Kant embraces the principle that it is wrong for us to treat others merely as means. This ‘Mere Means Principle’, as I refer to it, stems from one formulation of the categorical imperative, namely the Formula of Humanity: ‘So act that you treat humanity, whether in your own person or in the person of any other, always at the same time as an end, never merely as a means.’Footnote ¹ If an agent treats another merely as a means, claims Kant, then his action is morally impermissible. In contemporary Kantian ethics, the Mere Means Principle plays the role of a moral constraint: it limits what we may do, even in the service of promoting the overall good.

The Mere Means Principle allows of no exception, and it thus might fail to square with our considered moral views. Perhaps no matter how we specify the notion of treating others merely as means, we are able to imagine an extreme scenario in which treating them in this way will not seem to us to be wrong. For example, what if millions of people will die in a nuclear explosion unless, in order to prevent it, John fatally shoots an innocent person – someone who would survive the explosion and is begging for his life?Footnote ² In killing the innocent person, John would, many of us think, be treating him merely as a means. But we resist the conclusion that he would be acting wrongly.

In the end, philosophers attracted to the Mere Means Principle might be forced to revise it by making it non-absolute. They might argue, for example, that although we always have significant moral reason not to treat others merely as means, once we reach a certain threshold of good that can be promoted only if we do so, this reason gets overridden. Treating others merely as means is always wrong pro tanto, but, in (presumably rare) cases in which other, weightier, moral reasons apply, not wrong all things considered. Such a revision might allow philosophers to maintain something like the Mere Means Principle even in the face of extreme cases.

However problematic extreme cases might be for this principle, those of us inclined to defend it as a valid moral constraint (absolute or non-absolute) face a more serious problem. On straightforward interpretations, the Mere Means Principle generates implausible results in relatively ordinary, everyday cases. On one interpretation, for example, a tennis player treats his opponent in a tournament merely as a means and thus wrongly when he tries, through defeating him, to win first place. An unintended implication of another interpretation is that a police officer treats a white-supremacist merely as a means if, in order to prevent race-based attacks on law-abiding citizens, he arrests him. In neither case is it plausible to conclude that the action was morally impermissible. These results are unfortunate, for they discredit a principle that seems to capture an important aspect of everyday moral thinking and to constitute a promising candidate for a valid moral constraint.

This article aims to present an interpretation (or reconstruction) of the Mere Means Principle that has more plausible implications regarding relatively ordinary cases than do the rival accounts examined. It presents a sufficient condition for an agent's treating another merely as a means.

Kant gives us little guidance for understanding the notion of treating another merely as a means. So it is not surprising that philosophers have focused on one of the few passages in which he seems to shed light on the topic, namely his attempt in the Groundwork of the Metaphysics of Morals to show how a duty not to make false promises stems from the Formula of Humanity.Footnote ³ I follow the practice of invoking this passage. Against the background it provides, I focus mainly on two different accounts of what it means to treat another merely as a means. Both the Ending Sharing and the Possible Consent accounts specify a procedure for determining whether in treating another in some way, an agent is treating her merely as a means. But both accounts have implausible implications in relatively ordinary cases, I argue. By modifying and combining elements of the two, I try to develop one that is more philosophically satisfying than they are. I call it the Reinforced Hybrid account. I am not here claiming that Kant himself is committed to this account. While it was available to him, and it even meshes quite well with much of what he says in the passage on false promising, I do not offer it as an interpretation of his text. At the end of the article, I explore an objection to the Reinforced Hybrid account that reveals one of its limitations.

Before turning to Kant's passage on false promising, some clarificatory remarks are in order. First, the Formula of Humanity forbids agents from treating ‘humanity’ merely as a means. As readers of Kant are well aware, he employs ‘humanity’ interchangeably with ‘rational nature’.Footnote ⁴ In doing so he suggests that having humanity involves having certain rational capacities. Among them are the capacities to set and pursue ends and to act autonomously, that is, (roughly) to conform to self-given moral imperatives purely out of respect for these imperatives.Footnote ⁵ The main aim here is to specify conditions for beings with humanity treating other such beings merely as means. Thus the terms ‘agent’, ‘person’, and ‘other’ (as, for example, in the expression ‘treating others’) each refer here to beings with humanity.

