How Do Reasons Transmit to Non-Necessary Means?

ABSTRACT

Which principles govern the transmission of reasons from ends to means? Some philosophers have suggested a liberal transmission principle, according to which agents have an instrumental reason for an action whenever this action is a means for them to do what they have non-instrumental reason to do. In this paper, we (i) discuss the merits and demerits of the liberal transmission principle, (ii) argue that there are good reasons to reject it, and (iii) present an alternative, less liberal transmission principle, which allows us to accommodate those phenomena that seem to support the liberal transmission principle while avoiding its problems.

KEYWORDS:

• reasons for action
• normative transmission principles
• instrumental reasons
• practical reasoning
• acting for a reason
• Too Many Reasons problem

1. Introduction

Some of our reasons for action are explained by the fact that the action in question is a means to something else that we have reason to do. For example, your reason to visit the dentist is explained by the fact that doing so is a means to avoiding future toothaches. But how, exactly, do reasons for actions transmit to reasons for taking the means to these actions? The following transmission principle seems plausible and widely accepted:

Necessary Means Transmission. If A has a reason to φ, and ψ-ing is a necessary means for A to φ, then A has a reason to ψ.¹

This principle can account for many of those normative phenomena that transmission principles should make sense of, but it is questionable whether it can account for all of them. In particular, it seems that taking means to an action favoured by a reason is often something that we have reason to do, even if none of these means is necessary. In light of this and similar observations, some authors have proposed transmission principles that are much more liberal than Necessary Means Transmission, and that apply even when ψ-ing merely facilitates φ-ing. Here we will focus on the following principle, which best captures the spirit of various similar proposals discussed in the recent literature:

Liberal Transmission. If A has a final reason to φ, and ψ-ing is a means for A to φ, then A has a reason to ψ.²

A final reason is, roughly speaking, a non-instrumental reason; we will say more on this in section 2. Whether or not Liberal Transmission holds true is not only relevant for those who take an intrinsic interest in the correct principles of practical deliberation; the principle also figures in other important philosophical debates. One prominent example is the debate over whether the ‘wide-scope’ interpretation of the principle of instrumental rationality can avoid implausible ‘bootstrapping’ of reasons.³ Further contexts in which Liberal Transmission plays a crucial role include Schroeder’s Humean account of reasons and Way’s argument for scepticism about so-called reasons of the ‘wrong kind’.⁴

While Liberal Transmission has explanatory virtues, it has also been argued to have counterintuitive implications. Our aim in this paper is to assess the merits and demerits of Liberal Transmission and to suggest more plausible alternative principles. We start with some preliminary remarks about terminology and the exact content of Liberal Transmission (section 2). Subsequently, we discuss the objection that Liberal Transmission entails ‘too many reasons’. We strengthen the case of those who reject the principle due to its implausible implications, by casting doubt on a common pragmatic strategy to which Schroeder and others appeal in order to explain away the intuitions that the reasons implied by Liberal Transmission do not exist (section 3). In the paper’s second part, we examine three important arguments in favour of Liberal Transmission and argue that they fail to make a convincing case for it (sections 4–6). In the course of the discussion, we also present alternative, less liberal, transmission principles. We argue that these allow us to accommodate the phenomena that have been argued to support Liberal Transmission, without being vulnerable to the ‘too many reasons’ problem. We close by considering the question of whether the phenomenon of instrumental transmission can be captured by one single principle (section 7).

2. Preliminaries

Let us start with some clarificatory remarks about terminology, the exact content of Liberal Transmission, and the aims of this paper. First, we use the expression ‘ψ-ing is a means for A to φ’ broadly, as roughly equivalent to ‘ψ-ing is something that A can do intentionally and that will help to bring it about that A φ-s.’ We understand this notion in a way that allows not only actions, but also omissions, to be means. What we say will be neutral between probability-raising and other accounts of what it is for something to be a means.⁵

Second, we call a reason to ψ an instrumental reason if and only if it is explained by the fact that ψ-ing is a means to something else that there is reason to do.⁶ A reason that is not explained in this way is a final reason, and a reason that is referred to in the explanation of an instrumental reason is a source reason. Given these definitions, a source reason (but not a final reason) might itself be an instrumental reason. However, as we have formulated Liberal Transmission, only final reasons can be source reasons for instrumental reasons generated by this principle.

This restriction is necessary in order to avoid what Bedke calls ‘the problem of subversion’,⁷ which can be illustrated by the following example. Suppose that you have a final reason to make your friend happy. A means to doing this is your buying him a present, and so Liberal Transmission entails that you have reason to do this. But, now, one means to buying your friend a present is to steal money from him and use it to buy the present. To make things worse, one means to steal money from your friend is actually to kill him. If we allow all kinds of reasons to give rise to instrumental reasons in this way, we end up deriving, in very few steps, a reason to kill your friend from a reason to make him happy. The problem is that we are deriving instrumental reasons from instrumental reasons and thereby deriving reasons that fail to serve any final reason. We can avoid this result by restricting the antecedent of a liberal transmission principle to final reasons. Since killing your friend is no means to making him happy, Liberal Transmission does not let us derive a reason to kill your friend from a reason to make him happy.

