Desire-Based Theories of Reasons and the Guise of the Good

1. Introduction

This paper motivates a novel account of desire that reconciles two apparently conflicting intuitions about the connection between desires and knowledge of practical reasons. The argument centres on a certain intuitive datum. The datum is that often when an agent feels a desire for P, she will seem to immediately and conclusively know that there is a reason to bring P about.¹ This intuitive datum concerns the epistemic role desires play in grounding knowledge of reasons. We can explain this connection in two directions: either the agent has a reason because of her desire, or the agent has her desire because she has a reason. Desire-based theories of reasons take the former approach, while the guise of the good takes the latter. Desire-based theories of reasons seem uniquely placed to explain the intuitive datum because they maintain that desires are the source of an agent’s practical reasons (Hubin 1991; 1999; 2003; Schroeder 2007b; Sobel 2001a; 2001b; 2005; Williams 1981).² Since desires generate reasons, feeling a desire for P can provide immediate and recognisably conclusive justification for the practical judgment that one has a reason to bring P about.³ The epistemic claim that S is an experiential source of recognisably conclusive justification is naturally opposed by the thought that S is a fallible source of justification. Any view which claims, or implies the claim, that desires confer recognisably conclusive justification to practical judgments also needs to account for how desires could be fallible. Desire-based theories face the objection that they make basic desires oddly infallible because, on such views, basic desires always guarantee the existence of a reason to bring P about.⁴ This conflicts with the intuition that sometimes our basic desires rationally ought to count for nothing. All else being equal, we want to explain both how desires sometimes ground conclusive knowledge and sometimes rationally count for nothing.

The guise of the good takes the opposing order of explanation. On a standard version of this view, a desire for P represents P as good in some respect (Alvarez 2010; Anscombe 1957; Boyle & Lavin 2010; Brewer 2009; Clark 2010; Hawkins 2008; Johnston 2001; Oddie 2005; 2009; 2015; 2017; 2018; Scanlon 1998; Schafer 2013; Smithies & Weiss 2019; Stampe 1987; Tenenbaum 1999; 2007; 2008; 2009; 2018).⁵ The guise of the good provides a satisfying explanation of the fallibility of desires. This view presupposes that reasons are based on value and are metaphysically independent of desires. The independence of desire and value allows for the claim that desires sometimes misrepresent value. A desire that misrepresents value thereby misrepresents one’s reasons, and so rationally ought to count for nothing. At first glance, this view seems to explain the intuitive datum. However, extant versions explain the fallibility of desires in a way that commits them to the claim that desires provide at best inconclusive justification for practical judgments. This commitment leads to a denial, or at least a significant revision, of the intuitive datum. We face an unpleasant trade-off: either explain the intuitive datum by denying the fallibility of desires or explain the fallibility of desires by denying the intuitive datum.

I propose an epistemological disjunctivist version of the guise of the good that reconciles both intuitions.⁶ The disjunctive account says that desires come in two categorically distinct epistemic kinds. A desire is either an awareness of value or an illusion of value. An awareness of the value of P provides the agent with recognisably conclusive justification to judge that there is a reason to bring P about. In favourable circumstances, the desiring agent can tell by reflection that she is aware of the value of P. This explains how an agent can immediately and conclusively know that she has a reason to bring P about on the basis of a desire for P. An illusion of value merely purports to be an awareness of value and the agent is entitled to discount such a desire as rationally counting for nothing. The possibility that desires could have two distinct epistemic values is explained in terms of a more basic capacity to discriminate value.⁷ A successful exercise of this capacity issues in an awareness of value, while a defective exercise issues in an illusion of value. The disjunctive account explains how desires can both ground conclusive knowledge of reasons and can sometimes be entirely wrong about one’s reasons. The disjunctive account promises to resolve a deep tension in our epistemic theorising about the rational role of desire.⁸ Of course, its ability to reconcile these intuitions does not establish its truth. Nevertheless, its promise of being the best of both worlds makes it worthy of further consideration in the debate around the rational role of desire.

2. The Intuitive Datum

Let me introduce the intuitive datum with the following case:

Long-Haul Flight. Ruth sits in an aisle seat on a long-haul flight. Her legs have been aching and she feels an intense desire to stretch her legs into the aisle to relieve the ache. Ruth immediately judges on the basis of her desire that “relief is a reason to stretch” and she stretches her legs in just the right way to relieve her ache. Ruth acts on her desire without any prior deliberation or discursive reasoning. Nevertheless, the action does not seem like a mere reflex to her. Ruth is unreflectively confident that her action does not need further justification.

In Long-Haul Flight, Ruth immediately, unreflectively passes the practical judgment that she has a reason to stretch and acts accordingly. Ruth’s judgment and action are directly based on her desire to stretch. Ruth is not simply moved by her desire like a marionette puppet, yet she is also not engaged in explicit, discursive reasoning about what reasons she has.⁹ Ruth unreflectively, but still rationally, takes at face value the way her desire makes it seem that she has a reason to stretch. Ruth’s epistemic conduct seems appropriate, and it is natural to regard her as having acquired knowledge of a reason for action. Long-Haul Flight captures the following intuitive datum about practical agency: often when an agent desires a prospective state of affairs P, she will seem to immediately and conclusively know that she has a reason to bring P about.¹⁰ Cases similar to this are discussed by advocates of desire-based theories (see Chang 2004: 58–70; Hubin 2003; Schroeder 2007b: 1–22; Sobel 2005: 437–41) and by advocates of the guise of the good (Hawkins 2008: 249–50; Johnston 2001: 188–94; Oddie 2005: 54–57; Scanlon 1998: 41–50; Schafer 2013; Smithies & Weiss 2019; Stampe 1987: especially 344, 375–76). I take Long-Haul Flight to be representative of an intuition that is central to the debate around the connection between desires and reasons. I will now comment on how we ought to understand this intuitive datum.

Most importantly, the intuitive datum concerns the role that a desire plays in acquiring knowledge of reasons. The datum is not in the first instance about the conditions that must be satisfied for the existence of a reason. I will not use the datum to directly adjudicate that issue. Rather, it will serve as a constraint on our epistemological account of how desires ground knowledge of reasons.¹¹ Due to this epistemological concern, I will often discuss how an agent introspectively regards her desires as grounds for judgment.¹² I should emphasise that the intuitive datum involves a practical judgment and action being directly justified by a desire, not by an agent’s introspective awareness of her desire. There are another four features of the intuitive datum I would like to highlight.

First, the intuitive datum concerns felt desires rather than dispositional desires.¹³ A felt desire is something that is experienced, a mental state with a phenomenal character. The intuitive datum involves an agent judging that she has a reason directly on the basis of her desire. Only a felt desire can be the direct object of introspective awareness and thereby form the basis for judgment. It might be possible for an agent to deduce that she has a reason on the basis of a dispositional desire. However, Long-Haul Flight involves Ruth recognising, not deducing, that she has a reason on the basis of her desire. I will be narrowly concerned with explanations of the intuitive datum and will only use the term “desire” to pick out felt desires. So, not everything that we might ordinarily call a “desire” will be covered by my discussion.

