Loving truly: An epistemic approach to the doxastic norms of love

If you love someone, is it good to believe better of her than epistemic norms allow? The partiality view says that it is: love, on this view, issues norms of belief that clash with epistemic norms. The partiality view is supposedly supported by an analogy between beliefs and actions, by the phenomenology of love, and by the idea that love commits us to the loved one’s good character. I argue that the partiality view is false, and defend what I call the epistemic view. On the epistemic view, love also issues norms of belief. But these say simply (and perhaps surprisingly) that you should adhere to epistemic norms in forming and maintaining beliefs about loved ones. I offer two arguments for the epistemic view. The first appeals to the emotional responses of love, which, when sensitive to what the loved one is really like, can make love great and be morally transformative. The second is a new argument for why caring for a loved one requires true beliefs about him. We see that there may be some boundaries, such as stuffy traditions, that love is right to defy, but that epistemic boundaries are not among them.


Introduction
Evidence is mounting that Queen Guenever, the wife of King Arthur, is having an affair with Sir Launcelot. The only person in Camelot who still believes that Guenever is faithful, short of King Arthur himself, is the loyal knight Sir Gawain. Gawain even refuses to believe ill of Guenever when a posse of knights ambushes her and Launcelot in her bedchamber. Instead, Gawain forms the charitable belief that Guenever must have summoned Launcelot to thank him for rescuing her so often; the privacy of her bedchamber is advantageous given the court's propensity to gossip.

3
Unfortunately, Guenever's tactful plan backfired, and she now seems guiltier than ever. Gawain comforts King Arthur by reasoning that "ofttimes we do many things that we [think] it be for the best, and yet peradventure it turneth to the worst. For I dare say…my lady, your queen, is to you both good and true" (Malory, 1485, Book XX, Chapter 7).
It is a common trope that excellent love, be it for a friend, family member, romantic partner, or someone else, involves thinking well of your loved one-indeed better than epistemic norms permit. Gawain's naiveté is exaggerated, but some philosophers defend a more realistic version of this idea. According to the doxastic-partiality view, love comes with a norm governing the beliefs that you form about loved ones. This norm says that it is good, from the viewpoint of love, when these beliefs are positive rather than negative, even if they violate epistemic norms (Keller, 2004;Stroud, 2006;Hazlett, 2013). For example, the person who loves well takes her 1 loved one's actions to be motivated by kindness rather than cynicism, even given strong evidence to the contrary.
Several considerations favor the partiality view. First is the claim that love commits you to a positive view of the loved one (Stroud, 2006). Second, believing partially about loved ones goes along with treating them partially (Hazlett, 2013). Third, we supposedly want to believe well of loved ones and for them to believe well of us (Keller, 2004;Stroud, 2006;Hazlett, 2013). Finally, since love is allegedly based on loved ones' qualities, inflated beliefs about those qualities can safeguard it (Stroud, 2006).
But I argue that the doxastic-partiality view is false: love does not issue norms of belief that conflict with epistemic norms. This is not to deny that love issues norms of belief. I argue that it does-but that they are standard epistemic norms promoting accurate belief. That is, from the standpoint not only of epistemology but of love too, it is good to obey epistemic norms when forming beliefs about a loved one-especially, I'll argue, when you do so motivated by a concern to believe accurately about them. The reason is that doing this promotes an accurate view of your loved one, and an accurate view-even including negative beliefs-in turn promotes excellent love. I call this the epistemic view of excellent love.
I offer two arguments for the epistemic view. First, love comes with emotional responses, and these make love great to the extent that they fit the way your loved one really is. Fitting responses, however, require accurate beliefs. Second, we care for those we love-but to do so well, we need accurate beliefs about them.
These arguments are situated in an ongoing discussion. Kawall (2013) and Arpaly and Brinkerhoff (2018) criticize doxastic partiality on similar grounds to me, but we'll see that their arguments do not go far enough. Crawford (2019) argues that excellent love (or friendship) involves responding to the other person's "perceived features" (1595). But rather than abandon doxastic partiality, she remains open to a weak version that I criticize too. Mason (2020) draws on (Murdoch, 1970) to develop a view of love's nature that is congenial to my view of its normativity.
1 3 opportunity for secure attachment and vulnerability. Another is that excellent love involves caring for the other party; this is a matter of facilitating his flourishing, at least to the extent that he welcomes this.
Section 2 presents the partiality view and Sect. 3 presents the epistemic view. Section 4 defends the epistemic view by appealing to the emotional responses proper to love, and Sect. 5 does so by appeal to our care for loved ones. Section 6 argues, by addressing an objection, that adhering to epistemic norms promotes accurate beliefs in love. Section 7 addresses the arguments for the partiality view, and Sect. 8 concludes.

Doxastic partiality
The idea that love issues norms of belief may seem strange. For beliefs have norms of their own, epistemic ones. Epistemic norms often support positive beliefs about loved ones. But sometimes they support negative beliefs-after all, everyone typically manifests at least some negative characteristics.
So in a world like ours, the partiality view is headed for a conflict with epistemic norms. When epistemic norms point toward negative beliefs about a loved one, the normativity of love supposedly points the other way. Here is a summary of the partiality view: The Partiality View: An excellent-making feature of love is (i) the forming and maintaining of positive, or at least minimally negative, beliefs about your loved one, even if this requires violating epistemic norms; where (ii) your love is a significant factor motivating such beliefs.
The normativity here is not prescriptive: love does not require you do anything. It is rather evaluative: it concerns what makes a given case of love excellent. This distinction matters, because the prescriptive claim is stronger than the evaluative one. The claim that love requires you to form positive beliefs entails that forming positive beliefs makes a given case of love excellent, but the reverse does not hold: that forming positive beliefs makes love excellent does not entail that love requires you to form them. After all, there may be cases where two things that make love excellent cannot be jointly realized-and there is little reason to think that love would require one at the expense of the other. Proponents of the partiality view do not always specify whether their version is prescriptive or evaluative. 4 I will argue against the evaluative version of the partiality view, and thus a fortiori against the prescriptive one. For I want to show that love issues no normative pressure whatsoever to defy epistemic norms-and that it issues normative pressure, instead, to obey them.
