Someone displays instrumental rationality insofar as she adoptssuitable means to her ends. Instrumental rationality, by virtually anyreckoning, is an important, and presumably indispensable, part ofpractical rationality. However, philosophers have been interested init for further reasons. To take one example, it has been suggestedthat instrumental rationality, or some tendency toward it, is partlyconstitutive of intention, desire, or action. To take another, moreimportant, example, it has been argued that instrumental rationalityis not only a part, but a special part, or even the whole, ofpractical rationality. This thesis appears to threaten the“rational authority” of morality. It seems possible thatacting morally on some occasion might not be a suitable means to anagent’s ends. If so, then according to this thesis, it would notbe irrational for her to refuse to act morally on such anoccasion.

We should note two restrictions on the scope of this article. First,this article focuses on contemporary discussions of the subject, apartfrom some brief discussions of Hume and Kant. Notable omissions, forexample, are Aristotle’s discussion of the practical syllogismand his claim that deliberation is of means (NicomacheanEthics, Book III, Chapter 3).

Second, this article doesn’t focus on decision theory, whichsome may view as containing, or just being, a theory of instrumentalrationality: what means to take to achieve certain ends (thesatisfaction of preferences, or the realization of valuable states ofaffairs) given credences, or beliefs, or information, about theprobabilities that various means will achieve those ends.

• 1. Rational Coherence and Reason
• 2. Instrumental Transmission
• 3. Instrumental Coherence
  • 3.1 Independent Requirements of Reason
  • 3.2 Requirements of Reason to be Instrumentally Coherent As Such
  • 3.3 Instrumental Incoherence without Violating Requirements of Reason
  • 3.4 “Cognitivist” Requirements on Belief
• 4. The Status of Instrumental Rationality
  • 4.1 The Only Kind of Practical Rationality?
  • 4.2 The Relevance of Carroll’s Paradox
  • 4.3 Hume on Instrumental Rationality
  • 4.4 Kant on Hypothetical Imperatives
• Bibliography
• Academic Tools
• Other Internet Resources
• Related Entries

1. Rational Coherence and Reason

Someone displays instrumentally rationality insofar as she adoptssuitable means to her ends. But how should “rationality”be understood? Here it is important to distinguish between two thingsthat the casually phrased question, “What would it be rationalfor an agent to do or intend?” could mean:

1. By doing or intending what would the agent make her responses(i.e., her attitudes and actions)cohere with one another?Call the answer, stipulatively, what it would berationallycoherent for the agent to do or intend.
2. What does the agent have reason, or ought she, to do or intend?This is the question that an agent characteristically asks when she isdeliberating about what to do or intend, or when someone else isadvising her what to do or intend. (Note that claims about reasons areweaker than claims about “oughts”. One could have a reasonto do something without it being the case that one ought to do it, aswhen the reason is outweighed by competing reasons. However, it isgenerally thought that if one ought to do something, then one has areason to do it.)

As with any philosophical distinction, this contrast between“rational coherence” and “reason” might bequestioned. But it is accepted by many of the authors who have writtenon the subject, and it will help to organize the discussion of thesubject matter.

In principle, the answers to these questions might perfectly coincide:that what agents have reason, or ought, to do just is what it would berationally coherent for them to do, and vice versa. In several ways,however, the answers might be expected to diverge.

First, even if what one ought to do is just to make one’sresponsesglobally coherent, what it takes to makeone’s responseslocally coherent might differ from whatit takes to make them globally coherent. According to the SubjectiveDesire-Based Theory, for example, what agents have reason, or ought,to do or intend is just what, given what they believe theircircumstances to be, would best satisfy their strongest, presentintrinsic desires taken as a whole. Suppose that the agent’sstrongest, present intrinsic desire is for health. Nevertheless, outof habit, he intends to have a smoke, believing that lighting up is anecessary means. According to the Subjective Desire-Based Theory, itis not the case that he ought to intend to light up. And it might besaid that if he were globally coherent, the agent would not intend tolight up. But if he does form an intention to light up, he at leastachieves a kind of local coherence.

Second, what the agent has reason, or ought, to do or intend maydepend not on what shebelieves her circumstances to be, buton something more “objective.” There are severalpossibilities. What an agent has reason, or ought, to do, might bewhat the evidence (where this depends on something other than herattitudes) available to the agent suggests about her circumstances,what the evidence of the person making the reason- or ought-claimsuggests about the agent’s circumstances, what the evidence ofthe person assessing the claim suggests about the agent’scircumstances, or all of the relevant facts about the agent’scircumstances. Consider the Objective Desire-Based Theory, which saysthat agents have reason, or ought, to do or intend just what, givenwhat their circumstances actually are, would best satisfy theirstrongest, present intrinsic desires taken as a whole. Suppose theagent’s strongest, present intrinsic desire is to drink a ginand tonic, and she so intends. However, she mistakenly believes thatthe stuff in this bottle is gin, when it is in fact petrol. So shebelieves that mixing the stuff with tonic is a means to drinking a ginand tonic. According to the Objective Desire-Based Theory, she doesnot have reason to intend to mix the stuff with tonic and drink it.But if she does so intend, she might be said to have achieved a kindof rational coherence, both local and global.

Third, one might hold not a Desire-Based Theory, but a Value-BasedTheory, according to which whateverultimate ends an agenthas reason, or ought, to achieve depend not on what shedesires orwills, but instead on what is ofindependent value. Suppose the madman’s strongest, presentintrinsic desire is to set off a nuclear war, and he so intends.Moreover, the madman knows that intending to press this button is anecessary and sufficient means to setting off a nuclear war. It mightbe said that, in intending to press this button, the madman wouldachieve a kind of coherence, both local and global. And according tothe Desire-Based Theories, the madman ought so to intend. According toa Value-Based Theory, however, it presumably is not the case that heought so to intend.

There are several reasons, therefore, to expect at least somedivergence between what one has reason, or ought, to do or intend, andwhat it would be rationally coherent for one to do or intend. But thatis perfectly compatible with partial convergence, even of anon-accidental character. As we will discuss below, some suggest thatamong the things that agents have reason, or ought, to do or intend isprecisely to make their responses rationally coherent. Just as weought not to torture, or ought to care for our children, we ought tobe rationally coherent.

2. Instrumental Transmission

We’ll begin by discussing how means-end relations matter forwhat one ought, or has reason, to do or intend. (For an opinionatedsurvey, see Kolodny 2018.) But we should first consider how“means” is understood. While “means” are oftentaken to be tools or resources, for our purposes, we will view meansas actions, intentions, choices, decisions, etc. that might make useof such tools or resources. It is natural to think of means to an end,so understood, as actions that help tocause the end tooccur. But often philosophers count as means to an end whatever mayhelp to bring it about.M may bring aboutE bycausing it, as flicking this switch causes the illumination of thisroom, butM may also bring aboutE by helping toconstitute it, as playing this measure partially constitutes theplaying of this sonata, or fulfilling preconditions for it, asregistering for this conference fulfills preconditions for attendingit, or preventing things that might prevent it, as vaccinationprevents disease that might prevent successful completion of theschool year. Means are also often probabilistic. Something may alsocount as a means to an end even if it is not guaranteed to help tobring about the end, but only has some chance of doing so. Usage of“end” is also often ambiguous. Sometimes “end”is used to refer to anaction, such as the action ofilluminating this room. But sometimes “end” is used torefer to astate of affairs that is either desired orvaluable, such as the state of affairs of this room’s beingilluminated. We’ll here use “end” to refer to anaction.

Reasons for action are often said to “transmit” from endsto means, in the sense that if there is reason for an end, then,because of this, there is reason for the means. This idea is capturedin Joseph Raz’s Facilitative Principle:

[w]hen we have an undefeated reason to take an action, we have reasonto perform any one (but only one) of the possible (for us) alternativeplans that facilitate its performance. (Raz 2005b)

Insofar as what one ought to do depends on what one has reason to do,one might expect a similar transmission of “ought” fromend to means. If one ought to achieve an end, then one ought to takethe means, or at least means of certain kinds.

Principles of instrumental transmission of this kind are compatiblewith different views about what ends we have reason to achieve. Forexample, it may be accepted by either a Desire-Based, or aValue-Based, Theory.

What the correct principle, or principles, of instrumentaltransmission are is controversial. Most commonly, principles describetransmission to necessary means. However, some speak of sufficientmeans, and others speak of neither. For example, Raz’sFacilitative Principle speaks of “facilitating plans.”Some transmission principles say that “oughts” transmitfrom end to means, others say that the full force of the reason forthe end transmits to the means, others say that the strength of thereason for the end that transmits to the means is proportional toother factors, and others, like Raz’s Facilitative Principle,say only that some reason for the end transmits to the means.

To illustrate the problems facing an account of instrumentaltransmission, it may be useful to work with some very simpleprinciples, which lack many of the qualifications and nuances ofRaz’s Facilitative Principle and other principles discussed inthe literature:

Ought Necessity:

If one ought toE, andM is a necessary means toE-ing, then one ought toM.

Strong Reasons Necessity:

If one has reason toE, andM is a necessarymeans toE-ing, then that is as strong a reason toM.

Weak Reasons Necessity:

If one has reason toE, andM is a necessarymeans toE-ing, then that is some reason toM.

Ought Sufficiency:

If one ought toE, andM is a sufficient meanstoE-ing, then one ought toM.

Strong Reasons Sufficiency:

If one has reason toE, andM is a sufficientmeans toE-ing, then that is as strong a reason toM.

Weak Reasons Sufficiency:

If one has reason toE, andM is a sufficientmeans toE-ing, then that is some reason toM.

There are several problems that principles of instrumentaltransmission need to avoid. First, means to an end can becostly, or otherwise objectionable. Suppose that I ought totake a job in Alaska. One (sufficient) means is to fly there; another(sufficient) means is to take free passage on a merchant ship. Butflying costs more than I would earn there. Presumably, I ought notfly. It may be said that I have some reason to fly, but that it isoutweighed. This is an argument against Ought Sufficiency.

However, with other kinds of costs, or objectionable features, it maysound wrong to say even that one has reason for the means. Suppose Ihave reason to alleviate pangs of hunger. One sufficient means to thisis eating a sandwich; another sufficient means is killing myself(Broome 2005b). Some will resist the conclusion that I have any reasonat all to kill myself. This is an argument against Strong ReasonsSufficiency and Weak Reasons Sufficiency. Or suppose I have reason toimprove college policy. But a necessary means to this is to kill theold don who resists any change (because unanimous approval isrequired). Some will resist the conclusion that I have any reason atall to kill the don. This is an argument against Strong ReasonsNecessity and Weak Reasons Necessity. However, it may not be anargument against Ought Necessity. It may be said that the fact thatimproving policy requires killing the don is itself a strong reasonagainst improving policy, in light of which it is not true that oneought to improve it. In that case, Ought Necessity does not imply thatone ought to kill the don, which seems intuitive. In any event, theforce of the intuition on which the argument rests may be resisted.Granted, it sounds odd to say that one has reason to kill oneself. Butthis may have a pragmatic explanation. Why bother to identify anyreason to kill oneself when it is vastly outweighed by the reasonsagainst doing so? (Raz 2005a; Schroeder 2005a, 2007; for objection tothe pragmatic explanation, see Gertken and Kiesewetter 2021.)

Second, problems arise fromrepeated applications oftransmission principles. For one thing, meansto means to anend may actually undermine the end. Suppose, to take an example ofRyan Millsap’s, Kenny has reason to bake both a loaf of breadand a cake for a display on Marie Antoinette. A (necessary) means tothat end is baking a loaf of bread. Plausibly, he has reason to bake aloaf of bread. A (sufficient) means tothat means is baking agiant loaf, which uses all of the flour. But that would prevent himfrom baking a cake, and so prevent him from achieving the end. So,presumably, he does not have reason to bake a giant loaf, even thoughit is a means to something that he has reason to do. For anotherthing, if we accept both Weak Reasons Necessity and Weak ReasonsSufficiency, then repeated applications of these principles may implythat if there is reason for something, then there is reason foreverything. To take another example from Ryan Millsap, suppose thereis reason toE, and letF beanything. IfE-ing orF-ing is a necessary meanstoE-ing, and ifF-ing is a sufficient meanstoE-ing orF-ing, then, repeated application ofthose principles implies that there is reason toF. Note thatsimilar problems arise if we accept both Ought Necessity and OughtSufficiency.