Second, in Long-Haul Flight, Ruth takes herself to have a normative reason for her action rather than simply a motivating reason.¹⁴ That is, Ruth takes her reason to justify her action and not merely explain its occurrence. Furthermore, there is a reason to stretch, and Ruth grasps how this reason speaks in favour of judgment and action. So, the intuitive datum involves three parts: there is in fact a reason for action, the agent knows that there is a reason for action, and she also knows what her reason is a reason for.¹⁵ It involves an agent successfully recognising reasons she, in a sense to be specified, “objectively” has. Applied to the case, we can say that relief is a reason to stretch, Ruth knows that she has a reason to stretch, and she has a grip on how relief speaks in favour of stretching.

Third, Ruth’s knowledge has a particular form. Ruth passes the practical judgment that “relief is a reason to stretch” on the basis of her desire. Ruth’s desire provides her with a practical reason and, in doing so, also provides her with an epistemic reason to pass the practical judgment that “relief is a reason to stretch.” The general form of such judgments can be characterised as either pro tanto or all-out judgments, and the claim contained within the judgment is either hedged or outright. I will now explain each piece of terminology in turn.

Pro tanto judgments are about possessing pro tanto reasons, that is, judgments about possessing a reason or some reason (Davidson 2001: 97–100; Tenenbaum 2007: 261–62).¹⁶ By contrast, all-out judgments concern the best reasons or strongest reasons, all-things-considered. I do not want to take a stand on how exactly an agent arrives at all-out judgments. Desire-based theories of reasons and the guise of the good are plausibly associated with different theories of the behaviour and interplay of reasons.¹⁷ Adopting a substantive theory of the behaviour and interaction of reasons at this stage risks prejudicing the dialectic. Fortunately, the intuitive datum is explicable in terms of pro-tanto judgments, which allows us to sidestep this issue. I assume that pro tanto judgments can be sufficient for action, at least under favourable conditions. I also leave open the possibility that pro tanto judgments might automatically qualify as all-out judgments under certain conditions. Plausibly, Ruth’s judgment is a pro tanto judgment. An awareness of a reason, namely the pleasure of relief, seems sufficient to account for her judgment and action, so there is no obvious need to attribute to Ruth an all-out judgment.

The content of a judgment about reasons can be hedged or outright. Hedged judgments have a qualified, inconclusive character, such as “relief is probably a reason to stretch” while outright judgments do not contain a hedge, simply asserting “relief is a reason to stretch.” In Long-Haul Flight, Ruth seems to pass a pro tanto, outright judgment with the content “relief is a reason to stretch.” And, by generalisation, any instance of the intuitive datum involves an outright, pro tanto judgment. Intuitively, Ruth would only introduce the hedge that she “probably” has a reason if she had epistemically motivated doubts about her reasons. So, the intuitive datum involves a practical judgment that has the form of a pro tanto, outright judgment.

Fourth, the intuitive datum involves Ruth’s desire directly conferring justification to her practical judgment and action. Ruth’s practical judgment can, broadly speaking, be justified conclusively or inconclusively.¹⁸ Ruth’s judgment is inconclusively justified when her desire raises the likelihood that there is a reason to bring P about but does not rule out the possibility of error.¹⁹ The relevant sense of justification is “objective” in that it tells us what an agent is rationally entitled to believe, not how she takes herself to be rationally entitled. If Ruth’s desire provided her with inconclusive justification, she would only be rationally entitled to a hedged judgment. Ruth might, in ordinary discourse, express her judgment as the outright judgment “relief is a reason to stretch.” However, if based on inconclusive justification, such justification does not guarantee the truth of her judgment, so her rational entitlement would more accurately expressed as a hedged judgment that “relief is probably a reason to stretch.” Consequently, I will talk as if inconclusive justification only rationally warrants a hedged judgment.

By contrast, conclusive justification guarantees the truth of a corresponding judgment.²⁰ If Ruth acquires conclusive justification from her desire, then it guarantees that relief is a reason to stretch or, equivalently, it rules out the possibility that there is no such reason. Conclusive justification is also an “objective” notion. Conclusive justification guarantees the truth of a corresponding judgment even if the agent is not in the epistemic position to recognise that she has such justification or simply fails to recognise this, despite being in the position to do so. Such justification rationally entitles an agent to an outright judgment. If Ruth’s desire provides conclusive justification, then it settles the matter of whether relief is a reason to stretch and entitles her to an outright judgment that “relief is a reason to stretch.” I wish to make some clarifying comments on the relationship between conclusive justification and certainty.

Conclusive justification can rationally entitle an agent to a particular kind of certainty.²¹ Suppose that Ruth’s desire guarantees that relief is a reason to stretch. Intuitively, Ruth would be rationally entitled to a high level of confidence in her judgment. However, she would never be entitled to an absolute confidence that involved ignoring all possible counterevidence to her judgment.²² Such absolute certainty is irrational because Ruth knows herself to be fallible and would have made mistakes in similar circumstances. Notice that there are two factors determining her confidence while there is only one factor determining her justification. The level of confidence she is rationally entitled to is the result of (at least) the justification conferred by her desire and her knowledge of her own fallibility. Ruth’s justification for her judgment is determined by what she has taken for her basis for judgment, namely, her desire. Confidence in a judgment has a higher-order component that one’s justification for judgment need not have. So, the possession of conclusive justification can, under the right conditions, rationally entitle an agent to non-absolute certainty. This level of certainty involves a confident, outright judgment while being open to entertaining possible counterevidence.

The intuitive datum is most naturally characterised in terms of Ruth having conclusive justification (and in similar cases where things go wrong, seeming to have conclusive justification). I consider, in turn, three major motivations for this claim.

The first motivation is that Ruth has a high level of confidence in her judgment. Ruth’s epistemic conduct seems appropriate. This suggests that Ruth is correct to take herself as possessing a strong form of justification. Such rationally warranted confidence contrasts with the level of confidence rationally warranted by inconclusive justification. Suppose I arrive home and see my spouse’s car in the driveway and judge that “my spouse is home.”²³ Even though the judgment is expressible as an outright judgment, I do not mentally “assert” it with a comparably high level of confidence. After all, I know that my spouse only drives to work three of the five working days in a week. My evidence raises the likelihood that my spouse is home but by no means rules out the possibility of error. I am, strictly speaking, only rationally entitled to a hedged judgment such as “my spouse is probably home.” Correspondingly, I am not entitled to a high level of confidence in my judgment. In Long-Haul Flight, Ruth’s confidence in her judgment is very high, which suggests a correspondingly high strength of justification. This contrast makes it natural to think that Ruth possesses conclusive justification.