Here are some characteristic statements of the partiality view (italics added): 5 Friendship places demands not just on our feelings or our motivations, but on our beliefs and our methods of forming beliefs….
[T]his epistemic partiality is contrary to the standards of epistemic responsibility and justification held up by mainstream epistemological theories. (Stroud, 2006, 499) [F]riendship requires (being disposed towards) an unreliable doxastic practice, namely, what I have called "partiality bias." (Hazlett, 2013, 97) [S]ometimes, being a good friend can mean failing to be good in another important respect; it can mean failing to be a good believer. Sometimes, to be a good friend, you need to compromise your epistemic integrity. (Keller, 2007, 25) [T]he specific differential practices and beliefs the good friend adopts … seem to lead her into a distorted conception of reality….
[F]riendship involves, if not a blind spot, at least less than perfect vision where your friends' sins and flaws are concerned; the good friend's set of beliefs is to that extent necessarily out of kilter. As a good friend, your belief set is slanted: you actually believe your own spin. (Stroud, 2006, 513) Although the partiality view lauds epistemic violations, it is not draconian. Love's supposed victory over epistemic norms is not winner-takes-all. Epistemic violation admits of degree, and the partiality view is compatible with greater or lesser ones. For example, if you saw your loved one do something apparently rude, excellent love need not involve flat-out disbelieving this. You might instead regard instances of rudeness as minor in your "overall portrait" of her character (Stroud, 2006, 508), or give high but not certain credence to a more charitable explanation than epistemic norms allow. The partiality norm leaves room for negotiation, as long as you flout epistemic norms somewhat. The extent of their flouting depends on which version of the partiality view is operative. On Stroud's version (2006), the clash between love and epistemic norms is strong; those who love well flout epistemic norms frequently and egregiously. To the extent that the other person displays negative qualities, "the good friend's set of beliefs is necessarily out of kilter" (513, emphasis added). On Keller's view (2004), by contrast, the situations in which excellent love involves epistemic violation are occasional and less egregious; he says that "sometimes, the norms of friendship clash with epistemic norms" (330). Hazlett (2013) is in the middle. On the one hand he says that partiality does not apply when evidence against your beloved is unmistakable (92); on the other hand he insists that excellent love is doxastically unreliable (97, 102). 6 1 3 One argument for the partiality view draws on the idea that love (or friendship) involves commitment, and that this includes commitment to the goodness of the other person's character (Stroud, 2006, 511-512).
A second consideration draws an analogy to ethics. We should act partially toward loved ones, for example by helping them in ways that we would not help strangers. Hazlett claims that there is "no reason to expect" belief to be any different (2013,93).
A third argument concerns phenomenology. When we love someone, we strongly incline to believe good of him, and supposedly feel anything else to be disloyal (Stroud, 2006, 503-504). And when someone loves us, we supposedly want her to believe well of us, even despite evidence to the contrary (Keller, 2004, 338;Hazlett, 2013, 90).
A fourth argument comes from the quality view of love. If love is based on your beloved's positive characteristics, then positive beliefs are crucial to preserving it (Stroud, 2006, 511).
A fifth argument appeals to thought experiments. In a case from (Stroud, 2006), elaborated by (Crawford, 2019), a third party says that your friend Reggie led someone on and knowingly broke her heart. Stroud's intuition is that you are disloyal if you do not resist believing this supposed slandering of Reggie. In a case from (Keller, 2004), Eric is a poetry connoisseur attending a reading by his friend, Rebecca. He is unfamiliar with her poetry, but knows that most poems read at this venue are substandard. So if Rebecca were not his friend he would expect hers to be poor. But Keller's intuition is that the good friend Eric will do his best to believe well of her poem.
I'll address these arguments in Sect. 7. For now a comment on the thought experiments. Many authors have pointed out that they can be read in epistemically benign ways (Jollimore, 2011, chapter 3;Kawall, 2013;Hawley, 2014;Piller, 2016;Arpaly & Brinkerhoff, 2018;Goldberg, 2019;Crawford, 2019). One way is to construe epistemic norms as permissive. That is, a given body of evidence does not license merely one doxastic attitude (e.g., a belief or credence) toward a specific proposition; rather, it may allow a limited range of mutually incompatible ones. A lover could obey epistemic norms, forming the most positive belief that they permit (Kawall, 2013;Hawley, 2014;Piller, 2016).
However, permissive epistemic norms are neither necessary nor sufficient for deflecting the partiality view. They are not necessary because, even if permissivism is false, so that each body of evidence licenses only one doxastic attitude toward a given proposition, that attitude might be a charitable one (Goldberg, 2019). And permissive norms are not sufficient. First, they concern only what you do with evidence that you have, saying nothing about how you gather new evidence. Yet doxastic partialists are equally adamant that the good lover's evidence-gathering also facilitates positive beliefs (Goldberg, 2019). Second, epistemic norms can be reliabilist or virtuetheoretic, not involving evidence evaluation at all. 7 Permissive epistemology will not counter the partiality view.
But there are less theoretically laden ways to read such cases as respecting epistemic norms. First, one might argue that lovers simply have more evidence about their loved ones' positive qualities than others do (Jollimore, 2011, chapter 3). Second, certain forms of doxastic partiality are epistemically acceptable. In the heartbreaking case about Reggie, there is nothing wrong with scrutinizing the storyteller's honesty. In the poetry case, it is epistemically good for Eric to listen for signs of literary quality (Jollimore, 2011, chapter 3;Kawall, 2013). What is epistemically crucial is that the protagonists scrupulously avoid bias or wishful thinking.
But that these conciliatory suggestions would leave many doxastic partialists cold. As we saw, they insist that unreliability is inevitable (Hazlett, 2013, 97, 102) or distortion necessary (Stroud, 2006, 513). 8 The idea seems to be that love is expressed by defying certain normative restrictions (Baron, 1991, 856). If a system of norms says "Don't", then great love-which is "unto itself a higher law" 9 -often says "I'll prove my love by doing it". Restricting positive beliefs to those which obey epistemic rules misses the point.
I'll now present and defend the epistemic view.

