How we know what we intend

Abstract

How do we know what our intentions are? It is argued that work on self-knowledge has tended to neglect the attitude of intention, and that an epistemological account is needed that is attuned to the specific features of that state. Richard Moran’s “Authorship” view, on which we can acquire self-knowledge by making up our minds, offers a promising insight for such an account: we do not normally discover what we intend through introspection. However, his formulation of the Authorship view, developed primarily with the attitude of belief in mind, is found wanting when applied to intention. An alternative account is proposed for knowledge of one’s own intentions that gives a central role to the mental act of deciding what to do. It is argued that we can come to know what we intend by making a decision about what to do and self-ascribing the content of that decision as our intended action.

I would like to ask a very specific question about self-knowledge: how do we know what we intend?

It is not usual in doing epistemology of mind to focus on intention in isolation from other mental states. The question of how we know of our intentions is often treated as a mere instance of the general puzzle of how we know our own minds, where it is supposed that the answer to the general puzzle will account in more or less the same way for all mental states. This blanket approach is paradigmatically taken by Neo-Cartesian views, on which we know our minds through a kind of direct, unmediated inner observation (e.g. Chisholm 1981; Chalmers 2003), as well as many introspection-based views that posit a kind of “inner sense” or “self-scanning” capacity (e.g. Armstrong 1981; Lycan 1996). The Rationality Model (e.g. Shoemaker 1996), according to which the capacity for self-knowledge is a condition of possibility for the rationality we in fact exhibit, likewise purports to apply in essentially the same fashion to all mental state types.

Some approaches do make a broad distinction between the self-knowledge of phenomenal states and that of propositional attitudes.¹ These accounts offer a different story for intention than for pains and tickles, but lump it together with beliefs, desires, wishes, and hopes. And then there are views that have been developed primarily with the attitude of belief in mind, but that are claimed to extend to intention in a parallel way.² These views single out belief and intention as having a different epistemic status from other mental state types, but suppose that these two attitudes are sufficiently similar in the relevant respects to merit the same epistemological story.

This is certainly not an exhaustive taxonomy of approaches to self-knowledge, but it suffices to motivate a general point. The leading theories were not developed with intention specifically in mind, and I do not think they well account for it. The tendency of theories of self-knowledge either to treat all mental state types indiscriminately or to focus on belief and suppose that a similar story must apply to intention has led to the epistemology of intention getting short shrift.

I can best explain why I think this is a problem by articulating what it is I take the puzzle about the self-knowledge of intention to be. On one hand, it is commonly held that first-personal self-ascriptions of mental states carry a certain default authority.³ One is generally entitled to report on the state of one’s own mind without being challenged for one’s grounds for this report or contradicted by third-personal evidence. This authority is not infallible; there are circumstances in which we are inclined to override a first-personal report on the basis of third-personal observational evidence. But ordinarily it is granted that a subject is in a better position to know his own mind than anyone else, and that this knowledge is not based on the interpretation of observed data in the way third-personal ascriptions of mental states typically are.

On the other hand, the conditions under which the ascription of a mental state (whether first- or third-personal) comes out true have appeared to philosophers of mind to be extremely difficult to articulate. It is currently more or less standard to think of mental states in broadly functionalist terms, individuated by the set of causal relations in which they stand with respect to sensory stimulation, behavioral dispositions, transitions to other mental states, and so forth. But it is no simple matter to make these functional roles explicit or to identify the boundary conditions between different states. A great deal of philosophical ink has been spilled over the question of what kind of state an intention is, for instance: a kind of belief, a kind of desire, some combination of the two, or some kind of sui generis state? Much of this debate turns on the conditions under which one counts as being in the state of intending. To what extent must the subject believe he will succeed in order to count as intending a given outcome, rather than merely desiring it or hoping for it?⁴ If that subject knowingly aims at two incompatible outcomes, is he simply irrational, or does he actually fail to count as intending those outcomes?⁵ And to further complicate matters, some think that mental states like intention can only be ascribed holistically as elements of the best global interpretation of the subject, and that such questions cannot be made sense of when states are considered in isolation.⁶

The challenge of investigating self-knowledge is to explain how a subject can come to know in a specially first-personal authoritative way of a given mental state that he is in that state, where this is at least a matter of making a true self-ascription of that mental state. But given the foregoing understanding of what mental states are—diachronic functional states with unique dispositional roles and vague, controversial boundary conditions—this feat begins to look quite extraordinary, and the blanket approach to self-knowledge appears inadequate. A true, justified self-ascription of a given mental state must involve some kind of sensitivity to the particular conditions on being in a state of that kind, and this should lead us to think that a successful epistemological account must make reference to those state-specific features and our sensitivity to them.

My aim in this paper is to make headway on such an account of knowledge of one’s own intentions. I do not suppose the suggestions I offer are complete or exhaustive of the ways we may come to know what we intend; there may well be a number of ways. But I will focus on a route to self-knowledge that I consider to be a central and particularly valuable case: the knowledge acquired by making up one’s mind about what to do. The key idea is that a subject can come to know what he intends by making a decision about what to do, thereby both forming an intention and acquiring knowledge of it. Although this kind of “Authorship” view of self-knowledge has been defended primarily with respect to the case of belief, I will argue that it is far better as an account of the self-knowledge of intention than of belief. We are in a much more legitimate sense the authors of our intentions than we are of our beliefs.