The second motivation is that is it natural to assume that Ruth could appeal to her desire as a regress-stopping source of justification.²⁴ Suppose we asked Ruth a series of “why?” questions about her justification for her action. If asked why she stretched, Ruth might reply “I just wanted to.” If pressed further about why her desire justifies, she might register bafflement at the question and reiterate that she simply wanted to stretch. The implication would be: what stronger grounds could one have to believe that one has a reason to stretch other than a desire to do so? The search for the justification of Ruth’s action naturally terminates by citing her desire. This suggests that her desire guarantees that there is a reason. Inconclusive sources of justification do not have this authority. Suppose I judge that my spouse is home on the basis of seeing their car in the driveway. When asked to justify my belief, I might reply that I saw their car. When pressed about why that entitles me to such a judgment, I would need to elaborate on why I regard the presence of their car as an indication of their being home. Simply citing the presence of their car does not end the regress of justification. This contrast is neatly explained by the idea that Ruth possesses conclusive justification, while seeing my spouse’s car only provides inconclusive justification.

The third motivation comes from an observation about an agent’s retrospective attitudes toward her desires.²⁵ Suppose that Ruth stretches her legs and discovers that she was wrong—she had no reason to stretch her legs. Perhaps stretching her legs did not provide her with relief or that she got relief, but it was not pleasurable in the envisaged way. We naturally imagine Ruth as being surprised and regarding her desire as misleading. Furthermore, this retrospective attitude is epistemically appropriate. We can explain the appropriateness of this attitude in terms of conclusive justification. Ruth is surprised that she is in error because her justification seemed to rule out the possibility of error. Consider, by comparison, how we ordinarily respond to erroneous judgments based on inconclusive evidence. Suppose I discover that my spouse is not home after seeing their car in the driveway. I am not surprised, and I do not regard my basis for judgment as misleading. Clearly, my basis for judgment is compatible with the possibility of error. An agent’s retrospective attitudes to their basis for judgment in cases of error tell us about the strength of the justification that the agent took herself to have, assuming that the agent is epistemically responsible. Since it is natural to imagine, in the event of error, Ruth being surprised and regarding her desire as misleading, we have good reason to say that Ruth seemed to possess conclusive justification. And, since in Long-Haul Flight, Ruth is epistemically successful, it looks like she genuinely has conclusive justification.

Let us take stock of what we have covered so far. I have motivated the following interpretation of the intuitive datum. The intuitive datum represents the intuition that desires can provide recognisably conclusive justification for practical judgment and action. The intuitive datum involves an agent passing a conclusively justified pro tanto, outright judgment that she has a normative reason to bring the desired state about, for example, “relief is a reason to stretch.” This is, of course, not the only way to understand such cases. However, this interpretation does seem to be a natural one and, in any case, is a dialectically fair place to begin the comparison of desire-based theories and the guise of the good.

3. Desire-Based Theories of Reasons

A natural way to explain the intuitive datum is to say that Ruth has a reason to stretch because she has a desire to stretch. Desire-based theories of reasons take this order of explanation. On this view, desires are the source of practical reasons (Hubin 1991; 1999; 2003; Schroeder 2007a; 2007b; Sobel 2001a; 2001b; 2005; Williams 1981). A desire for P generates, or partly constitutes, a reason to bring P about. A consideration, C, becomes a normative reason for an agent when that agent has a desire for P and C promotes the satisfaction of the desire for P (Hubin 2003: 329–32; Schroeder 2007b: 5–21; Sobel 2001b; Williams 1981). Desire-based theories come in two basic forms which disagree about how desires generate reasons. Simple versions maintain that reasons are generated by an agent’s actual desires (Schroeder 2007a; 2007b; Sobel 2001b; 2005). Counterfactual versions hold that reasons are generated by the desires an agent would have under idealised conditions (Hubin 2003; Smith 1995; Williams 1981). Simple and counterfactual views also disagree about the rational significance of basic desires. Simple versions say that basic desires always generate reasons (Schroeder 2007b: 84–101). Counterfactual views say that only the basic desires an agent would have under idealised conditions generate reasons (Hubin 2003; Smith 1995; Williams 1981). All desire-based theories agree that non-basic desires may fail to generate a reason, for instance, due to instrumental irrationality. Let us now consider how desire-based theories explain the intuitive datum.

Desire-based theories say that Ruth’s desire to stretch generates a reason to do so. The consideration that stretching provides relief becomes a normative reason because it promotes the satisfaction of her desire for relief by stretching. This order of explanation seems uniquely placed to capture how Ruth’s desire confers recognisably conclusive justification. Her desire to stretch is the very state that makes it true that she has a reason to stretch. Her desire guarantees the truth of her judgment and is reflectively accessible to her as a source of conclusive justification. This vindicates Ruth’s epistemic conduct. She would be entitled to a high level of confidence in her judgment, she could cite her desire as a regress-stopper, and it would be appropriate for Ruth to be surprised in the event of error. Desire-based theories provide a satisfying and non-revisionist explanation of the intuitive datum.

Crucially, these epistemic payoffs depend on the metaphysical thesis that desires generate reasons. How else could a desire rule out the possibility of error other than by being the source of reasons? This thought is an expression of a more general assumption that an experience E can only guarantee the truth about judgments within a domain D when there is a relationship of (at least partial) constitution between E and the elements of D.²⁶ On this assumption, constitution could work in two directions: reasons could constitute desires, or desires could constitute reasons. Only the latter option seems plausible. Insofar as we assume that conclusive justification requires a relationship of constitution between desires and reasons, desire-based theories of reasons look like our only option. However, this very metaphysical commitment is a source of trouble for desire-based theories. And, as we will see below, we can account for desires as providing conclusive justification without making the constitution assumption.

4. Over-Generating Reasons and Fallibility

The thought that S is an experiential source of conclusive justification is naturally opposed by the thought that S is a fallible source of justification. All else being equal, we should want to capture the possibility of conclusive justification in such a way that captures fallibility. Desire-based theories succeed in saying that desires are a source of conclusive justification, but in a way that renders them unable to capture the fallibility of desires. The source of the problem is that simple desire-based theories generate reasons where, intuitively, there are none (Hubin 2003; Parfit 2011: 58–82; Schroeder 2007b: 84–110). Basic desires always generate reasons, which implies that a basic desire for P will always provide justification for the judgment that there is a reason to bring P about. This clashes with the intuition that sometimes our basic desires rationally ought to count for nothing (see Quinn 2012). To capture this intuition about fallibility, we need to claim that basic desires are sometimes entirely in error about our reasons, but that is a claim flatly inconsistent with the basic commitments of simple desire-based theories.

