The Primacy of the Practical

Abstract

According to Action-First theorists, like Jonathan Dancy, reasons for action explain reasons for intentions. According to Intention-First theorists, like Conor McHugh and Jonathan Way, reasons for intentions explain reasons for action. In this paper, I introduce and defend a version of the Action-First theory called “Instrumentalism.” According to Instrumentalism, just as we can derive, using principles of instrumental transmission, reasons to ψ from reasons to ϕ (provided there’s some relevant instrumental relation between ψ-ing and ϕ-ing), we can derive reasons to intend to ϕ from reasons to ϕ (provided there’s some relevant instrumental relation between intending to ϕ and ϕ-ing). After providing some defense of Instrumentalism, I turn to two recent, important arguments for the Intention-First theory advanced by McHugh and Way, and I argue that neither of them succeed. I conclude that we should reject the Intention-First theory and that we have grounds for optimism about the Action-First theory.

1. Introduction

The weather is nice outside. That’s a reason for me to take a walk this afternoon. It’s also a reason for me to intend to take a walk this afternoon. But we might wonder about the direction of explanation: Is the nice weather a reason to intend to take a walk because it’s a reason to take a walk? Or is it a reason to take a walk because it’s a reason to intend to take a walk? Or neither?

The first of these three possibilities would be an instance of what Jonathan Dancy refers to as “the primacy of the practical”—the idea, roughly, that reasons to act explain reasons to intend to act (Dancy, 2018, 4). Dancy does not present any defense of this thesis, perhaps because he thinks “it seems just obvious” that it’s true. I agree with Dancy that this thesis, or at least a version of it, is true—and my hope in this paper is to go some ways toward convincing you of this—but I do not think it’s obviously true. It’s worth noting that several philosophers instead have endorsed views very close to the second of the three possibilities above—the idea, roughly, that reasons to intend to act explain reasons to act—including Allan Gibbard, T.M. Scanlon, Michael Smith, Douglas Portmore, and Daniel Whiting.¹ Additionally, Conor McHugh and Jonathan Way have presented some important critiques of the primacy of the practical and have also argued in defense of the second possibility above, specifically defending the view that “reasons for action are derivative from reasons for intention.”²

Introducing some terminology, we’ll say that Action-First theorists, like Dancy, think that reasons for action explain reasons for intention, whereas Intention-First theorists think that reasons for intention explain reasons for action. In this paper, I aim to articulate and defend a version of the Action-First theory which I call Instrumentalism.

Like any Action-First theory, Instrumentalism holds (putting aside “wrong kind of reasons” cases, for reasons that will be explained shortly) that reasons to intend to ϕ are explained by reasons to ϕ. But what’s distinctive about Instrumentalism is that it takes the nature of the explanation to be broadly instrumental in character.³ Instrumentalism starts with the idea that one’s intending to ϕ will often promote, or be necessary for, one’s ϕ-ing. And just as we can derive, from our reasons to ϕ, reasons to perform other actions which promote, or are necessary for, ϕ-ing, we can derive, from our reasons to ϕ, reasons to intend to ϕ on the same basis. Although there’s some debate about the precise formulations of the principles of “instrumental transmission,” there’s a wide consensus that instrumental derivation is a familiar and acceptable way in which some reasons can be derived from other reasons. And the central strategy of Instrumentalism is to understand the way in which reasons for intention are explained by reasons for action within this familiar framework.

In this paper, I’ll aim to set forth an initial case for Instrumentalism. But I’ll also aim to deflate the case for the Intention-First theory. This will involve both arguing against the Intention-First theory directly and responding to the arguments for the Intention-First theory advanced by McHugh and Way. Here’s the road map. I start (§2) by introducing Instrumentalism, showing how recent work on instrumental transmission lends support to it. I then (§3) offer replies to two objections to Instrumentalism. I then (§4) turn to two recent arguments from McHugh and Way in support of the Intention-First theory, showing how neither one of them succeeds. Lastly, I conclude (§5) that we should reject the Intention-First theory and be optimistic about the prospects for the Action-First theory.⁴