My strategy will be to motivate the positive proposal by first critically discussing the Authorship View of self-knowledge as proposed by Richard Moran, which I think provides a key insight into the nature of our relationship to our own intentions. However, Moran’s emphasis on the so-called “transparency” that purportedly explains authoritative knowledge of one’s beliefs does not apply straightforwardly to intention in the same way. Crudely put, I will argue that intentions just aren’t transparent enough. I will then propose an alternative way of understanding the idea of acquiring knowledge of one’s own intentions through authorship of them, via making a decision about what to do.

1 Authorship and transparency

The primary motivation for the Authorship account is the need to explain two features of first-personal self-ascriptions of certain mental states, among these belief and intention: their authority and their immediacy. We seem to occupy a privileged position with respect to knowing our own beliefs and intentions, and our knowledge of them usually does not appear to depend on evidence. Indeed, a thinker’s self-ascriptions of belief or intention tend to carry with them a special authority lacked by third-personal ascriptions that are based on evidence.

The Authorship account, most prominently defended by Moran, claims that this special authority could not simply be a matter of having more or even exclusive perceptual access to one’s own attitudes. Introspection alone cannot account for it, for a model on which we are limited to discovering our own attitudes cannot say what is wrong with finding out that one has an attitude one sees no reason to have. A subject who learns through perceptual means that he has a belief or an intention for which he sees insufficient justification will not be in a position to report on this attitude in a first-personally authoritative way, for it will not seem to him to be an attitude he himself ought to have. There is a sense in which he is “alienated” from that attitude, as if it belonged to some creature that only contingently happened to be himself. (Moran 2001, p. 33). This insight inspires the Authorship proposal that we are able to speak authoritatively about our own beliefs and intentions not because we learn about them in a uniquely reliable way, but because we are in a unique position to make up our minds about what to believe and intend. But this is not to say that we can simply inscribe whatever beliefs and intentions we wish onto the blank page of the mind. According to Moran’s version of the view, authorship is a matter of making a self-ascription of an attitude that obeys what he calls the Transparency Condition.⁷

The sense of transparency Moran intends is an idea sometimes attributed to Wittgenstein and expanded upon by Evans (1982): that we often do know our own minds not by looking to some inner mental theater but by looking outward to the world. Evans points out that the question of whether I believe that P, asked from the first-person perspective, gives way to the question of whether P is true: I can answer the former by getting myself in position to reach an answer to the latter, by looking to the world to find out whether P. We can call this the ability to treat one’s own beliefs as “transparent” to the objects of those beliefs, and to answer the question of what one believes by answering the question of what is true (by one’s own lights).⁸ On Moran’s view, to obey the Transparency Condition in self-ascribing a belief is to report the belief on the basis of coming to a deliberative conclusion about what one has reason to believe. When the question of what I believe explicitly arises, I obey the Transparency Condition when I consider not if I believe P, thereby granting authority to whatever antecedent fact of the matter I should discover, but rather whether to believe P. And the deliberative question of whether to believe P is answered by reflecting on reasons for belief, to ascertain whether P is true.⁹

Though Moran focuses primarily on belief, he suggests that the Authorship view applies in an analogous way to intention. Just as in the case of belief, he claims that we usually need not conduct a theoretical investigation of what we intend; we can treat this question transparently as well: “My knowing what I will do next is not based on evidence or other reasons to believe something, so much as it is based on what I see as reasons to do something.”¹⁰ The suggestion seems to be that one’s own intentions can be treated as transparent to reasons for action. When the question arises of what one intends to do, the parallel view would be that one can seek to answer this question by reflecting on what is to be done and avowing the conclusion as one’s intended action.

Unfortunately, Moran is not entirely clear about what the deliberative standard is by which the question of what to do should be settled. Nishi Shah offers a potential answer: he argues that the deliberative question of whether to intend to ϕ must as a conceptual matter be settled by answering the normative question of whether one ought to ϕ (Shah 2008). More generally, the claim is that in explicitly asking oneself the question of what to intend, one is asking a question that can only be answered by coming to a conclusion as to what one ought to do. There are indications that Moran also thinks of the question of what one intends as transparent to the normative question of what one ought to do. In discussing the special case of knowing one’s intention in action, for instance, he depicts intention as transparent to what is good to do:

The description under which an action is intentional gives the agent’s primary reason in so acting, and the agent knows this description in knowing his primary reason. This description is known by him because it is the description under which he conceives of it in his practical reasoning. It is the description under which the action is seen as choiceworthy by him, as aiming at some good to be achieved. The agent takes the question of what he is doing to be answered by his decision as to what is worth pursuing… (Moran 2001, p. 126).

The analogy with belief thus seems to be that from the first-person perspective, we can come to know what we intend by coming to a conclusion about what is desirable, choiceworthy, or worth pursuing. Whether treating one’s intentions as transparent to normative reasons for action is a conceptual requirement or merely a rational ideal, the idea is that insofar as one upholds it, one can self-ascribe an intention by answering for oneself the question of what is choiceworthy or worth pursuing.

