The epistemic demands of friendship: friendship as inherently knowledge-involving

Many recent philosophers have been tempted by epistemic partialism. They hold that epistemic norms and those of friendship constitutively conflict. In this paper, I suggest that underpinning this claim is the assumption that friendship is not an epistemically rich state, an assumption that even opponents of epistemic partiality have not questioned. I argue that there is good reason to question this assumption, and instead regard friendship as essentially involving knowledge of the other. If we accept this account of friendship, the possibility of epistemic partialism does not arise.

In the recent literature on friendship, various 'epistemic partialists' have suggested that there are important norms that conflict with epistemic norms. In particular, the suggestion has been that norms deriving from valuable relationships such as friendship sometimes demand things that conflict with epistemic demands on us. 1 It would be disconcerting if there were the kind of conflict between friendship and epistemic norms that the epistemic partialist claims there is. Such conflict would imply that in acting well in one regard (as a friend or else as an epistemic agent) we can be systematically precluded from being a good agent in another respect (as an epistemic agent or else as a friend). There would be no way of getting it right overall: it would be impossible to be both a good friend and a responsible epistemic agent. 2 In response to this pessimistic thought, sceptics about epistemic partiality have sought to find ways to understand the demands of friendship and epistemic agency such that they are consistent. Usually, they have suggested both that epistemic norms are less restrictive than epistemic partialists assume, and that friendship is less demanding than epistemic partialists assume. Such sceptics conclude that epistemic norms permit the doxastic states or actions characteristic of friendship, and that friendship fortuitously turns out not to require anything epistemically dubious after all. 3 In this paper, I will be exploring an alternative response to epistemic partialism inspired by Iris Murdoch's work on love and knowledge. 4 On Murdoch's view, love is (or at least involves) knowledge of the individual, such that to fail to see the beloved truly is not only an epistemic failure but also a failure of love. On her view, insofar as one's knowledge of the other is lacking, so is one's love. Given the plausible assumption that friendship involves love, then a Murdochian view of friendship will imply that the person who fails to see their friend as they truly are, far from being a good friend, necessarily falls short of the intrinsic demands of friendship.
A Murdochian view of friendship thus undermines a key assumption of epistemic partialism that has been accepted by proponents and sceptics alike. This assumption is that friendship is independent of epistemic goods such as knowledge and truth, such that friendship is not itself an epistemically rich state. Keller (2018) explicitly makes this claim, and it is an assumption that epistemic partialists must make if the norms of friendship and epistemic norms are to constitutively conflict. On the Murdochian account, the concerns that underlie the norms of friendship are at least in part epistemic concerns, thus ruling out the possibility of constitutive tension between the two kinds of demand. I will therefore be arguing that a Murdochian view of friendship provides a novel way of rejecting epistemic partialism.
A Murdochian account of friendship and epistemology is not, however, merely of exegetical interest; I will also be suggesting that it casts light on some common intuitions about friendship. It makes sense of intuitions regarding the impact of mistaken beliefs on apparent friendships-namely that deep mistakes about our friends seem to undermine friendship. It also makes sense of the ways in which we would expect good friends to behave-namely that we expect them to be responsive to the actual 2 It is worth noting that on this view, one would fall short either as an epistemic agent or a friend, whether or not one encounters an actual clash between the demands of friendship and those of epistemology. One might be fortunate enough to have friends who truly are wonderful, and thus it might require no false beliefs to think good things of them. Nonetheless, one kind of norm ('believe what the evidence points to' or 'believe good things of one's friends', for example) will in fact guide one's belief-formation, since the two could come apart, and in that case one would be forced to choose which norm to follow. This is sufficient to think that whether or not one encounters an actual clash, one falls short with regard to one kind of norm. 3 Hawley (2014), for example, suggests that given background commitments to epistemic permissivism (the denial of the claim that given one's total evidence there is a unique rational doxastic attitude that one can take to any proposition) and the thought that friendship requires less of us than some assume, the norms of friendship will conflict with those of epistemic agency rarely, if ever. Kawall (2013) and Morton and Paul (2018) also draw on theories of epistemic permissivism and pragmatic encroachment to suggest that there is no conflict between the two sets of norms. 4 In particular, see Murdoch (1959Murdoch ( , 1970 character and needs of their friends. I will thus be arguing that there is reason to take the Murdochian view of friendship seriously.