Third, a means to an end might have a verylow probability ofhelping to bring that end about. Suppose, to take an example fromJackson and Pargetter 1986, Professor Procrastinate has been asked toreview a book, which he both ought, and has reason, to do. A(necessary) means to reviewing it is accepting the commission toreview it. However, even if he accepts the commission, he is veryunlikely to review it. Professor Dispatch, in a parallel universe, isin the same situation, except that he is very likely to review it ifhe accepts. Perhaps Procrastinate, like Dispatch, has some reason toreview it. But Procrastinate may not have as much reason as Dispatchto accept; not all of the reason for the end may be transmitted tothis (necessary) means. And it may not be the case that Procrastinateought to accept. This is an argument against Strong Reasons Necessityand Ought Necessity, since even if one takes a necessary means, theremay still be a low probability of helping to bring that end about.However, it is not an argument against the Sufficiency principles,assuming that sufficient means, if taken, are sure to help to bringthe end about.

In response to this problem, one might argue that Procrastinate bothought to write the review and ought to accept the invitation. Afterall, if we follow Jackson and Pargetter in thinking that Procrastinateboth ought to [write the review and accept the invitation]and ought to decline the invitiation, then it would no longerbe possible for Procrastinate to do everything that he ought to do. AsBenjamin Kiesewetter (2015) argues, we would then have to reject

Joint Satisfiability:

IfA ought tox andA ought toy, then it is possible forA to [x andy].

which might prove costly. It may be that such counterexamples to OughtNecessity (and Strong Reasons Necessity) rely on such costly views.(For further discussion, see Kiesewetter 2015, White 2017, andKiesewetter 2018.)

The fourth and final problem is that a sufficient means may besuperfluous to an end, where this means that the end wouldstill be achieved, even if those means were omitted. Suppose that Drug1 is a sufficient means to relieving the patient’s pain. Adoctor is sure to give Drug 1. Giving Drug 2, in addition, would atfirst neutralize the effect of Drug 1, but then recombine with it toproduce a third drug, which will relieve the patient’s pain.Giving Drug 2 would be superfluous to relieving the patient’spain, although it would help to bring it about, by being part of thecause of the relief of the patient’s pain. Does the doctor thenalso have any reason to give Drug 2? At first, one might be tempted tosay that the doctor does indeed have reason to give Drug 2, but thatthis reason is outweighed by the cost of giving Drug 2. But observethat no matter how low we make that cost, it still outweighs anyreason to give it. This suggests that there is no reason at all togive it. Note that necessary means cannot be superfluous, since ifnecessary means were left out, the end would not be achieved.

3. Instrumental Coherence

Now we consider instrumental rationality understood as a matter ofrational coherence. It is often said that someone is being locallyinstrumentally incoherent insofar as he intends an end, believes,roughly, that something is a means to it, but refuses to intend thatmeans. More precisely, it seems that this is so only when the relevantbelief is aboutnecessary means. Suppose that I intend to goto Alaska, but believe that flying there is a sufficient, but notnecessary, means to going to Alaska. Am I being instrumentallyincoherent in refusing to intend to fly there? Not obviously: perhapsI believe that going by sea is cheaper, and I intend that.

Some suggest that this focus, on intentions and beliefs aboutnecessary means, inspired by Kant’s initial discussion ofhypothetical imperatives (see§4.4 below), is overly narrow (see Smith 2004 and Wedgwood 2011). Notsimply intentions, but also desires, should be considered, and notsimply beliefs about necessary means, but also beliefs aboutnon-necessary means should be considered. And others have developedtheories of instrumental rationality focused on (intentional) actions,as opposed to intentions (see Tenenbaum 2021).

More precisely still, the relevant belief must be thatintending the means is itself necessary. Suppose your dentistintends to remove your tooth, and believes that a necessary means toremoving your tooth is causing you pain. Is she being instrumentallyincoherent in refusing to intend to cause you pain? Not necessarily:she might believe that she will remove your tooth even if she does notintend to cause you pain. But she will, presumably, believe that shewill remove your tooth only if sheintends to use herinstruments. So, though she wouldn’t be irrational in notintending to cause you pain, she would be irrational in not intendingto use her instruments (Kamm 2000, 2007; Searle, 2001; Broome2003).

Moreover, the relevant belief must be that intending the meansnow is necessary. Suppose a beginning student intends to earna doctoral degree and believes that, when the dissertation iscompleted, intending to submit the final paperwork is necessary forearning the degree. It is not clear that she is being instrumentallyincoherent in refusing to intend, now, to submit that paperwork. Shemay trust that, when the time comes, she will intend it (Setiya 2007).This suggests, as Raz notes, that opportunities for instrumentalincoherence may actually be very rare.

Even with the content of the relevant belief being restricted in theseways, worries might still arise given that (all-out) belief fallsshort of certainty. Suppose I believe the only way to arrive on timeis to steal my neighbor’s car, which I’m unwilling to. ButI also think there’s a very small chance I could bike in andarrive on time. (This is compatible with my all-out believingI’ll arrive on time only if I steal my neighbor’s car.)One might argue that it’s rationally permissible to intend toarrive on time in such a case (Nuñez 2020). However, perhaps weshould instead say here that it’s rationally permissible to havethegoal, oraim, of arriving on time, not theintention to arrive on time.

There seems to be considerable agreement, then, on what we might callthe:

Violation Claim:

If one is instrumentally incoherent—i.e., one intends att toE, believes att that one willE only if one intends att toM, but doesnot (or, more restrictively, refuses to) intend att toM—then one violates some requirement.

But it is unclear exactly what requirement one violates. Additionally,it is unclear whether the requirement violated would be“normative” in the sense that one ought not to violate it.In other words, it is unclear whether, in being instrumentallyincoherent, one violates arequirement of reason, in thesense of “reason” identified earlier. In the remainder ofthis section, we’ll consider some possible answers to thesequestions.

First, (§3.1) we might say that there isn’t any requirement of reason to beinstrumentally coherentas such, but whenever one isinstrumentally incoherent, one violates someindependentrequirement of reason. Second, (§3.2) we might say that there is a requirement of reason to beinstrumentally coherent as such. Third, (§3.3) we might deny that one necessarily violates a requirement of reasonin being instrumentally incoherent; perhaps the normativity ofinstrumental coherence is merely apparent, or the requirement merelyevaluative (it may be a standard of proper functioning without givinganyone reason to comply with it). Fourth, (§3.4) one might think, given the ways in which intention is thought toinvolve belief, that in being instrumentally incoherent, one violatesa requirement oftheoretical rationality.

There are, of course, other answers that could be given. (Forinstance, we briefly discuss a constitutivist approach to instrumentalincoherence, attributed to Kant, in§4.4) But consideration of these approaches will give a reasonablycomprehensive picture of the recent attempts to explain the ViolationClaim.

3.1 Independent Requirements of Reason

One answer is that when one is instrumentally incoherent one violatessome independent requirement of reason. That is,either onehas somespecific attitude (or pattern of attitudes)that one ought not to haveor onelacks somespecific attitude (or pattern of attitudes) that one ought to have.The source of the problem, according to this answer, isnotthat when one is instrumentally incoherent, one violates somerequirement of reason to be instrumentally coherentas such.The source of the problem is instead that when one is instrumentallyincoherent, then it is guaranteed that one’s having (or lacking)one attitude (or pattern of attitudes) violates someindependent requirement of reason. (See Kolodny 2007b, 2008a;Kiesewetter 2017, esp. Ch. 10; Lord 2018, esp. Ch. 2. For criticism,see Brunero 2020, Ch. 4; Fogal and Worsnip 2021; Worsnip 2021, Ch. 3,Lee 2021.)

Certainly, this can happen in some cases, at least accidentally.Suppose our madman intends to start World War III, believes that hewill start World War III only if he intends to launch a nuclearmissile, but refuses to intend to launch a nuclear missile. TheViolation Claim might be explained in this case by pointing out that,whatever his other attitudes may be, he ought not to have theintention to start World War III, but does have that intention. Butwhy should we suppose that, in general,whenever one has aninstrumentally incoherent set, oneeither has an attitude (orpattern) that one ought not to have,or lacks an attitude (orpattern) that one ought to have?

This would be so, in general, if the following were the case:

Ought Pattern 1:

In any situation, either

• one ought not to believethat one willE only if oneintends toM, or
• one ought not to intend toE, or
• one ought to intend toM.

It would also be so, in general, if the following were the case:

Ought Pattern 2:

In any situation, either

• one ought not to believethat one willE only if oneintends toM, or
• whatever pattern of intention one ought to have, it is notoneintends toE and does not intend toM; i.e.,either
  • one oughtnot to intend toE and not to intend toM, or
  • one oughtnot to intend toE and to intend toM, or
  • one oughtto intend toE and to intend toM, or
  • one ought eithernot to intend toE and not to intendtoM ornot to intend toE and to intend toM, or
  • one ought eithernot to intend toE and not to intendtoM orto intend toE and to intend toM, or
  • one ought eithernot to intend toE and to intend toM orto intend toE and to intend toM, or
  • one ought eithernot to intend toE and to not tointend toM ornot to intend toE and tointend toM orto intend toE and to intendtoM.

Ought Pattern 2 speaks of “oughts” for patterns ofintention, rather than for single intentions. The reason for this isthat one’s having an intention toX can affectone’s reason for having an intention toY. This canhappen in the following way. First, one’s reason to intend toY plausibly depends, among other things, on how likely one isto succeed in it. Second, how likely one is to succeed in it in turndepends on how the future is likely to be. I am more likely to succeedin drying the laundry if there is no rain. Third, how the future islikely to be can depend on how some person is likely to act. I am morelikely to succeed in drying the laundry if the children are unlikelyto be playing with the hose at the same time. Fourth, the relevantperson may be the agent himself. I am more likely to succeed in dryingthe laundry if I am unlikely to be at the same time watering thegarden. Finally, the fact that an agent intends toX can makeit more likely that the agent willX in the future. The factthat I intend to water the garden can make it more likely that I willbe watering the garden. Thus, the fact that an agent intends toX can make it more likely that the agent will not succeed inan intention toY, and so that the agent has less reason tointend toY. Given that one’s intending toXcan affect one’s reason toY and vice versa, thequestions, “What reason do I have to intend toY? Whatreason do I have to intend toX?” may be insufficientlydeterminate. Better questions may instead be “What reason do Ihave (to intend toX and to intend toY)? Whatreason do I have (to intend toX and not to intend toY)? What reason do I have (not to intend toX and tointend toY)? What reason do I have (not to intend toX and not to intend toY)?”

This phenomenon, or apparent phenomenon, that the presence of oneintention can affect one’s reason to have another intention,should be distinguished from the phenomenon, or apparent phenomenon,that the presence of one intention and the presence or absence ofanother can make it the case that one violates or satisfies a rationalrequirement of instrumental coherence as such. First, what matters forrational requirements is what webelieve about the relationsbetween the intentions, whereas what matters for reason would seem tobe how they arelikely to affect one another, which may beindependent of what we believe. Second, what matters for rationalrequirements is intending toX as such, whereas what mattersfor reason is what intending toX means for the future. Insome cases, intending toX (e.g., intending to become King ofFrance in 2023), maynot affect the likely future. In othercases, conditions other than intending toX (e.g., myself-trust that I willX even if I don’t intend it now,or my knowledge that I will be tortured or coerced intoX-ing) may affect the likely future in thesame way.Finally, rational requirements are usually taken to be wide-scope, aswe will see below. That is, one never needs to have or lack anyparticular attitude in order to satisfy them. By contrast, one usuallyneeds to have or lack a particular attitude in order to conform toone’s reason to have a certain pattern of intentions. If oneought, for example, (to intend toX and not to intend toY), then one can comply with this requirement of reason onlyby intending toX and not intending toY.

One might argue for Ought Pattern 1 in the following way. Suppose thatit is not the case that one ought not to believe that one willE only if one intends toM. Then by:

Epistemic Strictness:

If one believesP or is considering whetherP,and it is not the case that one ought not to believeP, thenit is the case that one ought to believeP.

one ought to believe it. Suppose further that it is not the case thatone ought not to intend toE. Then by:

Practical Strictness:

If it is not the case that one ought not to intend toE,then one ought to intend toE.

one ought to intend toE. Suppose further that:

Ought Intention:

One ought to intend toE only if one ought toE.