According to Moran, attitudinal self-ascriptions that obey the Transparency Condition amount to an importantly authoritative kind of self-knowledge. Only transparent self-ascriptions capture the specially first-personal relationship one has to one’s own mind, where one regards what one believes and intends as ultimately up to oneself—determined by what one sees oneself as having reason to believe and intend. And insofar as a thinker takes responsibility for letting his assessment of his reasons settle what he believes and intends, coming to a deliberative conclusion will amount to his having the relevant attitude:

…it is because the deliberator declares the authority of reason over his thought and action that at the conclusion of his thinking there is no further thing he does to make that conclusion his actual belief or his intention…. In the case of practical deliberation, reaching a conclusion means making a decision, hence forming an intention; reaching a conclusion in theoretical deliberation means achieving conviction in the form of a belief about what is the case (p. 131).

The privilege enjoyed by first-personal attitude reports is explained by the fact that the thinker is in a unique position in being able to take this responsibility for his attitudes and authoritatively ascribe the deliberative conclusion as his believed or intended answer. The consequence is that one need not conduct an investigation of one’s mind in order to discover what one believes or intends; one can come to know at any point by reflecting on what is true, or what ought to be done, and ascribing the answer as one’s attitude.

To be clear, the Authorship account does not claim that we can only come to know what we believe and intend through this kind of deliberative self-ascription on the basis of reasons, nor that our beliefs and intentions are always or even usually formed through explicit deliberation. The thought is rather that when the question arises for a thinker as to what he believes or intends he can then take responsibility for conforming his attitudes to the reasons for them. Regardless of what they were before, it is open to him at any point to make up his mind about what they should be and thereby gain knowledge of what he believes or intends. Authorship is taken not to be the only route to self-knowledge, but an especially authoritative one and one deeply connected with our status as rational agents.¹¹

2 Intentions: not transparent enough

I think the Authorship View gets a great deal right about our knowledge of our intentions. We could not be limited to finding out what we intend by peering within and discovering what they are if we are to see them as our intentions, and the resulting actions as our actions. We often come to have intentions by making up our minds about what to do, and our capacity to know our minds in this respect must be linked to this ability. However, I do not think the account of transparency developed primarily with belief in mind applies so straightforwardly to intention as Moran suggests.

Let us ask why reflecting on reasons for belief might be a dependable guide to what oneself believes—which is, after all, quite a different subject matter. The answer seems to have to do with the concept of belief. Beliefs are commonly said to “aim at the truth,” in the sense that it is correct to have the belief that P if only if P is true (e.g. Wedgwood 2002). Correspondingly, there is a single rational standard for whether one ought to believe that P: whether one’s evidence sufficiently supports taking P to be true. And in the ideal, theoretical evidence is more or less commensurable; if sufficient evidence is available, correct doxastic deliberation will in principle afford a determinate, univocal answer for what to believe. I do not mean to deny that there are questions here as to the permissibility of disagreement about a body of evidence, or to assume that there is a unique rational doxastic attitude one can take given one’s total evidence.¹² But whether or not there is rational leeway here in responding to evidence, the point remains that there is a single standard according to which having a belief is correct—that it is true—and thus that a judgment that P is true leaves open no further question from the first-person point of view as to whether it is correct to believe that P.¹³

A thinker may only justifiably self-ascribe this answer as his belief insofar as his beliefs truly tend to mirror his deliberative conclusions about what is the case. If his beliefs tended to be impervious to the influence of his judgment, reflection on reasons for belief would bear very little on the question of what he actually believes. And indeed, this kind of imperviousness may occur; a judgment that the evidence sufficiently supports P may fail to impact the thinker’s existing doxastic structure in the way requisite for him to qualify as believing that P. But the concept of belief as an attitude that constitutively aims at the truth ensures this could not be the rule in a subject that counted as a Believer at all. A subject whose judgments about what is true consistently failed to guide his reasoning, action, and assertive dispositions would not be a subject to whom beliefs could be attributed. Whatever this creature is in the business of doing, it would not be the business of forming beliefs, for Believers are disposed to regulate their doxastic representations in accordance with their judgments about what is true. In fact, I will argue in Sect. 4 that the possibility of a breakdown between judgment and belief is serious enough to pose a threat to the proposed entitlement. However, it seems to me that the plausibility of the Authorship account with respect to belief is tied to the fact that the extent of such breakdowns is restricted by the kind of attitude belief is.

The problem for the Authorship account is that I do not think the parallel holds in the case of intention. I do not think the concept of intention can be understood as containing an internal standard according to which one can derive a univocal answer as to what it is correct to intend. Importantly, this claim does not depend on rejecting the thesis that all intending must be done under the “Guise of the Good,” or in light of some desirability characteristic the agent sees in the intended action (however misguided he might be). Even if the Guise of the Good thesis is true,¹⁴ it does not suffice to bind intention to practical judgment in the way needed to justify the self-ascription of the deliverance of practical judgment as one’s intended action. This is because there will generally be a great many incompatible things the agent sees as in some respect desirable or reasonable to do. Unlike truth, bonum est multiplex: A and B may both be actions worth doing, though one cannot do them both. And though there may be a normative requirement in favor of regulating one’s intentions in accordance with a concern for intending only what is best to do, this does not seem to me plausibly understood as a conceptual requirement on counting as an Intender. On the contrary, sane people who possess the concept of intention blithely and regularly intend actions they know are not supported by the balance of their practical reasons, or are at odds with what they have reason to do, or that they have no reason to prefer over other available options, or in light of no reason whatever. Acting in these ways may be morally, prudentially, or otherwise flawed, but there is nothing conceptually incoherent about knowingly having such intentions. Someone who uttered Moore’s famous paradox ‘It’s true that it’s raining, but I don’t believe it’s raining” might be justly accused of incoherence about the concept of belief, but if he said “It’s true that I ought to quit smoking, but I intend to keep on smoking,” he is merely our familiar clear-eyed akratic: vicious but not necessarily conceptually confused.¹⁵