In Sect. 1 of this paper I further outline why epistemic partialists think that norms concerning friendship conflict with epistemic norms. In Sect. 2 I argue that underlying the appeal of this thought is the assumption that good friendship is independent of epistemic goods such as knowledge and truth. In Sect. 3 I argue that Murdoch's work suggests a novel way of responding to this apparent problem by insisting that friendship involves knowledge, thus ruling out the possibility of constitutive conflict between the norms of friendship and epistemology. In Sect. 4 I argue that the Murdoch-inspired response best makes sense of some important intuitions we have about friendship. Finally, in Sect. 5 I consider the implications of this for accounts of friendship, before concluding that discussions of epistemic partiality go awry in taking for granted a substantive and questionable conception of friendship.

Partiality and epistemic responsibility
At any one point, humans are subject to various different kinds of norm: epistemic norms, prudential norms, moral norms, and social norms, for example. These norms sometimes come into conflict. Some conflicts between two kinds of norm are merely coincidental-they arise simply because of the particular circumstances in which one finds oneself. For example, if someone promises to greatly financially reward you for believing something utterly trivial but patently untrue, then in that particular instance what you prudentially ought to do and what you epistemically ought to do conflict. Such a conflict of demands is of course unfortunate. However, it does not indicate any deep-seated, inherent conflict between the two kinds of norm; it does not indicate that there is any constitutive conflict between prudence and good epistemic agency. Epistemic demands and those of friendship can clearly conflict in this coincidental sense.
The controversial claim made by epistemic partialists is that there is inherent or constitutive conflict between the norms of friendship and epistemology. Their claim is that being a good friend is partially constituted by forming positive beliefs about one's friends or by adopting belief-forming strategies that are likely to lead to such positive beliefs. 5 In other words, they suggest that being a good friend can be partially constituted by biased beliefs, or biased belief-forming mechanisms. On the other hand, epistemic norms entail that one should form beliefs only on the basis of considerations that bear on epistemic goals: the truth of the claim in question, or one's knowledge. 6 Such beliefs, or belief-forming mechanisms, will by definition be unbiased; epistemic norms demand lack of bias. The potential conflict between the demands of friendship and epistemic demands is thus no mere coincidence, but inherent to those norms them-5 There are important differences between these ways that biases can enter, which I discuss in Sect. 1.1. 6 Exactly what kinds of reason make for good epistemic reasons is disputed, but on any plausible account of epistemic reasons, the fact that someone is one's friend will not count as a good epistemic reason. Hazlett (2016) notes that 'epistemic' can be used in two distinct senses, one meaning relating to belief , and the other meaning relating to knowledge. I am using the term in the latter sense. selves. 7 Epistemic partialists claim that the sets of norms themselves prescribe things that are incompatible, since one kind of norm demands bias, whereas the other prohibits it. Stroud (2006) thus claims that "[f]riendship constitutively involves belief-forming practises which are irrational or otherwise objectionable by the light of mainstream epistemologies" (Stroud 2006: p. 500).
On Stroud's view, being a good friend involves believing, or being disposed to believe, good things about one's friends: "the good friend is prepared to take her friend's part both publicly and, as it were, internally" (Stroud 2006: p. 505). Stroud notes that friendship is often thought to be in some sense based on one's view of one's friend's character and merits, and thus that "it makes sense that we need, as it were, to maintain a favorable opinion of our friend's character" (Stroud 2006: p. 511). The suggestion is that it is a norm of friendship, inherent to friendship itself, that one believe good things about one's friends, or be disposed to form such beliefs. And this, she suggests, is the case simply because they are one's friends rather than because there is some epistemic explanation available (such as that one is likely to know one's friends well and therefore have plenty of evidence about their good qualities). The norm that one should believe good things of one's friends, she suggests, thus clashes with epistemic norms, since epistemic norms require that one should be responsive only to epistemic considerations. 8 Stroud thus suggests that epistemic norms require lack of bias, whereas friendship demands bias. 9 To illustrate what epistemic partiality might require, Keller (2004) puts forward the following example: Rebecca is scheduled to give a poetry reading at a café and has invited her friend Eric along. Having attended many such readings at the café, Eric has accumulated strong reasons to believe that the poetry is never more than mediocre, and often positively bad. Keller suggests that prior to knowing who will be performing Eric would have "good evidence" for the belief that their poetry will be unimpressive (Keller 2004: p. 332). Were Eric assessing the likelihood that a stranger would give a good poetry reading at the café, Keller claims that he ought to believe that they won't, since he has plenty of evidence for this conclusion. However, since Rebecca's identity as a poet is important to her, and he's there to support her, Keller claims that Eric ought not to form such a belief about Rebecca. This is despite having no extra knowledge of her general competence, aesthetic capacities, or anything similar that might count as evidence for that belief: [S]eeing as he's Rebecca's good friend, though, and seeing as he is there to offer her support, he ought not [to]… have those beliefs about her (Keller 2004: 332).