2. Instrumentalism

A strong version of an Action-First theory would hold that every reason to intend to ϕ is explained by a reason to ϕ. I have no interest in defending any such view here. There would be difficulties caused by the “wrong kind of reasons” to intend. A standard example of a wrong kind of reason to intend is provided by Kavka’s toxin puzzle: The eccentric billionaire will deposit $1M in your bank account right now if he detects your intention to drink some toxin tomorrow, but you know that by tomorrow, the money will already be (or not be) in your bank account and you know you can change your mind (Kavka, 1983). It seems you have a reason to intend to drink the toxin (you’ll get $1M) that is not a reason to drink the toxin. Of course, some philosophers are skeptical about the existence of wrong kind of reasons to intend; they’ll hold that the billionaire’s incentive is not really a reason to intend to drink the toxin.⁵ But the complicated debate over the existence of the wrong kind of reasons is one we cannot hope to settle here. Instead, we will be working with a weaker formulation of the Action-First theory: Putting aside cases of wrong kind of reasons to intend, reasons to intend to ϕ are explained by reasons to ϕ. This formulation would still hold that the practical is primary—Intention-First theorists will still want to insist that it gets the direction of explanation backward—but it sidesteps worries about the wrong kinds of reasons.

Instrumentalism is a version of the Action-First theory. Instrumentalism starts from a familiar thought about how some reasons can be derived from other reasons: Reasons for means can be derived from reasons for ends. As Thomas Nagel writes:

We might say that if being thirsty provides a reason to drink, then it also provides a reason for what enables one to drink. That can be regarded as the consequence of a perfectly general property of reasons for action: that they transmit their influence over the relation between ends and means (Nagel, 1970, 34).

Much of the recent literature on “instrumental transmission” has focused largely on transmission to necessary means. For instance, Michael Bratman defends the following principle:

Transmission Reasons: If R is a practical reason in favor of X, X is attainable by the agent, and M is a necessary means to or necessary constitutive element of X, then R is a practical reason in favor of M (Bratman, 2009, 424).⁶

And Mark Schroeder defends:

Reason Transmission: If X has an objective reason to do A and to do A X must do B, then X has an objective reason to do B of equal weight to X’s objective reason to do A (Schroeder, 2009, 245).

Other philosophers have defended principles that apply to non-necessary means. For instance, Joseph Raz has defended the following principle, applicable to undefeated reasons:

Facilitative Principle: When we have an undefeated reason to take an action, we have reason to perform any one (but only one) of the possible (for us) alternative plans that facilitate its performance (Raz, 2005, 5–6).

And Niko Kolodny has defended:

General Transmission: If there is a reason for one to E, and there is positive probability, conditional on one’s M-ing, that one’s M-ing, or some part of one’s M-ing, helps to bring it about that one E’s nonsuperfluously, then that is a reason for one to M, whose strength depends on the reasons for one to E and the probability, so long as the reason for one to E is not explained by an application of General Transmission to reasons for one to achieve some distinct E’ (Kolodny, 2018, 752–753).⁷

There are subtleties to each of these principles, and their various clauses, that we need not explore here. Nor do we need to argue for any formulation. Such arguments can be found elsewhere, and we are mentioning the principles only to illustrate how several philosophers have endorsed and attempted to refine, Nagel’s thought that reasons transmit from ends to means. Even if there are debates about the details, instrumental transmission is a familiar way in which some reasons can be derived from other reasons.

Instrumentalism aims to understand the relationship between reasons for actions and reasons for intention within this familiar framework. Just as we can use principles of instrumental transmission to derive reasons for actions from other reasons for action (provided that the relevant instrumental relation obtains), we can use principles of instrumental transmission to derive reasons for intention from reasons for action (again, provided that the relevant instrumental relation obtains).

For instance, let us consider Bratman’s transmission principle above, according to which we can derive reasons for M from reasons for X when M is a “necessary means to or necessary constitutive element of X.” For instance, when buying a ticket is a necessary means to seeing the exhibit, and there’s a reason to see the exhibit (and this is attainable), there’s a derived reason to buy the ticket. My main suggestion here is that we should be able to plug “intending to X” in for M and derive reasons to intend to X in the same way—perhaps deriving a reason to intend to go to the exhibit from a reason to go to the exhibit. It’s worth noting that Bratman himself, in applying the principle, takes the formation of intentions to count as a necessary means.⁸ And, more importantly, he seems to be right to do so. There does not seem to be any justification for privileging non-psychological over psychological barriers to action: My failing to intend to go to the exhibit is just as much an obstacle to my going to the exhibit as my failing to buy the required admission ticket, and so it would seem arbitrary to allow that we can derive a reason to buy the ticket when doing so is necessary, but not a reason to intend to go to the exhibit when doing so is necessary.