Even a flawless practical deliberator will not be in a position to self-ascribe intentions solely on the basis of practical reasoning. This is in part because there is no single kind of reason bearing on how we should act, and I take it that the reasons favoring various courses of action are not always commensurable with one another. There are a variety of goods our actions can reasonably aim at, and reason will not always tell one which to give priority. A committed utilitarian or the like will reject this claim, but many will agree with the likes of Bernard Williams (1981) and Taylor (1982) that there is just no one currency by which to reckon the tradeoffs between many competing values, or the actions that aim to promote these values. Second, all should agree that there are Buridan’s Ass situations in which there is reason to choose between options one judges to be equally desirable in the relevant aspects, but no clear reason to choose one over any of the others. In these situations of incommensurable or indistinguishable value one must simply plump for a given option. We clearly have the ability to do this; people do not starve to death in front of a display of twenty identical Campbell’s soup cans, or dither endlessly between different valuable career choices. But reflection on reasons for action will not tell the agent which option he intends over the others, for reason delivers no unique answer.

A third problem is that transparency to reasons cannot be a route to knowledge of one’s unreasonable or spontaneous intentions, which we surely sometimes have in a first-personally authoritative way. In cases of akratic intending, one judges that all things considered it would be best to do A and then promptly forms the intention to do B. However we understand this puzzling phenomenon, it is clearly a case in which one cannot acquire knowledge of such intentions by self-ascribing the conclusion of one’s practical reasoning. But surely the drunk knows that he intends to go on a bender tonight even though reflection on the damage this will do to his health, career, and relationships inexorably dictates that he should not. And in cases of spontaneous intentions, one simply opts for a given course of action without reflecting on reasons one way or the other. Running down an unfamiliar trail, I might spontaneously form the intention to take the right fork rather than the left, and know which way I intend to turn, without any deliberation to support this self-ascription.

The general point is that while in the case of belief, the entitlement to self-ascribe one’s conclusions about what is true as one’s beliefs is arguably contained in the very concept of belief, I do not think this holds of intention.¹⁶ Our actions are conditionally dependent on our intentions in a way that the truth is not dependent on our beliefs, and this gives us a latitude in forming intentions that we do not have in the case of belief. Awareness of this fact allows for coherent thoughts of the form ‘I ought to do A, but I intend to do B’ and ‘A and B are equally good, so I’ll do A’. Intentions need not dependably mirror some single standard of what is choiceworthy, and this means that we are not entitled to make the transition from ‘A is best to do’ to ‘I intend to A’. What it is conceptually permissible for us to intend is not limited to the deliverance of practical judgment, and transparency to practical reasons is therefore inadequate to explain how we know of our intentions.

3 Deciding and intending

The discussion has aimed to demonstrate that the capacity to form intentions is not identical to the capacity to assess reasons for action. Even the most virtuous agent must transcend reason to form the intention to take one of several equally good routes to his destination or to decide whether to put loyalty to country or to family first. And most of us are not such virtuous agents; we sometimes knowingly intend the worse of our options. What is at operation here is the faculty of the will, or in less nebulous terms, the ability to decide what to do. In the cases at issue, it is an act of decision that takes one from an indeterminate or uncompelling set of practical options to the state of intending. And given that it is possible for us to know of these intentions, this should lead us to think that the act of deciding what to do must play a role in the epistemological account.

I propose that one can come to know what one intends by self-ascribing the content of a decision about what to do as one’s intended action. That is, we are entitled to take ourselves to intend those things we have decided to do. Let us call this route to self-knowledge ‘decision-based self-ascription of intention’. The claim is not that decision-based self-ascription of intention is the only way of coming to know what one intends, but merely that it is an available and justified way. To defend this claim, I will need to first explain what I mean by the mental act of decision. I will then address the question of why the act of making a decision to ϕ provides sufficient grounds for taking oneself to be in a state of intending to ϕ.

A decision is paradigmatically a conscious mental act. This is as distinguished from intention, which is a diachronic mental state and plausibly one that does not essentially require any conscious activity on the part of the agent. A decision is a discrete event rather than a state, and one with respect to which we are normally active. It is the act of determining one’s will, of settling the question for oneself about what to do by arriving at an answer. This is not to insist that the question never gets settled without the conscious participation of the agent. Sometimes we do simply discover that we are already settled on a course of action when we attempt to treat it as an open question and find that we cannot. I therefore do not want to rule out the possibility of unconscious events that have many of the features of a decision. However, it seems to me that we understand these unconscious events as “decisions” only by reference to the ordinary way of exerting one’s executive capacity: by consciously settling on ϕ-ing as one’s answer to the question of what to do. Thus, I will hereafter mean ‘decision’ to imply ‘conscious decision’.