That is, despite the fact that all of the evidence available to Eric points strongly towards the conclusion that the poetry will be bad, Keller thinks that as a friend Eric ought not to believe that Rebecca's poetry will be bad. 10 Similarly, Stroud suggests that friendship gives us reason to disbelieve plausible testimony about the ethically bad behaviour of our friends, even if the friendship does not provide evidence for the falsity of the testimony.

Two kinds of epistemic partiality
Epistemic partialists, then, hold that the demands of friendship constitutively conflict with epistemic demands. There are two distinct claims, however, that epistemic partialists make: one is that friendship demands partiality on the level of belief, and another is that it demands partiality on the level of doxastic dispositions. These suggest two different kinds of conflict between epistemic reasons and those of friendship. Epistemic partialists have tended to accept that friendship gives us reason to be epistemically partial in both of these ways, although it would be possible for them to come apart.
The first claim made by epistemic partialists ('direct epistemic partialism') is that friendship provides direct reasons for belief. Epistemic partialists sometimes claim that the fact that someone is one's friend can be a direct reason to form a belief about them that is more positive than the belief the evidence points to. As a result, these friendship-based reasons for belief sometimes directly conflict with epistemic reasons for belief. Direct epistemic partialism is suggested, for instance, by Keller: [W]hen good friends form beliefs about each other, they sometimes respond to considerations that have to do with the needs and interests of their friends, not with aiming at the truth, and that's part of what makes them good friends (Keller 2004: 330).
On this account, friendship directly provides a reason to believe positive things about one's friends, and it is constitutive of good friendship that one is responsive to such reasons.
The second claim made by epistemic partialists ('indirect epistemic partialism') is that friendship provides reason to cultivate doxastic dispositions that will yield beliefs about one's friends that are positively biased. 11 In this case, the conflict between epistemic and friendship-based reasons occurs at the level of epistemic agency; on this conception of epistemic partiality, the epistemic reasons we have to cultivate unbiased doxastic dispositions conflict with the friendship-based reasons to form biased doxastic dispositions. Stroud, for instance, writes: Friendship positively demands epistemic bias, understood as an epistemically unjustified departure from epistemic objectivity. Doxastic dispositions which violate the standards promulgated by mainstream epistemological theories are a constitutive feature of friendship (Stroud 2006: 518).
Keller also discusses this possibility when he talks about "belief-forming strategies" being constitutive of friendship (Keller 2018: p. 32). Here, Stroud and Keller speak as if friendship gives us a reason to cultivate doxastic dispositions rather than a reason to form beliefs themselves.
The core of epistemic partialism is thus the claim that the norms of friendship and epistemology constitutively conflict in one (or both) of the above ways. Epistemic partialists claim that good friendship is in part constituted either by forming positive beliefs about one's friends, or by developing doxastic dispositions that lead to forming positive beliefs about one's friends. On either version, such reasons will clash with epistemic reasons, which have only to do with epistemic goods such as evidence, truth or knowledge, and which require forming unbiased beliefs and doxastic dispositions. 12

The independence assumption
The idea that the norms of friendship could constitutively conflict with epistemic norms in either of these ways depends upon the assumption that norms of friendship are inherently (positively) independent of epistemic norms. This assumption is necessary if there is to be the possibility of constitutive tension between the two. That is, friendship can only constitutively demand things that are inconsistent with epistemic demands if it is not itself knowledge-involving. If good friendship involves knowledge, or a propensity towards knowledge, then the demands of friendship would themselves involve epistemic demands, and thus could not constitutively conflict with epistemic demands. Keller (2018) explicitly notes this, and argues for the independence of friendship from epistemic considerations: Considering the very different respective sources of epistemic norms and the norms of friendship, there is no reason to presume that you will always be able to meet both kinds of norms at the same time. My needs for support and acknowledgment (and so on) are products of my changing human moods, insecurities, anxieties, and circumstances. Your epistemic reasons are products of facts about the evidence before you and about which belief-forming strategies are most likely to lead you to the truth. Whether my needs are nicely coordinated with the evidence in front of you appears to be an utterly contingent matter. Suppose that you are at the grocery store deciding what food to buy. Some choices would allow you to eat healthier food; some would allow you to eat tastier food. Suppose that I tell you that you do not need to decide between healthy food and tasty food, because whatever turns out to be the tastiest diet is guaranteed also to be healthy. There are many ways to eat healthily, I tell you, so considerations of health do not restrict you to one specified diet … The problem is that the considerations that determine whether food is healthy are deeply different from the considerations that determine whether food is tasty. Healthiness and tastiness in food have very different sources. As a result, there can be no guarantee that the tastiest diet for you will also be the healthiest diet for you, even if there are many different ways to eat healthily (Keller 2018: 32).