The epistemic view
The epistemic view agrees with the partiality view that love issues a norm of belief, but disagrees about what this norm is. On the epistemic view, love is not made excellent by violating epistemic norms, but by adhering to them: 10 The Epistemic View: An excellent-making feature of love is (i) adherence to epistemic norms in forming and maintaining beliefs about your loved one, even if this requires believing negatively about her; where (ii) your love is a significant factor motivating you to form accurate beliefs on important matters concerning your beloved.
Concerning clause (i), the epistemic norms might be internalist or externalist, as we'll see. Clause (ii) specifies that love-rather than, say, intellectual pride-plays an important role in motivating your beliefs. As for important matters, these include your loved one's character, what she cares about, certain significant actions of hers, and so forth.
The epistemic view is diametrically opposed to the partiality view. The partiality view declares the excellence of love-motivated positive beliefs, even if these violate epistemic norms, whereas the epistemic view extols love-motivated adherence to epistemic norms even this yields negative beliefs.

3
The epistemic view, like the partiality view, is normative in the evaluative sense: adhering to epistemic norms is not required, but rather excellent-making. This leaves open the possibility of situations where the all-things-considered best option, from the viewpoint of excellent love, is to decline to adhere to epistemic norms in forming beliefs about a loved one. What the epistemic view insists on, however, is that the love in such a case is not perfectly excellent.
The epistemic view might seem strange. After all, excellent lovers do form beliefs about loved ones differently than about others, and there would be something wrong with them if they didn't (Marušić, 2015). But the epistemic view takes this difference on board-it just accounts for it differently than the partiality view. To see how, consider clause (ii) of the epistemic view. This says that your love is a significant factor motivating you to form beliefs in the way it deems excellent. This is already one way in which you form beliefs about loved ones differently than about third parties: you are motivated by love.
But what difference does love make? One is that, because you care more, you pay closer attention (Jollimore, 2011;Pettigrove 2012). This is not a self-interested care (as with surveillance of an enemy), but a tender, other-directed care (see Sect. 5). The partiality view agrees. But whereas the partiality view says that the excellent lover pays attention to the loved one's positive features, the epistemic view says that he pays attention to the loved one-irrespective of whether the latter's features are positive or negative. In other words, the partiality view says that you avoid allowing her negative features to affect your beliefs, whereas the epistemic view says that you avoid allowing her negative features, which you notice because you notice her, to affect your love for her.
What about the fact that lovers tend to be biased to perceive the positive and overlook the negative in their loved ones? The partiality view zeroes in on this descriptive observation to fashion its normative account. The epistemic view takes a broader lens. However accuracy-conducive love's natural inclinations happen to be, love is excellent to the extent that we strive for greater accuracy yet. Consider an analogy. A child may be naturally inclined, or not, to share their toys. Either way, we teach them moral excellence by encouraging the virtue of generosity. This is not to say that love's natural inclinations to focus on the good are bad from the viewpoint of love. It is to say that this is not the whole story-that this inclination, perhaps thoroughly laudable, must be complemented with the inclination to know, and to love, the other party as she really is.
These considerations can help us clarify what sort of epistemic norms the epistemic view is committed to: norms that promote accuracy. We may debate which sorts of norms these are specifically (see Sect. 6)-for example whether they are internalist (focusing on evidence available from the agent's perspective) or externalist (focusing on the reliability of the agent's belief-forming processes). But to preempt misunderstanding, a word about one thing that they are not. To the extent that the epistemic norms do emphasize evidence, they must, to promote accuracy, take account of the special sorts of evidence that we get about loved ones. This includes experiences that we have of loved ones to which other people are not party, and things that we notice that others, because they lack the biased perception of love, overlook (Jollimore, 2011;Pettigrove, 2012). Any epistemology that neglects this sort of evidence in favor of merely third-personal considerations is not apt to be accuracy-conducive. 11 Two other accounts bear some resemblance to the epistemic view. 12 First, Mason's (2020) "Murdochian" account (cf. Murdoch, 1970) says that loving someone entails knowing important things about her. But whereas the Murdochian view concerns love's nature (descriptively), the epistemic view concerns love's normativity. And the Murdochian view focuses on having knowledge of the loved one, whereas the epistemic view focuses on adhering to epistemic norms out of a care for such knowledge.
Second, the "vision view" construes love as a way of perceiving your loved one that involves paying special, devoted, attention to him (Jollimore, 2011 chapter 3;Pettigrove, 2012, chapter 5). This doxastic practice of love, far from detracting from accuracy, helps you accurately perceive features of your beloved that others cannot. Yet this practice does not perfectly promote accuracy, says the vision view, since loving perception does risk distortion (Jollimore, 2011, 72). My main point of contention with the vision view concerns cases in which love's characteristic belief-forming practices, though often accurate, go astray. The vision view attributes such cases to inevitable imperfections in love's otherwise sound practices. The epistemic view, in contrast, says that excellent love strives to avoid errors altogether. Excellent love is epistemically perfectionist: to the extent that the natural belief-forming process of love mislead, excellent love seeks to correct them.
Here is a summary of my argument for the epistemic view: The Argument from Accuracy 1. Accurate beliefs (on important matters) about those we love are an excellentmaking feature of love. (The Accuracy Norm of Love) 2. Abiding by the epistemic view-that is, forming beliefs about loved ones by (i) adhering to epistemic norms, where (ii) our love is a significant factor motivating us to form accurate beliefs on important matters concerning loved ones-is conducive to accurate beliefs about those we love. (The Accuracy-Conduciveness Claim) 3. Therefore, the epistemic view holds: forming beliefs about loved ones by (i) adhering to epistemic norms, where (ii) our love is a significant factor motivating us to form accurate beliefs on important matters concerning loved ones, is an excellent-making feature of love.