Third, we understand Liberal Transmission in such a way that the instrumental reasons that it generates cease to exist once A conforms to the relevant final reason. This is already reflected in the above formulation, for once A conforms to the final reason to φ, what have been means for A to φ before are no longer means for A to φ. It follows that Liberal Transmission does not license the generation of instrumental reasons from reasons to which one has already conformed.⁸

Fourth, some authors seem to suggest further restrictions. Sometimes the source reasons of Liberal Transmission are restricted to undefeated final reasons, and sometimes the means are restricted to sufficient means.⁹ We focus on the unrestricted Liberal Transmission principle formulated above, since most of the arguments brought forward in support of a liberal transmission principle (see sections 3–5) suggest this unrestricted version. However, the considerations advanced here against Liberal Transmission also apply to more restricted versions of this principle. So, even though Liberal Transmission will be our primary focus in what follows, the relevant considerations both for and against it can be generalised.

Finally, throughout this article we bracket the question of how much of a reason’s weight gets transmitted to the means. Although this is a question of great interest for a theory of instrumental transmission, addressing it here would complicate issues in a way that is not necessary for the purposes of this paper.

With these clarifications in mind, we now present what we take to be the most important objection to Liberal Transmission, before discussing the considerations that seem to speak in its favour and then presenting our alternative principles.

3. Too Many Reasons

Liberal transmission principles have counterintuitive implications. As John Broome [2005: 7] points out, it does not seem plausible to say that one has reason to kill oneself as a means to avoiding a feeling of hunger if one could simply have lunch instead. Similarly, Simon Rippon [2011: 17] objects that liberal transmission principles license the inference from ‘I have a reason to tell a joke’ to ‘I have a reason to tell a racist joke.’

Since it might be questioned whether these examples are based on final source reasons, we will introduce two further cases to illustrate the problem. Our examples are based on source reasons that we take to be relatively uncontroversial candidates for final reasons: the first is a reason to keep your promise; the second is a reason to avoid pain. Regarding the first of these, suppose that you have promised your friend a surprise on his birthday. One way to keep your promise is to cut off your hand and give it to your friend as a birthday present, and thus Liberal Transmission licenses the conclusion that you have a reason to do this. Regarding the second final reason, suppose that you—the president of a powerful country—have a mild headache and that you could get rid of it either by taking a headache pill, or by launching a nuclear missile that would immediately kill you and hundreds of thousands of other people. Since this is a means to avoid being in pain, Liberal Transmission entails that you have reason to launch the nuclear missile. More generally, if you have a final reason to φ, Liberal Transmission implies that you have an instrumental reason to ψ, for any ψ that is an almost ineffective, highly inefficient, or otherwise extremely objectionable means to φ-ing. It seems, however, that, in many of these cases, the mere fact that these actions are a means to doing what we have final reason to do is not enough to give us a reason to perform them.

There are two responses available for proponents of Liberal Transmission. The first is to throw Liberal Transmission into reverse. You think that you have no reason to launch the nuclear missile? Perhaps, then, it is not strictly speaking true that you have a final reason to avoid pain, but rather that you have a final reason to avoid pain in ways that do not involve unnecessary burdens for yourself and others. By the same token, one might reply that it is not strictly speaking true that you have a reason to keep your promise, but rather that you have a reason to keep it in a way that does not bring misery upon yourself or others. Once the favoured action is sufficiently specified, a lot of implausible reason claims will not follow anymore—or so the proponent of Liberal Transmission might argue.¹⁰

Even if we are willing to accept such a redescription in particular cases, the response does not seem to work as a general strategy for dealing with all of the potential counterexamples. Liberal Transmission implies that there is a reason to ψ for any action ψ that is an almost ineffective, highly inefficient, or otherwise objectionable way of helping to bring about what one has final reason to do. Hence, potential counterexamples of this type can always be generated unless one maintains that, strictly speaking, specifications of actions favoured by final reasons must always be qualified by the complex phrase ‘in a way that neither involves nor requires taking ineffective, highly inefficient, or otherwise objectionable means’. This requirement strikes us as ad hoc and artificial. It leads to implausible hyper-specified descriptions of final reasons for actions, and it conflicts with many pretheoretically intuitive judgments about what we have reason to do.

The second and more common response to the counterexamples is to accept the implications and to argue that intuitions to the effect that the relevant reasons do not exist are not to be trusted. According to this strategy, our intuitions about whether there is a reason for an action in a particular context do not reliably differentiate between circumstances in which we have no reasons at all for a given action and circumstances in which we have massively outweighed reasons for this action, which are reasons that are very weak in comparison to those reasons that suffice to decide the case under consideration.