Keller here notes that foods are not tasty in virtue of their healthiness, nor healthy in virtue of their tastiness. Healthiness as such does not contribute to tastiness, nor vice versa. So, it is possible for the two considerations to pull apart in inconsistent directions.
Keller's suggestion is that, like the tastiness and healthiness of food, epistemic norms and those of friendship are unconnected. Since they are unconnected, it is unsurprising that they should pull apart. The two kinds of norm, like those deriving from health and tastiness, are independent of one another, and thus any coincidence between the two would be wholly fortuitous (as in cases where healthy foods are also tasty). His suggestion is that knowledge as such does not contribute to friendship, so there is no reason to expect epistemic demands to coincide with those of friendship. That friendship and epistemic goods are independent ('the independence assumption') is thus a necessary condition for constitutive conflict between the two kinds of demand. 13 Need the epistemic partialist assume that the norms of friendship and those of epistemology are wholly independent? Could they not instead claim that some of the norms of friendship are independent of those of epistemology? That would allow partialists to say that even if friendship involves some epistemic demands, it also involves some demands that are inconsistent with epistemic demands. That is, they might claim that friendship involves inconsistent demands, only some of which are independent of epistemic norms. However, epistemic partialists in fact seem to accept that the norms of friendship are consistent, and thus that the two sets of norms are wholly independent: they standardly claim that the norms of friendship conflict with epistemic norms, not that the norms of friendship are themselves conflicting. Moreover, there is good reason for them to make this claim. Ultimately, there would seem to be something wrong or incoherent about practises with inherently inconsistent demands. Friendship does not seem to be such a confused or incoherent practise-and regarding it as such would motivate a much deeper reassessment of friendship than that suggested by partialists. There is thus good reason for partialists to accept the full independence assumption. 14 The claim that friendship is independent of epistemic goods has generally gone unquestioned. Even sceptics about epistemically partiality have raised no challenge to the idea that the norms of friendship are inherently independent of epistemic norms, nor to the idea that friendship itself is independent of knowledge as such. 15 As I noted above, they have rather suggested that friendship does not require us to be as partial as epistemic partialists assume, or that epistemic norms are less restrictive than might be thought (and often both of these). I will now give a fuller overview of these approaches.
Some philosophers sceptical about epistemic partiality take a piecemeal approach to making the demands of friendship consistent with epistemic demands. They suggest that since many particular goods of friendship depend on negative or accurate beliefs about the other, friendship cannot require general positive bias. For example, some note that friendship often requires one to break hard truths to the other. Others note that friendship can require giving advice to another, and that such advice will be better insofar as it is accurate. 16 That is, they identify particular goods that arise in friendship and suggest that they are such that the norms of friendship that arise from them are consistent with those of epistemology. 17 On this approach, the norms of friendship and epistemology are demonstrated to be consistent in a piecemeal fashion: norms arising from particular goods of friendship are identified, and those norms are shown to be individually consistent with our epistemic obligations, since knowledge is necessary for the realisation of those goods. The greater the number of norms that are consistent, the less likely it is that friendship and epistemic obligations will conflict. However, the possibility of such conflict remains, since the requirements of one good might conflict with that of another (in much the same way that one might find that eating tasty food is consistent with eating healthily, since lots of individual healthy foods are tasty, despite tastiness and healthiness being wholly independent considerations). The assumption that friendship as such does not involve epistemic goods remains unquestioned by this approach.