If this thought is along the right lines, we already have in place a rather straightforward and powerful argument against the Intention-First theory. According to that theory, it holds in general that reasons for action are derived from reasons for intention. But our application of Bratman’s transmission reasons to intentions yields the result that in at least some cases, the explanation runs in the other direction: The reasons for intention are derived from the reasons for action. Proponents of the Intention-First theory must either (i) deny that Bratman’s principle is true or (ii), contra Bratman’s own understanding of it, hold that it does not apply to intentions. Neither option is particularly attractive.

It might be worth saying more about the understanding of “necessary means” at work in these principles. Bratman does not himself provide any definition of “necessary means,” but his principle is a development of Kieran Setiya’s transmission principle and Setiya does provide us with a definition. According to Setiya’s principle:

Transmission: If you should do E, all things considered, and doing M is a necessary means to doing E, you should do M, all things considered, too (Setiya, 2007, 652).

And Setiya offers us the following “general definition of a necessary means”:

…doing M is a is a necessary means to doing E, for a particular agent, A, just in case there is something A could do that is a means to doing E, and everything she could do that is a means to doing E involves doing M (Setiya, 2007, 657).

Setiya thinks that on this definition, A’s intention to ϕ could count as a necessary means (Setiya, 2007, 657).⁹ The restriction to necessary means is important because it allows for defenders of transmission to avoid many of the problems besetting a “deontic inheritance” principle. According to deontic inheritance, “whenever something is obligatory, everything that is a logical consequence is also obligatory” (McNamara & Van De Putte, 2022, §2, §6).¹⁰ This principle famously invites Ross’ paradox, since we can now infer from “you ought to mail that letter” that “you ought to mail the letter or burn it.”¹¹ But this case poses no problem for Setiya’s transmission, since mailing the letter or burning it is not a means to mailing the letter.

However, Setiya’s “general definition of necessary means” does not tell us how to define “means” (since that term itself appears in the definition), and one might object that we should construe “means” narrowly, so that intentions to ϕ are disallowed as means.¹² However, I’m not inclined to accept the narrow construal of “means.” First of all, for whatever it’s worth, many of those working with transmission principles, including Setiya and Bratman, have allowed for intentions to ϕ to qualify as means, so the narrow construal would be at odds with that tradition. Second, the narrow construal does not seem supported by our ordinary usage of “means.” Given that intending to ϕ seems to be a way of helping to bring it about that I ϕ, it’s perfectly natural to say, for instance, that my intending to go on vacation is a (perhaps necessary) means to going on vacation.¹³ Also, much of the discussion of the functional role of intention in the philosophy of action—I’ll say more about this in a moment—emphasizes the way in which intentions seem geared toward bringing it about that one performs the intended action.

A third and final point has to do not with our ordinary usage of “means” but the somewhat technical way it is used in other areas of philosophy. In addition to being central to discussions of instrumental transmission, the concept of a necessary means is central to discussions of instrumental rationality. ¹⁴ Instrumental rationality requires of us, roughly, that we intend the means we believe are necessary for our ends. As John Broome puts it, the instrumental requirement is the requirement that “you intend what you believe to be a means implied by an end you intend” (Broome, 2013, 159). But in explaining the instrumental beliefs that put that requirement in place, Broome does not disallow one’s intentions from counting as means implied by an end. Those beliefs are the following:

(2) N believes at t that m is a means implied by e.

(3) N believes at t that her herself’s then intending m is a means implied by m (Broome, 2013, 160).

The idea of a “means implied by” is explained as follows: “When I say a is a means implied by b, I mean that, were a not so, because of that b would not be so” (Broome, 2013, 160). Just as one may believe that were one not to m, because of that, one would not e, one may believe (roughly) that that were one not to intend to m, because of that, one would not m (and thereby also not e). In short, when it comes to formulating those beliefs that put in place a rational requirement to intend the necessary means to one’s intended ends, it seems fine to include both actions (m) and intentions (intending m) among the means implied by e. Discussion of the details of Broome’s formulation would take us too far afield. But my point is just that those who work with transmission principles employing a more expansive construal of “means” (allowing for intentions to figure among the necessary means) can take some comfort in the fact that a similarly expansive understanding of “means” is sometimes employed in discussions of instrumental rationality.