The Mere Means Principle stems from Kant's Formula of Humanity, but it does not exhaust the formula's content. An agent can act wrongly with regard to another even if he does not use the other merely as a means, according to Kant. An agent might, for example, express utter contempt for the other's intellectual capacities. Such an act might be an instance of failure to respect another as an end in himself, rather than one of treating him merely as a means.

But when can it be said that we are using another at all, whether it be merely as a means or not? When do we count as treating another in some way? Let me make a few observations. It does not seem sufficient for us to count as using another that we benefit from what another has done. If, on her usual route through the park, a jogger gets pleasure from a passing stranger's singing, she does not appear to be using the stranger.Footnote ⁶ Second, I have been employing (and will continue to employ) the terms ‘treating another in some way’ and ‘using another’ interchangeably. However, in ordinary English the notion of treating another in some way seems to have a wider extension. If someone smiles at you, for example, he treats you in some way, but he might not thereby be using you at all. He might simply be expressing affection. In this article the notion of treating another in some way does not have this wider extension. All cases of using another, or, equivalently, treating another in some way, are ones in which an agent intentionally does something to someone in order to secure (or as a part of securing) one of his ends. For example, I use a taxi driver if I hail his cab in order to get to the cinema; I use my spouse (or, a bit less awkwardly, treat her in some way) if I lie to her so that her birthday party will be a surprise; and I use a mugger (or treat him in some way) if I punch him in order to escape from his grasp. I sketch a sufficient condition for such cases counting as instances of using another merely as a means and thus as (at least pro tanto) instances of acting wrongly.

Finally, let me make a methodological point. I aim so far as possible to treat the Mere Means Principle as morally fundamental. In explicating its content, I hope for the sake of theoretical clarity and simplicity to avoid appealing to any further moral principles.

BEING ABLE TO SHARE AN END

Much discussion of what it means to treat others merely as means stems from a single passage in the Groundwork. Kant is attempting to demonstrate that his Formula of Humanity generates a duty not to make false promises:

We are looking for an interpretation of this passage that will help us to understand in general terms what it means to treat another merely as a means. According to Allen Wood, Kant is arguing here that making a false promise to another would treat the other merely as a means since it would express disrespect for his rational nature. ‘A false promise, because its end cannot be shared by the person to whom the promise is made, frustrates or circumvents that person's rational agency, and thereby shows disrespect for it.’Footnote ⁸ Apparently, according to Wood, when Kant says that a promisee cannot ‘himself contain the end’ of a false promisor's action, he is intimating that the latter cannot share the promisor's end. That interpretation seems reasonable enough. Borrowing from Wood here, we might try to construct a sufficient condition for using others merely as means. Although Wood does not do so himself, we might claim that if another cannot share an agent's end in treating her in some way, then the agent treats the other merely as a means.

Two agents presumably share a particular end if the following is the case: they are both trying, or at least have both chosen to try, to realize this end. If this is not the case, then they presumably do not share the end. But what, precisely, does it mean to say that two agents cannot share an end? Returning to the example at hand, what does it mean to say that the promisee cannot share the promisor's end? Wood is not helpful on this question. From the outset it is important to specify precisely which of the promisor's ends the promisee cannot share. It is presumably the promisor's end of getting money from the promisee without ever paying it back. For the promisor's ultimate end might be that of diminishing world hunger, and there seems to be no reason why the two cannot share that end. But it remains unclear just what sense of ‘cannot’ Kant is invoking (or should invoke) in suggesting that a promisee cannot share a false promisor's end.

On a reading suggested by Thomas Hill Jr., for Kant the promisee cannot share the promisor's end in that it is logically impossible for him to do so.Footnote ⁹ Suppose the promisor, a borrower, has the end of getting money from the promisee, a lender, without ever paying it back. The borrower makes a false promise in order to secure that end. At the time he makes a loan on the basis of this promise, the lender cannot himself share the end of the borrower's getting the money from him without ever paying it back, goes this reading. If the lender shared the borrower's end, then he would not really be making a loan. For according to our practice, it belongs to the very concept of making a loan, as opposed, say, to giving money away, that one believe that what one disburses will be repaid.

