not use others as means and thus cannot be said to use them merely as means, but we nonetheless act wrongly. The point of the label Sticker proposes is to distinguish a particular kind of wrongful conduct, that is, a particular way we might fail to respect others as ends in themselves.

It is possible to use another as a means when we engage in consensual killing or consensual disrespect. However, because these uses are consensual, Kleingeld’s ACA cannot register them as instances of using another merely as a means. The proposal considered above attempts to vindicate the ACA by stipulating that the wrongfulness of these forms of conduct can be explained in terms of a failure to respect persons as ends. This proposal effectively abandons the idea that the function of the label using merely as a means is to distinguish conduct that is wrongful instrumentalization, for it allows that there are cases of wrongful instrumentalization that are not instances of using another merely as a means. But then the question above reverberates: what is the point of the category using merely as a means?

If I am right in thinking that the function of the category using merely as a means is to distinguish instances of wrongful instrumentalization from other failures to respect persons as ends, then an account of when we use another merely as a means should be able to identify (at least some) consensual killing or consensual disrespect as instances of using another merely as a means. To do this, the account must be able to determine when another’s voluntary, informed and decisionally competent consent is not normatively transformative. This will require investigating the limits of what valid consent can transform in a Kantian system of duties, which is the sort of work I have begun above. Notably, this work requires us to consider more than just the descriptive facts of the case (e.g. whether X gave consent, whether X’s consent was given under the appropriate conditions, whether Y made X’s valid consent a condition of her use, etc.). Understanding the limits of what valid consent can transform requires appealing to normative content (e.g. whether it would be wrong for X to use herself in some way). I will return to this point in the final section of the article.

In this section I have demonstrated that while many instances of using another merely as a means are violations of consent-sensitive duties (e.g. theft, battery, rape), some instances of using another merely as a means are violations of consent-insensitive duties. I have argued that the duty to refrain from inflicting lethal harm on persons (or at least those who do not face the imminent loss of their personality) and the duty to respect others are both consent-insensitive duties, and that this is Kant’s considered view as evidenced by his later work. Consent-insensitive duties do not contain a consent condition. We wrong others (and if we use them, we use them merely as means) when we transgress these duties with or without their consent. ³⁰

Any acceptable set of necessary and sufficient conditions for using another merely as a means must reflect the fact that in some cases another’s consent to our use is highly relevant to determining whether our use is permissible, but in other cases another’s consent is not relevant at all. The following duty-based account of using another merely as a means does this.

X uses Y merely as a means if and only if (1) X uses Y as a means and (2) X’s use of Y violates either a consent-sensitive duty that X owes to Y or a consent-insensitive duty that X owes to Y.

The first clause is obviously a necessary condition; if you do not use someone as a means you cannot use them merely as a means. I find Kerstein’s account of when one agent uses another as a means to be an acceptable interpretation. Recall that this account, previously presented in section 2, stipulates that an agent uses another as a means ‘if and only if she intentionally does something to or with (some aspect of) the other in order to realize her end, and she intends the presence or participation of (some aspect of) the other to contribute to the end’s realization’ (Kerstein 2013: 58). The exposition in this section has established that satisfying (1) and either disjunct of (2) is sufficient to establish that you have used another merely as a means. In section 5, I will utilize an example from Pallikkathayil to demonstrate that clause (2) is a necessary condition for using another merely as a means. In the following section, I return to the cases discussed in section 2 in order to demonstrate how the duty-based account of using another merely as a means fares better than consent-based accounts.

4. Other consent-insensitive duties

Recall that Kerstein constructed the case of an unconscious jogger to illustrate a deficiency with O’Neill’s PCA. Insofar as the unconscious jogger cannot possibly consent to the passer-by performing CPR on him, O’Neill’s sufficient condition for using another merely as a means is satisfied. But it seems highly counter-intuitive to say that the agent who performs CPR uses the unconscious jogger merely as a means and thus acts wrongly according to the Kantian standard. In section 2, I argued that Kleingeld’s attempt to avoid this counter-intuitive assessment is unsuccessful and thus the unconscious jogger case is equally problematic for her ACA. Paul Formosa has attempted to deal with the unconscious jogger case in the following manner: ‘since she will die if I do not perform CPR on her, she rationally must consent because she has a duty to herself to preserve her own rational capacities and thus her life, and no other action will do that. This is sufficient to authorize me to undertake the action in this case’ (Formosa: 2014: 64–5). ³¹