In the very thin sense in which I am thinking of decisions, they are ubiquitous in our practical lives. The term is sometimes reserved for that which marks the end of an episode of practical deliberation with respect to a given practical question. This is undoubtedly a major role of decision; deliberation about what to do often has no endpoint that is endemic to the process, as one could continue indefinitely considering ever more reasons, weighing incommensurable values against one another, or run up against a Buridan situation in which reason alone cannot identify an answer. Deliberation must be cut off at some point and an answer chosen, and this is done by making a decision about what to do. But I do not mean to limit the term to these choice situations; we often decide what to do without any deliberation about reasons or without considering more than one option. As I am thinking of practical decision, then, it is the act common to concluding an episode of practical deliberation with an answer and the spontaneous selection of an answer without reflecting on reasons.¹⁷

But the crucial relevance of decision for the present inquiry is that the mental act of deciding to ϕ is a paradigmatic way of entering into the state of intending to ϕ. What distinguishes deciding to ϕ from merely entertaining the idea of ϕ-ing, fantasizing about it, imagining it, and the like is that decision involves the undertaking of a commitment to ϕ-ing (or more accurately, to initiating a ϕ-ing, since one cannot always commit oneself to succeeding). A decision is an act of settling on what to aim at and thereby committing oneself to exerting one’s agency toward that goal. It is part of the nature of this act that a thinker cannot decide to ϕ if he does not believe it to be reasonably possible for him to ϕ, since it is incoherent to commit to something one believes to be impossible.¹⁸ In committing to exerting his agency toward ϕ-ing, therefore, a thinker commits to doing what is in his power to ensure the possibility of ϕ-ing. Specifically, he undertakes to plan on and around ϕ-ing: to deciding at some point what means to take, to not intentionally doing anything incompatible with ϕ-ing, and to executing this action plan at the appropriate time. These elements are constitutive of the commitment to ϕ-ing in the sense that failing to uphold them amounts to undermining that commitment. Of course, one may fail because one is irrational, weak-willed, forgetful, or incapacitated, or one might simply decide to do something else and thereby abandon these commitments. The commitment to exerting one’s agency toward ϕ-ing and the undertaking of the requisite responsibilities is an engagement one may not uphold and that is conditional on not changing one’s mind. But the essential point is that in making a decision and thereby committing to ϕ-ing, one undertakes to shape one’s downstream thought and action with an eye to ϕ-ing in just those ways that are characteristic of the functional role of intending to ϕ. Deciding to ϕ is a way of forming an intention to ϕ.

I do not mean to claim either that making a decision to ϕ is the only way of coming to intend to ϕ, or that it is a failsafe way. With respect to it being the only way, I again see no reason to deny that there is such a thing as the subterranean settling on an intention without one’s ever deciding on it. We will often be inclined to call such states subconscious desires or sub-personal goals rather than intentions, but in some cases, when these states engender complex means-end reasoning and the like, they may well count as a state of intending. But while the odd subconscious intention is one thing, some have made the stronger claim that many or even most of our intentions are acquired non-actionally, either by making some other cognitive judgment—e.g. the judgment that it would be best, or at least acceptable, to ϕ—or out of a desire to ϕ that is not opposed by any competing desires or reservations.¹⁹ This I do not find persuasive; there is certainly the phenomenon of deciding that ϕ would be best to do, and the phenomenon of being led around by the nose by an unopposed desire to ϕ, but neither of these conditions is sufficient for intending to ϕ. The decision that ϕ-ing would be best is not the same as the intention to ϕ, as I have argued, nor does one necessarily count as intending to ϕ merely because one is overcome with desire.

Second, with respect to decision being a failsafe way of forming an intention, I think it is possible—though relatively rare, as I will shortly argue in more detail—to make a decision about what to do and fail to constitute an intention with this act. The kind of case I have in mind is one in which the agent’s motivations are overwhelmingly at odds with the action he decides on, so much so that it is implausible to think there was ever really an intention there. This is a lesson we can draw from one of the reductionist views just mentioned: that intention can be reduced to the agent’s predominant motivation, plus, perhaps, some other cognitive component (e.g. an expectation of acting, or the belief that an action is a means of satisfying a predominant desire).²⁰ I am no reductionist about intention, but I think an insight of this kind of view is that as an agent’s desires opposing an action get stronger and stronger, it is increasingly hard to say that the agent actually intends to ϕ even if he decides to.²¹ Take Anna, who upon reflecting on all the reasons why her love-affair with V is a terrible thing for her reputation, her children, and so forth decides to break it off with him. Someone with thorough knowledge of her motivational profile—her passion for V, her loathing of her husband, her feelings of suffocation in the role of political housewife—would conclude that her motivation to end the affair is far too weak to bring her even to try to break it off. Knowing this we may be tempted to say that Anna never really intended to break up with him. She believes she ought to and has even made the decision to but was at no point committed to the breakup in the right way to count as intending it.

That said, though failure to put in place an intention by making a decision to ϕ may be possible, it is far from the rule. What I will proceed to argue is that deciding to ϕ is ordinarily sufficient to count as intending to ϕ, and that the self-ascription of an intention to ϕ on the basis of having decided to ϕ is therefore justified. I will now turn to the question of why self-ascriptions of intentions on the basis of practical decisions count as knowledge.

4 Decision-based self-ascription and knowledge

I’ve argued that deciding to ϕ is a paradigmatic way of forming an intention to ϕ. The question now arises: how does making a decision afford knowledge of what one intends? Deciding to ϕ is not logically equivalent to intending to ϕ; further argument is required to establish the entitlement to self-ascribe the latter on the basis of the former.