Consider premise 1, the Accuracy Norm of Love. Three characteristics make beliefs accurate. First, an individual belief is accurate only if it is true, and a belief set is accu-11 What about permissive epistemology (discussed in the previous section)? I suspect that this can be accuracy-conducive, as long as the range of acceptable doxastic attitudes given a particular body of evidence is not too wide. 12 Crawford too (2019) might superficially be read as endorsing the epistemic view. But she declines to endorse clause (i), while arguing against it (1590)(1591)(1592)(1593). She leaves open the possibility that friendship might be made excellent by defying epistemic norms. Crawford does, however, endorse clause (ii), on the grounds that lacking at least the motivation to believe accurately about your friends would amount to "inauthentic engagement" with them (1589). rate only to the extent that it contains true beliefs. But accuracy is more than truth. Second, a belief set is accurate to the extent that it is representative-it does not distort the big picture by omitting important information. Third, a belief set is accurate to the extent that the beliefs within it coherently connect in a web of understanding. 13 And accurate beliefs, says premise 2, tend to result when you abide by the epistemic view. I thus defend the epistemic view on instrumental grounds: abiding by it is accuracy-conducive and this makes love excellent. 14 The bulk of this paper argues for the Accuracy Norm of Love, premise 1 (Sects. 4 and 5). A briefer argument for the Accuracy-Conduciveness Claim, premise 2, comes in Sect. 6.

Love's emotional responses
My first argument for the Accuracy Norm turns on the emotional responses proper to love. These make for excellent love, I will argue, to the extent that they fit the way the loved one really is. 15 If emotional responses are to fit the loved one, the lover typically needs accurate beliefs about him. Whether this is because beliefs partly constitute emotions, or simply because they contribute causally to them, accurate beliefs about a person strongly promote fitting emotional responses-which in turn, I argue, make love excellent.
I'll assume that emotions can fit or fail to fit their objects. And I will restrict discussion to cases in which the object of an emotion-here, your loved one or his features, including his actions-brings about that emotion in a causally straightforward way.
I'll suppose further that a person's features can be positively or negatively valenced. This valence might be moral, but need not be-it could also (for instance) be aesthetic. I will focus for simplicity on features with mind-independent valences, such as kindness or cruelty, as opposed to subjective ones, such as finding a habit annoying or endearing. 16 There can be multiple, even incompatible, fitting responses to a given feature, depending on the context and the person doing the responding. For example, if your loved one betrays you, anger might be fitting, but so might be compassion on her circumstances (Pettigrove, 2021).
I'll focus on two types of fitting emotional response. A fitting type of response to positive features is delight, which I will use as an umbrella term for a range of positive emotions from joy to fond amusement. Yet most loved ones have negative features too, and I will use the umbrella term grief to refer to responses ranging from minor dismay or irritation to inconsolable sorrow. For minor flaws, or flaws that are not the loved one's fault (such as those resulting from childhood trauma), other responses than grief, such as compassion, or learning to ignore them, may be fitting. I consider only cases in which grief is fitting. Both delight and grief are directed not just at the feature, but also at the beloved's having it.
One might agree that delight is proper to excellent love, but object that grief is selfish and so diverges from excellence. This may sometimes hold. But there is also a pure kind of grief, arising from your desire to enjoy relationship with your beloved (Stump, 2010), to identify with her (Jollimore, 2011), and for her to be the best version of herself (Kawall, 2013); negative features can impede these aims. As Velleman notes, the emotional responses of love "need not be exclusively favorable. Love also lays us open to feeling hurt, anger, resentment, and even hate" (1999,361).
Let's consider fitting delight and fitting grief in turn. Fitting delight, I'll argue, makes love excellent for the lover and loved one alike. To see this for the lover, imagine a situation in which much of his delight responds to unrealistically rosy beliefs about his beloved. To the extent that his beliefs miss their target, so will his emotional responses. In the extreme case his attitude might not even count as love, but rather idolatry: directing your love at something other than the loved one herself (Jollimore, 2011, 47;Kawall, 2013, 361;Mason, 2020). 17 By contrast, when the lover's view of his loved one is accurate, he can delight in her positive features even more because they (and she) are real as he envisions them. 18 Now consider the loved one. There are two reasons why it is good to be delighted in for who one is and for the features that one has, instead of in response to exaggerated or invented beliefs. First, as Jollimore (2011) argues, people generally deserve to be celebrated for the unique value that they have; if your loved one is instead falsely idolized, he misses out on what he deserves. Second, being idolized instead of delighted in for who one is courts the imposter syndrome, in which a person fears that she is not as good as people think she is, and that her social or professional life will collapse if they realize it. A love in which delight is exaggerated in response to exaggerated beliefs sets the loved one up for this-indeed in the very place that should be emotionally the safest. If delight is to make love excellent, it must respond to the loved one as she is.
The same holds for grief over negative features. Love is made excellent when the lover has accurate beliefs about, and fitting responses of grief to, his beloved's negative features-and loves her anyway. (At least, this holds when the loved one is not abusive to the lover; I make no claims about such cases.) Kawall defends a similar claim ( , 356-357, cf. Mason 2020. He says that.
among the elements of friendship are such things as acceptance, hope, encouragement, and forgiveness. With our best friends we hope and expect that they accept us, flaws and all; they see enough value in our other traits, or enough potential to change that they stand by us. They will hope that we can improve, and will presumably encourage us to do so. They will also typically be more forgiving than others when we do fall short. But all of these important aspects of friendship only come into play as we recognize our friend's shortcomings, and they ours. Indeed, it is often taken as a mark of best friends that they are quite aware of our flaws yet they still find us worthy of love; they recognize our strengths and potentials.
I endorse Kawall's claim, but his argument for it is vulnerable to partialist objections. We can divide his argument into two premises: (1) Being a good friend involves certain attitudes, including "acceptance, hope, encouragement, and forgiveness." (2) Having these attitudes requires 19 being aware of (certain of) your friend's shortcomings. Good friendship, Kawall concludes, requires awareness of (certain of) your friend's shortcomings.
It is premise (2) that partialists will dispute. One objection applies to three of the four attitudes that Kawall mentions: acceptance, hope, and encouragement. These attitudes, it says, do not require a de re awareness of specific shortcomings, but only de dicto awareness that the loved one, like any person, has shortcomings. That is, you can accept her no matter what, without committing to the existence of any particular "what", and you can hope for and encourage her without thinking that her current state is sub-par. The only attitude that Kawall mentions for which this partialist response does not hold is forgiveness, which can only be proffered for a specific offense.