Other philosophers suggest that reconceiving of the nature of epistemic norms allows us to see that they are consistent with the demands of friendship. They thus suggest that the norms of epistemology are not as demanding or restrictive as partialists assume. Such sceptics about epistemic partialism turn to theories such as epistemic permissivism or pragmatic encroachment to suggest that our epistemic duties are not as stringent or as wholly divorced from our wider duties as partialists assume. 18 This approach generally speaks only to a narrow range of the cases where epistemic par- 15 Crawford (2019) is the exception to this. She claims, as I shall, that "there is something about friendship itself that explains why friendship cannot generate its own reasons for doxastic partiality" (Crawford 2019(Crawford : p. 1586. She argues that authentic friendship involves internalist epistemic constraints on belief formation: the good friend cannot take the friendship itself as a reason to believe positive things about their friends. The Murdochian account I will offer differs significantly from this in that it entails that there are externalist epistemic constraints on friendship: I will argue that friendship requires knowledge of the friend. The arguments I offer in Sect. 4 support these specifically externalist epistemic constraints on friendship. 16 See Kawall (2013), Arpaly and Brinkerhoff (2018), and Morton and Paul (2018). Arpaly and Brinkerhoff argue that although in some cases epistemic partiality might help friendship, many particular goods of friendship depend on accuracy. They suggest that overall we are better off not developing epistemically partial doxastic dispositions. 17 Kawall, for example, notes that only accurate beliefs will be truly reassuring (2013: p. 359), supportive (2013: p. 360), or meet our desire to be loved for who we are (2013: p. 359). 18 See Kawall, Hawley (2014), Piller (2016) and Morton and Paul. For a general defence of the idea that epistemic norms are not distinct from moral and prudential norms, see Rinard (2018). tialists think bias is required. 19 It also depends upon accepting substantive epistemic commitments, and thus speaks to a narrow potential audience. 20 Proponents of this approach, too, have not questioned the assumption that friendship is inherently independent of one's epistemic standing regarding one's friends.
I will argue that the norms of friendship are not independent of those of epistemology, but this does not entail rejecting either of the above approaches. My explanation suggests that importantly mistaken beliefs about one's friends themselves undermine friendship. It thus rules out the very possibility of conflict between epistemic demands and those of friendship, and therefore provides an overarching explanation as to why the demands of friendship so fortuitously turn out not to require bias. 21 The Murdochian response I shall offer is also consistent with epistemic theories such as permissivism and pragmatic encroachment but does not require them.

Friendship as an epistemically rich state
Although the assumption that the norms of friendship are inherently independent of epistemic norms is widely taken for granted, there is reason to be hesitant about accepting it. On some plausible conceptions of friendship, the norms of friendship and epistemic agency are not independent. On the assumption that friendship involves and is partially constituted by love, the Murdochian conception is one such. On Murdoch's account of love, love is not something whose demands or norms could be in constitutive conflict with epistemic demands or norms. And this is because love, as she conceives of it, is an epistemically rich state: love is partly constituted by epistemic goods such as knowledge and openness to knowledge. Thus, since friendship is partially constituted by love, a Murdochian conception of friendship will entail that friendship, too, is partly constituted by epistemic goods such as knowledge, or a propensity to knowledge.
In Sect. 2 I argued that Keller's claim that the norms of friendship and epistemology are entirely independent is underpinned by the idea that epistemic goods are unrelated to the goods of friendship. Friendship, he states, fundamentally involves goods such as "support, openness, encouragement, and the assurance that someone is on my side" (Keller 2018: p. 32). 22 None of these goods, he claims, involves epistemic goods. As such, the norms relating to them would plausibly be independent of epistemic norms.

By contrast, Murdoch conceives of love (and hence plausibly friendship) as having an inherent epistemic dimension:
Love is the perception of individuals. Love is the extremely difficult realisation that something other than oneself is real. Love … is the discovery of reality (Murdoch 1959: 51). Love is knowledge of the individual (Murdoch 1970: 28).
These suggest that on Murdoch's account love is not simply a matter of realising goods such as assurance, support and encouragement. Rather, love involves a kind of knowledge or perception, where these are meant to be factive states. On Murdoch's account, then, love is not simply a matter of acting in a certain way or feeling a certain way, but an epistemically rich state; she suggests that loving is partly constituted by epistemic goods.