Given our aim of arriving at a plausible account of treating others merely as means, this interpretation of the promisee's inability to share the promisor's end is unhelpful. First, it just does not seem to be logically impossible for the lender to share the borrower's end. The borrower, let's say, is trying to secure the end of his getting money from the lender without ever repaying it in order ultimately to enjoy a vacation in Tahiti. The lender is also trying to realize this end – not so that the borrower can enjoy a vacation in Tahiti, but so that he, the lender, who despises the borrower, can revel in the demise of the borrower's reputation. The joy the lender would experience at the borrower's loss of reputation would more than compensate for his loss of the money, the lender might think. The lender shares the borrower's end. As far as I can tell, nothing in this case entails that in making the loan, the lender fails to believe that what he gives out will be repaid. It is easy to imagine him reflecting, with regret, that the borrower will probably pay him back and that his ultimate aim of enjoying the demise of the borrower's reputation will probably remain unfulfilled. It might not be a common occurrence, but it is logically possible for a lender to share a borrower's end of getting money from the lender without ever paying it back.

Second, this interpretation of the false promising case leads naturally to the view that a sufficient condition for an agent's treating another merely as a means is that it is logically impossible for the other to share the end the agent is pursuing in treating her in some way. However, not only this case but also other paradigmatic cases of treating another merely as a means fail to involve such logical impossibility. Take, for example, a loiterer who threatens an innocent passer-by with a gun in order to get $100. The sort of sufficient condition for treating another merely as a means that we seek should allow us to conclude that the loiterer is treating the passer-by merely as a means, for he is mugging her. But the sufficient condition on the table does not do this. It is improbable, but still logically possible, that the passer-by shares the loiterer's end of his getting $100.

One might reply that if the loiterer is to count as mugging the passer-by and thereby treating her merely as a means, it must not be the case that the passer-by shares the loiterer's end. For if she does, then he simply does not count as mugging her. It follows from the concept of mugging that the victim does not have the same end that the mugger is pursuing.

But this reply fails. Suppose the loiterer is the passer-by's nephew whom she hasn't seen in a few years and that on this dark, foggy night the two do not recognize one another. The aunt (the passer-by) has the end of her nephew (the loiterer) having $100, which is coincidentally precisely the amount of cash she has in her purse. She was planning to give the $100 to him for his birthday the next day. In these circumstances, it is surely possible for him to mug her in order to realize his end of getting $100. How else would we describe his waving a gun in her face and shouting to her to give him her purse? It is logically possible for two agents to share an end even in cases paradigmatic of one's using the other merely as a means. So the logical possibility account of end sharing is unhelpful when built into a sufficient condition for an agent's treating another merely as a means.

According to a different account, namely an interpretation of Kant offered by Christine Korsgaard, another cannot share the end an agent pursues in treating her in some way if how the agent behaves ‘prevents [the other] from choosing whether to contribute to the realization of that end or not’.Footnote ¹⁰ The lender in our example cannot share the borrower's end of getting money without ever repaying it; for the borrower's false promise obscures his end and thus prevents the lender from choosing whether to contribute to it. Therefore, implies Korsgaard, the borrower is in Kant's view treating the lender merely as a means.

This account of possible end sharing has unacceptable implications. Consider an agent, namely a customer in a restaurant seated at a table not far from the bar. She has the end of getting a clear view of the attractive bartender who is making drinks. The other is a large waiter who, seated between the customer and the bartender, is busy doing paperwork. Suppose that the customer orders a drink from the waiter just to get him to move out of her line of sight. The customer realizes that she could, but she chooses not to, tell the waiter that she is ordering a drink to get a good view of the bartender. She just goes ahead and orders the drink. The way she acts prevents him from choosing whether to contribute to the realization of her end. If Korsgaard is correct, then Kant is forced to embrace the following implausible view: since the waiter cannot share the customer's end, she is treating him merely as a means and thereby acting wrongly.

One might respond that the customer's behaviour does not really prevent the waiter from choosing whether to contribute to her end; it simply fails to facilitate his doing so. But if the customer knows, and it is reasonable to assume she would, that the waiter can find out what she is up to only if she tells him, then the way she acts does for all practical purposes prevent him from choosing to contribute to her end. Sometimes in pursuing an end an agent uses another without the other's being aware of his doing so. And, contrary to Korsgaard's Kant, there need be nothing morally problematic about this lack of awareness, even when it is within the agent's power to eliminate it.