There is reason to be dissatisfied with Formosa’s analysis of this case as well. If the source of our authorization to intervene is the fact that the agent rationally must consent in virtue of her moral obligation to preserve her life, then it would seem also to authorize us to execute potentially life-preserving interventions on those who are conscious and actively dissent. This is a familiar problem for rational consent accounts; they abandon the notion of consent as a discretionary normative power allowing rational commitments to do the bulk of the justificatory work. However, in the case of actual dissent, it looks like we are infringing on the freedom of another for the sake of forcing her to comply with the demands of morality. But this is precisely one of the features that is supposed to distinguish ethical duties from juridical duties – others may coerce me to fulfil the latter but not the former (MS, 6: 219–21; 6: 231–3).

A better analysis of the unconscious jogger case acknowledges that we owe different things to persons whose agency is impaired. Indeed, we cannot respect impaired agents in the same way that we respect unimpaired agents. We cannot obtain even a token of consent from an unconscious person, and we should not regard a drunk person’s consent as normatively transformative. Persons with impaired agency cannot exercise their normative powers. The standard for engaging with persons who cannot exercise their normative powers must be different from the standard for engaging with persons who can. It might be tempting to say that we should simply refrain from interfering with impaired agents at all, but this position would render us problematically indifferent to the harm that can come to impaired agents, including death.

Kant does not have much to say about unconscious persons, and he was not in a position to contemplate whether and when it is appropriate to perform CPR. ³² Kant, however, was familiar with the state of drunkenness. In the Doctrine of Virtue he writes, ‘A human being who is drunk is like a mere animal, not to be treated as a human being’ (MS, 6: 427). Kant’s comparison between a drunk person and a mere animal is problematic, and I suspect he would not have condoned treating drunk persons like animals. A better comparison is between a drunk person and a young child, as neither can function as a competent adult though both (unlike non-human animals) retain the potential for this. ³³ If a drunk person is like a child, then it makes sense to treat the drunk person in a manner comparable to how we should treat children: paternalistically.

When it comes to persons with impaired agency (agents who cannot give valid consent), treating them as ends in themselves seems to require that we provide care that aims to protect them from immediate harm and restore their agency (when possible). This obligation must be a consent-insensitive duty because the ones to whom we owe the duty cannot give valid consent. Some individuals have stringent role-based obligations to provide this care (e.g. immediate responders), however, all agents have this duty to some extent. An agent who comes upon an unconscious person while out for a walk may not permissibly continue on her way. At a minimum, she must call for help and perhaps also remain with the incapacitated person until help arrives. Our end in these cases is the restoration of the other’s agency with minimal harm to their person.

The agent who performs CPR on the unconscious jogger acts in accordance with the consent-insensitive duty to come to the aid of impaired agents. And though she uses the unconscious jogger as a means to her end of rendering aid, she violates neither a consent-sensitive nor a consent-insensitive duty. The duty-based account of using another merely as a means, unlike O’Neill’s PCA and Kleingeld’s ACA, does not produce the verdict that the agent performing CPR uses the jogger merely as a means. The duty-based account also does better than a possible consent account like Formosa’s insofar as it does not lead us to the conclusion that we may interfere with competent adults against their wishes for the sake of preserving their lives. What the duty-based account cannot do is register the wrong of failing to render aid to the unconscious jogger; however, this is not a defect. The duty-based account I offer articulates conditions for using another merely as a means – which is but one way of acting wrongly. ³⁴

I want to avoid over-simplifying the duty to care for impaired agents. Like the duty to refrain from lethal harm, the particular contours of this duty require specification. The existence of mechanisms designed to project our agency into a future when we cannot exercise agency (e.g. advance directives and do not attempt resuscitation (DNAR) orders) introduce important questions about whether and how we can meaningfully give prior dissent or consent. Like the previously considered questions about consent to lethal harm, I cannot treat these biomedical questions with the depth they deserve in this space. It is a virtue of the duty-based account that it requires us to think through the particular contours of what we owe others under particular conditions and not merely whether another gave consent or could give consent.