It’s also worth noting that nothing about the more expansive construal requires that we reject the idea that when it comes to transmission principles, both the “source” and the “destination” have to be such that they are supported by (normative) reasons. The Instrumentalist suggests that just as we can derive reasons for actions from reasons for actions, we can also derive reasons for intentions from reasons for actions. None of this is problematic, since there are normative reasons for actions and normative reasons for intentions. But one might take note of the fact that a necessary part of my performing some particular action might be my hand muscles flexing in a particular way, or my neurons firing a particular way.¹⁵ However, nothing forces us to derive (normative) reasons for my hand muscles flexing a particular way, or my neurons firing a particular way. (There are, of course, explanatory reasons; we can explain why my hand muscles flexed the way they did and why my neurons fired.) Only those actions or attitudes capable of being supported by normative reasons are candidates for inclusion in the principles of instrumental transmission, at either the “source” end or the “destination” end. There are, of course, questions about why intentions and (intentional) actions are capable of being supported by normative reasons, and hand muscle flexing and neuronal firings are not. But this is not the place to take up those questions. Here, we need only note that nothing about Instrumentalism requires that we abandon this restriction.

Bratman and Setiya’s transmission principles apply only to necessary means (and necessary constitutive elements), and so have a rather narrow application. However, as we noted above, there are other principles that apply to non-necessary means, such as Kolodny’s General Transmission. We could appeal to such principles to explain how reasons for intention are derived from reasons for action in cases in which an intention to ϕ would promote, but not be necessary for, one’s ϕ-ing. Suppose there’s a lot to be gained by throwing a chess match, and so I intend to lose the match. However, intending to lose is not necessary for losing; provided that my opponent is sufficiently skilled, she could beat me even if I intended to win. But intending to lose significantly increases the probability of my losing. Here, applying Kolodny’s General Transmission, we could derive, from my reason to lose the match, a reason to intend to lose the match, even when intending to lose is not necessary for losing.¹⁶

We’ve seen how some recent work in normative philosophy on transmission principles could lend support to the Action-First theory. But it’s worth observing that there’s support available in the philosophy of action as well. Those writing about the functional role of intentions tend to highlight the ways in which intentions play a role in the production of action. For instance, most famously, Michael Bratman argues that intentions are “conduct-controlling” pro-attitudes which enable to us save deliberative resources by (defeasibly) cutting off deliberation in advance of action—thereby enabling us, for instance, to decide on Monday what we’ll do on Friday, without having to deliberate throughout the week (Bratman, 1987, 15–18).¹⁷ Intentions enable this by being (again, defeasibly) resistant to reconsideration, stably persisting until the intended action is completed. Bratman also observes that agents who intend to ϕ will be disposed to engage in further sub-planning that facilitates their ϕ-ing, and they will also be disposed to take means to ϕ-ing (Bratman, 1987, 2–3). In short, it seems that at least one important role of intentions is their instrumental role in bringing about action. In being resistant to reconsideration, disposing us to engage in further sub-planning, and disposing us to take means, the formation of intentions is normally a good means to the production of the relevant actions. And the Instrumentalist just adds the further claim that we do not have any good basis for privileging non-psychological over psychological means: Just as we can derive reasons to ψ from reasons to ϕ when ψ-ing is (roughly) a good means to ϕ-ing, we can derive reasons to intend to ϕ from reasons to ϕ when intending to ϕ is (roughly) a good means to ϕ-ing, as it very often is.

3. Two Objections to Instrumentalism

I’ve noted above that intending to ϕ is very often a means to ϕ-ing. But what about cases in which intending to ϕ will make no difference to whether or not one ϕs—perhaps because one will ϕ anyway, regardless of whether one intends to do so? Are we forced to conclude that there’s no derived reason to intend to ϕ in that case? And, if so, is that a problem for Instrumentalism?

McHugh and Way have presented an objection along these lines. They ask us to consider a case in which “you are going to ϕ regardless of whether you intend to”: Suppose you intend to walk to work and know that by doing so you’ll get in your exercise, but you do not intend to get in your exercise (McHugh & Way, 2022, 156, 162).¹⁸ There are, in McHugh and Way’s view, still reasons for you to intend to get in your exercise. In particular, those features which make exercising worthwhile are reasons for you to intend to get in your exercise, and you may form such an intention for those reasons. But it seems that Instrumentalism would be unable to explain why there would be reasons for you to intend to get in your exercise in this case, since the intention is not necessary for, and does not promote, the relevant action.