Perhaps we should take from the false promising passage a third interpretation of possible end sharing, namely the view that the promisee cannot share the promisor's end in the sense that, in typical cases, it would be practically irrational for him to share this end. In typical cases, it would be irrational for the promisee to try to realize the end of making a loan that is never repaid. For this end's being brought about would prevent him from attaining other ends he is pursuing, ends such as buying new rose bushes, saving money for college, and, of course, just getting his money back.

The notion of irrationality at work here is familiar. In the Groundwork, Kant seems to embrace what Hill calls ‘the hypothetical imperative’, namely a principle that goes roughly like this: if you will an end, then will the means to it that are necessary and in your power, or abandon the end.Footnote ¹¹ Kant implies that the hypothetical imperative is a principle of reason: all of us are rationally compelled to abide by this principle. An agent would act contrary to the hypothetical imperative and thus irrationally by willing an end yet, at the same time, willing another end, the attainment of which would, he is aware, make it impossible for him to take the otherwise available means to his original end. An agent would violate the hypothetical imperative, for example, by willing to buy a car yet, at the same time, willing to use the money he reserved for the down payment to make a gift to his nephew. The Kantian hypothetical imperative implies that it is irrational to will to be thwarted in attaining ends that one is pursuing. In typical cases, if a promisee willed the end of a false promisor, she would be doing just that.

Against the background of this example, let us flesh out an account of treating another merely as a means. Let us specify the End Sharing account to be the following:

The notion of reasonableness at work here is non-moral. What it is reasonable for an agent to believe is roughly what the evidence available to the agent favours, given the information he has, his education, his upbringing, and so forth. On this usage, it might be reasonable for an agent to believe that another can (or cannot) share a particular end even if experts, say a group of leading psychologists, would believe the opposite.

Unfortunately, the End Sharing account has serious shortcomings. Suppose that Pete and Andre are competing in the men's singles final at the US Open tennis tournament. At stake is the number one ranking for the year, which each player has, and has announced publicly, as his goal. Pete is treating Andre as a means; for he is intentionally doing something to Andre, that is, trying to beat him, in order to secure his goal of being number one. But, according to the present account, Pete is also treating Andre merely as a means. In pursuing the end of Pete's being number one (e.g. by purposefully throwing the match) Andre would be willing to be thwarted in attaining his end of being number one. Andre cannot share Pete's end in the sense that it would be practically irrational for him to do so. Moreover, it would obviously be unreasonable for Pete to think that Andre can share his end. Pete knows that Andre too aims to be number one. So on the present account, it turns out that Pete's treatment of Andre is morally impermissible. In general terms, the account has the following implication. Suppose an agent is pursuing an end in a competition. If it is unreasonable for the agent to believe that his competitor can, rationally speaking, both pursue the agent's end and strive to secure his own end, the agent's action is morally impermissible. The End Sharing account entails that competitive behaviour of a common sort is wrong.

One might reply that though this implication initially seems to discredit this account of treating others merely as means, reflection reveals otherwise. If Pete's end was not to be number one, but rather to develop his capacities as a tennis player, then he would not be treating Andre merely as a means. For it is reasonable for Pete to believe that this is an end that Andre can share. (Even if Andre himself was playing for the number one ranking, he could, without practical irrationality, pursue at the same time the end of developing Pete's skills. After all, in the course of losing Pete might grow as a tennis player.) Competitors often aim ultimately at developing their own capacities, rather than at being number one. When they do strive to be the sole victors in a competition, it is not implausible to hold that they are treating their opponents merely as means and thus wrongly, concludes the reply.

In my view, Pete and Andre would in some sense be more virtuous if each could share the other's end. Competitors who each have as an end to develop their own capacities seem morally more attractive than ones who each have as an end to defeat their rivals. There is something admirable in holding that, ultimately, one is ‘competing against’ oneself. However, I think that most of us would find unacceptably strong the judgment that it is morally wrong to act as Andre and Pete do in the original example. Granted, some competitors do aim ultimately at improving their own skills, but others (and, it seems to me, a lot of others) aim ultimately to be victorious. To conclude that these others are thereby treating their opponents in a morally impermissible way seems far-fetched.