Then one ought toE. Then by:

Variant of Ought Necessity:

If one ought toE, and one ought to believe that one willE only if one intends toM, then one ought to intendtoM.

one ought to intend toM. In other words, either one oughtnot to believe that one willE only if one intends toM, or one ought not to intend toE, or one ought tointend toM.

One problem is that the objections to Ought Necessity that we saw in§2 seem to apply to Variant as well. Another is that PracticalStrictness seems untenable (Way 2012a; Cheng-Guajardo 2014). It mightbe the case neither that one ought not to intend toE, northat one ought to intend toE; intending toE mightbe merely permitted. Suppose I ought to take a can of beans. I cantake the can of beans on the left, or I can take the equally good canon the right. It is neither the case that I ought not to intend totake the can on the left, nor that I ought to intend to take the canon the left.

But perhaps in such cases one’sintending to take thecan on the left provides a reason that tips the scales, thereby makingit the case that one ought to intend the means to that end. (SeeSchroeder 2009, §4.1; For criticism of the view that intentionsprovide reasons in this way, see Broome 2001 and Brunero 2007.) Ifthat’s true, then these cases of “picking”wouldn’t pose a special problem for Ought Pattern 1; in all suchcases, one ought to intend toM.

One might argue for Ought Pattern 2 in the following way. Again,suppose that it is not the case that one ought not to have the belief.Then, by Epistemic Strictness, one ought to have the belief. For thetime being, suppose that one does not intend toM. Thenby:

Required Self-Knowledge:

If one does not intend toM, then one ought to believethat one does not intend toM,

one ought to believe that one does not intend toM. Thenby:

Reverse Closure:

If one ought to believe notQ and one ought to believeP only ifQ, then one ought to believe notP,

one ought to believe that one will notE. Then by:

Effectiveness:

So long as one ought to believe that one will notE, onehas no reason to intend toE, since the evidence suggeststhat doing so is pointless.

one has no reason to intend toE. In other words, so long asone does not intend toM, one has no reason to intend toE. If we also accept:

Cost:

There is reason not to intend toE,

then, so long as one does not intend toM, one has no reasonto intend toE and reason not to intend toE. Thismakes it plausible that there is more reason (not to intend toM and not to intend toE) than (not to intend toM and to intend toE). This doesnot meanthat one ought (not to intend toM and not to intend toE) all things considered. For example, it might be the casethat one ought (to intend toM and to intend toE).But it does mean that whatever pattern of intention one ought to have,it is not to (not to intend toM and to intend toE). There is a superior alternative.

One question, of course, is whether we should accept theseassumptions, in particular Required Self-Knowledge.

Another problem is that even if this explains Ought Pattern 2, and sothe Violation Claim, it does not explain the:

Satisfaction Claim:

One satisfies a requirement by avoiding or escaping instrumentalincoherence (of a kind which one would not satisfy by entering orremaining in instrumental incoherence) no matter how one does it.

Suppose that one ought to have the means-end belief and one ought (tointend toE and to intend toM). However, one avoidsincoherence by dropping the means-end belief, not intending toE, and not intending toM. Intuitively, one hassatisfied some requirement. This is what the Satisfaction Claimreflects. But one hasn’t adopted the particular attitudes (orpatterns) that one ought to adopt. How, then, are we to explain theSatisfaction Claim?

A quick answer might be this. One might suggest, first, that one oughtto be instrumentally coherent, because it is a necessary means to, orat least a necessary condition of, having the particular attitudes (orpattern) that one independently ought to have, or not having theparticular attitudes (or pattern) that one independently ought not tohave. Suppose that one ought to intend toM, but does not.Just as (P orQ) is a necessary condition ofP, so too a necessary condition of intending toM is(either not to intend toE, or not to believe that one willE only if one intends now toM, or to intend toM). Now suppose that instead of intending toM, onedrops one’s intention toE. While one hasn’tintended toM, which one ought to do, one has fulfilled anecessary means, or at least a necessary condition, of that. AssumeOught Necessity of§2, perhaps slightly strengthened: that one ought to take any necessarymeans to, or at least fulfill any necessary condition of, what oneought. In that sense, by dropping one’s intention toE,one has done something that one ought, and so satisfied a requirement.Analogously, if one ought to post the letter, then, by similarreasoning, one ought (either to post the letter or to burn it). So, ifone burns the letter, one has done something that one ought, and sosatisfied a requirement.

Note that this response assumes that either Ought Pattern 1 or OughtPattern 2 is correct. Otherwise, we have no guarantee that, in anygiven case, a necessary condition of having the particular attitudesthat one ought to have, or not having the particular attitudes thatone ought not to have, is (either not to intend toE, or notto believe that one willE only if one intends now toM, or to intend toM). So it inherits all of thechallenges in defending Ought Pattern 1 or Ought Pattern 2.

Another problem with this response is that it is not obvious thatOught Necessity is true, as the Professor Procrastinate example of§2 suggested.

A third problem is that necessary conditions of intending toM are easy to come by. Another necessary condition is (eitherintending toMor being instrumentally incoherent)!Therefore, in exactly the same way, one satisfies a requirement byremaining incoherent. But the intuitive Satisfaction Claim isprecisely that in escaping incoherence one satisfies a requirement ofa kind which one would not satisfy by remaining instrumentallyincoherent.

3.2 Requirements of Reason to be Instrumentally Coherent As Such

The Violation and Satisfaction Claims would be explained by arequirement of reason to avoid or escape instrumental incoherenceas such. Just as we ought not to torture, or ought to carefor our children, we ought to be instrumentally coherent as such.

One might instead suggest a weaker view: we have a reason to beinstrumentally coherent as such. But this weaker version faces adifficulty in explaining the “strictness” of the ViolationClaim. One could have a reason toX without being under anyrequirement toX, as when that reason toXis outweighed. But, according to the Violation Claim, whenever one isinstrumentally incoherent, one violates some requirement. This“strictness” would thus remain unexplained by the weakerview (Broome 1999). Of course, the weaker view—that we have areason to be instrumentally coherent as such—could be true eventhough it fails to explain the Violation and Satisfaction Claims.(However, some philosophers have raised important challenges to the“strictness” of rational requirements. Most notably, Fogal2020 suggests replacing a “requirements-based” account ofstructural rationality with a “pressure-based”account.)

So, let’s work with the stronger view, according to which weought to be instrumentally coherent as such. One worry about thisanswer is “bootstrapping.” Suppose that this requirementto be instrumentally coherent as such is stated in“narrow-scope” form: if one intends toE andbelieves that one willE only if one intends toM,then one is required to intend toM. Now suppose that oneintends to stay in power and believes that one will do so only if oneintends to kill one’s rival. If one ought to be instrumentallycoherent as such, it would follow that one ought to intend to killone’s rival. But that seems implausible. So, one may conclude,we cannot understand “one is rationally required” asimplying “one ought.”

3.2.1 Wide-Scoping and its Problems

A natural reply, however, is that there isanother way to beinstrumentally coherent: namely, to abandon the intention to stay inpower. Thus, a requirement of instrumental coherence as such shouldtake “wide-scope.” It should say: one is required toeither revise one’s intention toE, or reviseone’s belief that one willE only if one intends toM, or intend toM. (See especially Broome 1999 andWallace 2006. For earlier defenses of the wide-scope view, see Hill1973 and Darwall 1983.) So even if “one is rationallyrequired” is taken to imply “one ought,” all thatfollows is that one ought to either revise the intention to stay inpower, or revise the belief that one will do so only if one intends tokill one’s rival, or intend to kill one’s rival. It doesnot follow that one ought to intend to kill one’s rival.

However, there are some problems facing this approach, which weoutline in the supplemental documentProblems for Wide-Scoping. Some of those problems concern the allegedly implausible consequencesthat emerge when we apply some of the transmission principles wediscuss in§2 above—specifically, Ought Necessity, Weak Reasons Necessity,and Weak Reasons Sufficiency—to the relevant wide-scope ought incertain cases. Other problems concern whether the wide-scopeformulation is adequate as a formulation of the rationalrequirement.

3.2.2 What is the Reason?

There are other challenges facing the view that the Violation andSatisfaction Claims are explained by a requirement of reason to beinstrumentally coherent as such. The proponent of this view,presumably, should be able to saywhat the reason is to beinstrumentally coherent as such (see Kolodny 2005, §2).

Keep in mind that it won’t be enough to state a reason to bedisposed to be instrumentally coherent. It doesn’tfollow from one’s having a reason to have a disposition that onehas a reason to manifest that disposition on every particularoccasion. And so pointing to a reason to be disposed to be coherentwouldn’t yield a reason to be instrumentally coherent as such.(As we’ll see in§4.4, a similar problem confronts the Kantian constitutivist view thatcontinued instrumental incoherence threatens our status as agents.Since our agency isn’t at risk with every instance ofinstrumental incoherence, this account won’t yield a reason tobe instrumentally coherent on every particular occasion; it would onlyprovide a reason to do so when one’s incoherence is extremeenough to threaten one’s status as an agent.) Furthermore, it isnot clear what reasons there are to be disposed to be instrumentallycoherent. It might be thought that being disposed to be instrumentallycoherent will make one more likely to conform to reason over the longrun. But given that instrumental coherence may just as soon be gainedby revising one’s attitudes against reason as in conformity toreason, one might doubt that this is so (Kolodny 2008b).

There’s a reason to be skeptical that such a reason could beprovided. A reason to be instrumentally coherent as such would have toapply at once to beliefs and intentions. But one might think that thiscannot be the case: that reasons for belief and reasons for intentionare fundamentally different in kind. Reasons for belief areconsiderations that show that the belief is likely to be true, whereasreasons for action are considerations that show that the action wouldrealize one’s desires, or promote or respect something ofvalue.

Some philosophers have attempted to meet this challenge. Jonathan Way(2012a) has developed an account which points to the costs necessarilyinvolved in intending toE without also intending toM, specifically the psychological resources that would bewasted (one’s monitoring one’s progress towardsE-ing, being disposed to take means towardsE-ing,etc.). And Michael Bratman (2009c) has suggested that we have a reasonto be instrumentally coherent as such because doing so is a necessary,constitutive component of self-governance, and we have an intrinsicreason to govern ourselves. In his view, just as we have a reason notto torture, and a reason to care for our children, we have a reason togovern ourselves. And one’s being instrumentally incoherent onany particular occasion would be a way of failing to govern oneself.Also, Alex Worsnip (2021, Ch.8) has argued that consideration aboutcoherence constitute (non-derivative) reasons tostructuredeliberation in certain ways — in particular, theyconstitute reasons to treat certain options as“off-limits” in one’s deliberations. On this view,we would have reason to treat the instrumentally incoherentcombination [intending to X, not intending to Y, believing intendingto Y is necessary for X-ing] as off-limits. Why is there such areason? For Worsnip, it’s simplyfitting to stucturedeliberation in a way that respects coherence constraints.

3.3 Instrumental Incoherence without Violating Requirements of Reason

A third approach to instrumental incoherence holds that one need notviolate any requirement of reason in being instrumentally incoherent.One version of this view (§3.3.1) holds that instrumental coherence is to be explained in terms ofappropriate responses to one’sbeliefs aboutindependent requirements of reason, where these beliefs could befalse. Another version of this view (§3.3.2) holds that instrumental incoherence is nothing more than an indicatorthat one is not functioning well as an agent, where this evaluationdoesn’t entail any particular conclusions about what one ought,or has reason, to do.

3.3.1 Responsiveness to Beliefs about Independent Requirements of Reason

On this first approach to explaining the Violation and SatisfactionClaims, what one does “wrong” in being instrumentallyincoherent is violate one’s own judgments about one’sreasons (or fail to be attentive enough to form such judgments in thefirst place), and what one does “right” in becominginstrumentally coherent is respond appropriately to one’sjudgments about one’s reasons.

Suppose one is instrumentally incoherent. Plausibly, one is in aposition to know this, that one is instrumentally incoherent.Plausibly, one is also in a position to know Ought Pattern 2: thegeneral fact that, when one is instrumentally incoherent, there issome specific change that one ought to make. So, one is in a positionto know that there is some specific change that one ought to make. So,plausibly, one ought to give some thought to what this specific changeis. In other words,

Self-Monitoring:

If one is instrumentally incoherent, then one ought to arrive atsome belief about which attitude (or pattern of attitudes) one oughtto have.