Let us consider a case to make the objection vivid. Suppose Ruth feels a desire to drink a can of paint and she is baffled by this desire.²⁷ Suppose further that Ruth regards the prospect as utterly undesirable: she does not want to make herself sick, she knows that paint tastes awful, and so on. Intuitively, Ruth has no reason to drink paint and she completely lacks justification for the judgment that she has a reason to drink paint. Ruth seems to recognise that her desire is simply wrong about what she has reason to do. Yet, simple desire-based views maintain that her desire cannot be wrong about there being some reason to drink paint. Simple desire-based theories make basic desires an oddly infallible source of reasons.

Some desire-based theorists, like Mark Schroeder, bite the bullet and say that basic desires always generate reasons (2007b: 92–98). Schroeder (2007b: 92–98) argues that although the attribution of a normative reason may sound strange it is strictly speaking correct (see also Hubin 2003). Ruth really does have a reason to drink paint, but it is massively outweighed by competing reasons. In support of this line of response, Schroeder (2007b: 92–98) advances an error theory about our negative existential intuitions about outweighed reasons for action. Schroeder argues that we naturally say that there is “no reason” when we strictly speaking mean that there is a negligibly small, outweighed reason. For instance, Schroeder (2007b: 95–96) tells us that we all have reason to eat our own cars. This sounds strange at first. But presumably we all have a desire to maintain our iron levels and so we have normative reason to eat our cars to acquire their iron content. The claim now sounds a little less bizarre when put in this context. Schroeder (2007b: 96) further insists that the reason to eat one’s car “is of about as little weight as any reason could possibly be” and that one would be deeply irrational to consider acting on it. Schroeder generalises this strategy to account for all cases of apparent over-generation of reasons.

Schroeder’s theory could be used to respond to the worry about fallibility like this. Schroeder could say that Ruth’s desire is wrong about her reasons because it misleads about the weight of her reason to drink paint. Plausibly, Ruth has a basic desire to be healthy, a basic desire to avoid unpleasant tastes, and so on. Such basic desires could outweigh or override the reason generated by her basic desire to drink paint but the psychological salience of her basic desire for paint presents the wrong view of her reasons. Ruth would be irrational to act on her desire because she would be knowingly acting on a desire that presents her reason as stronger than it really is.

The problem with this approach is that it depends on the idea that the presence of overriding reasons makes the existence of the overridden reason somehow more palatable. From a third-personal perspective, we are not ordinarily inclined to ascribe to Ruth a reason of any weight to drink paint. Furthermore, from within the desiring agent’s reflective perspective, the desire to drink paint is not likely to pre-theoretically register as a source of any reason. Consider things from Ruth’s perspective: in her deliberation about what to do, her awareness of her basic desire to drink paint would not even seem to provide her with a reason for doing so (see Brewer 2009: 23–24; Quinn 2012). Intuitively, such a desire to drink paint does not provide a reason but rather presents the question of whether, from Ruth’s own perspective, she has a reason to go along with how her desire inclines her to act (Brewer 2009: 24).²⁸ There is no obvious reason why we should attribute a reasons-generating significance to Ruth’s desire other than a prior commitment to the general thesis that desires generate reasons. Of course, this might be the correct view when all is said and done. However, that it cannot capture the intuition that our basic desires sometimes count for nothing is a significant strike against the prima facie plausibility of this approach to explaining the fallibility of desires. Desire-based theories have other options for explaining fallibility which we should explore before considering the guise of the good.

5. Counterfactual Desire-Based Theories

Counterfactual versions of desire-based theories try to avoid over-generating reasons by placing constraints on the way basic desires generate reasons. Consequently, they promise to explain the intuition that desires can sometimes be entirely wrong about what reasons we have. While desires are necessary for the existence of a reason, desires generate normative reasons insofar as they pass some counterfactual test that filters rationalising desires from non-rationalising desires. Only desires which are (or approximate) those an agent would have under idealised conditions successfully give rise to reasons for action. There are a variety of options of how to conceive of an agent’s idealised conditions (see Schroeder 2007b: 84–86). Some accounts require that the content of an agent’s first-order desires must cohere with the content of her second-order desires (Lewis 1989)²⁹ or that an agent’s desires must cohere with her values (Hubin 2003). More complex accounts hold that an agent’s desire should be able to survive reflective deliberation (Williams 1981) or require that one should obey the advice of one’s maximally rational counterpart (Smith 1995). The details of any one account need not detain us. In the present dialectical setting, the generic claim that the rationalising power of desires is subject to counterfactual constraints is enough to explain how counterfactual versions seek to explain fallibility. This generic claim is also enough to show how counterfactual views force an unwelcome revision to the intuitive datum.

Counterfactual views seem, at first glance, to explain why Ruth’s desire to drink paint should be rationally impotent. Her desire is entirely misleading about what reasons she has because it fails to meet a certain counterfactual constraint. Although Ruth desires to drink paint, she plausibly would not continue to do so in certain idealised conditions. Ruth’s desire to drink paint is, or ought to be, rationally impotent because the desire clashes with Ruth’s values, or would not survive reflective deliberation, and so on. If any of these proposals are plausible, it looks as if desire-based theories can avoid over-generating reasons and so explain how desires could be a fallible source of reasons. Let us now turn to consider some problems with this approach.

6. Problems with Counterfactual Versions

The first problem to mention is that it is not at all clear that counterfactual versions avoid over-generating reasons.³⁰ This would mean that counterfactual versions would still make basic desires oddly infallible. It seems possible to always adjust cases such that a bizarre basic desire meets any possible counterfactual constraint. Consider again the case in which Ruth desires to drink paint. We could tweak the case so that Ruth’s second-order desires all concern wantonly pursuing her first-order desires, or make it that she valued consuming non-food items over her health. This makes Ruth’s psychology quite unusual but, nevertheless, the sheer possibility of such bizarre reason-generating desires is an unwelcome result. I cannot see how counterfactual views could handle such cases in a principled manner other than biting the bullet. The idea would be that such cases are possible but, it just so happens, they are very unlikely. Counterfactual versions would be little better than simple versions of desire-based theories with respect to explaining our intuition about the fallibility of desires. I will not place too much weight on this objection. It might be possible to develop the details of a particular counterfactual constraint in a way that rules out such cases. Even if the counterfactual view avoids this problem, it faces another serious challenge.