One initial, defensive remark is that this objection need not spell doom for the Action-First theory. Even if there is not an instrumentally derived reason to intend to get in your exercise, it’s open to the Action-First theorist to hold that the reason is derived in some other way. However, I will not explore this strategy, since I do think there’s an instrumental derivation applicable to McHugh and Way’s example. In their case, the intention to walk to work is what ensures that you’ll get in your exercise, rendering the intention to get in your exercise unnecessary. However, it’s a familiar fact about us that we are capable of changing our minds, especially in response to new information. Perhaps once the clouds move in, you’ll decide to drive to work instead of walking. Knowing this about yourself, you might now form both an intention to walk to work and an intention to get in your exercise, in case the former intention does not persist. The latter intention functions as a “backup means” to getting in your exercise.

The idea of a “backup means” is a familiar one. I might know, but not be certain, that I’ll get into State University on an athletic scholarship, but also focus on my academic abilities, in case I get sidelined by an unexpected injury. In the same way, I might know, but not be certain, that I’ll get in my exercise based on my intention to walk to work, but also intend to get in my exercise, in case I change my mind about walking. In both cases, given that there’s a chance of needing the backup means, having it in place would raise the probability of achieving the relevant end (getting into state; getting in your exercise), and so we could allow that the reasons for the ends transmit to reasons for means, perhaps again employing Kolodny’s General Transmission.

It’s open to McHugh and Way to reply that I’ve misunderstood their case. They could claim that their case is one in which it’s stipulated that you are going to get in your exercise regardless of your intentions, and so having a backup means would not raise the probability of your getting in your exercise. If they make this reply, they will probably also have to change the example, since, ordinarily, it’s up to you to change your mind about how you get to work, and your choosing driving over walking would put your exercise in jeopardy. Perhaps we could instead suppose that you are in a harsh prison where you will be forced to get your steps in as part of the labor you must perform. Here, it’s clearer that an intention to get in your exercise would not raise the probability of your doing so.¹⁹ But it’s also less clear that there’s a reason for you to intend to get in your exercise, any more than there would be a reason for me to intend to do other things that I’m ordinarily guaranteed to do (like, say, move my limbs at some point over the course of the day).

In short, if we understand the case such that whatever is ensuring that you’ll get in your exercise is subject to change (as their own example suggests), then there’s room to appeal to the idea of a “backup means” to explain why there’s a reason to intend to get in your exercise, but if we stipulate that you’ll be getting in your exercise regardless, then it’s less clear that there really is any reason to intend to get in your exercise. For that reason, I’m not persuaded by the objection.

Let us turn to a second objection to Instrumentalism. One might worry that instrumentally derived reasons to intend come out to be state-given reasons. Ulrike Heuer has advanced a claim along these lines.²⁰ In her “Chocolate Hazard” example, Carl is on a diet and has conclusive reason to lose weight and hence has conclusive reason to refrain from eating chocolates, but he often goes on autopilot and eats chocolates anyway. Heuer thinks that Carl has a reason to intend not to eat chocolates, but argues that this reason will come out to be a state-given, as opposed to an object-given, reason:

Carl has such a reason because forming the intention not to eat the chocolates may stop the autopilot response and prevent him from eating them. Carl has a kind of instrumental reason for forming the intention: having the intention not to eat the chocolates helps him to comply with his reason to lose weight. But the instrumental reason to form the intention is a peculiar one, at least within the dialectic of the current debate on reasons to intend: it is a reason to intend not because there is a reason to act, but because there is value in having the intention. It is, in that sense, a state-given reason (Heuer, 2018, 878–879).

Such a result would be worrisome, especially if we also endorse the view that state-given reasons are wrong-kind reasons.²¹ In that case, skepticism about wrong-kind reasons would yield skepticism about Carl’s reason to intend not to eat chocolates.

I agree with Heuer that there’s instrumental value in Carl’s having the intention, but I deny that this makes the reason a state-given reason. Heuer herself presents a definition of state-given reasons, on which the reason would not come out to be state-given. She writes: “Reasons of this kind are sometimes summarily called ‘state-given reasons’ since it is the value of being in the state of intending (or believing) which provides the reasons independently of the value of acting as intended” (Heuer, 2018, 866–867, emphasis added). In Carl’s case, however, the value of his intending to refrain from eating chocolates is not independent of the value of acting as intended; indeed, it seems to be entirely explained in terms of the value of his not eating the chocolates (which is valuable in that it will lead to his losing weight). So, I do not think we are forced into thinking that instrumentally derived reasons to intend are state-given reasons.