The other counter-example mentioned in section 2 involved an arresting officer. This case also involves a consent-insensitive duty. Agents acting in official state capacities – police officers, judges, jailers, law makers – have consent-insensitive duties to carry out the responsibilities of their offices. The judge does not need the criminal’s consent to make punishing him permissible. The authority to act comes from the legitimate juridical role particular individuals take on, not from another’s consent. So long as the arresting officer is acting within the constraints of her position (e.g. not using excessive force), the duty-based account, unlike O’Neill’s PCA and Kleingeld’s ACA, will not yield the verdict that the officer has used the suspect merely as a means.

5. The normative complexity of using others merely as means

In this final section, I will complete the defence of my duty-based account of using another merely as a means by utilizing the following example introduced by Pallikkathayil. Imagine you are trying to determine, before leaving home, whether you will need a coat to be comfortable outside. You look out your window to see if the pedestrians on the street are wearing coats. You are using others as a means to your end, determining how to dress, but it is implausible to think that you use the pedestrians in a morally impermissible manner insofar as you deny them an opportunity to refuse to be used in this way (Pallikkathayil 2010: 127).

Pallikkathayil’s example will be problematic for consent-based accounts of using another merely as a means. Insofar as the pedestrians do not give their consent to your use of them and you do not make their consent a condition of your use, Kleingeld’s ACA yields the verdict that you use the pedestrians merely as a means and thus act impermissibly. Similarly, insofar as the pedestrians cannot avert or modify your use of them by withholding their consent or cooperation, O’Neill’s PCA will also yield the verdict that you use them merely as means. ³⁵ Pallikkathayil’s case is importantly different from the ones previously considered insofar as it does not obviously involve either a consent-sensitive or a consent-insensitive duty. Rather, it is designed to illustrate that relationships between agents ‘can fail to be cooperative without being manipulative’ (Pallikkathayil 2010: 129).

Recall that my duty-based account of using another merely as a means specifies the following necessary and sufficient condition:

X uses Y merely as a means if and only if (1) X uses Y as a means and (2) X’s use of Y violates either a consent-sensitive duty that X owes to Y or a consent-insensitive duty that X owes to Y.

For the duty-based account to yield the verdict that the observer has used the pedestrians merely as means in Pallikkathayil’s example, it would need to be demonstrated that the observer violated a duty she owed to the pedestrians. The trouble with the consent-based accounts, as previously noted, is that they wrongly assume that consent (or the possibility of consent) is always a morally relevant feature. Unlike consent-based accounts, the duty-based account will not necessarily produce the verdict that the pedestrians in Pallikkathayil’s case have been used merely as means. It will instead prompt us to consider whether persons in public spaces have privacy claims that the observer fails to respect. I take Pallikkathayil’s example to demonstrate that clause (2) is a necessary condition for using another merely as a means.

It is expected that some will be disappointed or put off by the fact that the duty-based account makes it impossible to ascertain whether we have used another merely as means without introducing additional normative content. Rather than apologize for this feature, I maintain that this is an important insight that can pave the way to more fruitful investigations. What failed consent-based accounts show us is that using another merely as a means is normatively complex in the way that wrongful coercion is normatively complex. ³⁶ To determine whether a particular form of conduct (e.g. a speech act) is an instance of wrongful coercion one must understand the relevant rights and obligations of the parties involved. Understanding only the descriptive facts of the case, for instance, that X threatened Y, and that Y felt like she had no choice but to acquiesce, is insufficient. ³⁷ We cannot ascertain whether X wrongfully coerces Y without knowing what X owes to Y. In a similar manner, I am arguing that determining whether one agent uses another merely as a means requires understanding their particular rights and obligations at the time of use. My proposal is consistent with Pallikkathayil’s ground-breaking observation that specifying which forms of conduct constitute using another merely as a means will require determining what another’s rights are, which, in turn, depends on actual political decision-making (Pallikkathayil 2010: 141–2). The duty-based account of using another merely as a means forces us to employ other Kantian resources, like Kant’s political philosophy and the concept of dignity, and to ask difficult questions about what we owe to each other (and ourselves) under different conditions. This is not such a bad thing. After nearly four decades of failed consent-based accounts, it is time to consider alternative approaches to demarcating conduct that uses another merely as a means.