Counterfactual views imply that an agent’s justification for her practical judgment cannot be recognised as conclusive.³¹ This, in turn, makes such views unable to explain the intuitive datum. On counterfactual views, whether a desire succeeds in generating a reason depends on that desire meeting certain counterfactual constraints. The success or failure of a desire in meeting such constraints is not something that is immediately evident from experiencing the desire or upon reflection on the experience of a desire. The desiring agent cannot know, just by reflecting on her desire, whether her desire generates a reason. So, an agent will need to have some standing reason to regard her desires as a reliable source of reasons for it to be epistemically responsible for that agent to take her desires at face-value as providing a reason. Perhaps the agent could acquire knowledge of reliability through reflection on her epistemic “track record” or perhaps this knowledge is something possessed by default.³² The problem is that the agent would not be rationally entitled to regard her desire as a source of conclusive justification. Although some of her desires guarantee reasons, she is not in the position to tell by reflection that a particular desire guarantees a reason. After all, the desiring agent is not entitled to view her desires as perfectly reliable. Rather, she can at best take a desire as raising the likelihood that she has a reason without thereby ruling out the possibility of error. Despite the relationship of constitution between desires and reasons, an agent cannot ever tell just by reflection on her desires that her justification is conclusive. So, Ruth would not be rationally entitled to her high level of confidence in her practical judgment, simply citing a desire could not end the regress of justification, and it would be irrational to be surprised if her desire misled. The counterfactual view forces significant revisions to the intuitive datum.

Of course, none of the objections I have considered falsify desire-based theories. For all I have said, there might be some ingenuous way for a desire-based theorist to resolve all of these issues. However, I hope I have placed enough prima facie pressure on the idea that the desire-based order of explanation can succeed in reconciling conclusive knowledge with fallibility to warrant exploring other options.

7. The Guise of the Good

Desire-based theories are attractive because the claim that desires generate reasons can account for how a desire could confer recognisably conclusive justification. Yet, this very commitment makes desire-based theories struggle to explain the fallibility of desires. Another way to capture the fallibility of desires begins with the thought that desires and reasons are independent. Among the most plausible ways of developing this thought are that reasons are based on value (Maguire 2016; Quinn 2012; Raz 2010) or that reasons are primitive (Scanlon 1998). For my discussion of the guise of the good, I will assume the truth of a value-based account of reasons. I will remain neutral on the precise metaphysical relationship between the two and I will not present an account of the behaviour or interaction between reasons. What is most important is simply the thought that desires and reasons can come apart.

The guise of the good explains how desires can make us aware of metaphysically independent reasons. A standard version of the guise of the good holds that a desire for P represents P as good in some respect (Alvarez 2010; Anscombe 1957; Boyle & Lavin 2010; Brewer 2009; Clark 2010; Hawkins 2008; Johnston 2001; Oddie 2005; 2009; 2015; 2017; 2018; Scanlon 1998; Schafer 2013; Smithies & Weiss 2019; Stampe 1987; Tenenbaum 1999; 2007; 2008; 2009; 2018).³³ The notion of value in this context is understood broadly to encompass moral, prudential, hedonic, and aesthetic value (Alvarez 2010: 85; Johnston 2001: 181–83; Orsi 2015: 715). On this view, a desire that successfully represents the value of P will, under the right circumstances, make the desiring agent aware of a reason to bring P about. This explains how an agent can immediately transition from a desire to a practical judgment. Furthermore, this view has a neat explanation of fallibility: desires sometimes simply misrepresent value. For example, Ruth’s desire to drink paint rationally counts for nothing because it simply gets her reasons wrong.³⁴ The guise of the good is attractive in our dialectical context because its explanation of fallibility is built into its explanation of how desires confer justification.

Let us consider how this view attempts to explain Long-Haul Flight. In this case, Ruth’s desire represents the prospect of relief as hedonically valuable. The goodness of the pleasure of relief provides Ruth with a reason to stretch and the evaluative content of Ruth’s desire shows her how relief speaks in favour of stretching. The way her desire casts stretching in a positive light, by representing the pleasure of relief, directly justifies her practical judgment and action.³⁵ Since the content of her desire is veridical, has the right aetiology, and Ruth’s epistemic conduct is appropriate, Ruth seems to be entitled to the knowledge that relief is a reason to stretch. This broad-strokes explanation of the intuitive datum is promising. However, I argue below that, despite its initial promise, extant versions of the guise of the good cannot explain the intuitive datum because they cannot capture the claim that Ruth’s justification is recognisably conclusive. For now, let us examine how this approach tries to explain the intuitive datum in more detail.

In Long-Haul Flight, Ruth’s practical judgment and action is immediately justified by her desire. Advocates of the guise of the good explain the immediacy of such justification in terms of the phenomenology of desire. A standard claim is that desires represent value in a manner analogous to perceptual experience (Hawkins 2008; Johnston 2001; Oddie 2005; Schafer 2013; Smithies & Weiss 2019; Stampe 1987).³⁶ The desired prospect will appear or seem good in a distinctively experiential way, quite unlike an evaluative belief.³⁷ In Long-Haul Flight, the prospect of stretching presents itself as good in an immediate, vivid, and epistemically forceful manner (Hawkins 2008: 262–63; Johnston 2001: 189–90, 205–13; Oddie 2005: 47–80; Schafer 2013: 276–78; Smithies & Weiss 2019: 38–40; Stampe 1987: 356–57). The phenomenology of desire suggests to Ruth that she is aware of the value of stretching. Consequently, Ruth’s desire rationally disposes her to believe that she has a reason to stretch. Taking her desire at face value is rational because her desire purports, by its experiential character, to be an awareness of value. Let us consider some important epistemic consequences of this analogy between desire and perception.

Consider first how the phenomenology of perception is involved in perceptual judgment. An agent who sees P is rationally inclined to form the belief that P obtains and can reflectively recognise that she has grounds to believe that P obtains just on the basis of her seeing P. By contrast, an agent who imagines P is not inclined to believe that P obtains and can reflectively recognise that she has no grounds, just on the basis of her imagining, to believe that P obtains. One plausible explanation of this contrast is that perception involves the phenomenology of P’s presenting itself as true (Chudnoff 2012; 2013: 173–203; Dorsch 2016; 2018; Frey 2013; McDowell 1995; 2008: 378–82; 2009; 2010; 2018; Schafer 2013; Sturgeon 1998: 181–82). The character of P’s presenting itself as true intimates to the agent that she is aware of P. A first-personal feature of perceptual experience—P’s presenting itself as true—justifies a judgment about the third-personal structure of experience—that the agent is perceptually aware of P. An agent’s judgment that she is perceives P is grounded in the phenomenology of her perceptual experience.

Advocates of the perceptual analogy make analogous claims about the phenomenology of desire (Hawkins 2008: 258–64; Johnston 2001; Oddie 2005: 47–80; Schafer 2013: 276–78; Smithies & Weiss 2019: 38–40; Stampe 1987: 356–57). An agent who desires P is rationally inclined to believe that P is good and can tell by reflection that her desire provides her with grounds to believe that P is good. By contrast, an agent who imagines that P is good is not provided with grounds to believe that P is good just from an introspective awareness of that imagining. Again, a plausible explanation of this asymmetry is that the passivity, immediacy, and epistemically forceful character of the phenomenology of desire intimates to the agent that she is aware of the value of P. A first-personal feature of Ruth’s desire—P’s presenting itself as good—provides grounds for a judgment about the third-personal structure of her experience—that she is aware of the value of P. Ruth’s transition from a desire to a practical judgment can simply be an endorsement of the way her desire makes things seem (Schafer 2013: 277). If we suppose that the evaluative content of Ruth’s desire is veridical, holds the right causal connection with value, and possesses the phenomenology of presenting the prospect of stretching as good, it looks like that should rationally entitle her to the immediate judgment that relief is a reason to stretch.