Why should this be so? According to a popular proposal put forward by Mark Schroeder, we can systematically explain away intuitions about the non-existence of reasons by appealing to pragmatic factors that concern whether it is appropriate to call certain facts reasons in a conversational context. Schroeder argues that when people say that there are reasons to do something, they usually mean to imply that these reasons have significant weight and are deliberatively relevant by playing a role in determining what one should do. According to Schroeder, this in turn explains why it seems unnatural to assert that there is a reason to perform a certain action if the relevant reason is a (comparatively) very weak reason that does not play a significant role in determining the overall balance of reasons in a certain situation. In Gricean terms, the crucial idea is that asserting the existence of a reason standardly carries the conversational implicature that the reason is not massively outweighed. That it seems odd to assert the existence of a reason is thus taken to be compatible with there being such a reason. Schroeder [2007: 92–7] maintains that this suffices to debunk intuitions about negative reason existentials, which are taken to essentially depend on the felt oddity of making certain reason-claims. Applying the debunking strategy to potential counterexamples to Liberal Transmission, one might thus argue that what makes it unnatural to assert, for example, that there is a reason to launch the nuclear missile because one has a headache is simply the fact that the reason is massively outweighed.¹¹

In reply to such attempts to debunk intuitions about negative reason existentials, we first want to point out that it is very often not misleading to call massively outweighed reasons ‘reasons’ in standard conversational contexts (that is, contexts in which the conversational implicatures of reason statements have not been cancelled). We take it that the reasons for saving one thousand valuable paintings from being destroyed, for example, massively outweigh the reasons for saving one other valuable painting, but, contrary to what the debunking strategy seems to imply, it does not seem odd to assert that we have a reason to save the single painting as well in standard conversational contexts. Since this observation does not depend on exceptional features of the chosen example, it casts doubt on the assumption that asserting the existence of a reason standardly carries the pragmatic implicature that this reason is not massively outweighed or that it makes a significant contribution to what one ought to do in the case under consideration. We are thus sceptical that the intuition to the effect that (for example) you have no reason to launch the nuclear missile can be explained in the way that Schroeder’s approach suggests.

Perhaps the idea that reason statements standardly carry an implicature of deliberative relevance can be defended if one avoids spelling out deliberative relevance in the specific (comparative) terms that Schroeder suggests. However, it is far from obvious that a suitable alternative criterion is available, and it is the proponents of Liberal Transmission who bear the burden of providing one. More importantly, we believe that, even if we grant (for the sake of the argument) that there is a plausible way to spell out a deliberative relevance implicature for reason statements, we should hesitate to assume that this allows us to explain all relevant intuitions about negative reason existentials. Since the (alleged) standard conversational implicature of ‘There is a reason to φ’ could be cancelled, it is possible for us to address the question of whether some agent has a reason to (for example) launch a nuclear missile in a conversational context in which we make sure that the suggested implicature is cancelled. Thus, we might say ‘Please let it be understood that, when we talk about a reason in what follows, we do not mean to imply that it is sufficiently strong to play a significant role in practical reasoning’, and we might then go on to ask whether one has a reason to launch a nuclear missile, given that doing so would put an end to one’s headache. That there is a reason to launch a nuclear missile still strikes us as highly counterintuitive in such a context. It is difficult to see why intuitions about whether or not one has such a reason should not be reliable under these circumstances, when the suggested conversational implicatures that might lead us astray in our judgment have been explicitly cancelled.

Hence, the pragmatic debunking explanation for why it seems so plausible to deny that there are instrumental reasons to take certain means in the cases under discussion is much less straightforward than it might seem at first glance, and the ‘too many reasons’ problem for Liberal Transmission retains its bite. Although it does not amount to a full-fledged refutation of this principle, we think that a commitment to Liberal Transmission and thus a commitment to debunking (for example) the intuition that one does not have a reason to launch a nuclear missile just because one has a mild headache, or that one does not have a reason to cut off one’s hand just because one has promised to surprise a friend, constitutes a considerable disadvantage for a theory. Since both of our examples appealed to undefeated source reasons and sufficient means, the same applies to liberal transmission principles that are restricted in these ways.

It would be premature to reject Liberal Transmission, however, before looking into the considerations that seem to support this principle. As noted above, some philosophers believe that there are normative phenomena of which we cannot make sense without committing ourselves to Liberal Transmission, and, in what follows, we will address whether this is indeed the case by examining three such suggestions in more detail.

4. First Argument: Instrumental Reasons without Necessary Means

In practical reasoning, it seems that we need to suppose that reasons transmit to more than just necessary means. Very often, there is a variety of ways to conform to a final reason, rather than a single necessary one, and we seem to have reasons to pursue such ways at least in some cases. It is thus natural to assume that there is some general principle that allows the transmission of reasons to non-necessary means and that Necessary Means Transmission is not enough to account for all of the instrumental reasons that we have. Of course, it would not strictly follow from this that Liberal Transmission is true; nonetheless, it challenges those who deny Liberal Transmission to offer some viable alternative.