According to the End Sharing account, Kant suggests that an agent uses another merely as a means if it would be unreasonable for the agent to believe that the other can share the agent's end in using him, that is, pursue it without practical irrationality of the sort we have highlighted. Unfortunately, this interpretation yields counterintuitive results. So it makes sense to look elsewhere for a reading of Kant's false promising passage.

POSSIBLE CONSENT

According to Onora O'Neill, Kant suggests a different account of treating another merely as a means, namely (what I call) the Possible Consent account:

The account incorporates O'Neill's particular reading of possible consent. She implies that an agent can consent to a course of action only if it is possible for him to dissent from it. It is possible in the relevant sense for someone to dissent from a course of action, she says, only if he ‘can avert or modify the action by withholding consent and collaboration’.Footnote ¹³ According to O'Neill, if an agent deceives or coerces another, then the other's dissent is ‘in principle ruled out’, and thus so is his consent.Footnote ¹⁴ Suppose, for example, that an auto mechanic tricks a customer into authorizing an expensive repair. The customer does not really have the opportunity to dissent with the mechanic's action. For he does not know what her action is, namely one of lying to him about what is really wrong with his car. (If he did know what her action was, then he wouldn't be deceived.) Or suppose that a mugger approaches you on a dark street, points a gun at your torso, and tells you with chilling candour that unless you give him your wallet, he will kill you. He gives you no opportunity to avert or modify his action: you must either give him your money or forfeit your life. So in O'Neill's sense, you cannot consent to his action.

Is it a strength of the Possible Consent account of treating another merely as a means that, contrary to the End Sharing account, it does not imply that one need be doing so in striving to be the sole victor in a competition? Again, suppose that Pete is trying to defeat Andre in a tennis match in order to promote his aim of being the year's number-one-ranked men's player. At the time Pete's action takes place, Andre probably cannot avert or modify it. (If Andre refuses to play, for example, then Pete defeats him in a walkover.) So it appears that the Possible Consent account also implies that Pete is doing something wrong.

But it is, I believe, in the spirit of O'Neill's account to notice here that Andre has been able to consent to the competitive order inherent in the world of professional tennis. It has been open to him to remove himself from this order. He has had the ability to avert his being subject to rules according to which it is obviously legitimate and anticipated that his opponents on the court will try to maximize their ranking by defeating him. I believe that on O'Neill's considered view, Andre's possession of this ability entails that he can consent to Pete's treatment of him and that Pete is not treating him merely as a means (and thus wrongly). If an agent can consent to a set of rules according to which a particular kind of treatment of him is obviously both legitimate and anticipated, then, for purposes of the Possible Consent account, he counts as being able to consent to the treatment itself.

Although the Possible Consent account allows us to avoid the implausible conclusion that it is morally impermissible to pursue in a competition the goal of defeating another who is also striving for victory, it suffers from two difficulties.

An initial problem is easy to illustrate with an example. Suppose that I hail a cab. In order to get to the airport on time, I ask the driver to take a short cut through the city. But, unknown to me, the driver cannot refuse my request; he cannot avert or modify my use of him. For he has been hypnotized into being unshakably convinced that he cannot turn down any request I make. In such a scenario, although the driver cannot avert or modify my use of him by withholding consent and collaboration, it would be implausible to think that I have acted wrongly in my use of him. It is not I, but rather the hypnotist who has (presumably) acted wrongly in his use of the driver. It would be easy to multiply cases such as this in which an agent is unaware that another cannot avert or modify the agent's treatment of him but in which it seems implausible to conclude that the agent is treating the other wrongly. In such cases, the agent reasonably believes that the other can avert or modify his treatment of him. Here again reasonable belief is an epistemic rather than a moral notion. We will need to have this criticism in mind as we try to arrive at a plausible account of treating others merely as means below.

Another, far more serious, difficulty is the following. It is possible for a person not to be able to consent to the way an agent is treating her, yet nevertheless to be able to share the agent's end in treating her this way. Sometimes when this possibility is realized, it is implausible to claim, as does the Possible Consent account, that a person is being treated merely as a means and therefore in a morally impermissible fashion.