While this standard version of the guise of the good can explain the immediacy of Ruth’s knowledge of reasons, it is not clear that it can explain the conclusive character of Ruth’s justification. Recall that it is often assumed that for a source of experiential justification S to confer conclusive justification to judgments about a domain D, there must be a relationship of constitution between S and the elements of D. The guise of the good maintains that desires and reasons are metaphysically independent. A desire has the same fundamental metaphysical character when it is an awareness of a value and when it is not. This is often taken to mean that desires provide the same kind of reflectively accessible justification in both cases. I will call this position the common kind assumption.³⁸ Although not usually explicitly stated, the common kind assumption is part of the standard position held by advocates of the guise of the good. Advocates of this view frequently qualify their epistemic claims by insisting that the justification provided by desires is defeasible, prima facie, or otherwise inconclusive (Oddie 2005: 40–44, 54–58, 213–16; 2009: 132–34; 2015: 72–73; 2017: 29–30, 35; 2018: 241–42; Smithies & Weiss 2019: 33). Cases of epistemic success (the “good case”) and failure (the “bad case”) both provide at best inconclusive justification for practical judgment.

The common kind assumption is attractive because it seems like the only way to explain the fallibility of desires. More specifically, it seems to be required to explain the fact that an agent might mistakenly judge that she has a reason on the basis of a desire that is epistemically unsuccessful. I will call a desire that falls short of being an awareness of value an “illusion of value.” An agent who mistakenly judges that she has a reason on the basis of an illusion of value seems to be, at least in some minimal sense, just as rational as when she acquires knowledge of reasons on the basis of an awareness of value.³⁹ One explanation of this rationality is that an illusion of value might make things seem the same as an awareness of value. If an agent were to experience an illusion of value, then immediately transition into an awareness of value, she might not be able to introspectively distinguish between the two experiences.⁴⁰ The best explanation of these observations seems to be that both desires provide the very same kind of reflectively accessible justification for practical judgment.⁴¹

The common kind assumption renders the guise of the good unable to explain the intuitive datum.⁴² An important consequence of the assumption is that an agent can never tell, just by reflection on her desire, that she is aware of value. An agent would have the very same reflectively accessible grounds for the judgment that she is aware of value in both the good case and the bad. So, she could not take such grounds as sufficient for the truth of the judgment that she is aware of value. Ruth’s entitlement to an immediate, outright judgment that relief is a reason to stretch depends on her implicit recognition that her basis for judgment conclusively justifies. But if an agent could never tell, just by reflection on her desire, that she is aware of value, then she would never be in the epistemic position to recognise that her desire rules out the possibility of error. So, Ruth would only ever be entitled to a hedged judgment. We cannot vindicate Ruth’s epistemic conduct. On the common kind assumption, Ruth’s confidence in her judgment rationally ought to be lower, surprise in the event of error would be epistemically inappropriate, and she would not be entitled to cite her desire as a regress-stopper. The standard version of the guise of the good must deny, or at least significantly revise, the intuitive datum.

Of course, we may want to accept these revisions after all theoretical considerations have been surveyed. However, there are a couple of reasons to hesitate. First, such revisions put the guise of the good in a dialectically weak position vis-à-vis desire-based theories of reasons. If the guise of the good cannot explain the intuitive datum as it stands, desire-based theories can plausibly claim to be the best way, perhaps even the only way, to do justice to the connection between desires and reasons that seems to be evident in the intuitive datum. This would be a major explanatory advantage for desire-based theories. Advocates of the guise of the good should not concede this point before exploring all their options. Second, advocates of the guise of the good should prefer, all else being equal, a theory that can explain both our intuitions about fallibility and explain the intuitive datum as it stands. However, it is not obvious how they reconcile these intuitions. We seem to face an unpleasant trade-off: capturing the intuition that desires recognisably conclusively justify rules out capturing the intuition that desires are fallible, and vice versa. I will now propose an account of desire that promises to reconcile these intuitions.

8. An Epistemological Disjunctivist Version of the Guise of the Good

The common kind assumption forces us to deny, or significantly revise, the intuitive datum. I propose that we seriously consider rejecting the common kind assumption.⁴³ Instead, we could adopt an epistemological disjunctivist version of the guise of the good.⁴⁴ On such a view, desires come in two categorically distinct epistemic kinds.⁴⁵ The first kind of desire is an awareness of value. Such a desire provides conclusive justification for practical judgment. Furthermore, an awareness of value has a phenomenology that enables the desiring agent to tell by reflection that she is aware of value. The conclusive justification conferred by an awareness of value is reflectively accessible as conclusive justification, at least under the right conditions.⁴⁶ The second kind of desire is an illusion of value. Such a desire merely seems to provide conclusive justification. An illusion of value has a phenomenology that is liable to mislead the desiring agent into the mistaken judgment that she is aware of value. So, any given desire will belong to only one side of the following disjunction: either the desire is an awareness of value, or it is an illusion of value (hence the name “disjunctivism”). In sum, this epistemological disjunctivist version of the guise of the good is committed to three central theses.⁴⁷ First, it claims that desires come in two categorically distinct kinds. Second, an awareness of value confers conclusive justification on desire-based practical judgments and actions. Third, the conclusive justification conferred by an awareness of value is reflectively accessible as conclusive justification because of its phenomenology. These theses are supported by a background account of fallibility which I discuss below.

I will briefly indicate how this account promises to be the best of both worlds before elaborating in a bit more detail. The disjunctive account explains the intuitive datum like this. In Long-Haul Flight, Ruth enjoys an awareness of the value of stretching. An awareness of value consists in direct epistemic access to an evaluative fact. Furthermore, the justification conferred by an awareness of value is reflectively accessible as conclusive. If Ruth’s desire did indeed provide her with conclusive, reflectively accessible justification, she would be rationally entitled to an immediate, outright judgment that relief is a reason to stretch. Since her epistemic conduct is faultless and her epistemic terrain is favourable, her judgment amounts to knowledge. When Ruth feels a desire to drink paint, she is subject to a value-illusion.⁴⁸ Ruth’s desire misrepresents the value of drinking paint, and she is in the position to recognise her desire as misleading. Ruth is rationally entitled to regard her desire as counting for nothing. The disjunctive account delivers satisfying, non-revisionist explanations of both cases and the intuitions behind them.