Heuer’s definition of state-given reasons here aligns with the standard understanding of the object-given/state-given reasons distinction, according to which object-given reasons are provided by the object of the attitude, whereas state-given reasons are provided, not by the object of the attitude, but by one’s being in the state of having the attitude. For instance, Derek Parfit (who introduced the terminology) explains state-given reasons for desires as follows: “Many people assume that we can also have state-given reasons to have some desire. Such reasons would be provided by certain facts, not about some desire’s object, but about our state of having this desire” (Parfit, 2011, 50). Applied to reasons for intention, state-given reasons would be provided not by facts about the intention’s object, but by facts about the state of having the intention. But the object of Carl’s intention to not eat chocolates would be his not eating chocolates. And it’s the value of his not eating chocolates (which is a means to weight loss) which explains why he has a reason to intend not to eat chocolates.²² So, Carl’s reason would count as an object-given reason.²³

4. Arguments for the Intention-First Theory

Let us take stock. In §2, we introduced Instrumentalism as a version of the Action-First theory, noting that it’s supported by both recent work in normative philosophy on instrumental transmission and the standard view of the functional role of intentions in the philosophy of action. In §3, we defended Instrumentalism against two objections. However, we have not yet seen what’s to be said in favor of the Intention-First theory. In this section, we’ll turn to the two main arguments that McHugh and Way advance in favor of their Intention-First view. I’ll argue that neither argument succeeds.

4.1. The response constraint argument

The Response Constraint is the idea, roughly, that reasons must be such that we are capable of responding to them. Applied to reasons for action, the Response Constraint holds that “if p is a (normative) reason for you to ϕ, you can ϕ for the motivating reason that p” (McHugh & Way, 2022, 157). McHugh and Way assume the truth of the Response Constraint (while acknowledging that it’s controversial) and use it to argue for the conclusion that “any reason to ϕ is also a reason to intend to ϕ.”²⁴ And they then argue that this claim is best explained by their particular Intention-First view.

Let us review their argument. As noted above, they start with the Response Constraint, according to which our practical reasons must be such that we can act for them. But what’s involved in acting for a reason? McHugh and Way claim that “you act for a reason only if you execute an intention held for (based on) that reason” (McHugh & Way, 2022, 157). (I’ll call this “the Intention Execution Claim.”) They provide some examples to illustrate. Assume the fact that the reading group meets this afternoon is a reason for you to take your book with you to campus. If you were to intend, based on the fact that the reading group is meeting this afternoon, to take the book with you to campus, and you were to execute that intention, you would thereby be acting for that reason. But if you were to take the book to campus without executing that intention, you would not act for a reason. For instance, if you did not know that the book slipped into your backpack, and you took your backpack to campus, you would not have taken the book to campus for that reason. In their view, “acting for a reason entails acting from an intention held for that reason” (McHugh & Way, 2022, 158).

McHugh and Way then argue that whenever you act from an intention to ϕ held for a reason, you treat that reason as a (normative) reason to intend to ϕ. Moreover, it’s appropriate for you to do so. And they then argue that “if it’s appropriate to treat something as a reason to intend to ϕ, then it is a reason to intend to ϕ” (McHugh & Way, 2022, 158). Thus, every reason to ϕ would also be a reason to intend to ϕ. Here’s their summary of the argument so far:

Reasons for action can be acted on. You act on a reason only if you intend for that reason. In doing so you treat it as a reason for intention, and appropriately so. But a consideration that can be appropriately treated as a reason for intention is one. So a reason to ϕ is also a reason to intend to ϕ (McHugh & Way, 2022, 158).

They then go on to argue that the fact that a reason to ϕ is also a reason to intend to ϕ is best explained by their particular version of the Intention-First theory according to which “reasons for action are fundamentally reasons for intention.”²⁵

However, it’s not clear to me that the Intention Execution Claim—that is, the claim that “you act for a reason only if you execute an intention held for (based on) that reason”—is true. Some cases involving foreseen, unintended consequences seem to be counterexamples. Consider the following illustration from Gilbert Harman, concerning a sniper who intends to kill the target, but foresees that doing so will alert the enemy to his presence:

It is a mistake to suppose that whenever someone does something intentionally, he intends to do it. Things someone does as foreseen but unintended consequences of what he intends, for example, are sometimes things he does intentionally. In firing his gun, the sniper knowingly alerts the enemy to his presence. He does this intentionally, thinking that the gain is worth the possible cost. But he certainly does not intend to alert the enemy to his presence (Harman, 1976, 433).

Harman’s case appears to be a counterexample to the Intention Execution Claim: The sniper acts for a reason (he alerts the enemy, and his reason for doing so is that the gain is worth the possible cost), but he does not execute an intention to alert the enemy; rather, he intends to kill the target.