First there is the matter of the knowledge of one’s decisions: does one necessarily know that one has decided? One might object that knowledge of deciding demands its own equally substantive epistemological story, and thus that the proposed account makes no progress. But one need not embrace a full-blown Cartesian infallibilism about occurrent thought to accept that knowledge of what one is consciously thinking at a time is far less problematic than knowing one is in a certain diachronic functional state. Decisions are discrete, conscious mental events and as such are normally known to the thinker at the time of making the decision; if one does not know whether one has decided, one generally has not. This is in part due to the nature of decision understood as undertaking a set of commitments to action, which dictate that decision cannot be thought of simply as one kind of conscious experience. The thought by which one undertakes to direct one’s agency toward ϕ-ing is a performative type of thought, self-conscious and self-constituting: to take a conscious thought of the form ‘It is settled, I shall ϕ’ to be a decision just is to make a decision. Even here, there may be a space for doubt; if a thinker immediately finds himself continuing to consider the question of whether to ϕ without any pause in deliberation, this might be reason to deny that he really ever made a decision. But it would be a mistake to conclude from the bare possibility of error here that further grounds are required in the good case in order to know that one has made a decision. The entitlement to apprehend occurrent thought as being of a certain type without further grounds is part of being the self-conscious creatures that we are.

It follows from the understanding of decision as self-conscious mental act that the content of the decision will be known to the thinker at the time of deciding. We cannot make sense of someone who claims at t to have decided what to do but does not know at t what he had decided on. Of course, one may well forget what one has decided, even very shortly after one has made the decision; knowledge of one’s decisions is potentially ephemeral. One must merely know at the moment of decision what one has decided. More precisely, the thinker will know what he has decided under the description by which he represents the course of action to himself. This does not ensure that he knows of everything he has in fact decided to do, if there are consequences or alternative descriptions of his action he has not considered in his practical deliberation. Decisions are intensional; one merely knows what one has decided on as one conceives of it.

I take it that if deciding to ϕ is ordinarily sufficient for coming to be in the state of intending to ϕ, the self-conscious making of that decision justifies the self-ascription of the relevant intention. At the least, if deciding is normally sufficient for intending, taking oneself to intend what one has decided to do would be a truth-conducive method of ascribing intentions to oneself. Further, this connection would be no accident. The fact that making a decision to ϕ is a way of forming an intention to ϕ is contained in the very concepts of intention and decision. The point of making a practical decision is to engage oneself to pursue a given course of action and thereby to settle on an intention; this is part of what makes it the act that it is. And part of what defines the state of intention is the fact that intentions are paradigmatically formed by making a decision about what to do. Thus, a thinker who possesses these concepts, however implicitly, will be disposed and willing to ascribe to himself an intention to ϕ on the basis of having decided to ϕ when the question of what he intends arises.²² A thinker who knew he had decided to ϕ but considered it a serious open question whether he intended to ϕ would be misapplying the concepts of decision and intention—at least in the absence of evidence of an abnormal psychological breakdown between deciding and putting an intention in place.

But I have granted that breakdowns between deciding and intending are possible; we may call into the “vasty deeps” of our will by making a decision but fail to constitute an intention.²³ So why think such breakdowns abnormal enough that the proposed entitlement is not defeated? It might be that we presuppose in making a decision that we have thereby formed an intention but are routinely mistaken in this presupposition. If deciding to ϕ is frequently insufficient for counting as intending to ϕ, self-ascriptions on the basis of such an event would be dubious candidates for knowledge.

I will try to illustrate the full force of this challenge by arguing that it does in fact pose a problem for a parallel account of how we know what we believe. Consider Christopher Peacocke’s account of the knowledgeable self-ascriptions of the belief that P on the basis of making a conscious judgment that P (Peacocke 2003). He argues that a thinker’s mastery of the concept of belief involves the disposition and willingness to self-ascribe the belief that P when he makes a conscious judgment that P, for the reason that he made the judgment, and when the question arises (p. 99). Making the conscious judgment rationalizes the self-ascription, Peacocke claims, and the thinker grasps this fact insofar as he grasps the concept of belief. Since making a judgment is the fundamental way to form a belief, a self-ascription on this basis will be correct (when all is working properly). He concludes:

In summary, when all is working properly, knowledgeable self-ascriptions track the property of belief for this reason: the very means by which they are reached are ones whose availability involves the thinker’s having the relevant belief. When all is working properly, these means would not be available were he not to have the relevant first-order belief (p. 102).

In a similar vein, Nishi Shah and J. David Velleman claim that there is no problem with moving from a conscious judgment that P to the attitude of belief that P:

Exactly how one accomplishes this transition is of course ineffable, but it is a perfectly familiar accomplishment, in which a proposition is occurrently presented as true in such a way as to stick in the mind, lastingly so represented. Affirming that P typically induces an affirmative attitude toward P (p. 503).

I believe these treatments of the transition between judgment and belief are too quick. It is not obvious that the conscious judgment that P cannot be made unless the thinker truly has the first-order belief. In fact, it seems to me not uncommon that a conscious judgment is made without the thinker having the right dispositional structure to count as believing that P. On a plausible analysis of belief, being disposed to think and act as if P is the case is an important constitutive element of believing that P.²⁴ If a thinker does not guide oneself as though P is true, especially when the stakes are high, this is good reason to deny that he believes that P. And while being inclined to make the conscious judgment that P is one relevant disposition, this often occurs without there being sufficient further dispositional configuration to be described as believing that P. One may make the sincere, conscious judgment that P, at a given moment and in a given social or evidential context, and then proceed to think and behave in ways that are completely inconsistent with P, revealing that one is not being guided by a representation of P as true.