The disjunctive account might seem to present a puzzle. How is it possible for desires to have two distinct epistemic values? In short, the explanation is that all desires issue from a more basic capacity to discriminate value.⁴⁹ This capacity can be successfully exercised and issue in an awareness of value, or it can be defectively exercised and issue in an illusion of value. Although both types of desire might share the same content, they acquire their content from distinct types of processes, and each type of desire places the agent in distinct types of epistemic relations to value.⁵⁰ In the good case, a successful discrimination of value results in evaluative content that enables a relation of awareness to value. This capacity-based view of desire explains why it is consistent to say that desires have categorically distinct epistemic values while belonging to the same metaphysical kind. There is a metaphysical difference which explains the epistemic difference, but it is not to be located in the content of desires, but rather in a broader account of the genesis of desires and the relational structure of awareness.

An important epistemic consequence of the capacity-based view is that we do not need to claim that desires all share the same epistemic value because the possibility of error can be explained in terms of a more basic capacity to discriminate. This capacity-level conception of fallibility is crucial for the very intelligibility of the disjunctive account.⁵¹ Accordingly, I will elaborate on the explanation of fallibility before explaining how an awareness of value provides recognisably conclusive justification.

9. Fallibility

As we saw above, the thought that S is an experiential source of conclusive justification is naturally opposed by the thought that S is fallible. I want to suggest that these two thoughts do not need to be understood in a way that conflicts. The thoughts conflict when S is viewed as needing to simultaneously rule out the possibility of error and be liable to error. The thoughts can be harmonised by saying that S₁ rules out the possibility of error and S₂ is liable to error. Let us consider how this plays out for the case of desire.

The motivation for the common kind assumption is that it explains our intuitions about fallibility. It allows us to say that a desiring agent is liable to error because all her desires are always liable to error. All desires must “accommodate” the possibility of error. So, all desires at best provide inconclusive justification to allow for cases in which things go wrong. The disjunctive account denies that all desires must be liable to error. Rather, the possibility that things can go wrong is explained in terms of a more basic capacity for discrimination.⁵² Fallibility is located one step further back, so to speak. The idea is that desires issue from a capacity to discriminate the value of a state of affairs.⁵³ Such a capacity can be successfully exercised or unsuccessfully exercised. This capacity is liable to error because its defective exercise results in an illusion of value. Since fallibility has been explained at the level of a capacity for discrimination, there is no need to say that all desires are liable to error. An awareness of value does not need to “accommodate” any liability to error and so can guarantee the truth of a corresponding practical judgment.

The capacity-level explanation of fallibility has important consequences for how an agent is rationally entitled to regard her desires. The common kind assumption implied that an agent rationally ought to regard all desires as potentially misleading, so her reflectively accessible justification is at best inconclusive. On the disjunctive account, it is false that all desires are potentially misleading. An awareness of value, by its very nature, cannot be in error. The desiring agent is rationally entitled to regard an awareness of value as providing conclusive justification for a practical judgment. Of course, a variety of conditions must be satisfied for the agent to be so entitled. Her desire must be an awareness of value, she must exhibit epistemically virtuous conduct in the way that she bases her judgments and actions on her awareness of value, and she must occupy a broader epistemic context that is conducive to knowledge. If these conditions are met, an agent is not under any obligation to regard her desire as providing anything less than conclusive justification. This account of rational entitlement is crucial to capturing the intuitive datum. It allows us to say that, under the right circumstances, an agent can be perfectly rational in regarding her desire as simply settling the question of whether she has a reason to bring P about.

Of course, several factors can interfere with an agent’s attempt to acquire knowledge on the basis of her desire, even when she enjoys an awareness of value. Perhaps the agent has, or seems to have, higher-order defeaters concerning her own epistemic conduct or her broader epistemic context.⁵⁴ For example, Ruth might mistakenly believe that she is subject to an illusion of value when she enjoys an awareness of value. In such a case, Ruth would be rationally required to discount her awareness of value as counting for nothing. Nevertheless, Ruth’s awareness of value continues to provide first-order justification for a corresponding practical judgment, regardless of her apparent reasons to form the higher-order judgment that she is subject to an illusion of value. Ruth would be rationally disbarred from actually basing a judgment on her awareness of value due to higher-order considerations. So, the capacity-based conception of fallibility vindicates our intuitions about the conditions under which agents fail to acquire knowledge due to higher-order defeaters, despite epistemic success at the level of desires.

This alternative capacity-based conception of fallibility puts us on the right path to explaining the intuitive datum. We want to say that Ruth’s grounds for judgment are sometimes recognisably conclusive, like in Long-Haul Flight. At the same time, we want to allow for the possibility that a desire could be entirely wrong about an agent’s reasons. We can explain the agent’s fallibility in terms of a defective exercise of a capacity to discriminate value, while also maintaining that an awareness of value provides conclusive justification. The next step is to explain how agents know that they are in the good case.

10. Conclusive Justification

The disjunctive account maintains that an agent can tell by reflection on her desire that she enjoys an awareness of value. Yet, it also maintains that an illusion of value can be mistaken for an awareness of value. How could it be possible for an agent to reflectively recognise that she is in the good case, if she can be mistaken in the bad case?

There are at least two ways of understanding this question.⁵⁵ I will call them the “comparative” and “non-comparative” versions. The comparative version means to ask how an agent can tell between an awareness of value and an illusion of value. Suppose an agent introspectively considers two desires, D₁ and D₂.⁵⁶ She then compares them in the attempt to “sort” each desire into the “awareness of value” basket or the “illusion of value” basket. The non-comparative version asks how an agent can succeed in telling whether this desire is an awareness of value. Suppose she introspects on an occurrent, or recently occurrent, desire and tries to reach a judgment about its nature. The disjunctive account must concede that in many cases an agent simply cannot tell between an awareness of value and an illusion of value.⁵⁷ There is no introspectively available “marker” that indicates which is which. Usually, when an agent is successful in telling between the two, she relies on background knowledge about her broader epistemic context and her own psychological dispositions. But epistemological disjunctivism about desire does not need to say that agents are always able to tell between an awareness of value and an illusion of value.⁵⁸ By analogy, it is consistent to say that an agent may not be able to visually tell between (say) a zebra and a cleverly painted mule, while maintaining that she can sometimes straightaway know that she sees a zebra (see McDowell 2008: 387; Pritchard 2012: 78–81).⁵⁹ In our context, the occasional inability to tell between an awareness of value and an illusion of value does not mean that an agent can never straightaway tell that she is aware of value.