But a second partialist objection covers forgiveness. It says that you need not have accurate beliefs about every detail of an offense in order to forgive the loved one for it. (Compare: one might be aware of a tree in the distance, without knowing whether it is an oak or a maple.) You might know, for example, that your loved one did a specific wrong thing, but you need not know that she was motivated by a flaw in her character, rather than (say) by extenuating circumstances or by a positive trait taken to an extreme. A general idea of the deed suffices to forgive it.
In response to these partialist objections, we need to explain why excellent love involves accurate beliefs about specific shortcomings, where this accuracy reaches a certain threshold of specificity. I will argue that beliefs that are accurate in these ways, by generating fitting grief, promote three goods of excellent love: a feeling of mutual safety, secure emotional attachment, and moral transformation.
To see how, consider the following passage from Edith Wharton's The Mother's Recompense. The long-term bachelor Fred Landers loves the middle-aged Kate Clephane. Yet Kate has a dark secret that she fears will prompt Fred to abandon her. As a result, she cannot relax in his love-she needs to confess her secret to him. Here is what happens after she does (Wharton, 1925, Book 3, Chapters XXIX and XXX): [Kate] sat down and waited. Curiously enough, she was less unhappy than for a long time past. [Fred's] pain and his pity were perhaps what she had most needed from him: the center of her wretchedness seemed the point at which they were meant to meet …. [B]ut to thank him for what he was suffering [because of her confession] would have seemed like mockery; and she could only wait and say to herself that perhaps before long he would go….
After a while he lifted his head and slowly got to his feet. For a moment he seemed to waver; then he crossed the space between them, and stood before her….
"The time will come," he said, "when all this will seem very far off from both of us. That's all I want to think of now." She looked up, not understanding. Then she began to tremble in all her body…. He was still looking at her, and she saw the dawn of the old kindness in his [eyes]. He seemed to have come out on the other side of a great darkness…. He had overcome his strongest feelings, his most deep-rooted repugnance; he had held out his hand to her, in the extremity of her need, across the whole width of his traditions and his convictions.
…. It seemed to her that for the first time in her life she had been picked up out of the dust and weariness, and set down in a quiet place where no harm could come. 20 Two things happen. One is that Kate bravely trusts Fred: she makes herself vulnerable by exposing a feature of hers to which grief-we will suppose for the sake of argument-is a fitting response. She confesses a specific secret, in enough detail for Fred to fully understand its repugnance. The second is that Fred responds by experiencing grief, but then showing Kate grace despite it. Grace, which Pettigrove characterizes as "an intentional act of unmerited favour" (2012, 127), might include forgiveness, but is a broader category (ibid., 142-150), including a charitable attitude to negative features extending beyond personal offenses. Fred shows Kate grace by remaining loyal despite his fitting grief, and declining to give her guilt. Note that it is Fred's fitting grief that makes his grace significant. If he were not emotionally moved, his grace would be cheap. Kate's making herself vulnerable by exposing her negative feature to Fred, and his responding with grief and then grace, demonstrate three ways in which vulnerability and grace can make love excellent: First, they can cultivate a feeling of safety for both parties. Consider Kate. On one level she gains an assurance of safety in the relationship: Fred has demonstrated his commitment to her. But on a deeper level the safety for Kate is existential. In being both seen and loved for who she is-despite feeling unlovable-Kate feels safe in life itself, indeed for the first time. Now consider Fred. In trusting him as she has, Kate has demonstrated how much he matters to her, making an assurance of emotional safety available for him too.
Second, and in this case partly as a result of the increased sense of mutual safety, vulnerability and grace can deepen the secure attachment proper to excellent love. Where Kate was braced for rejection, Fred offers her renewed commitment. And Fred, in learning that Kate does not want to hide things from him, is invited thereby to trust and draw closer to her. Attachment, if fed by brave vulnerability and grace, is apt not to be insecure (whether avoidant of intimacy, or anxious about abandonment), but rather secure. 21 The reason is that it arises from, and encourages, emotional honesty on the part of both parties. Kate signals that she wants to be real with Fred even if it means inflicting grief on him. As for Fred, acknowledging his grief frees him to release it and move beyond it toward grace.
Third, vulnerability and grace can morally transform lover and loved one. For the loved one, there are two ways in which making herself vulnerable and being responded to graciously can be morally transformative. First, it can transform her moral-epistemic faculties by enabling her to perceive human value. She comes to perceive her own value: she sees herself as the lover sees her, namely as well worth the grief that he undergoes for her sake. 22 And she comes to recognize her lover's value, as it awakens compassion for the grief that he experiences for her sake. Second, being shown grace in response to vulnerability can help the loved one become a better person more generally: it can motivate her to live up to the value that her lover has helped her perceive in herself and in him, and thus to become-for her own sake and as a gift back to him-someone who needs less grace.
As for the lover, there are two ways in which being trusted with the loved one's vulnerability, and responding graciously, can be morally transformative. First, being trusted can be part of what empowers you to respond graciously to begin with, for it holds up a picture of yourself as being the sort of person who will respond graciously. Second, showing grace can cultivate relational and personal virtues, such as patience and magnanimity (cf. Pettigrove, 2012, 91). These can help you love better: "the recognition of a modest flaw [in one's beloved] can chafe against love's native impulse", so that "learning to deal with this sort of chafing while continuing to love the other is part of maturing as a lover" (ibid., 96).
There are three things that I am not claiming. First, I am not claiming that revealing one's negative features to a lover will always yield these positive results. 23 What I am claiming is that it will do so to the extent that the love is excellent. Second, I am not claiming that it is always a good idea to strive for excellent love if doing so might destroy whatever love there already is: the perfect can sometimes be the enemy of the good. Third, it is not the case that lovers should always make themselves vulnerable by revealing deep secrets. Doing so is a gift of intimacy for certain times and places, and for a certain maturity of the love relationship. 21 For discussion of secure and insecure attachment see (Bowlby, 1980). 22 And in this way she gains new self-knowledge, often a catalyst for healing. Thanks to an anonymous reviewer for this point. 23 See for example the story of Maria von Herbert in (Langton, 1992).