Suppose one arrives at some such belief. And suppose one revisesone’s attitudes in accordance with that belief. If one becomesinstrumentally coherent inthis way, then, even if onerevises one’s attitudes in the wrong way (i.e., against reason),one still satisfies:

1. Self-Monitoring, and
2. Krasia: If one believes that reason requires one toX (ornot toX), then one is rationally required toX (ornot toX).

This would explain the Satisfaction Claim.

It might be thought that it’s not much of an improvement toexplain the requirement to be instrumentally coherent in terms ofSelf-Monitoring and Krasia, since we are still appealing to arequirement of coherence: namely, Krasia. The only difference is thatit requires us not to beinstrumentally coherent, but insteadto bekratically coherent: i.e.,to follow one’sjudgment about one’s reasons. Indeed, if we thought that itwere a requirementof reason to comply with Krasia, thensimilar challenges to those we discussed in§3.2. would reappear, and so it would not be much of an improvement.

But, the view considered here naturally suggests a new strategy:presenting a Transparency Account according to which the normativityof instrumental coherence is merely apparent (Kolodny 2005, §5).According to the Transparency Account of rational requirements ingeneral, when we point out to people that rationality requiressomething of them, we aren’t presenting them with somerequirement of reason; rather, we’re pointing out to them whatthey already think reason requires of them. And, so, when we point outto someone that rationality requires they be instrumentally coherentwith respect to some end, we’re pointing out to them that (onthe assumption that Self-Monitoring is satisfied) they already thinkthat reason requires some specific change in their attitudes. Ofcourse, the instrumentally incoherent agent could be mistaken in thesethoughts; perhaps he thinks reason requires him to intend the meanswhen reason actually requires that he abandon the end instead. It isthis possibility for false belief that restricts us to saying that thenormativity of rational requirements is merely apparent.

However, one might challenge whether the Satisfaction Claim isadequately explained by an appeal to Self-Monitoring and Krasia.Suppose an instrumentally incoherent agent knows that she ought togive up her intention to kill her boss, but instead she comes tointend to poison the office coffee, which she believes to be anecessary means to killing her boss. She clearly violates arequirement of reason and she clearly violates Krasia, but it stillseems that she satisfies some requirement in coming to intend themeans to her immoral end. The appeal to Self-Monitoring and Krasiacan’t explain our thinking that she satisfies some requirementin proceeding this way.

3.3.2 Proper Functioning

Perhaps a violation of a requirement of instrumental coherence ismerely an indicator that an agent is not functioning well in theexercise of her rational capacities, where this evaluationdoesn’t entail that she ought, or has reason, to beinstrumentally coherent on any particular occasion (see Raz 2005a,b).In holding this view, one could also say that there is value in beinga properly functioning agent, and that this value is explained by theways in which functioning properly as a agent tends to bring one intoconformity with what reason requires. But, it doesn’t followfrom this that there is value in functioning properly, or that one hasany reason to function properly, in every particular occasion.(Similar things could be said about the value of havingwell-functioning perceptual capacities.) For instance, when one hasthe unalterable intention to kill one’s boss (for more onunalterable ends, seeSection 1 of the supplementary documentProblems for Wide Scoping), there is nothing valuable in one’s functioning properly, so faras instrumental rationality goes, on that particular occasion, and onewould have no reason to do so (see Raz 2005a, especially pp.10–13, and Bratman 1987).

But one might worry about how well this account explains the Violationand Satisfaction Claims. On this view, the requirements violated, orsatisfied, are simply standards of proper rational functioning. Thesestandards can be used to evaluate how well particular agents arefunctioning, in the same way that standards of what makes for awell-functioning perceptual apparatus can be used to evaluate how wellparticular perceivers are functioning. But, rational requirements alsoseem tocall for some response. For instance, when we adviseothers who are in violation of rational requirements, such as thosewho are instrumentally incoherent, we say that some change in attitudeis called for. And, when we ourselves are instrumentallyincoherent, we feel some “normative pressure” to changeour attitudes in some way (see Kolodny 2005, §4). It’sunclear how these features of rational requirements could be explainedby a view that takes them to be merely evaluative.

Another related view attempts to explain instrumental incoherence interms of thesuccess conditions of the relevant individualattitudes. If your belief that intending to Y is necessary for X-ingis successful (true), then your failure to intend to Y ensures thatyour intending to X will not be successful (in that it won’tresult in your X-ing). In other words, your being instrumentallyincoherent ensures that either your instrumental belief is false, orthat your intention will not be successfully executed (Brunero 2020,Ch.7). Some philosophers have recently argued that the impossibilityof joint attitudinal success can explain coherence requirements ingeneral, not just instrumental rationality (Fullhart and Martinez,forthcoming; Fink, forthcoming).

3.4 “Cognitivist” Requirements on Belief

A fourth approach to explaining the Violation and Satisfaction Claimsargues that the requirement violated, or satisfied, is, or isexplained by, requirements onbelief exclusively. There aretwo components to this “cognitivist” approach: first,there is some thesis about how intention involves belief, and, second,there is some account of how the requirements governing those involvedbeliefs constitute, or explain, the requirement to be instrumentallycoherent.

Here are some candidates for the first component:

• Strong: If one intends toE, then one believes that onewillE.
• Medium: If one intends toE, then one believes that it ispossible that one willE.
• Weak: If one intends toE, then one does not believe thatone will notE.

(Cognitivist accounts developed with Strong, or a variant of it, arefound in Harman 1976, 1986; Velleman 1989, 2007; Setiya 2007; Broome2009; and Archer 2018. Earlier defenses of Strong can be found inHampshire and Hart 1958 and Grice 1971. Wallace 2006 develops acognitivist account with Medium, and Binkley 1965 does so withWeak.)

3.4.1 Strong

Let’s begin with Strong. Suppose I am instrumentallyincoherent—that is, I intend toE, believe I willE only if I intend toM, but don’t intend toM. According to Strong, I believe

1. I willE.

When we pair this with my instrumental belief

2. If I willE, then I intend toM,

and apply the theoretical rational requirement

Closure:

Rationality requires that [if one believesP, and onebelievesP → Q, then one believesQ].

we get the result that I am required to believe

3. I intend toM,

or drop either (1) or (2). If I drop (1), I would cease to intend toE (according to Strong) and thereby cease to beinstrumentally incoherent, and if I drop my instrumental belief (2), Iwould also cease to be instrumentally incoherent.

But is coming to believe (3) enough to cease to be instrumentallyincoherent? What if my belief is false? There are two options for thecognitivist at this point. The first would be to defend

Self-Fulfillment:

If one believes that one intends toM, then one intendstoM.

But we might doubt that we are infallible in our beliefs about our ownintentions (Bratman 2009b, §4,5). After all, it seems that we canbe mistaken in our beliefs about other attitudes we have (seeSchwitzgebel 2010, §4).

A second option for the cognitivist would be to say that having falsebeliefs about one’s own intentions ispossible, buttheoreticallyirrational. (On this option, the cognitivistapproach to instrumental coherence appeals totwo theoreticalrequirements: Closure plus some other requirement not to have falsebeliefs about one’s own intentions.) However, there is usuallynothing necessarily incoherent about one’s having afalse belief. So, the challenge for cognitivists is toexplain why, in this particular case, there is. (This strategy ispursued by Setiya 2007, 670–671. For criticism, see Bratman2009a, §4 and Brunero 2009, §2.)

Additionally, one might worry that Strong is too strong. Suppose youintend to go shopping on Thursday but don’t believe you intendthis and instead believe that you intend to go shopping on Friday.Maybe you’ve temporarily forgotten that you’ve recentlychanged your mind, deciding to go on Thursday instead of Friday(Bratman 2009a, §5). In this case, you might not believe you willgo shopping on Thursday. In light of this example, John Broome (2009,§7) suggests the following

Variant of Strong:

If onebelieves one intends toE, then onebelieves one willE.

But this wouldn’t help the cognitivist if she aims to accountfor the application of instrumental coherence to all intentions,including those we don’t believe we have (Bratman 2009b,§2). Additionally, both Strong and Variant of Strong seemvulnerable to counterexamples. Suppose I intend, and believe I intend,to carry out a difficult task. Knowing the difficulty involved, Icould intend to carry out the task but remain agnostic about whether Iwill do so. Here’s another counterexample: suppose I intend tostop by a bookstore on the way home, but know that once I’mriding on my bicycle, I tend to go on “autopilot.” Awareof my absentminded tendencies, I might intend to stop by the bookstorewithout believing I will (Bratman 1987, 38–39; also discussed inHolton 2008, §2). Since such intentions make applicable arequirement of instrumental coherence, but don’t involve abelief in success, the cognitivist cannot explain why requirements ofinstrumental coherence apply to these intentions.

3.4.2 Medium

Next, let us consider Medium. Suppose again that I am instrumentallyincoherent—that is, I intend toE, believe I willE only if I intend toM, but don’t intend toM. By

Self-Knowledge:

If one does not intend toM, then one believes that onedoes not intend toM.

I believe:

1. I do not intend toM.

Let us now suppose that the instrumental belief takes this form:

2. If I do not intend toM, then it is not possible for metoE.

If I satisfy Closure, then I believe:

3. It is not possible for me toE.

However, since I intend toE, then, by Medium, I believe:

4. It is possible for me toE.

Thus, in believing (3) and (4), I violate:

Belief Consistency:

Rationality requires that [if one believes notP, thenone does not believeP]

Thus, if one satisfies Closure, then one violates Belief Consistency(Wallace 2006).

One might object to Self-Knowledge, arguing that we aren’talways aware of what we fail to intend. One might falsely believe oneintends the means (Bratman 2009a) or one might simply fail to noticethat one hasn’t intended the means (Brunero 2005a). One responseavailable to the cognitivist would be to say that such lapses inself-awareness arepossible, but are theoreticallyirrational. And so instrumental coherence would be explainedby Closure, Belief Consistency, and the requirement to be aware ofone’s absent intention in this context. While it wouldn’tbe plausible to claim that rationality requires, in general, awarenessof what one does not intend, it might be plausible to say thatawareness of what one does not intend is rationally required in thisspecific context (Wallace 2006, 118).

One might also object that if one is going to be a cognitivist aboutpractical requirements in general, one must account for bothinstrumental coherence and

Intention Consistency:

One is required [if one intends not toE, then one doesnot intend toE].

But Medium, on its own, does not seem sufficient to account forIntention Consistency. Such a conflict of intentions would involve thefollowing beliefs:

5. It is possible for me not toE,
6. It is possible for me toE,

But unlike those beliefs involved according to Strong, namely,

7. I will notE,
8. I willE,

(5) and (6) are perfectly consistent with one another. I couldconsistently believe it’s possible that I not go to the gametonight, and possible that I go to the game tonight. So, we are stillin need of an explanation of why I’m rationally required to notboth intend not to go and intend to go (Bratman 2009b, §3).

One might also raise doubts about Medium. Suppose I know Ihaven’t yet investigated whether the stores are closed for theholiday on Thursday, and so I haven’t yet formed any beliefsabout whether I can go shopping then. Couldn’t I still intend togo shopping on Thursday? That might be irrational or careless planningon my part, but it seems possible. (The example might prove even moreconvincing if we add the feature of Bratman’s example above:suppose, forgetting about my change of mind, I believe I intend to goshopping on Friday, not on Thursday.) Perhaps doubts could even beraised, not just about whether intending toE involvesbelievingE is possible, but also about whether intending toE involvesnot believingE is impossible.Richard Holton (2008), drawing on Anscombe’s example fromIntention (§52) of a man who is certain he’llbreak down under torture but is determined not to break down, hasargued that one could intend some end while believing that endisn’t a “live possibility.” Such combinations,Holton argues, may be irrational, but aren’t impossible.

3.4.3 Weak

Next, let us consider Weak. Suppose again that I am instrumentallyincoherent—that is, I intend toE, believe I willE only if I intend toM, but don’t intend toM. By Self-Knowledge, I believe:

1. I do not intend toM.

When this belief is paired with the contrapositive of my instrumentalbelief

2. If I do not intend toM, I will notE,

Closure requires that I believe

3. I will notE,

or cease to believe (1) or (2). But since I intend toE, itfollows, by Weak, that I do not believe (3). Were I to comply withClosure by forming the belief (3), I would, according to Weak, therebycease to intend toE. In ceasing to believe (2), I wouldcease to be instrumentally incoherent. And in ceasing to believe (1)(which, if Self-Knowledge is true, would mean that it is no longer thecase that I do not intend toM) I would also cease to beinstrumentally incoherent.