We believe that this challenge can be met. The first point to note here is that those who reject the idea that reasons transmit to just any means might still embrace the idea that reasons transmit to the best means or the optimal means:

Best Means Transmission. If A has a final reason to φ, and ψ-ing is a necessary part of every optimal sufficient means to φ-ing, then A has a reason to ψ.

Optimal Means Transmission. If A has a final reason to φ, and ψ-ing is a necessary part of an optimal sufficient means to φ-ing, then A has a reason to ψ.

These principles raise the question of the conditions under which a sufficient means can be said to be optimal. Although we cannot offer a definition, we suggest that the relevant intuitive notion of a ‘good means’ involves considerations not only of effectiveness in securing the relevant end, but also of efficiency in using agential resources and of conduciveness to conformity with final reasons more generally. A less efficient means may be, on the whole, better than a more efficient means if the latter precludes a valuable option that the former does not preclude.

To have a useful label, let us call Necessary Means Transmission, Best Means Transmission, and Optimal Means Transmission ‘conservative’ principles of instrumental transmission. Conservative transmission principles can vindicate a number of intuitive judgments about instrumental reasons for taking non-necessary means, but they do not seem to capture all important judgments to this effect. In particular, it seems that agents can be said to conform to instrumental reasons even if the means that they take are suboptimal. However, it is important to note that this neither implies nor requires for its explanation that, whenever ψ-ing is a means to an action one has final reason to perform, one also has a reason to ψ. It is enough to assume that, whenever we have final reason to φ, we have reason to perform an action of the type taking a means to φ-ing. That is, it is enough to appeal to the following transmission principle for reasons, which we suggest as an alternative to Liberal Transmission:

Generic Instrumental Reason. If A has a final reason to φ, then A has a reason to take means to φ-ing.¹²

To clarify, Generic Instrumental Reason does not state that we have a reason for each action ψ that is a means to an action φ that we have final reason to perform. It merely claims that we have reason to perform the act of taking a means to φ-ing in such cases. It might nevertheless be objected that Generic Instrumental Reason entails Liberal Transmission rather than being an alternative to it. This would be so if Generic Instrumental Reason implied, for each means to φ-ing, a reason to take it. But whether this is so depends just on the truth of Liberal Transmission (or close relatives of this principle). If ψ-ing is a means to φ-ing, then ψ-ing is also a means to the act of taking a means to φ-ing, which, according to Generic Instrumental Reason, is an action that one has a reason to perform. But we cannot conclude from this that one also has a reason to ψ without relying on Liberal Transmission, which obviously cannot be presupposed at this stage of the discussion.¹³

Quite generally, we believe that, over and above the conservative transmission principles, Generic Instrumental Reason is a natural and plausible alternative to Liberal Transmission that can play a crucial role in accounting for the phenomena that might be taken to support Liberal Transmission. This is why, in what follows, Generic Instrumental Reason will also figure in our criticisms of the second and third arguments in favour of Liberal Transmission.

5. Second Argument: More Reason to Take the Means than to Take No Means

One might still contend that there are facts about instrumental reasons that specifically support Liberal Transmission and cannot be accounted for by our alternative principles. One argument to this effect is that, for many suboptimal means, it seems true that we have more reason to take these particular means than to take no means at all. This seems to presuppose that we must have some reason at least for the particular suboptimal means. ¹⁴

Suppose, for that matter, that you can rescue a person on an island that is miles away either by taking a boat, which would bring you to the island in a minimal time, or by swimming, which would take several hours and be extraordinarily exhausting. In this case, it seems compelling to judge that you have more reason to swim to the island than to take no means to saving the person. It is not obvious how to make sense of this merely on the grounds of the conservative transmission principles and Generic Instrumental Reason.

Let us consider this idea in more detail. The assumption under consideration must be stronger than the claim that, other things being equal, we have more reason to perform the act of taking a means to what we have final reason to do than to take no such means, since Generic Instrumental Reason provides a straightforward explanation of this assumption. What is at issue is rather the claim that, other things being equal, if one has a final reason to φ, then, for every action ψ that is a means to φ-ing, one has more reason to ψ than reason to take no means to φ-ing.

This claim contains an important ‘other things being equal’ clause: it is clearly not always the case that one has more reason to ψ than reason to take no means to φ-ing if one has a final reason to φ and ψ-ing is a means to φ-ing. It is not the case if the reasons against ψ-ing are stronger than the reasons in favour of φ-ing: for example, you do not have more reason to launch the nuclear missile than to take no means to avoid being in pain. It is also not the case when the reasons against φ-ing outweigh the reasons in favour of φ-ing: if, all things considered, you ought not to rescue the person on the island (for example because doing so would prevent you from rescuing your own children on the mainland), then you also do not have more reason to swim to the island than to take no means to rescuing the person on the island. Hence, the assumption on which the second argument for Liberal Transmission rests needs to be restricted as follows:

More Reason Claim (= MRC). If A has an undefeated final reason to φ, and if ψ-ing is a means for A to φ, and if A does not have reasons against ψ-ing that are at least as strong as A’s reasons for φ-ing, then A has more reason to ψ than reason to take no means to φ.