Murdoch is not the only proponent of a conception of friendship on which it involves knowledge of one's friends. Aristotle, too, assumes that friendship involves knowledge of one's friends. Indeed, he takes this to be part of what makes friendship valuable, since on his account knowledge of one's friends aids self-knowledge. 23 However, I will focus on Murdoch's discussion here, since her position regarding the connection between love and knowledge is more explicit than Aristotle's.
Murdoch associates love with a certain kind of open-endedness in one's vision of an object. Love, as she understands it, involves not only a static set of true beliefs about the other, but also progressive movement towards deeper, truer knowledge of the other. The lover, she states, is confronted with an "endless task", an "ideal limit of love or knowledge which always recedes" (Murdoch 1970: p. 28). The person who attends to some object with love, on Murdoch's conception, is continuously involved in a progression towards an understanding of the object that does greater justice to its full complexity. 24 Murdoch discusses love in general rather than friendship in particular, and she conceives of it in a broad sense. For instance, as she conceives of it, one can direct loving attention towards not only people but also inanimate objects and even concepts. 25 Her general view of love, then, may not tell us all that there is to know about the more specific phenomenon of friendship. However, it is plausible that love is necessary for, and a central aspect of, friendship-at least for the kinds of friendship that are deep and important, and that are thus candidates for influencing our epistemic obligations. 23 This depends on the idea that a friend is 'another self' or a 'separated self', an idea Aristotle discusses in the Eudemian andNicomachean Ethics (2013, 2014). Aristotle writes: [T]he friend tends to be a sort of separated self. So perceiving a friend is, necessarily, a matter of perceiving oneself, in a way, and knowing oneself, in a way (EE 1245a35-37).
There is much discussion of how this argument is supposed to work. See Cooper (1977), Kraut (1989), and Veltman (2004) for discussion of the Aristotelian conception of friendship. 24 Chappell (2014) suggests that this is true of all 'objectual knowledge': "one striking characteristic of objectual knowledge is its exploratory nature" (Chappell 2014: p. 287). Knowing a friend seems to fit well within her conception of objectual knowledge (in that, for example, it essentially involves acquaintance with the friend). 25 Murdoch writes "why not consider red as an ideal end-point, as a concept infinitely to be learned, as an individual object of love" (Murdoch 1970: p. 29). Any feature of love will thereby be a feature of friendship, even if it does not tell us all there is to know about friendship. The feature of her account of love that is significant for this discussion of friendship is her claim that to love someone entails having knowledge of them, and progressing towards deeper and fuller knowledge. This, she suggests, is a constitutive feature of love.
On Murdoch's conception, love is not a state one can be in regardless of one's epistemic standing. It is not merely some kind of feeling that one can experience regardless of one's relations to reality. Love, as she conceives of it, involves knowledge of the object of love, which in the context of friendship is another person, the friend. To love someone involves knowing who they are as a person. Not all knowledge about another person will matter for friendship: for example, knowing the exact time that a friend's bus leaves for work seems irrelevant to the quality of one's friendship. But on the Murdochian account, knowledge of a friend's character traits or personal qualities, deeply held beliefs, and values is essential to love. 26 Such knowledge plausibly contributes to one's knowledge of them as a person.
The assumption that norms of friendship and epistemic norms are inherently independent of one another is thus precluded by a Murdochian account of love. The Murdochian account of love implies that the demands of love (and thereby of friendship) could not constitutively conflict with epistemic demands, because knowledge-and the deepening of such knowledge-is an integral aspect of friendship. On the Murdochian account, friendship is an epistemically involved, knowledgeinvolving, state, so the norms of friendship could not constitutively conflict with epistemic norms. If the Murdochian view is right, the parallel Keller draws concerning the independence of considerations of tastiness and healthiness is mistaken. Rather, friendship and epistemic considerations would be related in a way akin to tastiness and textural variety: whilst tastiness is not identical to texture, good texture contributes to tastiness.
On the Murdochian account, forming unjustified or inaccurate beliefs about a friend's character would thus detract from one's attempts to be a good friend. Since forming epistemically unjustified beliefs about who they are is a bad way to try to gain knowledge of the other person, and since on the Murdochian account friendship involves such knowledge, it would also be a bad way to try to be a good friend. On such a view, it could not possibly therefore be a requirement of friendship that one form beliefs (or belief-forming dispositions) about the character of one's friends in an epistemically questionable manner. The Murdochian conception of friendship thus rules out both direct and indirect epistemic partiality.