This is particularly common with value judgments, but it also happens with ordinary descriptive judgments. We are all familiar with the ways that conscious judgments can be radically disconnected with broader dispositions of thought and action: people judge that “he’s never going to leave her,” but wait for decades for it to happen; they judge that women and men have innately equal musical talent, but hire a much higher proportion of male candidates for their orchestra than they would if auditions were blind; they judge that tipping well is the right thing to do, but never tip well; they judge that heaven is far superior to this terrestrial existence, and that it is easier for a camel to pass through the eye of a needle than for a rich man to get into heaven, but do everything they can to avoid departing for the afterlife and to amass a fortune in this one. In fact, Lawlor (2003) points out that transparent reflection on reasons for belief as emphasized by Moran can make us particularly prone to reaching conscious judgments that will do nothing to control future behavior, if we are led to focus only on a subset of the relevant reasons. Bluntly, conscious judgments often simply happen at a moment, in response to social and internal pressures other than evidence of what is true or to a certain framing of the evidence, and fail to be effectively incorporated into the thinker’s web of doxastic commitments or to alter his associated dispositions.

The trouble is that lacking the right dispositional structure to behave as if P is true weighs heavily in favor of the conclusion that one does not really believe that P. In ascribing beliefs, it does not suffice to consider only what a person is inclined to judge at a time and in a certain framed context; we must also consider how his thoughts, choices, and actions will be guided over time. A thinker’s consistent failure to behave in accordance with a conscious judgment that P is thus strong reason for us to deny that he ever really believed that P. The prospects for an account of self-ascriptions of belief that are based solely on conscious judgment at a time therefore do not look good. If making the conscious judgment that P is not sufficient for counting as believing that P, then it is not a sufficient ground for the self-ascription of that belief. Essentially, Peacocke, Shah, and Velleman make too little of the difference between the mental act of judging that P and the mental state of believing that P. If the same problem faces the account of decision-based self-ascription of intention, it will turn out to be an equally beleaguered account.

However, I want to argue that the parallel account for intention does not face this problem to anything like the same degree. Decision and intention are not logically identical, but deciding to ϕ is normally sufficient for counting as intending to ϕ. Let’s start by noting a curious fact: there are some philosophers who have claimed that we can’t fail to know what we intend. That is, some who have thought a great deal about the matter have argued that we cannot have intentions we do not know of or believe we have intentions we do not have.²⁵ For example, Wallace (2001) defends both claims:

Someone who is entirely unaware of their alleged intention to do x, or who has completely forgotten that this is what they intend to do, cannot really be described as having the intention to do x any longer…. By the same token, it is equally difficult to imagine a scenario in which you sincerely believe that you intend to do x, while utterly failing to have this intention in fact (p. 22).²⁶

I do not think this is true on either count; it seems to me that in the above example, Anna falsely believes she has the intention to break it off with V and intends to continue the affair without knowing it. But what these claims indicate is that it is difficult enough to find a clear-cut case of never actually intending to ϕ when one thought one did that reflection on the matter leads some to believe there aren’t any. I think there is a reason why. In the belief case, I argued that there is a tension between judging that P and subsequently failing to reason and act as though P is true which in some cases should lead us to deny that the thinker ever believed that P. In contrast, the lack of any downstream result of a decision to ϕ need not belie the presence of the original intention. In the vast majority of cases in which a self-ascribed intention never leads to action we simply take the agent to have changed his mind, sometimes at the very last moment. Even with Anna it is difficult to find clear grounds on which to insist that she really never intended to break off the affair, rather than concluding that the rush of emotion brought on by seeing V causes her to decide at the last minute not to break it off. Alternatively, as long as her intention does not specify a particular time or way of fulfilling it, the standard description of her situation is that she is instrumentally irrational or simply procrastinating. Unless we know an unusual amount about Anna’s motivational psychology and know that the break-up would be well-nigh impossible for her even to attempt, the ordinary thing to say is that she intends to end the affair but keeps putting it off.

Changing one’s mind about what to do or procrastinating in taking the means to one’s ends may leave one open to criticisms of weakness, irresoluteness, cowardice, ineffectiveness, or other negative evaluations. However, these are not grounds for denying that one ever had the intention one thought one did. Essentially, my claim is that whereas the proof of the believing is to a significant extent in the downstream behaving, an intention does not depend for its existence on there being any follow-up. This is certain to sound odd, for intention even more than belief is constitutively connected with action—how can it be that failure to act is not in conflict with the original ascription of the intention? But the answer, again, lies in the relation between intention and action and its dissimilarity to the relation between belief and truth. We have seen that in the case of action, there are a multitude of things one could reasonably do and a variety of means one could take to do them. Reason constrains but often does not dictate these choices. Consequently, it is frequently rationally open to the agent to change his mind about what to do, thereby abandoning the previous plan and adopting a new one. Given the variety of reasonable actions available, one may legitimately change one’s mind without new reasons coming to light or specific evidence that one’s prior deliberation was flawed. It may be inadvisable from the point of view of efficiency to continually change one’s mind about what to do, but it is not fundamentally in conflict with the concept of intention.