3
The three benefits that I have described-a sense of safety, secure attachment, and moral transformation-arise when both parties have accurate beliefs about the loved one's specific shortcoming and its gravity. They do so, first, by making grief, and thus genuine vulnerability on the part of the loved one, and meaningful grace on the part of the lover, real possibilities; and second, by eliciting the lover's grace, which is significant in proportion to his fitting grief. The de dicto knowledge that the loved one has shortcomings is too unspecific to merit such a response; and too shallow an awareness will not merit a response that fits the extent of the shortcoming. But a fitting emotional response to a specific shortcoming can, through the brave vulnerability of the loved one and through the grief and grace shown by the lover, contribute to love's excellence.

Care
My second argument for the Accuracy Norm supposes that excellent lovers care for their loved ones in appropriate circumstances. 24 Like any actions, caring ones tend to accomplish your aims to the extent that they are well chosen in response to the way the world is. Caring well for a person thus requires accurate beliefs, including about them and their needs. As Pettigrove puts it, "[p]romoting the other's interests requires an adequate recognition of those interests" (2012, 89;cf. Tronto 1993, 137;Baier 1982, 274;Mason 2020). Excellent love thus requires accurate beliefs about the loved one.
A version of this argument is given by Kawall (2013) and Arpaly and Brinkerhoff (2018). But we'll see that it does not go far enough. I will supply the missing step.
Their argument appeals to cases in which true, but negative, beliefs about a loved one would serve her better than false but positive beliefs. Consider the amateur poet Rebecca and her poetry-connoisseur friend Eric. Were Rebecca not his friend, Eric would think that her drafts need serious improvement. But she is, so he manages to believe that they are good, or at least less rough than they are. He returns the drafts to Rebecca with few suggestions for improvement. Far from helping Rebecca get published, Eric's behavior will deprive her of much-needed feedback and encourage false hope (Kawall, 2013, 357;Arpaly & Brinkerhoff, 2018, 43). One might think that Rebecca should know better than to rely on friends for honest criticism. But honest criticism would further her aims, and if her friends will not help her do this, then what sort of friends are they?
It is hard to see how Stroud's strong partiality view could respond. Stroud argues that it is constitutive of friendship to "believe your own spin" (2006,513), without qualifying her view to accommodate different contexts. Keller's and Hazlett's weaker views, on which partiality is local and context-sensitive, do better (respectively 2004 and 2013).They may grant that Eric should obey epistemic norms when offering feedback on Rebecca's drafts, while insisting that at the poetry reading, where she 1 3 needs his sincere support, he should violate epistemic norms and believe well of the poem that she reads. Their weak partiality view, Keller and Hazlett will say, achieves the desired result in the feedback and the reading scenarios, whereas the epistemic view achieves it only in the former.
But their view is problematic too. Jollimore (2011, 50) argues that "belief is a holistic matter-the content of one's beliefs depends, in part, on what other beliefs one accepts….The elimination or suppression of true beliefs changes the character of one's other beliefs and can lead to the formation of beliefs that are straightforwardly false" (2011,50). And Arpaly and Brinkerhoff (2018, 47) argue: Counter-evidential beliefs easily propagate and are bound to become risky, sometimes unexpectedly, when they meet other beliefs and desires or when new situations come along in which accuracy is of the essence. There is no safe dose of epistemic irrationality-once you have allowed it into your life, you are never safe from its repercussions, and neither are your friends.
The idea is that partial but false beliefs that we form now may come back to haunt us later in the form of faulty background evidence. This argument shows that even the weak partiality view has a cost: it can promote unwanted error or distortion, hampering you in caring for loved ones when doing so requires accuracy.
Yet this conclusion does not go far enough to justify rejecting the partiality view in favor of the epistemic view. For we saw reason to think that the epistemic view has a cost too: a biding by it will hamper you in caring for the other party when doing so requires positive beliefs. The two views, then, seem to be at an impasse.
I will resolve this impasse in favor of the epistemic view. Three considerations speak for it. One comes from the previous section. We saw that accurate beliefs promote fitting emotional responses, which can yield personal and moral benefits for loved ones, things we surely want to give them.
A second consideration arises when we look more closely at cases that seem to support the partiality view: where we seem to care best for loved ones by believing positively yet inaccurately about them. I will argue, in contrast, that believing accurately, even if negatively, enables us to care for them better still.
To see this, imagine that your loved one is struggling to recover from addiction, and your evidence speaks against his succeeding. Stacked against his success are the statistics for people in relevantly similar situations, his previous failed attempts, and still-operative factors that previously sapped his willpower. When he asks whether you think he will recover, it might seem that the most caring response is a convinced "Of course you will!" You can give him the gift of encouragement that you really stand by, even if your evidence speaks against it. For if anything will mobilize your loved one to make the needed changes, however improbably, it is the belief of those who love him.
The epistemic view may seem cruel for denying this. In fact, it furnishes an even more caring response. This response is not to believe the improbable "You will definitely recover." It is to believe-and communicate-the conditional proposition, "You definitely could recover if you address what previously kept you from recovering. And because I love you, if you let me I'll support you in doing that." When and how you say this is a sensitive matter. But assuming that recovery is your loved one's aim, violating epistemic norms about his prospects, given the obstacles he faces, will help him much less than maintaining the clear vision toward his goal that he himself may lack (cf. McGeer, 2008;cf. Kawall, 2013, 363-364). 25 A similar point holds for the poetry case. If Eric is in a position to realize that Rebecca's drafts are unlikely to impress the agent, the most caring thing he can do is save her from a professional blunder, while promising to help her: "Great poetry is 1% inspiration and 99% perspiration. You can do great work if you put your mind to it, and I'll help!" But what if, as in Keller's original case, Eric does not read the drafts, and has only his knowledge of the venue to go on? Here it is not clear that such circumstantial evidence would prescribe believing that Rebecca's poetry is substandard (cf. Jollimore, 2011, 54). But suppose the case is strong (e.g., Eric has regularly attended this venue for the past decade). In this case, I suggest that the caring response is not to believe that her poetry will be good, but to mentor her in advance, or (if time does not allow), gently counsel her to postpone the reading until she will truly impress the agent. In general, even when positive yet inaccurate beliefs might be caring, accurate beliefs-with a commitment to help-are more caring still.