For example, suppose that during a run through a park you pass a man in jogging clothes lying at the side of a path. You are dismayed to find on closer inspection that he is neither conscious nor breathing. You call out for help, but no one responds. You give him CPR and, after some very tense moments, get his breathing going again. The unconscious jogger cannot consent to your way of treating him; he cannot avert or modify your action of giving him CPR in order to save his life. But he is very likely to be able to share your end. It is very likely that if he pursued the end of having his life saved, he would not thereby be willing to be thwarted in his other projects. Partly for that reason, it seems wildly implausible to contend that your attempt to save him was morally impermissible. The point is not that if a person can share the end an agent has in treating her in a way to which she cannot consent, then the agent's action must be morally permissible, but rather that it might be.

Someone might reply to the collapsed jogger case by claiming that what O'Neill really intended (or should have intended) to maintain was that a person treats another merely as a means if the very nature of the agent's action renders the other unable to consent to what the person does to him. The nature of your action, that is, trying to save the jogger's life, does not entail that the jogger cannot avert or modify it; what does that is simply that the jogger is unconscious.

Although this reply might be effective against the collapsed jogger case, it yields an account that still suffers from unacceptable implications. To cite a familiar example, if your spouse deceives you so that your birthday party will be a surprise, then the nature of her action (deception) renders you unable to avert or modify it. But you might be able to share your spouse's end in throwing the party, namely that of your enjoying your birthday. It seems counterintuitive to say that your spouse's action was morally impermissible. Or consider an action of deceiving a tipsy friend into believing that her husband already left with their car and that she must therefore have you drive her home. Your action itself prevents the friend from consenting to the way you are treating her, but she might be able to share your end – for example, if it is to get her home safely. Many of us would not judge your action to be morally impermissible. Finally, consider a scenario recently discussed by Derek Parfit: ‘I know that, unless I tell you some lie, you will believe truly that Brown committed some murder. Since you could not conceal that belief from Brown, he would then murder you as well.’Footnote ¹⁵ If I lie to you in order that you not be murdered, which is an end you can presumably share, then my action itself renders your consent impossible. But it is implausible to conclude that it would be morally impermissible for me to lie to you. That the nature of an action itself precludes consent does not entail that it is wrong.

HYBRID ACCOUNTS

We are trying to understand what the Kantian notion of treating someone merely as a means amounts to or should be taken to amount to. On both the End Sharing and Possible Consent accounts this notion generates counterintuitive results. The two key concepts Kant presents in the promising example are that of another's being able to share the end an agent is pursuing in using him in some way and that of another's being able to agree to the way an agent is using him. In some cases, such as that of the tennis competition, another cannot share an agent's end, but the other can agree to the way the agent is treating him. The End Sharing account implausibly implies that in all cases of this type the agent acts wrongly. The Possible Consent account avoids this difficulty, but also clashes with common sense. In some cases, such as that of an agent's lying to another in order to prevent that other from being murdered, another can share an agent's end but cannot agree to the way the agent is treating him. The Possible Consent account implausibly implies that in all cases of this type the agent acts wrongly.

In this section, I present an account of treating another merely as a means that, I believe, avoids the difficulties with the End Sharing and Possible Consent accounts. This account is somewhat complex. Before presenting it, it will be helpful to shed light on why a simpler, rather obvious response to these difficulties would be ineffective.

The obvious response is to combine the two accounts, after modifying them, of course, to meet some of our criticisms. An agent treats another merely as a means, we might say, if it is neither reasonable for him to believe that the other can consent to the agent's use of him nor that the other can share the end he is pursuing in using him.

Unfortunately, this Hybrid account also has unwelcome implications. Suppose a police officer has the end of preventing race-based attacks on law-abiding citizens. In pursuing this end, she arrests a white-supremacist, someone she believes (correctly) to be planning an attack on a church frequented by Asian Americans. According to the notion of possible consent we have adopted from O'Neill, the white-supremacist cannot consent; for he cannot ‘avert or modify’ the officer's action by ‘withholding consent and collaboration’. It is obviously reasonable for the officer to believe that he cannot consent. Moreover, in arresting the white-supremacist, the officer might be pursuing an end that it is unreasonable for her to believe the white-supremacist capable of sharing. Suppose, as the officer is aware, the white-supremacist's end in planning the attack was to get ‘revenge’ on a racial group that he thinks to be inferior to whites and thus undeserving of the rights and liberties its members possess. The officer would surely realize that it would be practically irrational for the white-supremacist to pursue his end of revenge and at the same time to will the officer's end of preventing race-based attacks on law-abiding citizens. For in willing the former, he would be thwarting his pursuit of the latter. Therefore, the Hybrid account forces us to embrace the counterintuitive conclusion that in making the arrest the officer is acting wrongly.Footnote ¹⁶