So, how could an agent tell, in a non-comparative way, that she enjoys an awareness of value? The disjunctive account says that this knowledge is based on the phenomenology of desire.⁶⁰ As discussed above, the phenomenology of desire makes it seem to the agent that she is aware of value. A subjective feature of experience—P’s presenting itself as good—justifies a judgment about the third-personal structure of experience—that one is aware of value. The phenomenology of desire makes a “claim” about its epistemic status as an awareness of value. The “claim” of the phenomenology of an awareness of value has a distinct epistemic significance, compared to the “claim” of an illusion of value.⁶¹ In the bad case, the phenomenology of desire misleadingly suggests that the agent is aware of value when she is not. However, what matters to an agent’s epistemic position is not merely how things seem to her. Rather, it also matters how things are. In the good case, the agent does not pass a judgment of her epistemic success on the way things merely seem to her in the good case. Rather, she judges that she is in the good case on the basis of her being in the good case. Consequently, an awareness of value provides a kind of knowledge-conducive evidence for the judgment that one is aware of value that an illusion of value simply cannot.

An awareness of value can immediately lead to knowledge that one is aware of value when the agent has no reason to believe that her epistemic conduct is defective, or her epistemic terrain is unfavourable. Insofar as she is in the good case, and no higher-order defeaters are in play, an agent can simply tell by reflection that she enjoys an awareness of value. The fact that an agent can get things wrong in the bad case does not rule out her ability to succeed in the good case.⁶² It is consistent to maintain both that agents sometimes mistake the bad case for the good case and that in each case the agent is provided with different introspectively available grounds for judgment. That an agent might confuse one for the other does not mean they are the same. After all, agents are fallible in their attempts at self-knowledge (see McDowell 1998; 2010; 2013b; 2018).

However, there might seem to be a deeper problem.⁶³ When an agent mistakes the bad case for the good case, she will judge that she has a reason with the same rational confidence she would have in the good case. Disjunctivism accepts that this parity is perfectly legitimate: in the bad case agents often are, in some sense, equally entitled to form a belief on the basis of merely apparent grounds (see McDowell 2002; Pritchard 2012; Shaw 2017). Agents are, in some sense, rational in having the same level of confidence in both cases. The problem is that it seems plausible to suppose that an agent’s rational confidence ought to be proportionate to the strength of an agent’s reflectively accessible justification.⁶⁴ Either the disjunctivist accepts or rejects this principle. Rejecting the principle is implausible. Intuitively, any change in the strength of one’s reflectively accessible justification warrants a change in rational confidence. Yet, it looks like accepting the principle has implausible results too. According to disjunctivism, the agent’s reflectively accessible justification differs across the good case and the bad, so accepting the principle generates the claim that an agent in the bad case ought to know better than to have a high level of confidence. At first glance, neither accepting nor rejecting the principle is attractive. This seems to be a significant strike against the plausibility of the disjunctive conception.

The disjunctivist can hold to the principle of proportionality without making implausible demands about the agent’s epistemic requirements in the bad case. The disjunctivist reply is that agents in the bad case cannot respect the principle of proportionality, yet agents cannot know better than to judge on merely apparent evidence because agents are “epistemically doomed” in the bad case (Shaw 2017: 112–13, 117). Agents are not epistemically vicious for failing to bring their epistemic conduct in line with the principle of proportionality (McDowell 2002: 99; Pritchard 2012: 42; Shaw 2017). An agent in the bad case might respond to her apparent evidence with all of the intellectual virtues we would expect from an epistemically responsible agent (Shaw 2017). Since the agent is epistemically doomed in the bad case, she does not have any reflectively accessible grounds to suspect that anything is amiss. If the agent’s rational confidence were lowered in the bad case (when she had no reason to suspect anything was wrong) she would be acting irrationally by taking herself to be subject to defeaters without any reflectively accessible grounds for doing so. So, while an agent is violating an epistemic norm by having a high rational confidence in the bad case, the only ways available to her to abide by this norm are all epistemically vicious. It is consistent with the disjunctive account to maintain that an agent who takes herself to have conclusive justification in the bad case is nevertheless exercising (perhaps defectively) her epistemic virtues despite violating the principle of proportionality. Consequently, the disjunctivist account does not generate implausible demands for the epistemic conduct of agents in the bad case.

Let us finally revisit, in light of our more detailed discussion, exactly how the disjunctive account captures both the intuitive datum and the fallibility of desires. In Long-Haul Flight, Ruth is in the good case. Ruth’s epistemic conduct is faultless, her epistemic terrain is favourable, and she has not been misled into believing that any higher-order defeaters are at play. Ruth’s desire presents the prospect of stretching as good in a way which intimates that she is aware of the value of stretching. Ruth is rationally entitled to immediately transition from her desire to the outright judgment that she has a reason because her desire provides reflectively accessible conclusive justification for that judgment. The epistemic authority of her desire also explains why she is entitled to judge with a high level of confidence, it explains why she would be entitled to cite her desire as a regress-stopper, and also why Ruth would regard her desire as misleading in the event of error. Ruth has, epistemically speaking, everything she needs to outright judge that she possesses a reason to stretch just from her desire. By contrast, when Ruth desires to drink a can of paint she is in the bad case. Ruth’s capacity to discriminate value has misfired and produced an illusory experience of value. Since Ruth’s desire is an illusion of value, she rationally ought to discount her desire in light of the known undesirability of drinking paint. The disjunctive account can capture the intuition that Ruth’s desire to drink paint rationally ought to count for nothing. The disjunctive account delivers a natural, satisfying, and non-revisionist explanation of both cases. Moreover, the resources used to capture one intuition are perfectly compatible with those used to capture the other intuition.

11. Conclusion

I have proposed an account of desire that captures two important intuitions about practical agency. I began with the intuitive datum that often when an agent desires P she will seem to immediately and conclusively know that she has a reason to bring P about. Desire-based theories of reasons provide a natural and compelling explanation of the thought that desires provide recognisably conclusive justification. A perennial problem for the claim that S is an experiential source of conclusive justification is to show how S is also fallible. Desire-based theories struggle to do this. They make basic desires an oddly infallible source of reasons or explain fallibility by revising the intuitive datum. These problems can be avoided by saying that desires and reasons are independent. The guise of the good explains how desires could provide epistemic access to such metaphysically independent reasons. It also has a neat explanation of fallibility: desires sometimes simply misrepresent one’s reasons. However, standard accounts are developed in terms of the common kind assumption which leads to unwelcome revisions to the intuitive datum. I proposed an epistemological disjunctivist account that denies the common kind assumption. On this view, a desire is either an awareness of value or an illusion of value. The intuitive datum involves an awareness of value which provides reflectively accessible conclusive justification for a practical judgment. Desires are fallible because an exercise of the capacity to discriminate value might misfire, and result in an illusion of value. This view gives us the best of both worlds. The explanation of the intuitive datum is perfectly consistent with the explanation of fallibility. Of course, my argument has not established the truth of the disjunctive account. It does, however, motivate us to take the view seriously and explore its prospect further.