In the case of belief on the other hand, the truth is not similarly conditionally dependent on what one decides to believe. We are resigned to attempting to track the truth with our beliefs, where we usually have no choice about what is the case, and therefore no rational choice about what to believe (self-fulfilling predictions aside). This means that we cannot understand a thinker whose behavior clearly belies his avowed judgment as simply having changed his mind about what to believe, because one cannot rationally decide to believe something else absent new evidence. Either one takes the evidence to show that P is true or one does not, and so failing to operate as though P is true (without some intervening event rationalizing a change in belief) is in logical tension with the claim ever to have believed that P in a way that does not hold with respect to intention.

I have acknowledged that what might ground the denial that there was ever an intention present is a fact about the agent at the time of making the decision—evidence of a preponderance of motivation against the action decided so strong that the action is effectively impossible for the agent to carry out. But short of this, a decision that never ends up influencing the agent’s downstream thought and action can still for all that amount to intending at the time, albeit a short-lived intention. The decision generally trumps, I think; deciding to ϕ is normally sufficient to count as intending to ϕ, until one changes one’s mind or runs out of time to procrastinate.²⁷

5 What value self-knowledge?

The proposal is that one can knowledgeably self-ascribe an intention on the basis of making a decision about what to do. I have not claimed that this is the only way one can acquire such self-knowledge, and indeed, it may seem to have only a narrow scope of application. Surely one need not always re-open the question of what to do and make a new decision in order to acquire knowledge of what one intends. Of what use then is this decision-based route to self-knowledge?

The first thing to be said is that this account can be straightforwardly extended to include self-ascription on the basis of the memory of a decision. Whatever account we are inclined to accept concerning the epistemology of memory should apply to the memory of having made a decision, allowing the subject to acquire knowledge of his intention by recalling a past decision: ‘I know I intend to vote for Obama because I decided to after the first presidential debate.’ Second, we can acknowledge that the role of decision in constituting intentions allows for epistemic shortcuts by kicking things up in their wake, so to speak, without the thinker recalling the original circumstances of the decision. We tend to track our goals over time, building them together into a planning system that provides a structure to our agency. And insofar as we keep track of our goals, this can be a shortcut to knowing what we intend; it is generally enough to know that ϕ-ing is part of one’s plans. Similarly, referring directly to knowledge of a standing policy or habitual way of acting may afford one knowledge of what one intends without referring back to whatever thought went into creating these policies. The epistemic warrant provided by the decision can be preserved in a chain of knowledge without the subject maintaining awareness of the original grounding of that warrant.

However, the value of having access to the original decision if necessary can be illustrated by reflecting on the so-called ‘norm of stability’ on intention. Bratman (1987) teaches us that there is rational pressure against reconsidering one’s intentions in the absence of new reasons. But encountering new reasons is not the only situation in which it is rational to reconsider one’s intentions. We may also do so when we find ourselves tracking a goal and discover that we are no longer sure why. Suppose I recall that I intend to build a squirrel house, and see that I have reasons for doing so, but also see that I have sufficient reasons for not bothering. All of these are reasons I have known of all along. However, it now seems mysterious to me why I intend to build the squirrel house rather than not. In this case, a natural thing for me to do is to try to recall having decided to build it. If I can recall this decision, I may well take this as sufficiently authoritative to retain the intention, even if I can no longer recall why I so decided. But if I cannot recall the decision, the intention will begin to seem alien to me; it is something I find myself with, but that does not seem to have its source in me. Here, it would be rational for me to reopen the question of whether to build a squirrel house, and to resolve anew. What this shows, I think, is that our decisions are important not merely for knowledge of our intentions, but for self-knowledge of them. When an intention is both opaque to one’s practical reasons and cannot be linked to a decision, it tends to lose its status as an intention of one’s own.

Going along simply tracking our standing policies, habitual routines, and practical reasons for action without ever asking questions about our own psychological states may be perfectly unproblematic in the main. But the question of what one intends may arise, and I have argued that this question cannot necessarily be answered simply by reflecting outward on reasons for action. The value of decision-based self-knowledge is that when one is uncertain about what one intends, one may always simply to deliberate about what to do and decide anew, thereby both forming an intention and acquiring knowledge of what it is. This procedure will not provide an answer to what one intended at the time the question was asked, if indeed there is reason to suppose one had any intention at all. In this sense, it is a route to self-knowledge that is essentially forward-looking rather than retrospective. But I suggest that in the case of intention, it is precisely forward-looking knowledge of what one now intends for the future that is of most value. The reason is that what one will do depends on one’s current intentions; past intentions are irrelevant to future action. Knowledge of what one now intends for the future provides epistemic access to what one will do, allowing one to plan accordingly. Self-knowledge acquired through the act of making up one’s mind is thus central to our agency, in a way that knowledge of past intentions that have been forgotten or replaced is not.

I claimed at the outset that our understanding of self-knowledge must pay sufficient attention to the unique features of the mental state in question. And I have argued that when we do this with respect to intention, we should be impressed by the capacity to form intentions by making up one’s mind about what to do. Whereas we do not simply have the latitude to change our minds about what to believe in the absence of new evidence, we may in many cases change our minds about what to do without irrationality. And when we lack knowledge of what we intend, we can acquire it by making a decision about what to do and ascribing the content thereof as our intended action. We are in a position to author our intentions, and thereby to know what they are.