Some of the objections mentioned above would also apply to thisaccount. There is the objection to Self-Knowledge. AndAnscombe’s example, exploited by Holton, also presents achallenge to Weak, since the torture victim intends not to break downbut believes that he will.

Additionally, one might raise questions concerning the rationality ofresponding to the violation of Closure in this context.Usually, if one discovers that one believesP,P → Q, but does not believeQ, the rational way to proceed depends upon one’sassessment of the relevant reasons for belief (where these areevidential, not pragmatic, reasons). Let us assume, for simplicity,thatP → Q is a fixed backgroundbelief. If one thinks there are really strong reasons for believingP and not strong reasons for not believingQ, thenit would be irrational for one to refuse to believeQ andinstead to drop one’s belief thatP. Now, according tothis cognitivist account, when I’m instrumentally incoherent, Ibelieve (1) and (2) but not (3). How should I proceed? If I currentlydo not intend toM, and if, as Self-Knowledge holds,I’m infallibly aware of what I do not intend, then it seems thatthe evidence I have for my belief that I do not intend toMis really quite strong. Let’s assume (2) is a fixed backgroundbelief. In revising in light of my assessment of the evidence, itseems that the revision could only proceed in one direction: bymodus ponens, coming to believe I will notE. Itwould be irrational for me instead to proceed bymodustollens, ceasing to believe that I do not intend toM.So, if we respond to violations of Closure in the usual way, we areleft with the counterintuitive result that there is only one way toescape a state of instrumental incoherence: by abandoning the end.

Finally, one might also cast doubt on the general cognitivist strategybeing employed here, whether it appeals to Weak, Medium, or Strong.Many of the same questions that we’ve raised about instrumentalincoherence could be raised about rational requirements on belief,including Belief Consistency and Closure. There are analogues of theViolation and Satisfaction Claims for these rational requirements thatwould need to be defended and explained. For example, Joseph Razquestions whether the analogue of the Violation Claim for BeliefConsistency is even true (see Raz 2005b, §3; Raz 2005a, §5).If one has a set of beliefs that is inconsistent, then at least one ofthe beliefs in that set is false. But, Raz argues, it does not followthat there is any revision in one’s beliefs that one ought tomake. (Reflection on logical paradoxes, like the sorities paradox,shows that having an inconsistent set of beliefs doesn’t implythat there is any belief in the set that one ought to give up.)Although Raz objects to the cognitivist’s appeal to BeliefConsistency in particular, similar questions might be raised about thecognitivist’s appeal to Closure.

4. The Status of Instrumental Rationality

Much discussion of instrumental rationality has taken place in thecontext of a debate about whether instrumental rationality is the onlykind of practical rationality or, alternatively, there are otherrequirements of practical rationality as well, such as requirements ofprudence and morality. Typically, the former view is called a“Humean” one and the latter a “Kantian” one.In this section, we’ll consider some possible understandings ofthe thesis that instrumental rationality is the only kind of practicalrationality (§4.1), and whether some version of this thesis is given any support by aninstrumental analogue of Lewis Carroll’s famous paradoxconcerningmodus ponens (§4.2). We’ll also consider the views of Hume (§4.3) and Kant (§4.4) on instrumental rationality.

4.1 The Only Kind of Practical Rationality?

It is sometimes claimed that instrumental rationality is the only kindof practical rationality. But what exactly is being claimed?

Is the claim about rational incoherence? Then it would seem to be theclaim that the only kind of practical, rational incoherence isinstrumental incoherence. But this claim would have to contend withthe idea that there are other kinds of practical, rationalincoherence. For example, it seems incoherent to refuse to intend whatone believes one ought to intend. It also seems incoherent to intendtoX and to intend toY when one believes that ifoneXs, one will notY.

Alternatively, the claim might be about reason: that all (practical)reason is, in some sense, instrumental. On one interpretation, theclaim would say that theonly true claim about reason is someprinciple of instrumental transmission. But this would seem toimply that there are no practical reasons at all. Unless there is someother true claim about reasons for “ends” understood asactions, there is nothing to be transmitted. (For arguments that therewould need to be another kind of rationality, see Nozick 1993;Korsgaard 1997; Hampton 1998; Hubin 1999; Lavin 2004.)

A different interpretation is thatall reasons areinstrumental. In other words, the reason for any action (orintention) can be expressed as: the action will help to bring aboutsome “end,” where “end” is understood not asan action, but instead as some valuable or desired state of affairs.The Desire-Based Theory is committed to this claim. But not allValue-Based Theories are. The reason for an action, for example, mightbe that it honors or respects something of value, even though it doesnot help to bring it about.

A final interpretation is thatall deliberation is of means.In other words, there is no place for significant evaluative ornormative reflection on the ultimate ends that one has reason to bringabout. The Desire-Based Theory seems committed to this claim, at leastto some extent. According to the Desire-Based Theory, reflection onthe ultimate ends that one has reason to bring about is justreflection on the non-normative matter of what one in fact desires,plus grasp of the Desire-Based Theory itself. On the other hand,Value-Based theories seem committed to rejecting this claim, eventhose Value-Based Theories that accept that all reasons areinstrumental in the sense discussed in the previous paragraph. Forwhether a certain state of affairs is valuable, and so an end that onehas reason to bring about, is itself the object of significantevaluative thought. (Note that it is a mistake to express the presentidea, that all deliberation is of means, as the idea that one cannotbecriticized for ends. According to the Desire-Based Theory,one can be criticized for what oneintends oractsto bring about, insofar as it is not the realization of one’sstrongest, present intrinsic desires. And even according toValue-Based theories, one cannot be criticized for what one hasreason to bring about.)

Notice that the Desire-Based Theory is committed to the claim thatpractical reason is exclusively instrumental, on these two lastinterpretations. This may explain why the Desire-Based Theory is oftensimply identified with “instrumental rationality,” and whythe claim that “the only kind of practical rationality isinstrumental rationality” is sometimes just meant as anassertion of the Desire-Based Theory.

4.2 The Relevance of Carroll’s Paradox

Lewis Carroll (1895) introduced a famous paradox concerningmodusponens. Some philosophers have argued that an instrumentalanalogue of this paradox shows there to be something special aboutinstrumental rationality, and perhaps even provides support for someversion of the thesis that instrumental rationality is the only kindof practical rationality.

Carroll imagined a conversation between Achilles and a Tortoiseregarding the following valid argument:

(A)

Things that are equal to the same are equal to each other.

(B)

The two sides of the Triangle are equal to the same.

Therefore,

(Z)

The two sides are equal to each other.

The Tortoise adopts a skeptical position, asking how he can belogically forced into accepting (Z) when he accepts (A) and (B).Achilles attempts to supply him with the missing premise:

(C)

If (A) & (B) are true, (Z) must be true.

But the Tortoise grants (C), and resumes his skepticism, asking how hecan be logically forced to accept (Z) when he accepts (A), (B), and(C). It soon dawns on Achilles that the mere addition of furtherpremises (like “If (A), (B), and (C) are true, (Z) must betrue”, and so on) won’t help resolve the Tortoise’sskepticism. One could keep on adding such premisesadinfinitum and yet not get the Tortoise to accept (Z).

Much has been written about what lessons to draw from Carroll’sparadox, but this much seems clear: the Tortoise refuses to accept (Z)because he refuses toapply modus ponens to the premises heaccepts. So, ifmodus ponens itself ismerelyaccepted by him as yet another premise, that won’t suffice toget him to accept (Z).

Now, imagine we confront someone who holds some endE,believes some meansM is necessary forE-ing, yetdoesn’t intend toM, and, like the Tortoise,doesn’t intend toM because he refuses toapply instrumental rationality to the ends he has, includingE. If we suggest that he remedy this defect by adopting yetanother end (namely, the end of doing what instrumental rationalityrequires), this wouldn’t help matters. Since his problem is thathe doesn’t apply instrumental rationality to the ends he has, hewouldn’t apply instrumental rationality to this new end as well,and so still wouldn’t intend toM (or drop theintention toE). So, the mere addition of instrumentalrationality as another end would be futile.

What should we conclude from this instrumental analogue ofCarroll’s paradox? Railton (1997) argues that it shows that acertain form of skepticism about rationality—specifically,one’s asking “Why should I comply with the norms ofpractical rationality?”—can be answered. First, the askingof this skeptical question reveals a commitment to at leastinstrumental rationality, since one is here questioning whether beingpractically rational wouldadvance one’s ends. And,second, the instrumental analogue of Carroll’s paradox showsthat we can’t think of this commitment to instrumentalrationality as merely another end the agent accepts and could dismissat will. If the skeptic thinks it is, Railton suggests that “wecan answer ‘No, on pain of regress, it cannot be just anotherend of yours’” (p. 317).

However, it’s not clear why the skeptic would have to think ofinstrumental rationality as just another end she has. Couldn’tshe think of it instead as (or also as) a disposition that shemanifests, or a capacity that she exercises, with respect to the endsthat she accepts? (If so, she would avoid the particular problem thatconfronts the character in the instrumental analogue ofCarroll’s paradox, who merely accepts instrumental rationalityas another end withoutapplying it to her ends.) And thencouldn’t she request some philosophical explanation of why sheshould exercise that capacity? That she already has the capacity is ofcourse no barrier to posing such a question (Wedgwood 2005). Indeed,as we’ve seen in§3.3, some philosophers havedenied that we always ought, or havereason, to be instrumentally coherent; they’ve argued thatrational requirements are only apparently normative, or are merelyevaluative.

Dreier (2001) concludes that the instrumental analogue ofCarroll’s paradox shows that instrumental rationality has aspecial status. In Carroll’s original paradox, the Tortoisewould be such that we couldn’t give him reasons to believe insome proposition. In the normal case, I can provide you with a reasonto believeQ by getting you to believeP and tobelieveP → Q. But this simplywon’t work for the Tortoise, who refuses to applymodusponens. Likewise, Dreier observes, we would be incapable ofgiving reasons to someone who refused to draw “practicalinferences” between ends and means. The acceptance ofinstrumental rationality is thus, he argues, a necessary condition forbeing receptive to reasons at all. And this supports his conclusionthat “the only ultimate sort of reasons are instrumentalreasons” (p. 43). This seems to be an endorsement of theDesire-Based Theory.

Does the conclusion follow? Even if a disposition to be instrumentallyrational is indispensable in the way Dreier suggests tohaving a practical reason, nothing follows aboutwhat one has practical reason to do. Additionally, it’sworth noting that it’s doubtful that Carroll’s originalparadox would license a restriction on the scope oftheoretical reasons analogous to the Humean restriction onthe scope of practical reasons that Dreier thinks the instrumentalanalogue of Carroll’s paradox supports. That is, it is notobvious that I have reason to believeQ only if I am disposedto come to believe it by applyingmodus ponens. Perhaps mymerely believing, with reason, bothP andP → Q provides me with a reasonfor believingQ. Or perhaps the mere fact the availableevidence supportsQ is enough (Jollimore 2005).

Korsgaard (2009) argues that we can draw a conclusion about themetaethical status of instrumental rationality by consideringan instrumental analogue of Carroll’s paradox. Specifically, sheappeals to it to argue against metaethical realism about instrumentalrationality (and about other normative principles as well).Let’s suppose that there is a requirement of reason to beinstrumentally coherent. A realist would claim that there is somenormative fact, according to which we ought to be instrumentallycoherent. But a skeptic might ask, “Why ought I apply thatfact—that I ought to be instrumentally coherent—to myactions?” The realist might answer, “Because—andhere’s another normative fact—it just is the case that youought to apply the fact that you ought to be instrumentally coherentto your actions.” But one might resume one’s skepticismand ask why one must applythat normative fact to one’sactions. One could keep adding normative factsad infinitumwithout providing an answer to the question, in the same way thatAchilles could keep adding premises without meeting theTortoise’s challenge.