Liberal Transmission can be used to provide an explanation of why MRC is true. However, one can also explain everything that needs to be explained with regard to MRC without relying on more than Generic Instrumental Reason. We contend that sentences of the form ‘There is more reason to φ than reason to ψ’ admit of two different interpretations. On the first reading, MRC turns out to be uncontroversial and can be given a straightforward explanation by appealing to Generic Instrumental Reason. On the second reading, MRC cannot be presupposed as a normative explanandum in the debate about Liberal Transmission.

To illustrate this ambiguity, suppose that you have a strong reason to help your neighbour Paul with his gardening work (you have promised to help him), and that you also have a weaker reason to help your neighbour George with his gardening work (your help would make it less wearisome for him). Suppose further that you have a very strong reason not to help Paul (your ex-spouse will be at his place, and your joint presence will cause a lot of trouble for everyone involved). Finally, assume that this reason outweighs the reason that you have in favour of helping Paul, and assume that there are no further relevant reasons involved. Do you have more reason to help George than to help Paul in this scenario?

Although there is some temptation to answer ‘no’, the most natural answer seems to be ‘yes’. This is easily explained by distinguishing between two ways of understanding the phrase ‘more reason to’. The most common way of using sentences of the form ‘A has more reason to φ than reason to ψ’, we take it, is to make comparative judgments about the overall balance of reasons that count for and against the various actions under consideration, which are comparative judgments of overall choiceworthiness. Understood in this way, there is more reason to φ than reason to ψ if, and only if, φ-ing is preferable to ψ-ing in light of all of one’s reasons. Call this the broad interpretation of ‘more reason’ statements. The ‘more reason’ phrase can also be used in order to express comparative judgments about the overall strength of just those considerations that specifically count in favour of the actions under consideration. According to this narrow interpretation, ‘There is more reason to φ than reason to ψ’ is equivalent to ‘The reasons in favour of φ-ing are stronger than the reasons in favour of ψ-ing’.

Following the broad interpretation, you have more reason to help George than to help Paul, since in light of all of the reasons for and against both courses of action, helping George is the preferable option. According to the narrow interpretation, however, you have more reason to help Paul than to help George, since the reasons in favour of helping Paul are stronger than the reasons in favour of helping George.

One crucial difference between these two interpretations is that only the narrow understanding of ‘There is more reason to φ than reason to ψ’ implies that there is a reason to φ. The broad interpretation allows us to say, of a pair of actions φ and ψ, that there is more reason to φ than reason to ψ even if there is no reason at all to φ, as long as there are suitable reasons against ψ-ing.

In order to see why this is so, consider actions that are such that there are no reasons either for or against them (‘perfectly neutral acts’, in what follows). Examples might include touching one’s own nose, or staring at the wall for five seconds (under normal circumstances). Given the broad interpretation, it is clear that one has more reason to perform a perfectly neutral act than to kill an innocent person, since, given the strong reasons against killing an innocent person, performing a perfectly neutral act is preferable to killing an innocent person in light of all of one’s reasons. Since, by definition, there are no reasons to perform perfectly neutral actions, the broad interpretation of ‘There is more reason to φ than reason to ψ’ does not imply ‘There is a reason to φ.’¹⁵

Since the ‘more reason’ phrase is ambiguous, MRC allows for both a broad and a narrow interpretation. On the broad interpretation, MRC seems beyond reasonable doubt. However, it is easy to account for this reading of MRC without going beyond Generic Instrumental Reason. Consider again the rescue case. If you have a final reason to save the person on the island, Generic Instrumental Reason implies that you also have a reason to take a means to saving that person, and this reason is plausibly also a reason against taking no means to saving the person.¹⁶ That the reasons for saving the person are stronger than the reasons against swimming to the island suggests that the reasons against taking no means to saving the person are also stronger than the reasons against swimming. This is enough to conclude that swimming is preferable to taking no means, in light of all reasons. And this is equivalent to the claim that there is more reason to swim to the island than to take no means to saving the person, on the broad interpretation of the ‘more reason’ phrase. Thus, even though swimming to the island is a suboptimal means to saving the person, we have explained why there is (on the broad interpretation) more reason to swim than to take no means to saving the person, without appealing to Liberal Transmission.

On its narrow interpretation, the More Reason Claim is equivalent to this:

MRC*. If A has an undefeated final reason to φ, and if ψ-ing is a means for A to φ, and if A does not have reasons against ψ-ing that are at least as strong as A’s reasons for φ-ing, then A has a reason to ψ, which is stronger than any reasons A might have for taking no means to φ-ing.