Fortunately it is not conceptually difficult to modify the account in order to meet the criticism. We balk at the idea that the officer treats the supremacist merely as a means and thus wrongly. A main reason we do so is, roughly, that what prevents the supremacist from being able to share the officer's end is his treating others, namely the churchgoers, merely as means. Our account needs to be sensitive to this sort of consideration. But the account would be circular if it invoked the notion of treating others merely as means; for this is the very notion it is supposed to clarify. In order to avoid circularity, the account is going to have to be somewhat intricate. Suppose an agent treats another in some particular way, that is, uses another. According to the Reinforced Hybrid account, the agent thereby uses the other merely as a means if it is reasonable for the agent to believe neither that:

1. (a) The other can consent to the agent's use of him or can share the end he is pursuing in using him. nor

2. (b) That which, rationally speaking, prevents the other from sharing the agent's end is the following: the other is himself using someone in pursuing an end, and it is reasonable for the other to believe neither that this person can consent to the other's use of him, nor share the end the other is pursuing in using him.Footnote ¹⁷

The term ‘rationally speaking’ in (b) is meant to invoke solely a requirement of prudential rationality discussed earlier, namely that expressed by the hypothetical imperative.

This account's application to the case of the white-supremacist is straightforward. It is reasonable for the officer to believe that what, in terms of prudential rationality, prevents the white-supremacist from being able to share her end of forestalling raced-based attacks is his pursuing the end of getting ‘revenge’ on a minority group. It would violate the hypothetical imperative for him both to pursue this end and to will the officer's end. Moreover, it is obviously also reasonable for the officer to believe that the white-supremacist knows full well that he is using members of the minority in pursuit of an end they cannot share and in a way to which they cannot consent. In effect, it is reasonable for the officer to believe that (b) is fulfilled. Therefore, we are not forced to say that she is treating the white-supremacist merely as a means and thus immorally.

The Reinforced Hybrid account also has the advantage of implying, quite plausibly, that certain acts of self-defence are morally permissible. Say that after being violently attacked, you use violence to repel your attacker. Your end is to escape without serious physical or financial injury; the attacker's end is to get your money. It may well be unreasonable for you to believe that the attacker can consent to the way you are treating him and that he can share the end you are pursuing in treating him in this way. However, it would be perfectly reasonable for you to believe that, by his own lights, what prevents him from sharing your end is his pursuing an end that someone he is interacting with (namely, you) cannot share and doing so in a way to which this person (you) cannot consent. So the account does not imply that in using violence to repel your attacker, you are treating him merely as a means and thus acting wrongly.

The Reinforced Hybrid account is not, of course, invulnerable to objection. One objection worthy of careful consideration stems from cases such as the following: you live out in the country with just one neighbour within miles. Your child has suffered a life-threatening injury, and you need to get him to the hospital right away. By far the fastest way to get him there is to drive him yourself. Unfortunately, your spouse is using your car on a business trip. So, you reason, the only way to get your son to the hospital in time is to take your neighbour's truck. You run over to his farm in order to ask him to lend it to you, but you cannot find him. He has left his keys in his truck, and you drive off in a frantic attempt to get your child the medical attention he needs.

On the Reinforced Hybrid account, it might turn out that you are treating your neighbour merely as a means and thus wrongly – at least if we assume that in taking his property, you are intentionally doing something to him. For it is not reasonable for you to believe that he can consent to your taking his truck, at least not on the understanding of possible consent we have been employing. And suppose that it is also not reasonable for you to believe that he can share your end of getting your child to the hospital as soon as possible. You are aware that his pursuing this end, say, by letting you use his truck, would prevent him from attaining another end he has set himself, namely that of getting top dollar for his produce. As you realize, if he misses the city market this afternoon, he will in all likelihood have to settle for less money. According to the account on the table, your taking your neighbour's vehicle would be morally impermissible.