A full discussion of the metaethical issues raised by this argumentwould take us too far afield (see Parfit 2006). But one strategy ofreply is to argue that the realist has already answered the question.What it is for such normative facts to benormative just isthat they require that you act in some way. If you’re asking ofnormative facts in general why you should act in accordance with them,then you don’t fully understand what normative facts are.

4.3 Hume on Instrumental Rationality

The thesis that instrumental rationality is the only kind of practicalrationality is often taken to be a “Humean” thesis,though, as we’ve seen (§4.1), there are several ways in which this thesis could be understood. Onecould take the thesis just to be an assertion of the Desire-BasedTheory. One could also take the thesis to specify that instrumentalcoherence is the only requirement of practical rationality. However,it is doubtful that Hume held either of these views.

There are some passages which might seem to suggest these views. InBook II of theTreatise (1739–40), Hume denies that aperson’s ends, however immoral or imprudent, can be unreasonable(“contrary to reason,” to use his exact phrase) ifthey’re not based on a false belief:

Where a passion is neither founded on false suppositions, nor chusesmeans insufficient for the end, the understanding can neither justifynor condemn it. ’Tis not contrary to reason to prefer thedestruction of the world to the scratching of my finger. ’Tisnot contrary to reason for me to chuse my total ruin, to prevent theleast uneasiness of anIndian or person wholly unknown to me.’Tis as little contrary to reason to prefer even my ownacknowledg’d lesser good to my greater, and have a more ardentaffection for the former than the latter. (p. 416)

Here he seems to claim that one’s choosing means insufficientfor one’s end can be deemed unreasonable, even though the enditself is not subject to such rational assessment.

However, this wouldn’t support the Desire-Based Theory, sincethe Desire-Based Theory allows that ends can be unreasonable when theydon’t promote the satisfaction of one’s strongest, presentintrinsic desires. For instance, choosing one’s total ruin, whendoing so runs counter to one’s strongest, present intrinsicdesires, even when doing so isn’t based on a false belief, canbe criticized as unreasonable on the Desire-Based Theory.

Soon after this passage, Hume is careful to note that itwouldn’t be one’sintending of means insufficientfor one’s ends that’s unreasonable, but the falsebelief on which this intention is based. For example, when Ifalsely believe that staying on 95N will get me to Memphis, which iswhere I intend to go, and so I intend to stay on 95N, it’s mybelief that 95N will get me to Memphis that’s unreasonable, notthe intention, based on this belief, to stay on 95N.

Why did Hume think this? Hume holds that “reason is thediscovery of truth and falsehood” (p. 458). Since beliefs aim tocorrectly represent the world—and the beliefs that succeed inrepresenting the world are true, while the beliefs that fail to do soare false—they can be assessed as reasonable or unreasonable.But passions, as well as intentions and actions, are “originalexistences.” They don’t purport to represent or“copy” some other “existence” (p. 415).Specifically, they don’t aim to correctly represent the world,and so they can’t be true or false. Hence, they can’t beassessed as reasonable or unreasonable.

In light of such considerations, many philosophers have argued thatHume doesn’t endorse an instrumentalist conception of practicalreason, but instead, more radically, is a skeptic about all ofpractical reason. (See, for instance, Hampton 1995; Millgram 1995;Korsgaard 1997. For objections to this skeptical interpretation ofHume, see Setiya 2004.) On this interpretation, Hume acknowledges theexistence of requirements oftheoretical reason, sincebeliefs can be assessed as reasonable or unreasonable, but isskeptical about the existence of requirements ofpracticalreason, including instrumental reason, since desires, intentions, andactions can’t be assessed as reasonable or unreasonable. (Even apreference for “one’s own acknowledg’d lessergood” wouldn’t be unreasonable, on his view.) Since, onthis interpretation, Hume was skeptical about practical reasoningeneral, we can’t appeal to his view to establish a specialplace for instrumental rationality among the requirements of practicalrationality.

Various challenges could be raised against Hume’s rather narrowconception of reason. But our primary interest here is in whether Humecan provide us with a plausible account ofinstrumentalrationality. On Hume’s view, our intentions are subject torational criticism when, and only when, they are based on falsebeliefs (and, again, it’s the belief that’s to blame,strictly speaking). As an account of instrumental rationality,Hume’s view would face two serious objections.

First, it seems that one can be instrumentally irrational withouthaving any intentions based on false beliefs (or lacking anyintentions because of false beliefs). John Broome provides an example:suppose I know the hotel is on fire and that the only escape involvesmy intending to go through the open window, and, though I intend tolive, I sit on the windowsill singing happily until the flames consumeme. This is a clear case of instrumental irrationality, despite my nothaving any false beliefs (Broome 1997). Indeed, one could fail tointend the means believed necessary for one’s ends because ofrage, depression, distraction, grief, or other causes whichdon’t necessarily involve any false beliefs (Korsgaard 1986).Hume would have to conclude either that such cases are impossible, orthat they involve no instrumental irrationality.

Second, it seems that one can have intentions based on false beliefswithout being instrumentally irrational, or irrational in any otherway. Instrumental rationality, as it’s usually understood, is arequirement ofcoherence governing one’s intentions andbeliefs—a requirement that is in place regardless of thetruth of those beliefs. The person who intends to get toMemphis, believes that he must intend to stay on 95N to get there, anddoes intend to stay on 95N, displays coherence in his attitudes,despite having a false belief. So, an accusation of instrumentalirrationality here would be misplaced. (Additionally, it would also beinappropriate to accuse him of anytheoretical irrationality,if we suppose that this belief isn’t inconsistent with any otherbeliefs he holds, and he also believes there’s conclusiveevidence for it, and so forth.) Hume would have to conclude eitherthat such cases are impossible, or that they involve instrumentalirrationality.

4.4 Kant on Hypothetical Imperatives

An alternative to the narrow instrumentalism associated with Hume isfound in the work of Immanuel Kant, who thought that rationalityrequired us not only to be instrumentally rational, but also to obeythe moral law. In hisGroundwork for the Metaphysics ofMorals (1785), Kant distinguishes two kinds of imperatives:

Now all imperatives command either hypothetically or categorically.The former represent the practical necessity of a possible action as ameans for attaining something else one wants (or may possibly want).The categorical imperative would be one which represented an action asobjectively necessary in itself, without reference to another end.(4:414—citations use volume and pages from the Akademie Edition)

For example, “If you intend to go to law school, take theLSAT” would be a hypothetical imperative since it commands anaction (taking the LSAT) as a means to some end (going to law school).If you didn’t have any ends that would be served by taking theLSAT, the command for you to do so would be withdrawn. Categoricalimperatives, such as “If someone is drowning nearby, offerassistance,” differ in two ways. First, the command to offerassistance applies to you regardless of whether offering assistanceserves any of your (desired or intended) ends. Second, you have areason to offer assistance regardless of whether doing so serves anyof your (desired or intended) ends. (Philippa Foot has argued thatthis second feature doesn’t follow from the first: a rule ofetiquette, or a club rule, may apply to you regardless of whethercompliance with it serves any of your ends, but you may nonethelesshave no reason to comply with it. See Foot 1972.)

By “one wants” in the passage above, Kant means “onewills”: something closer to “one intends” than“one desires.” (After all, it wouldn’t be plausibleto suggest that one is instrumentally irrational in not intending thenecessary means to what one merelywants. I want to own aBMW, but I don’t intend to work the kind of job that would allowme to buy one. But there’s nothing incoherent about mycontinuing to desire to own one.) According to Kant:

How an imperative of skill is possible requires no special discussion.Whoever will the end, wills (so far as reason has decisive influenceon his actions) also the means that are indispensably necessary to hisactions and that lie in his power. This proposition, as far as willingis concerned, is analytic. (4:417)

But if Kant’s view is to provide a plausible account ofinstrumental coherence, it should instead refer to the meansbelieved to be indispensably necessary, since there isnothing incoherent about not willing necessary means of which you areunaware, and something incoherent about not willing the means youbelieve are necessary, even when they actually aren’t (Korsgaard1997, 236).

Kant’s parenthetical remark in the passage above is important.For Kant, when one fails to will the necessary means to one’sends, then reason has failed to have decisive influence. In otherwords, one is irrationally incoherent. Less guardedly, Kant says“the proposition (if I fully will the effect, then I will theaction required of it) is analytic” (4:417). But it is puzzlinghow this proposition could be analytic. If it were analytic, then itwould beconceptually impossible to be instrumentallyincoherent since whenever one doesn’t will the means, then onedoesn’t will the end. (For defense of this conceptualimpossibility, see Finlay 2009 and Lee 2022.) And if it’sconceptually impossible to be instrumentally incoherent, we’d beunable to make sense of why Kant formulates his account ofinstrumental incoherence in terms ofimperatives, which, onKant’s view, are commands addressed to a will that is capable ofboth following and not following the commands (Kant 1785, 4:414; Hill1973, 430; Korsgaard 1997, 236). So, perhaps we should readKant’s claim as that if one wills the end,rationalityrequires that one will the means. (For an alternative reading,see Lee 2018.)

Even so, there is a debate over the interpretation of Kant’sclaim, particularly about the logical scope of “requires.”According to the wide-scope interpretation of Kant’s claim, whatreason requires is that oneeither will the meansornot will the end. In contrast, according the narrow-scopeinterpretation of Kant’s claim, if one wills the end, then whatreason requires isthat one will the means. The wide-scopeinterpretation posits a disjunctive requirement (that one either notwill the end or will the means) that, like Kant’s categoricalimperative,applies to every rational agent, regardless ofthe particular ends she wills. But it’s hard to square theexistence of such a universally applicable requirement with some otherclaims Kant makes about how hypothetical imperatives are to bedistinguished from categorical ones, and how hypothetical imperativescould be derived from an analysis of the concept of willing an end.(See Schroeder 2005b. For a defense of the wide-scope interpretationagainst Schroeder’s objections, see Rippon 2014. For a relateddiscussion of these issues, see Schwartz 2008.)

But the narrow-scope interpretation appears to leave Kant’s viewvulnerable to “detachment” objections. Suppose you willsome immoral end but, irrationally, don’t will the immoral meansyou believe necessary to carry out this immoral end. On thenarrow-scope interpretation, reason would require that you will theimmoral means. But that seems implausible. At the very least, it wouldlead to something Kant wouldn’t accept: reason issuingcontradictory commands. Your willing the immoral means would berequired by a hypothetical imperative but forbidden by the categoricalimperative (Hill 1973, 436).

Kant could avoid the detachment problem if he holds, as someinterpreters claim, that it is impossible to will (Wille)immoral ends (Schroeder 2005b, §3). If this interpretation isright, then reason would never require one to will immoral means. Butit would also leave Kant without an account of instrumentalincoherence, at least as we have been understanding it. After all,agents can display instrumental incoherence in the pursuit ofboth moral and immoral ends. To use Kant’s ownexamples, the doctor who doesn’t intend the means believednecessary to making his patient healthy is instrumentally incoherentin the same way as is the poisoner who doesn’t intend the meansbelieved necessary to kill his victim (4:415). So, this interpretationwould have Kant avoiding the detachment objection, but at the cost ofno longer presenting an account of instrumental incoherence that holdsregardless of whether the end is reasonable or good.

Let’s now turn to Kant’s explanation of the normativeauthority of hypothetical imperatives. While Kant presents acomplicated and intricate argument to explain the normative authorityof the categorical imperative, he claims that hypothetical imperativesrequire “no special discussion,” and spends comparativelylittle time discussing them (4:417). Kant’s approachhere—which might seem to play into the hands of instrumentalists(see§4.1 above) who, while skeptical of the rational authority of thecategorical imperative, follow Kant in thinking that the authority ofhypothetical imperatives is obvious and unproblematic—has beenchallenged. Although the “ought” of a hypotheticalimperative is contingent on an agent’s ends, that“ought” still needs to be explained. Additionally,it’s not clear that a hypothetical “ought” is anymore compatible with a naturalistic worldview than a categorical one.Specifically, some of the same metaethical challenges that have beenraised about the status of the categorical “ought” couldbe raised about normative claims in general, including thehypothetical “ought” (Hampton 1998, Ch. 4; Shafer-Landau2006, §6).