Generic Instrumental Reason is of no help in accounting for MRC*. However, it is difficult to see how proponents of Liberal Transmission could substantiate MRC* without presupposing what is at issue in the present context. For this reason, we do not believe that MRC* can be taken for granted as an explanandum in the debate over whether we need to go beyond Generic Instrumental Reason and the conservative transmission principles in order to give a satisfying account of instrumental reasons.¹⁷

To sum up, on its broad interpretation, the More Reason Claim can be accounted for by Generic Instrumental Reason. On its narrow interpretation, the More Reason Claim cannot serve as an explanandum in a debate concerning Liberal Transmission. The second argument for Liberal Transmission therefore fails to support this principle, too.

6. Third Argument: Intentionally Taking the Means Amounts to Acting for a Reason

The third and final argument for Liberal Transmission that we want to consider has origins in considerations proposed by Joseph Raz [2005b: 8–9], but we will present it in a slightly different way. The general idea is that intentionally and knowingly taking means to actions that one has final reason to perform itself amounts to acting for a reason, and that accepting this involves a commitment to Liberal Transmission. Here is a way to spell out this argument explicitly:

1. If A’s ψ-ing is guided by her awareness of the fact that ψ-ing is a means for her to do what she has final reason to do, then, in virtue of this, A can correctly be described as ψ-ing for a reason.

2. If A can correctly be described as ψ-ing for a reason in virtue of the fact that A’s ψ-ing is guided by her awareness of the fact that p, then p provides a reason for A to ψ.

3. Therefore, the fact that ψ-ing is a means for A to do what she has final reason to do provides a reason for A to ψ.

The conclusion can be avoided by restricting premise (1) to necessary, best, or optimal means. However, we take it that the premise has considerable plausibility in its present form. We propose to reject premise (2) instead.

One might object to (2) that A can be correctly described as acting for a reason in virtue of being guided by her awareness of p if she falsely believes that p provides a reason. But if this objection is valid, the argument could plausibly be rephrased in a way that assumes that A has no false beliefs, and hence we will take this assumption for granted in what follows.

Premise (2) entails that, for a consideration to be the one in virtue of which an action counts as being done for a reason, it must (if true) provide a reason for this action. Two aspects of this claim may reasonably be denied. For one thing, it seems that a true proposition may be the relevant guiding consideration without providing a reason. For example, that there is a reason to φ may be the guiding consideration in virtue of which an agent counts as φ-ing for a reason, but it is far from clear that this consideration provides a reason to φ. More importantly for our purposes, an action φ can be performed for a reason (in virtue of being guided by the awareness of some fact) without there being a reason that specifically counts in favour of φ-ing at all. As we will argue presently, all that needs to be the case is that φ-ing is a way of doing what one has reason to do.

What seems to drive premise (2) is the idea that, in order to understand the guiding role of that particular kind of deliberation about reasons for φ-ing and non-necessary means to φ-ing that is often involved in acting for a reason, we need to assume that such deliberation leads one to recognise reasons that count in favour of those actions that are means to φ-ing. If being aware of a reason to φ and of the fact that ψ-ing is a means to φ-ing does not also potentially disclose a reason that specifically favours ψ-ing, it might seem that this awareness cannot guide one all of the way towards ψ-ing. For, in this case, there remains a normative gap between what the agent is guided towards and what the agent has reason to do.

However, note that the same problem does not arise when we consider an agent’s reflection about means/end-relations and decisive reasons. A normative gap between what one has decisive reason to do and what one is guided towards if one reflects on one’s decisive reasons and the relevant means/end-relations is not problematic at all. In particular, one can be guided towards ψ-ing by recognition of the fact that one has a decisive reason to φ and the fact that ψ-ing is a way of φ-ing, even in cases where there are no decisive reasons to ψ, as long as ψ-ing is a way of doing what one has decisive reason to do. Buridan’s ass missed this point when starving to death in front of two equally attractive piles of hay. He could have easily avoided this by recognising that, even though he did not have decisive reason to eat one particular pile, doing so would still be a way of doing what he had decisive reason to do—namely, to eat a pile of hay.

Now, if the normative gap between φ-ing and ψ-ing does not prevent one from being guided towards ψ-ing by one’s awareness of decisive reasons for φ-ing and means/end-relations between φ-ing and ψ-ing, then we should assume that the same point applies to non-decisive reasons as well. That is, we should assume that a normative gap between φ-ing and ψ-ing does not prevent one from being guided towards ψ-ing by one’s awareness of pro tanto reasons for φ-ing and means/end-relations between φ-ing and ψ-ing.

Indeed, this is just what we should expect when revisiting the examples discussed above in connection with the Too Many Reasons problem. These examples represent cases in which it seems plausible to describe agents as ψ-ing for a reason, but not at all plausible to assume they have a reason to ψ. It is plausible, for example, that a president who launches a nuclear missile in order to get rid of his headaches is indeed acting for a reason—but it is not plausible to assume that he really has a reason to launch a nuclear missile just because this is one way of getting rid of his headaches.

These considerations support the view that intentionally ψ-ing for a reason, even for the person with all relevant true and no relevant false beliefs, does not require the existence of a reason that specifically counts in favour of ψ-ing, but merely entails that ψ-ing is a way of doing what one has reason to do. In light of these considerations, premise (2) needs to be revised as follows:

(2)* If A can correctly be described as ψ-ing for a reason in virtue of the fact that A’s ψ-ing is guided by her awareness of the fact that p, then, in virtue of p, ψ-ing is a way for A to do what A has reason to do.