It is worth noting here that the Murdochian view of friendship does not rule out all false beliefs about one's friends. On this account one might be mistaken about trivial or unimportant details about a friend, and still be a good friend (this is plausibly true of most friendships). But it rules out bias regarding important features of one's friends, features that touch on who they are as a person. Of course, these important features of individuals are exactly the features concerning which partialists think one should be biased. And it is precisely because these features of our friends are important that partialists think we're justified in forming biased beliefs about them: only beliefs about important things would say anything about the quality of the friendship. That is, it seems plausible that we care only about our friends' beliefs regarding features of us that are important, features of us that are worth caring about. 27 These features are exactly the features which, on the Murdochian account, the good friend must get right. So, although the Murdochian account does not rule out all possible bias, it does rule out the kind of bias that epistemic partialists favour. The Murdochian account of friendship thus rules out the kind of inaccuracy that epistemic partialists recommend.

Motivating the Murdochian account
Should we accept the Murdochian conception of friendship? In this section, I will offer two reasons in support of it: first, it makes sense of some important intuitions regarding who is a good friend, and second, it makes sense of our expectations regarding the behaviour of good friends. Moreover, I will suggest that the supporter of the Murdochian conception of friendship can make use of attempts to explain away the intuitions in favour of epistemic partiality.
The Murdochian conception of friendship gains intuitive support from the idea that if one has sufficiently mistaken beliefs about another, one fails to relate to them. If I claim to love you, and yet I find that my beliefs about you are radically mistaken, it seems plausible to think that I was not really loving you at all. Rather, I was loving an illusion, a person whom I thought existed, but who was never really real. 28 I therefore seem to fail to relate to you in the way that would be required for true friendship. Insofar as the person I believe myself to love does not correspond to the actual person confronting me, then I am not loving them-or at least, not as well as I could.
To illustrate this, think of Dorothea Brooke's feelings of love for Edward Casaubon in Middlemarch. At the beginning of the novel, Dorothea believes herself to love Edward. She trusts him, makes plans with him, and holds him in very high esteem. But despite this it would be odd to say that she ever truly loves him, and this is largely because she is mistaken about who he is. 29 Dorothea conceives of Edward as a genius and loves him as such. But he turns out to be utterly unlike her vision of him, and she later comes to realise that he is a dry and narrow-minded pedant with pitifully little to show for his life's work. In the early period Dorothea fails to see Edward as he truly is, instead creating a stereotyped image of a solitary male 'genius' that has little correlation to reality. In loving this mere fantasy, Dorothea fails to love Edward himself. Her failure to truly know him (a failure for which he is himself largely responsible) seems to indicate a failure to love him. 30 The Murdochian conception of friendship enables us to say that Dorothea's love for Edward is deficient, as well as to explain why that is. Dorothea's feelings, hopes and beliefs about Edward are not guided by the person that he really is. The Murdochian account of friendship vindicates the intuitive connection between Dorothea's lack of knowledge and her lack of love: it allows us to say that her love is lacking because she is failing to relate to Edward as he really is. 31 That a failure to truly know the other person is a failure of friendship also makes sense of our intuitions about how a good friend typically behaves. Consider the ways that the person with mistaken beliefs about the other will act towards the person about whom they are mistaken. Since they lack an accurate vision of the other person, they lack a good sense of who the other person is and their complex constellation of merits and flaws. As such, they are likely to end up behaving in ways that are uncharacteristic of friendship, since they will not be responsive to the actual features of the friend. For example, the good friend typically perceives one's failures as such, sympathises with them, and is likely to help one in the things with which one struggles. These actions require in-depth knowledge of a person's limitations. Similarly, a good friend would be unlikely to saddle their friends with responsibilities they are unable to meet and would typically be well placed to offer advice on which responsibilities they should accept. Again, these require an accurate conception of the friend's flaws and failings, yet standing ready to do these things seems necessary for good friendship.