It would be good, I think, to avoid this result. A modification of our understanding of what it means for another to be able to share an agent's end will help us to do so. Thus far we have stated that another cannot share the agent's end if the other cannot pursue this end without practical irrationality. The other cannot pursue this end without practical irrationality if his doing so would amount to his willing to be thwarted in attaining ends he is pursuing and thereby violate the hypothetical imperative. A quick look back at this imperative will help us to modify this position. If you will an end, commands the hypothetical imperative, then will the means to it that are necessary and in your power, or give up the end. Now suppose an agent has willed to do something. There are two things he can do which would bring his action into compliance with the hypothetical imperative. He can either take the means he believes to be necessary and in his power to the end (which, of course, would rule out his willing to be thwarted in attaining the end) or he can give up the end. This observation suggests a simple revision to our understanding of what it means not to be able to share an end. A person cannot share an agent's end, let us say, if, in light of knowledge of the agent's end and the means the agent is taking to attain it, she would be unwilling to give up some end of hers, namely one that it would be practically irrational for her to pursue at the same time she was pursuing the agent's end.

So, returning to our example, your neighbour cannot share your end of getting your child to the hospital, if in light of knowledge of your end and the means you are trying to take to it, that is, taking his truck, he would be unwilling to give up his end of getting top dollar for his produce at the city market. For if he would be unwilling to give up that end, then it would indeed be practically irrational for him to pursue your end of getting your child to the hospital in time. On this account of what it means to share an end, it would probably be reasonable for you to believe that your neighbour would be able to share your end. It would therefore be unlikely that you would be treating him merely as a means in taking his truck without his permission.

But the unlikely is not the impossible. What if you know your neighbour to be exceptionally greedy, cold and ill-disposed to you and to your family? Even if he realized that your child's life was at stake and that that was the reason you took his truck, he would be unwilling to give up his end of getting top dollar for his produce that afternoon, it is reasonable for you to believe. In that case, even on the modified account, it would not be reasonable for you to hold that he would be able to share your end. So you would be treating your neighbour merely as a means and thus in a morally impermissible way. To many of us, this seems to be an unwelcome result. Of course, it would not be difficult to construct further cases of this sort. What such cases have in common is the following: in pursuing an end, an agent is treating another merely as a means. But given the moral weight of the agent's end, the necessity in the effective pursuit of it of treating another merely as a means, and the relatively minor harm the other suffers in being so treated, it seems implausible to conclude that the agent is acting wrongly.

In order to avoid counterexamples such as this it might be necessary to transform the Mere Means Principle into a non-absolute constraint. One way of doing so would be to hold, along the lines sketched earlier, that an agent's treating another merely as a means is always morally wrong pro tanto, but sometimes, that is, in cases in which other, weightier moral reasons apply, not wrong all things considered. Taking your neighbour's truck would, on this modified account, presumably not be wrong all things considered. For though you have moral reason not to use your neighbour merely as a means in the way described, you have a greater moral reason to take the truck in order to save your child's life.

Of course, making this modification would prompt one to try to answer some very difficult questions that I cannot broach here. When, for example, would an agent's moral reason not to treat someone merely as a means get outweighed by some other moral reason he has? Would it get outweighed only when an agent's using another merely as a means would presumably not destroy the other's humanity, but would be necessary to preserve someone's humanity? What, then, of cases such as that described earlier in which someone's killing an innocent person is the sole means of saving millions?

In any case, the Reinforced Hybrid account tries to identify a certain kind of immoral (or at least pro tanto immoral) conduct, namely that of using others merely as means. The account does not aim to shed light on why, ultimately, it is wrong for an agent to use others in this way. But Kant suggests that the agent fails to respect the other's humanity, that is, his rational agency. For rational agency has the sort of value that demands respect.Footnote ¹⁸ If Kant's suggestion is on target, then this article has explored in detail two features of an agent's use of another that can contribute to a failure to respect his rational agency: the other's inability to consent to the agent's use of him and the other's inability to share the agent's end. If successful, the Reinforced Hybrid account interprets and combines these elements in a way that does a better job than its rivals of making plausible the claim that, in relatively ordinary circumstances, to treat another merely as a means is to treat him wrongly. The account re-establishes the credibility of the Mere Means Principle as a moral constraint.Footnote ¹⁹