In order to see how the retreat from (2) to (2)* successfully blocks an inference to Liberal Transmission, recall that Generic Instrumental Reason guarantees that there is always an action supported by a reason in cases where ψ-ing is a means to an action φ that one has final reason to perform: the relevant action just is the act of taking a means to φ-ing.

Hence, if A is guided by her awareness of the fact that ψ-ing is a means for her to do what she has final reason to do, and if A, in virtue of this fact, can truly be said to ψ for a reason, then (2)* implies that ψ-ing is a way for A to do what she has reason to do—which is true, since ψ-ing is a way of taking some means to φ-ing. Consequently, that A ψ-s for a reason whenever her ψ-ing is guided by the consideration that ψ-ing is a means to do what A has final reason to do does not imply that A has a reason that specifically counts in favour of ψ-ing. It only implies that ψ-ing is a way for A to do what she has reason to do—from which we cannot infer a reason to ψ without presupposing Liberal Transmission.

Assumption (2)* and Generic Instrumental Reason suffice to make good sense of the idea that the consideration that an action ψ is a non-necessary means to doing what one has final reason to do can guide us towards ψ-ing in a way that amounts to acting for a reason. We do not have to assume that such guidance requires reasons in favour of each particular means that we might deliberately take, as premise (2) implies. The suggested inference to Liberal Transmission can thus reasonably be resisted, by rejecting premise (2) in favour of (2)*

7. Conclusion and Outlook

We hope to have shown that a commitment to Liberal Transmission is not forced upon us by those phenomena that an account of instrumental reasons needs to accommodate, and that rejecting this transmission principle is a viable option for which good reasons can be given. In addition, we have argued that other principles of instrumental transmission can account for a number of ideas that seem to support Liberal Transmission in the first place. On the one hand, there are the more conservative principles that allow transmission from final reasons to reasons for necessary, best, or optimal means. On the other hand, there is the generic principle that allows transmission to a reason for the act-type of taking a means, but not for particular means.

It is natural to ask whether these principles can be unified or reduced to each other, and we will conclude by briefly considering this question. The first point to note is that all necessary means are trivially necessary parts of every optimal sufficient means, and necessary parts of every optimal sufficient means are trivially necessary parts of some optimal sufficient means. Necessary Means Transmission can therefore be derived from Best Means Transmission, which in turn can be derived from Optimal Means Transmission.

How does Generic Instrumental Reason fit into this picture? Can it be derived from one of the conservative transmission principles? This will be so, if, for any action φ, taking some means to φ-ing is a necessary means to φ-ing or a necessary part of taking an optimal sufficient means to φ-ing. Whether this is indeed the case depends on subtle questions concerning the proper understanding of the relevant notion of a means and, in particular, on whether it makes sense to say of an action that it facilitates its own performance. Such questions are beyond the scope of this paper, however. We think that there is at least some reason to be optimistic that Generic Instrumental Reason can be derived from the conservative principles, but this is an issue that we must leave aside for another occasion.

In any case, we hope to have shown that a theory of instrumental transmission that appeals to no more than the conservative principles and the generic principle is a serious alternative to theories appealing to liberal transmission principles.¹⁸

Disclosure Statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

Work on this article has been supported by the Deutsche Forschungsgemeinschaft (DFG Project ‘Principles of the Deliberative Ought’), project no.275667980, and by the Einstein Stiftung Berlin, project no.8611000199. We acknowledge support by the Open Access Publication Fund of Humboldt-Universität zu Berlin.

Notes

1 For similar transmission principles about both reasons and oughts, see, e.g., Darwall [1983: 16], Setiya [2007: 660], Bratman [2009: 424], Schroeder [2009: 234, 245], Way [2010: 225], Scanlon [2014: 85], Kiesewetter [2015, 2018], and Kolodny [2018: sec. 2]. Some of these authors maintain that reasons transmit with equal weight to necessary means, but this claim is more controversial. See White [2017] and Kolodny [2018: secs. 2, 3] for criticism, and Kiesewetter [2015, 2018] for a defence, of this stronger principle. Throughout the paper, we use ‘A’ and ‘B’ as variables for agents and ‘φ’ and ‘ψ’ as placeholders for action terms. We treat ‘A has a reason to φ’ and ‘There is a reason for A to φ’ as equivalent.

2 Compare ‘People have reason to do what will bring them into conformity with reasons which apply to them’ [Raz 2005a: 3]; ‘One has reason to take the means to what one has ultimate reason to do’ [Bedke 2009: 678]; and ‘If there is a reason to A, then the fact that B-ing facilitates A-ing is a reason to B’ [Way 2012: 494]. For similar proposals, see also Bedke [2017] and Kolodny [2018: sec. 6]. We borrow the name ‘liberal transmission’ from Rippon [2011: 6–7], who uses it for a related principle entailed by the one to which we here refer by that name.