Empirical results support this idea. We are most apt to succeed not when we believe that we are great, but when we believe that if we work hard we can perform greatly (Dweck, 2006), and when we are motivated to work hard by a passion for accomplishing our aim (Duckworth, 2017). Far from believing positively no matter what, you can best care for someone by encouraging these more constructive attitudes and signaling your commitment to stand by her.
Here is the third consideration supporting the epistemic view. I have granted that believing positively can be a form of care, while arguing that believing accurately is a superior form of care. But believing positively can also do unintended damage. If you express, or emphasize, your positive beliefs about a loved one, you can inadvertently change the emotional register from one of valuing them to one of evaluating them. This can make salient the risk that they will fail to measure up (Dweck, 2006). In the extreme it can promote the imposter syndrome. Consider the following real-life example. Bruce Brown and Rob Miller of Proactive Coaching LLC 26 conducted an informal survey over three decades, asking children about their worst memories of playing sports. The overwhelming response was, "The ride home with my parents." Why? Parents tended to analyze every move that their child made in the game. No matter that these analyses were mainly affirming. They turned the focus away from mutual enjoyment of the shared experience and onto the child's performance. These assessments, well intentioned as they are, do not seem to express care nearly as well as what Brown and Miller report grandparents saying, which is simply, "I love watching you play." Rather than emphasizing the performance, the emphasis here is on the child's intrinsic worth.
The difference between the parents' and the grandparents' responses illustrates the difference between the partiality and epistemic views. The partiality view builds positive evaluation of the loved one into its conception of excellent love; a fortiori, and however inadvertently, it builds in evaluation. The epistemic view, by contrast, emphasizes valuing the loved one, just as they are.
I have given two arguments for the epistemic view, particularly for premise 1, the Accuracy Norm. The previous section argued that accurate beliefs promote fitting emotional responses, which in turn promote feelings of safety, attachment, and moral transformation. This section argued that accurate beliefs help you care for loved ones-first, by promoting the goods just listed; second, by enabling us to further loved ones' aims; and third, by communicating not an evaluative attitude, but a valuing one.
We are now in a position to see how adhering to epistemic norms, far from constraining love, makes it excellent. Epistemic norms in love are thus nothing like stifling traditions. They are far more like norms of consent. Adhering to them opens up a space for love to be the best it can be. You can truly encounter, and thus know, respond to, and care for, the other party as she is, treating her as an infinitely valauble end in herself.

The objection from distortion
One might object to the epistemic view on the basis of something I have so far said little about: what sorts of epistemic norms are at issue. This objection can be posed as a dilemma. One horn threatens premise 2 of the Argument from Accuracy, and the other threatens premise 1. It goes like this. Either epistemic norms are internalist, meaning that what matters epistemically occurs in the believer's mind, for example whether she respects her evidence (understood as perceptions and beliefs). Or epistemic norms are externalist, meaning that they concern things that occur outside the person's mind, especially whether she formed her beliefs reliably. Horn 1: If the epistemic norms are internalist, then premise 2 (the Accuracy-Conduciveness Claim), is false, because love distorts the lover's evidence about the loved one, leading to inaccurate beliefs about him. Horn 2: If epistemic norms are externalist, then premise 1 (the Accuracy Norm of Love) is false, because, even though love's distorted perception makes for inaccurate belief-formation, we can have excellent love. In either case, the Argument from Accuracy fails. 27 Taken more slowly, let's consider horn 1, which works with internalist epistemic norms. A reasonable internalism will include, as evidence, intersubjectively available considerations (such as what your loved one said in public). But crucially, we saw that it will also include your individual experiences of the loved one that are unavailable to third parties (Conee & Feldman, 2004, 112); after all, these are among your most basic sources of information about him. Yet few would deny that love comes with biased perception: although it can be accurate, it can also obscure the loved one's negative features, or portray positive features as more positive than they really are, perhaps even distorting your perception entirely (Keller, 2004;Hazlett, 2013;Stroud, 2006;Jollimore, 2011, chapter 3). If this is so, we get the internalist objection from distortion: the perceptual bias of love distorts a person's evidence. And forming beliefs in accord with distorted evidence is not accuracy-conducive. So if the epistemic norms in question are internalist, says the objection, premise 2, the Accuracy-Conduciveness Claim, is false: adhering to them is not conducive to accurate beliefs about the loved one.
One response would be to construe epistemic norms as externalist. If this is so, then premise 2 would wind up true, since reliabilist epistemic norms promote accuracy by definition. But then horn 2 arises-the bump in the carpet reappears at premise 1, the Accuracy Norm of Love. We may call this the externalist objection from distortion: the perceptual bias of love in this case prevents the lover from forming reliably accurate beliefs about the loved one to begin with. 28 But if reliably accurate beliefs about the loved one are so hard to come by, then so, by premise 1, is excellent love! The objector maintains that this conclusion is absurd: of course excellent love is attainable. 29 Premise 1 is false. Externalist norms are no better for the epistemic view than internalist ones.
Both horns of the dilemma work from the premise that love comes with a distorting perceptual bias. I have two responses. First, it is not clear that love's perceptual bias is epistemically bad on the whole. Biases can be epistemically good: consider the expert's laden perception that enables them to distinguish real gold from false. And as we saw, the belief-forming processes typical of love can help you know your loved one better than a third party would (Jollimore 2011, chapter 3).
Second, however, let us grant for argument's sake that love's perceptual bias does tend to impugn our evidence or the reliability of our cognition. In response, such a tendency is not inevitable-it can be counteracted. To see this we need a closer look at clause (ii) of the epistemic view. This posits a love-motivated care for accuracy in forming beliefs about loved ones. This amounts to an affective orientation toward certain sorts of truths, which counteracts the distorting effects of love's perceptual bias. How? Like any affective orientation, it generates a bias of its own. We may call this an epistemic bias, because it inclines you to form beliefs in accuracy-conducive ways. For example, it may foster perceptual and intellectual habits that sensitize you to fine-grained realities. An epistemic bias can thus temper the distorting effects of love's perceptual bias, while supporting its accuracy-conducive ones. And an epistemic bias can itself be an expression of love, insofar as it arises from the desire to know the loved one as she is.