There are several ways one could respond to the objection. McHugh and Way anticipate the objection and respond by denying that the case involves action done for a reason.²⁶ However, it’s not clear how plausible this response is. Alerting the enemy is clearly something that the sniper does. Moreover, he does it knowingly. (The sniper case is unlike the case of the person who transports the book to campus, unaware the book was slipped into his backpack.) And it seems that the sniper can easily cite the reasons for which he acted. When his commanding officer inquires, “Why did you alert the enemy to your presence?” the sniper could reply, “Because the gain (killing the target) was worth the possible cost of being discovered.” It would be a theoretical cost, I think, to be forced to say that the sniper alerted the enemy to his presence for no reason whatsoever.

Another possible reply is to say, contra Harman, that since the sniper alerted the enemy to his presence intentionally, he intended to do so. This would be entailed by what Michael Bratman has called the Simple View, according to which if A intentionally ϕs, A intends to ϕ. Bratman develops an argument against the Simple View, based off of the rational consistency constraints on intentions.²⁷ And there are other difficult cases for the view. (For instance, Harman, immediately before discussing foreseen, unintended consequences, gives the example of the sniper shooting a soldier from a great distance, with slim chances of success. In Harman’s view, the sniper does not “flatly intend to kill the soldier, although, if he succeeds, he kills him intentionally.”²⁸). However, I lack the space here to consider arguments for and against the Simple View. Instead, I’ll just note that, if McHugh and Way make this reply, they would be forced to adopt a controversial assumption. Moreover, it’s an assumption which they explicitly decline to adopt in their paper. (They write: “Some philosophers hold, further, that intentionally ϕ-ing requires intending to ϕ, or at least some relevant intention. We will not assume this” (McHugh & Way, 2022, 154).)

A third reply would be to question my interpretation of the Intention Execution Claim. Consider two possible interpretations of their claim that “you act for a reason only if you execute an intention held for (based on) that reason.” On the strong interpretation, you ϕ for a reason only if you intend to ϕ based on that reason. That interpretation has trouble with Harman’s sniper. On a weaker interpretation, you ϕ for a reason only if either you intend to ϕ based on that reason or you intend to ψ based on that reason, and ψ-ing involves ϕ-ing. This interpretation does not have trouble with Harman’s sniper, so long as we allow that the sniper intends to kill the target based on a reason, and killing him involves altering the enemy.

The problem, however, is that the weaker interpretation would no longer provide sufficient support for their argument’s conclusion that “a reason to ϕ is a reason to intend to ϕ.” The weaker interpretation allows for the possibility of ϕ-ing based on a reason without one’s intending to ϕ based on that reason. Thus, it would allow for the possibility of ϕ-ing based on a reason without (appropriately) treating that reason as a reason to intend to ϕ. ²⁹ We could of course then concede that whatever is appropriately treated as a reason to intend to ϕ is indeed a reason to intend to ϕ. But we would have no guarantee that every reason to ϕ is a reason to intend to ϕ.

Each of these three attempts to save the Intention Execution Claim invites further difficulties. In my view, we should reject the Intention Execution Claim, and so we should reject the Response Constraint Argument, even if we happen to accept the (controversial) Response Constraint.

4.2. Unifying theoretical and practical reasoning

McHugh and Way provide a second argument for their view that “reasons for action are fundamentally reasons for intention” (McHugh & Way, 2022, 153). The argument looks to theoretical and practical reasoning and the way in which reasons figure into that reasoning. They note that in theoretical reasoning, “you aim to settle on a view of what is true—to form a belief” and you do so by considering the evidence for the truth of the relevant proposition (McHugh & Way, 2022, 159). And in practical reasoning, “you aim to settle on a course of action—to form an intention” and, likewise, you do so by considering the pros and cons of the relevant action (McHugh & Way, 2022, 160).³⁰ They note that for both theoretical reasoning and practical reasoning, there is “a distinction between a response to which the reasoning leads and the object of that response” which we could chart as follows.

They then argue that their view that reasons for action are fundamentally reasons for intention allows for an attractive, unified account of theoretical and practical reasoning. They write:

In the theoretical case, it is clear that what reasons support is fundamentally beliefs, not propositions. If we take their role in reasoning to be an important aspect of the nature of reasons, then, we should regard practical reasons as fundamentally supporting intentions rather than act types. In doing so, we get an attractive, unified picture of how (fundamental) reasons feature in reasoning: A reason for a response is a premise of good reasoning that concludes in that response, reasoning that is about the object of the response. By contrast, if we regard practical reasoning as fundamentally supporting act types, then we find deep differences between practical and theoretical reasons in respect of the way they relate, in reasoning, to the responses they support (McHugh & Way, 2022, 160).