So, what is Kant’s explanation of the normative authority ofhypothetical imperatives? According to Korsgaard (1997), Kant’sviews changed over the course of his philosophical development.Kant’s early “dogmatic rationalist” view (evident intheGroundwork) held that we should comply with thesehypothetical imperatives simply because they express what a perfectlyrational agent would do. But the mature Kantian view—whichKorsgaard endorses—explains the normative authority ofhypothetical imperatives in terms of the commitments that areconstitutive of willing. According to this view, what’s involvedinwilling some end (as opposed to merely desiring it) isbeing committed to taking the means to that end. It’s not thatcomplying with this commitment is constitutive of intention(this would render instrumental irrationality conceptually impossible)but thathaving this commitment is constitutive of intention.So, when one is instrumentally irrational, one fails to live up toone’s own commitments. Korsgaard writes:

…willing an end just is committing yourself to realizing theend. Willing an end, in other words, is an essentially first-personaland normative act. To will an end is to give oneself a law, hence, togovern oneself. … What is constitutive of willing the end is… the inward, volitional act of prescribing the end along withthe means it requires to yourself. (Korsgaard 1997, 245)

But is Korsgaard right that willing an end necessarily involvesprescribing that end to yourself? Korsgaard argues that towill some end, you must at least take that end to be good in somesense, and you must also think that there is a reason for pursuingthat end (246, 250–1). But it seems possible for one to havesome akratic end while failing to see anything evenpro tantogood about it, or any reason to pursue it (Wallace 2006, §1). Andit’s clear that one could display instrumental rationality orirrationality relative to these akratic ends. So, Korsgaard’sview seems unable to explain the authority of instrumental rationalitywith respect to such ends—ends that an agenthas butdoesn’tprescribe to herself.

Korsgaard offers a second Kantian explanation of the authority ofinstrumental rationality. She argues that incessant instrumentalirrationality threatens one’s status as an agent:

So the reason that I must conform to the instrumental principle isthat if I don’t conform to it, if Ialways allow myselfto be derailed by timidity, idleness, or depression, then I neverreallywill an end. Thedesire to pursue the end andthe desires that draw me away from it each hold sway in their turn,butmy will is never active. The distinction between my willand the operation of the desires and impulses in me does not exist andthat means that I, considered as an agent, do not exist. (1997, 247)

But this argument doesn’t provide a reason to conform to theinstrumental principle in every case in which the instrumentalprinciple is applicable. Rather, it provides a reason to do so onlywhen accompanied by enough other instances of non-conformity tothreaten one’s status as an agent (Kolodny 2005, 544; Shaver2006, 342). So, like the previous Kantian argument, and like theHume’s skepticism discussed in§4.3, this argument also fails to explain why instrumental rationalityapplies in every case in which it does.