3 See, especially, Broome [2005], Bedke [2009: 678–86], Rippon [2011], and Kiesewetter [2017: 92–8]. Raz seems to presuppose Liberal Transmission in his argument against the wide-scope account [2005b: 11–14], but Liberal Transmission should be distinguished from his ‘facilitative principle’ (see [ibid.: 5–6] and, for a more accurate statement, [2011: 148]), which is more restrictive. Raz [2005b: 13n18] notes himself that the inference on which his argument against the wide-scope account relies is not supported by the facilitative principle.

4 See Schroeder [2007] and Way [2012: 494]. See also Schroeder’s [2009: 246] conception of means/end-coherence.

5 See Bedke [2017: 7–12] for a helpful recent discussion.

6 Since instrumental reasons are reasons that are explained in a certain way, fully spelled-out principles of instrumental transmission have to contain a ‘because of that’ clause. We omit this clause throughout this article for reasons of simplicity.

7 See Bedke [2009: 679n12], who credits the point to James Dreier. Rippon [2011: 17] also argues that subversion is a problem for liberal transmission principles, but he does not consider the possibility of restricting the source reasons to final ones.

8 We take it that this is also what Raz has in mind when he states that ‘we have reason to perform any one (but only one)’ of the actions that facilitate conformity with the source reason [2005b: 5]. As Raz’s response to Broome [2005: 6–8] makes clear, he really intends this statement to mean that we have a reason to perform each of the facilitative actions, although we do not have a reason to perform more than one that is sufficient for conforming to the source reason (Raz [2005a: 3n8]; see also Rippon [2011: 5n13]).

9 Way’s means-end transmission principle [2010: 224] applies only to sufficient means (in contrast, however, to Way’s later [2012: 494] transmission principle, which also applies to insufficient means). Raz’s facilitative principle ([2005b: 5–6], restated in [2011: 148]) restricts source reasons to undefeated reasons. Raz’s general position on this question is not entirely clear, however. On the one hand, he maintains that ‘there is no reason to facilitate conformity with a defeated reason’ [2011: 145]. On the other hand, the inference that he uses [2005b: 12, 2011: 152] to argue against the wide-scope account of instrumental rationality relies on a liberal transmission principle that does not restrict source reasons to undefeated reasons.

10 Bedke [2009: 681] makes this point with respect to the specification of wide-scope reasons.

11 See Schroeder [2005: 7–9, 2007: 92–7]. The same strategy is used by Raz [2005a: 3–4], Bedke [2009: 684–5], Way [2012: 496n14], and Kolodny [2018: sec. 6].

12 Generalising, we also propose the following variant of Generic Instrumental Reason for sufficient means: if A has a final reason to φ, then A has a reason to take sufficient means to φ-ing.

13 Liberal Transmission can also be derived from Generic Instrumental Reason on the assumption of the principle that one has reason to ψ if one has reason to φ and ψ-ing is an instance of φ-ing. This latter principle is weaker than Liberal Transmission, but it faces the same Too Many Reasons objection and thus also cannot be assumed without begging the question.

14 This argument is inspired by one that Bedke [2009: 683–4] brings forward against the view that we do not generally have reasons to take very inefficient means to actions that we have reason to perform.

15 This argument does not presuppose that there are any perfectly neutral acts. Since, given the broad interpretation, it is at least coherent to judge that one has more reason to perform a perfectly neutral act than to kill an innocent person, ‘There is more reason to φ than reason to ψ’ does not imply ‘There is a reason to φ.’

16 This presupposes that, whenever ψ-ing is an incompatible alternative to φ-ing, a reason for φ-ing also counts (with equal strength) against ψ-ing. This assumption is not uncontroversial (see, e.g., Kolodny [2018: sec. 2] and White [2018: sec. 4]), but, as one of us has argued elsewhere [Kiesewetter 2015, 2018: sec. 3], the objections raised against it do not withstand scrutiny.

17 Depending on the relevant conception of an optimal means, one might argue that MRC* is entailed by Optimal Means Transmission—but in this case it would not provide support for Liberal Transmission.

18 This paper has benefited greatly from the feedback that we received on several occasions, including the ninth international conference of the German Society for Analytic Philosophy (GAP) at Osnabrück University, the 2015 meeting of the European Normativity Network (ENN) at HU Berlin, as well as research seminars led by Ulrike Heuer and Joseph Raz at UC London, Peter Schaber at the University of Zurich, Thomas Schmidt at HU Berlin, and Daniel Star at Boston University. We would like to thank the participants for stimulating discussions. We are especially grateful to Alex Gregory, Ulrike Heuer, Joseph Raz, Thomas Schmidt, Michael Smith, Justin Snedegar, Pekka Väyrynen, Jonathan Way, two anonymous referees, and the editor of this journal, Stephen Hetherington, for very helpful comments on earlier versions of this paper.