To see how love's perceptual bias and the epistemic bias complement each other, consider an analogy. A mathematics student grows up in a household of mathematicians. Because of this, she begins university with a love of math and already honed mathematical thought patterns. But she will succeed at math only to the extent that she complements these with practice and the discipline of double-checking her work.
Her love of math itself motivates her to take extra care in forming mathematical beliefs. Similarly, the person who loves excellently is motivated by his care for accuracy concerning his beloved: he harnesses love's perceptual bias insofar as it promotes accuracy, but tempers it insofar as it distorts the truth.
An epistemic bias borne of care for accuracy can mitigate both versions of the objection from distortion. First, it can attune your mental states to reality, yielding high-quality evidence. If epistemic norms are internalist, premise 2 is upheld: adhering to them with a care for accuracy can promote accurate beliefs about a loved one. Second, an epistemic bias makes cognition about your loved one more reliable. This makes it easier to adhere to externalist epistemic norms, in turn making excellent love more achievable than the objector thinks-premise 1 does not reduce to absurdity. The claim that we can have an epistemic bias is subject to empirical confirmation, but I take it to be at least prima facie plausible-enough so, for example, that Jollimore (2011, chapter 6) argues that love itself is generally accuracy-conducive, and Mason (2020) argues that love entails knowledge of the beloved.
I conclude that neither version of the objection from distortion threatens the epistemic view.

Addressing the arguments for the partiality view
We may now address the arguments for the partiality view mentioned in Sect. 2. One is that a person who loves excellently is committed to a positive view of the loved one's character (Stroud, 2006, 511-512). My response is that she is committed instead to her loved one. The best loving commitments are for better or worsewhich only makes sense when "worse" is a relevant category.
The second argument draws an analogy with action (Hazlett, 2013, 93). We treat our loved ones better than others whom we do not love, so surely we should believe better of them than of others. But it should now be clear that belief is disanalogous to action, not least since partial actions-such as supporting a loved one in recovering from addiction-can be better targeted if your beliefs are accurate.
The third argument draws on phenomenology. Loving excellently, it says, involves sticking up for loved ones, including in our own minds (Stroud, 2006, 503-504). In response, sticking up for a loved one is not a matter of believing well of him come what may, but of loving him come what may (and expressing this when need be to others), while holding up an image of what he could be. What about the claim that we feel excellently loved when we are thought better than those who love us have reason to believe (Keller, 2004, 338;Hazlett, 2013, 90)? I think that a love in which such a desire dominates, as opposed to the desire to be known as one is, is not entirely healthy. I agree with the weaker idea that love involves a desire to be thought positively of; but surely this should motivate one to become the sort of person of whom a lover can think positively without violating epistemic norms.
The fourth argument draws on the quality view, according to which love is based on your loved one's positive features. Inflated beliefs, the thought goes, help sustain love (Stroud, 2006, 511). One response is to use this claim as a reductio of the quality view: what kind of love can only be kept afloat by distorting its object to please the lover? A second response, however, is compatible with the quality view (cf. Mason 2020). Instead of burdening the lover to think her loved one better than epistemic norms allow, we might motivate the loved one to become the best person he can benot from fear of abandonment, but to fulfill the positive vision of himself reflected by his lover's value of him, and to bring his lover happiness along the way. At the same time the lover can be challenged to grow by showing the loved one grace. When we combine the quality view with the epistemic view, it promotes both parties' moral transformation; when we combine it with the partiality view, it motivates both parties to deceive the lover.
The epistemic view faces a common objection that, it should be clear by now, rests on a mistake. It is that this view construes excellent love as involving a neutral detachment, which has no place for the special glow with which lovers perceive and think about loved ones (Baron, 1991). For how else could lovers ensure that their beliefs are accurate? The objection would hold if evidence from personal experiences of the loved one were epistemically inadmissible. Far from it. As we saw, personal experience of loved ones helps us know them better than otherwise, provided that we match the bias of love with an epistemic bias (Sect. 6). The epistemic view facilitates seeing the beloved much more accurately than someone would who did not love her, indeed much more intimately.

Conclusion
The partiality view draws on strong intuitive and theoretical considerations-but it is false. Excellent love does not involve violating epistemic norms, even when epistemic norms point to negative beliefs about the loved one. On the contrary: as the epistemic view says, negative beliefs no less than positive ones, when accurate, are crucial for excellent love.
One reason is that both promote fitting emotional responses to your loved one, which help make love excellent. Delight in your loved one's features as they really are staves off idolatry (on your part) or the imposter syndrome (on his). Fitting grief at your loved one's negative features, or the real possibility of it, is necessary for vulnerability and for showing grace; these can in turn foster a sense of safety, secure attachment, and moral transformation. Another reason is that accurate beliefs about your loved one are crucial for caring for him: they enable a more caring response than positive but mistaken beliefs, and turn the focus from evaluating the loved one to valuing him.
We have seen that love comes with a perceptual bias that can promote accurate beliefs but can also distort. This is why excellent love involves not just adhering to epistemic norms, but doing so motivated by a countervailing epistemic bias, which counteracts the distorting effects of love's perceptual bias while supporting the accuracy-conducive ones.
What of the intuition that love by nature bursts certain constraints, including epistemic ones? Epistemic constraints, it turns out, are less like those of pointless convention or unhealthy beauty standards, which love does well to defy, and more like those of personal autonomy or the age of consent, which love is realized by respecting.
We need not worry that the epistemic view depletes love of its mystery or suspense. For love often involves a desire to be known for who you are (on the part of the beloved) and a rapt attentiveness (on the part of the lover) that meets this need. What makes love mysterious and suspenseful is that this journey of mutual discovery stretches over time, and the fact that-given human complexity-it can never be complete. This journey includes a quest, among many other things, for accuracywhich the partiality view impedes, but the epistemic view facilitates.