Bibliography

• Andreou, C., 2005, “The Voices of Reason,”American Philosophical Quarterly, 42(1): 33–45.
• –––, 2006, “Might Intentions Be the OnlySource of Practical Imperatives?,”Ethical Theory and MoralPractice, 9(3): 311–325.
• Anscombe, E., 1957,Intention, Oxford: BasilBlackwell.
• Archer, A., 2018, “Trying Cognitivism: A Defense of theStrong Belief Thesis,”Theoria, 84(2):140–156.
• Beardman, S., 2007, “The Special Status of InstrumentalReasons,”Philosophical Studies, 124(2):255–287.
• Bedke, M., 2009, “The Iffiest Oughts: A Guise of ReasonsAccount of End-Given Conditionals,”Ethics, 119(4):672–698.
• –––, 2017, “Ends to Means: AProbability-Raising Account of Means and the Weight of Reasons to TakeThem,”Journal of Ethics and Social Philosophy, 12(1):6–28.
• Bennett, J., 1983, “Positive and Negative Relevance,”American Philosophical Quarterly, 20(2): 185–194.
• Binkley, R., 1965, “A Theory of Practical Reason,”The Philosophical Review, 74(4): 423–448.
• Bratman, M., 1987,Intentions, Plans, and PracticalReason, Cambridge: Harvard University Press.
• –––, 2009a, “Intention, Belief, andInstrumental Rationality,” in D. Sobel and S. Wall (eds.),Reasons for Action, Cambridge: Cambridge University Press,pp. 13–36.
• –––, 2009b, “Intention, Belief, Practical,Theoretical,” in S. Robertson (ed.),Spheres of Reason: NewEssays in the Philosophy of Normativity, Oxford: OxfordUniversity Press, pp. 29–62.
• –––, 2009c, “Intention, PracticalRationality, and Self-Governance,”Ethics, 119(3):411–443.
• –––, 2009d, “Intention Rationality,”Philosophical Explorations, 12(3): 227–242.
• –––, 2018,Planning, Time andSelf-Governance: Essays in Practical Rationality, Oxford: OxfordUniversity Press.
• Broome, J., 1997, “Reason and Motivation,”Proceedings of the Aristotelian Society (SupplementaryVolume), 71: 131–146.
• –––, 1999, “Normative Requirements,”Ratio, 12(3): 398–419.
• –––, 2001, “Are intentions reasons? Andhow should we cope with incommensurable values?,” in C. Morrisand A. Ripstein (eds.),Practical Rationality and Preference:Essays for David Gauthier, Cambridge: Cambridge University Press,pp. 98–120.
• –––, 2003, “Practical Reasoning,” inJ. Bermùdez and A. Millar (eds.),Reason and Nature: Essaysin the Theory of Rationality, Oxford: Oxford University Press,pp. 85–112.
• –––, 2005a, “Does Rationality Give UsReasons?,”Philosophical Issues, 15:321–337.
• –––, 2005b, “Have We Reason to Do asRationality Requires?—A Comment on Raz,”Journal ofEthics and Social Philosophy, Symposium 1: 1–8.
• –––, 2008a, “Is RationalityNormative?,”Disputatio, 11: 153–171.
• –––, 2008b, “Reply to Southwood, Kearnsand Star, and Cullity,”Ethics, 119: 96–108.
• –––, 2009, “The Unity ofReasoning?,” in S. Robertson (ed.),Spheres of Reason: NewEssays in the Philosophy of Normativity, Oxford: OxfordUniversity Press, pp. 62–92.
• –––, 2013,Rationality ThroughReasoning, Oxford: Wiley Blackwell.
• Brunero, J., 2005a, “Instrumental Rationality andCarroll’s Tortoise,”Ethical Theory and MoralPractice, 8: 557–569.
• –––, 2005b, “Two Approaches toInstrumental Rationality and Belief Consistency,”Journal ofEthics and Social Philosophy, 1(1): 1–20.
• –––, 2007, “Are Intentions Reasons?”Pacific Philosophical Quarterly, 88(4): 424–444.
• –––, 2009, “Against Cognitivism aboutPractical Rationality,”Philosophical Studies, 146(3):311–325.
• –––, 2010, “Self-governance, Means-endCoherence, and Unalterable Ends,”Ethics, 120(3):579–591.
• –––, 2012, “Instrumental Rationality,Symmetry and Scope,”Philosophical Studies, 157(1):125–140.
• –––, 2014, “Cognitivism about PracticalRationality,”Oxford Studies in Metaethics (Volume 9),Russ Shafer-Landau (ed.), Oxford: Oxford University Press, pp.18–44.
• –––, 2020,Instrumental Rationality: TheNormativity of Means-Ends Coherence, Oxford: Oxford UniversityPress.
• Carroll, L., 1895, “What the Tortoise Said toAchilles,”Mind, 4(14): 278–80.
• Cheng-Guajardo, L., 2014, “The Normative Requirement ofMeans-end Rationality and Modest Bootstrapping,”EthicalTheory and Moral Practice, 17(3): 487–503; first publishedonline 2013. doi:10.1007/s10677-013-9455-y
• Cullity, G., 2008, “Decisions, Reasons, andRationality,”Ethics, 119(1): 57–95.
• Darwall, S., 1983,Impartial Reason, Ithaca: CornellUniversity Press.
• Davidson, D., 2004, “Incoherence and Irrationality” inProblems of Rationality, Oxford: Oxford University Press, pp.189–98.
• Dreier, J., 2001, “Humean Doubts about CategoricalImperatives,” in E. Millgram (ed.)Varieties of PracticalReasoning, Cambridge: MIT Press, pp. 27–49.
• Ellis, S., 2008, “The Varieties of InstrumentalRationality,”Southern Journal of Philosophy, 46(2):199–220.
• Fink, J., forthcoming, “The Essence of StructuralIrrationality: The Impossibility of Attitudinal Success,”Journal of Ethics & Social Philosophy.
• Finlay, S., 2009, “Against All Reason? Skepticism About theInstrumental Norm,” in Charles Pigden (ed.),Hume,Motivation and Virtue, New York: Palgrave MacMillan, pp.155–178.
• Fogal, D., 2018, “On the Scope, Jurisdiction, andApplication of Rationality and the Law,”Problema, 12:21–57.
• –––, 2020, “Rational Requirements and thePrimacy of Pressure,”Mind, 129(516):1033–1070.
• Fogal, D. & A. Worsnip, 2021, “Which Reasons? WhichRationality?”Ergo, 8(11): 306–343
• Foot, P., 1972, “Morality as a System of HypotheticalImperatives,”The Philosophical Review, 81(3):305–316.
• Fullhart, S., and C. Martinez, forthcoming, “Coherence asJoint Satisfiability,”Australasian Journal ofPhilosophy.
• Gertken, J. and B. Kiesewetter, 2021, “How Do ReasonsTransmit to Non-Necessary Means?”Australasian Journal ofPhilosophy, 99(2): 271–285.
• Greenspan, P., 1975, “Conditional Oughts and HypotheticalImperatives,”The Journal of Philosophy, 72(10):259–276.
• Grice, P., 1971, “Intention and uncertainty,”Proceedings of the British Academy, 57: 263–279.
• Hampshire, S. and H. L. A. Hart, 1958, “Decision, Intention,and Uncertainty,”Mind, 67: 1–12.
• Hampton, J., 1995, “Does Hume have an InstrumentalConception of Practical Reason?,”Hume Studies, 21(1):57–74.
• –––, 1998,The Authority of Reason,Cambridge: Cambridge University Press.
• Harcourt, E., 2004, “Instrumental Desires, InstrumentalRationality,” inAristotelian Society SupplementaryVolume, 78(1): 111–130.
• Harman, G., 1976, “Practical Reasoning,”Review ofMetaphysics, 29(3): 431–463.
• –––, 1986,Change in View: Principles ofReasoning, Cambridge: MIT Press.
• Hill, T., 1973, “The Hypothetical Imperative,”ThePhilosophical Review, 82(4): 429–450.
• –––, 1992,Dignity and Practical Reason inKant’s Moral Theory, New York: Cornell UniversityPress.
• Holton, R., 2008, “Partial Belief, Partial Intention,”Mind, 117(1): 27–58.
• Hubin, D., 1999, “What’s Special aboutHumeanism,”Noûs, 33(1): 30–45.
• –––, 2001, “The Groundless Normativity ofInstrumental Rationality,”The Journal of Philosophy,98(9): 445–468.
• Hume, D., 1739–40,Treatise of Human Nature. L.A.Selby-Bigge (ed.), 2nd. ed., revised by P.H. Nidditch, Oxford:Clarendon, 1975.
• Jackson, F. and R. Pargetter, 1986, “Ought, Options, andActualism,”Philosophical Review, 95(2):233–255.
• Jollimore, T., 2005, “Why Is Instrumental RationalityRational?,”Canadian Journal of Philosophy, 35(2):289–308.
• Kant, I., 1785,Grounding for the Metaphysics of Morals,transl. by James Ellington, Indianapolis: Hackett Publishing Company,1993. [Citations to this text are given using the standard volume andpage numbers from the Akademie Edition.]
• Kamm, F., 2007,Intricate Ethics: Rights, Responsibilities,and Permissible Harm, New York: Oxford University Press.
• Kamm, F., 2000, “The Doctrine of Triple Effect and Why aRational Agent Need Not Intend the Means to His End,”Proceedings of the Aristotelian Society (SupplementaryVolume), 74: 21–39.
• Kelly, T., 2003, “Epistemic Rationality as InstrumentalRationality: A Critique,”Philosophy and PhenomenologicalResearch, 66(3): 612–640.
• Kenny, A.J., 1966, “Practical Inference,”Analysis, 26(3): 65–75.
• Kiesewetter, B., 2015, “Instrumental Normativity: In Defenseof the Transmission Principle,”Ethics, 125(4):921–946.
• –––, 2017,The Normativity ofRationality, Oxford: Oxford University Press.
• –––, 2018, “Contrary-to-Duty Scenarios,Deontic Dilemmas, and Transmision Principles,”Ethics,129: 98–115.
• Kolnai, A., 1962, “Deliberation is of Ends,”Proceedings of the Aristotelian Society, 62:195–218.
• Kolodny, N., 2005, “Why be Rational?,”Mind,114(455): 509–563.
• –––, 2007a, “State or ProcessRequirements?,”Mind, 116(462): 371–385.
• –––, 2007b, “How does CoherenceMatter?,”Proceedings of the Aristotelian Society,107(1–Part 3): 229–263.
• –––, 2008a, “The Myth of PracticalConsistency,”European Journal of Philosophy, 16(3):366–402.
• –––, 2008b, “Why be Disposed to BeCoherent?”Ethics, 188(3): 437–463.
• –––, 2018, “Instrumental Reasons,”in D. Star (ed.),The Oxford Handbook of Reasons andNormativity, Oxford: Oxford University Press.
• Korsgaard, C., 1986, “Skepticism about PracticalReason,”Journal of Philosophy, 83(1): 5–25.
• –––, 1997, “The Normativity ofInstrumental Reason,” in G. Cullity and B. Gaut (eds.),Ethics and Practical Reason, Oxford: Oxford University Press,pp. 215–254.
• –––, 2009,Self-Constitution: Agency,Identity, and Integrity, Oxford: Oxford University Press.
• Lavin, D., 2004, “Practical Reason and the Possibility ofError,”Ethics, 114(3): 424–57.
• Lee, W., 2018, “Willing the End Means Willing the Means: AnOverlooked Reading of Kant,”Ergo, 5(16):419–444.
• –––, 2021, “The Independence of(In)coherence,”Synthese 199: 6563–6584.
• –––, 2022, “The Real Myth ofCoherence,”Erkenntnis 87: 1211–1230.
• Liberman, A. and M. Schroeder, 2016, “Commitment: Worth theWeight,” in E. Lord and B. Maguire (eds.),WeighingReasons, Oxford: Oxford University Press.
• Lord, E., 2011, “Violating Requirements, Exiting fromRequirements, and the Scope of Rationality,”PhilosophicalQuarterly, 61(243): 392–399.
• –––, 2014a, “The Coherent and theRational,”Analytic Philosophy, 55(2):151–175.
• –––, 2014b, “The Real Symmetry Problem(s)for Wide-Scope Accounts of Rationality,”PhilosophicalStudies, 170(3): 443–464.
• –––, 2017, “What You’re RationallyRequired to Do and What you Ought to Do (Are The Same Thing!),”Mind, 126(504): 1109–1154.
• –––, 2018, “The Explanatory Problem forCognitivism about Practical Reason,” in C. McHugh, J. Way, andD. Whiting (eds.),Normativity: Practical and Epistemic,Oxford: Oxford University Press.
• –––, 2018,The Importance of BeingRational, Oxford, Oxford University Press.
• Millgram, E., 1995, “Was Hume a Humean?,”HumeStudies, 21(1): 75–94.
• –––, 2001, “Practical Reasoning: TheCurrent State of Play,” in E. Millgram (ed.)Varieties ofPractical Reasoning, Cambridge: MIT Press, pp. 1–26.
• Morton, J., 2011, “Towards an Ecological Theory of the Normsof Practical Deliberation,”European Journal ofPhilosophy, 19(4): 561–584.
• Nozick, R., 1993,The Nature of Rationality, Princeton:Princeton University Press.
• Nuñez, C., 2020, “Requirements of Intention in Lightof Belief,”Philosophical Studies, 177(9):2471–2492.
• Parfit, D., 2006, “Normativity,” in Russ Shafer-Landau(ed.),Oxford Studies in Metaethics (Volume 1), Oxford:Oxford University Press, pp. 325–380.
• Price, A., 2008a,Contextuality in Practical Reason,Oxford: Oxford University Press.
• –––, 2008b, “Against Requirements ofRationality,”Proceedings of the Aristotelian Society,108: 157–76.
• Quinn,W., 1992, “Rationality and the Human Good,”Social Philosophy and Policy, 9(2): 81–95. Reprinted inW. Quinn, 1993,Morality and Action, Cambridge: CambridgeUniversity Press.
• Railton, P., 1997, “On the Hypothetical and Non-Hypotheticalin Reasoning about Belief and Action,” in G. Cullity and B. Gaut(eds.),Ethics and Practical Reason, Oxford: OxfordUniversity Press, pp. 52–79.
• Raz, J., 2003, “Numbers, With and WithoutContractualism,”Ratio, 16(4): 346–367.
• –––, 2005a, “Instrumental Rationality: AReprise,”Journal of Ethics and Social Philosophy,Symposium 1: 1–19.
• –––, 2005b, “The Myth of InstrumentalRationality,”Journal of Ethics and Social Philosophy,1(1): 1–28.
• Reisner, A., 2009, “Unifying the Requirements ofRationality,”Philosophical Explorations 12(3):243–260.
• Richardson, H., 1994,Practical Reasoning about FinalEnds, Cambridge: Cambridge University Press.
• Rippon, S., 2011, “In Defense of the Wide-Scope InstrumentalPrinciple,”Journal of Ethics and Social Philosophy,5(2): 1–21.
• –––, 2014, “Were Kant’s HypotheticalImperatives Wide-Scope Oughts?”Australasian Journal ofPhilosophy, 92(4): 783–788.
• Ross, J., 2009, “How to be a Cognitivist about PracticalReason,” in Russ Shafer-Landau (ed.)Oxford Studies inMetaethics, Vol. 4, Oxford: Oxford University Press, pp.243–282.
• Scanlon, T., 2007, “Structural Irrationality,” in G.Brennan, G. Robert, J. Frank, and S. Michael (eds.),Common Minds:Themes from the Philosophy of Philip Pettit, Oxford: ClarendonPress.
• Schroeder, M., 2004, “The Scope of InstrumentalReason,”Philosophical Perspectives, 18:337–364.
• –––, 2005a, “InstrumentalMythology,”Journal of Ethics and Social Philosophy,Symposium 1: 1–12.
• –––, 2005b, “The HypotheticalImperative?,”Australasian Journal of Philosophy,83(3): 357–372.
• –––, 2007,Slaves of the Passions,Oxford: Oxford University Press.
• –––, 2009, “Means-end Coherence,Stringency, and Subjective Reasons,”PhilosophicalStudies, 143(2): 223–248.
• –––, 2014a, “Hypothetical Imperatives,Scope, and Jurisdiction” inExplaining the Reasons We Share:Explanation and Expression in Ethics (Volume 1), Oxford: OxfordUniversity Press, 216–226.
• –––, 2014b, “Scope for RationalAutonomy” inExplaining the Reasons We Share: Explanationand Expression in Ethics (Volume 1), Oxford: Oxford UniversityPress, 227–239.
• Schwartz, J., 2008, “Do Hypothetical Imperatives RequireCategorical Imperatives?,”European Journal ofPhilosophy, 18: 84–107.
• Schwitzgebel, E., 2010, “Introspection,” in Edward N.Zalta (ed.),The Stanford Encyclopedia of Philosophy (Fall2010 Edition), [Schwitzgebel 2010 available online].
• Searle, J., 2001,Rationality in Action, Cambridge, MA:MIT Press.
• Setiya, K., 2004, “Hume on Practical Reason,”Philosophical Perspectives, 18: 365–389.
• –––, 2005, “Is Efficiency a Vice?,”American Philosophical Quarterly, 42(4): 333–339.
• –––, 2007, “Cognitivism about InstrumentalReason,”Ethics, 117(4): 649–673.
• Shafer-Landau, R., 2006, “Ethics as Philosophy,” in T.Horgan and M. Timmons (eds.),Metaethics after Moore, Oxford:Oxford University Press, pp. 209–232.
• Shaver, R., 2006, “Korsgaard on HypotheticalImperatives,”Philosophical Studies, 129:335–347.
• Shpall, S., 2013, “Wide and Narrow Scope,”Philosophical Studies, 163: 717–736.
• Smith, M., 2004, “Instrumental Desires, InstrumentalRationality,”Proceedings of the Aristotelian Society(Supplementary Volume), 78(1): 93–109.
• Sobel, D., 2005, “Instrumental Rationality—Not DeadYet,”Journal of Ethics and Social Philosophy,Symposium 1: 1–12.
• Southwood, N., 2008, “Vindicating the Normativity ofRationality,”Ethics, 119(1): 9–30.
• Tenenbaum, S., 2021,Rational Powers in Action: InstrumentalRationality and Extended Agency, Oxford: Oxford UniversityPress.
• Titelbaum, M., 2015, “How to Derive a Narrow-scopeRequirement from Wide-scope Requirements,”PhilosophicalStudies, 172(2): 535–542.
• Thompson, M., 2008,Life and Action: Elementary Structures ofPractice and Practical Thought, Cambridge: Harvard UniversityPress.
• Valaris, M., 2014, “Instrumental Rationality,”European Journal of Philosophy, 22(3): 443–462.
• Velleman, J.D., 1989,Practical Reflection, Princeton:Princeton University Press.
• –––, 2007, “What Good is a Will?,”in A. Leist and H. Baumann (eds.),Action in Context, Berlin:de Gruyter, pp. 193–215.
• Vogler, C., 2002,Reasonably Vicious, Cambridge: HarvardUniversity Press.
• Wallace, R. J., 2005, “Comment on Raz,”Journal ofEthics and Social Philosophy, Symposium 1: 1–4.
• –––, 2006, “Normativity, Commitment, andInstrumental Reason,” reprinted with a postscript inNormativity and the Will: Selected Essays on Moral Psychology andPractical Reason, Oxford: Clarendon Press.
• Way, J., 2010, “Defending the Wide-scope Approach toInstrumental Reason,”Philosophical Studies, 147(2):213–233.
• –––, 2011, “The Symmetry of RationalRequirements,”Philosophical Studies, 155(2):227–239.
• –––, 2012a, “Explaining the InstrumentalPrinciple,”Australasian Journal of Philosophy, 90(3):487–506.
• –––, 2012b, “InstrumentalRationality,”Routledge Encyclopedia of Philosophy.doi: 10.4324/9780415249126-L148-1.
• –––, 2013, “Intentions, Akrasia, and MerePermissibility,”Organon F, 20(4): 588–611.
• –––, 2018, “Reasons andRationality,” in D. Star (ed.)The Oxford Handbook ofReasons and Normativity, Oxford: Oxford University Press.
• Wedgwood, R., 2003, “Choosing Rationally and ChoosingCorrectly,” in S. Stroud and C. Tappolet (eds.)Weakness ofWill and Practical Irrationality, Oxford: Oxford UniversityPress, pp. 201–229.
• –––, 2005, “Railton on Normativity,”Philosophical Studies, 126(3): 463–479.
• –––, 2007,The Nature of Normativity,Oxford: Clarendon Press.
• –––, 2011, “InstrumentalRationality,”Oxford Studies in Metaethics, 6:280–309.
• White, S., 2017, “Transmission Failures,”Ethics, 127(3): 719–732.
• –––, 2018, “Intention and Prediction inMeans-Ends Reasoning,”American PhilosophicalQuarterly, 55(3): 251–266.
• Worsnip, A., 2015, “Narrow-Scoping for Wide-Scopers”Synthese, 192(8): 2617–2646.
• –––, 2018, “What is (in)coherence?”Oxford Studies in Metaethics, 13: 184–206.
• –––, 2021,Fitting Things Together:Coherence and the Demands of Structural Rationality, Oxford:Oxford University Press.

Academic Tools

	How to cite this entry.
	Preview the PDF version of this entry at theFriends of the SEP Society.
	Look up topics and thinkers related to this entry at the Internet Philosophy Ontology Project (InPhO).
	Enhanced bibliography for this entryatPhilPapers, with links to its database.

Other Internet Resources

[Please contact the authors with suggestions.]

Related Entries

action |Aristotle, General Topics: ethics |decision theory: causal |desire |Hume, David: moral philosophy |intention |Kant, Immanuel: account of reason |metaethics |moral motivation |practical reason |practical reason: and the structure of actions |reasons for action: internal vs. external |reasons for action: justification, motivation, explanation |weakness of will

Acknowledgments

The authors thank Justin Vlasits for help in preparing this entry.