In short, their view allows for an attractive, unified account of the way in which reasons, both practical and theoretical, figure into reasoning.

There are, as McHugh and Way recognize, a number of large and longstanding philosophical questions relevant to this argument, including questions about the conceptual connections between reasons and reasoning and questions about how to understand the premises and conclusion of practical reasoning. I will not explore those questions here. Instead, I want to raise a more basic methodological concern about their argument: It’s not clear to me that we should be looking to the theoretical domain to understand the practical domain; after all, there may be important differences between the two domains that we should acknowledge.

Let me try to make this methodological concern more concrete. If we look at the table above, which specifies the responses and objects of practical and theoretical reasoning, we can already see an important difference between the theoretical and practical domains when it comes to reasons. One of the quadrants is not like the others: While there are normative reasons for beliefs, intentions, and actions, there aren’t normative reasons for propositions. This raises the question of why we would try to look to theoretical reasoning for illumination in our debate over the primacy of the practical. Our debate is about the nature of the reasons in the bottom row of the table: Action-First theorists think, roughly, that reasons for action explain reasons for intention, while Intention-First theorists think, roughly, that reasons for intention explain reasons for action. It might make sense to look to the theoretical domain for guidance if we were looking at a domain involving two kinds of reasons. We could then examine which of the two kinds of theoretical reasons is explanatorily fundamental and then suggest that the analogous direction of explanation should hold in the practical case. But that’s not what’s going on here; while there are reasons for beliefs, there aren’t reasons for propositions.

Here’s another way to put my worry about McHugh and Way’s argument. If we are looking to arrive at a unified account of theoretical and practical reasoning, why would not we instead arrive at a wholesale skepticism about reasons for action? We could reason as follows: Just as there are no reasons for the objects of theoretical reasoning (propositions), only reasons for the relevant responses (beliefs), there should be no reasons for the objects of practical reasoning (actions), only reasons for the relevant responses (intentions). That seems to be the conclusion that’s supported by the analogy with theoretical reasoning. But of course that’s not the conclusion that McHugh and Way want to arrive at: They believe that there are both reasons for intentions and reasons for actions; they just think the former are explanatorily fundamental.

In short, there seem to be important differences between the theoretical and practical domains that make it inappropriate to look to the theoretical domain for guidance on how to understand the practical domain. McHugh and Way are right to observe that just as theoretical reasoning concludes in a belief, practical reasoning concludes in an intention (assuming the Aristotelian thesis is incorrect).³¹ But we should not expect this to tell us anything illuminating about the direction of explanation when it comes to reasons for intention and reasons for action, since there’s no relevant analogous structure to be found in the theoretical domain: There, we have only reasons for belief.

5. Conclusion

Let us sum up. In this paper, I’ve tried to provide some defense of the primacy of the practical—specifically, the thesis that (putting aside “wrong kind of reasons” cases) reasons to intend to ϕ are explained by reasons to ϕ. I’ve set forth a version of the Action-First theory that I call “Instrumentalism.” According to Instrumentalism, just as we can instrumentally derive one reason for action from another, we can instrumentally derive a reason for intention from a reason for action. I argued (§2) that work in normative philosophy on instrumental transmission, as well as work in the philosophy of action on the nature of intentions, lends support to Instrumentalism. I then (§3) defended Instrumentalism against two objections. First, I argued that we can instrumentally derive reasons to intend even when you have some other intention that will likely bring about the relevant action. Second, I argued that the derived reasons need not be state-given reasons. Lastly, I considered (§4) the two main arguments that McHugh and Way give for their Intention-First theory. I pointed out a way in which the Intention Execution Claim in their Response Constraint Argument is problematic. And I argued that their analogy with theoretical reasoning is inapt because we do not have, in the theoretical domain, an explanatory relation between two different kinds of reasons.

In conclusion, I think we should reject the Intention-First theory. Both of the main arguments for the theory fail (and I know of no other good arguments for it), and the application of principles of instrumental transmission suggests that at least sometimes the explanatory direction is the other way around. And I think we should be optimistic about the prospects of the Action-First theory. I do not think I’ve given anything close to a complete defense of the theory, but I have developed a particular version of it (Instrumentalism), and shown how it has some advantages and can avoid some objections. So, I think we have at least good grounds for optimism about the primacy of the practical.³²