Take the above case of Dorothea and Edward: what would friendship demand here? Edward is wholly wrapped up in his work, but highly (miserably) insecure about it. Given this, it might initially seem that he needs support, encouragement, and reassurance. 32 But it is clear that Edward does not in fact have the capacity to write the monumental work that he dreams of. A true friend to Edward, it seems, would encourage him to let go of his increasingly desperate-and futile-wish to write a work of genius and instead focus on other goods that he might realise. Or they might try to assist him in building up a sense of self-worth that does not reside wholly in his work, so that his insecurity would no longer drive him to see the value of his life as only 30 Arpaly and Brinkerhoff (2018) raise a similar line of thinking: "Consider a husband who puts his wife on a pedestal, so to speak: he sees her as an angelic doll who could do no wrong even if she tried. We judge the husband negatively for overestimating the wife in this way and for taking steps to maintain such a sterile view of her. The husband's take on the wife objectifies her. He fails to see her for what she is: a person who flails and mistakes and acts as people do, sometimes with bad intentions" (Arpaly and Brinkerhoff 2018: p. 44). It seems no coincidence that in both examples the agents who fail to properly love their partners depend on heavily stereotyped images of the other person. 31 Were Dorothea's vision of Edward at this point to nonetheless be progressing towards a more adequate knowledge of him, it perhaps would involve some element of love on the Murdochian conception. However, it is notable that at this point in the novel, Dorothea's conception of him is resistant to such progression. It is not a deepening of her initial knowledge that is needed in order to truly know Casaubon, but a rejection of her initial illusions about him. 32 These are identified by Keller as'goods of friendship', and it is along these lines that Dorothea initially conceives of her role in the marriage. redeemable through his writing. What the good friend would not do is to express a high opinion of Edward's work and genius, for doing so would clearly be contrary to his interests, feeding a deep insecurity in him that cannot be satisfied. Acting in this way would require great insight into his character, and especially into his shortcomings. Many true acts of friendship, it seems, thus require a deep knowledge of one's friends.
Moreover, there is reason to be sceptical about the intuitions in favour of epistemic partiality. Arpaly and Brinkerhoff (2018) and Kawall both offer explanations of those intuitions that undermine the appeal of epistemic partiality. They claim that friendship is frequently causally connected with overestimating one's friends, which might lead us to expect friends to overestimate a person's good qualities. Still, this causal connection would not entail that any constitutive connection holds between the two and thus would not suggest that friendship requires bias. Another reason to doubt the reliability of the intuitions in favour of epistemic partiality is that such intuitions may not distinguish between requirements to hold positive beliefs about friends and requirements to fulfil other supportive roles. For example, as Kawall argues, there are plausibly requirements of friendship to have other positive attitudes towards one's friends and their endeavours (hopes and desires, for example). Moreover, friends are plausibly expected to behave in supportive ways despite their beliefs: a good friend would plausibly not voice their belief that Rebecca's poetry will be bad or dwell on how bad it was afterwards, and they would attend the reading despite fearing that the poetry will be bad.
Taken together, these provide good reason to think that a Murdochian account of friendship is plausible. It can explain why deep mistakes about one's friends detract from one's friendship. It can also explain why certain actions are characteristic manifestations of friendship, whilst others are contrary to friendship. Furthermore, there is reason to think that the intuitions that initially seemed to favour of epistemic partiality are shakier than they first appear. We have, then, good reason to think that friendship requires knowledge, and that friendship thus could not require epistemic partiality.

Concluding remarks
In Sect. 1 I noted that epistemic partiality has sometimes been motivated by an overarching conception of friendship. As Stroud puts it, on that conception friendship is "in some important sense based on your friend's character and merits" (Stroud 2006: p. 511). The general line of thinking is that since friendship is based on one's friend's character and merit, good friendship requires maintaining a positive view of the friend, and that bias can thus contribute to being a good friend. 33 In some sense, this overarching conception of friendship seems right: friendship with someone in whom one could see no redeeming qualities seems impossible. I have rejected the idea that good friendship requires bias, suggesting that good friendship instead requires knowledge of the other. However, the overarching conception of friendship whereby it requires seeing merits in a friend is compatible with the rejection of epistemic partiality. We can accept that friendship is in some sense based on one's friend's character and merits, so long as we accept that it would be impossible to be friends with someone in whom there were no good qualities. After all, given this conception of friendship the only way to be friends with them would be to form biased or mistaken beliefs, but I have suggested that such beliefs undermine friendship. This seems like a generally plausible claim: it seems impossible to imagine a good friendship involving someone with no good qualities. Furthermore, it has historically had defenders. 34 Moreover, the Murdochian conception of friendship suggests that we can be mistaken about who our friends are. In this sense, the account entails that friendship is a factive state. One might, for instance, like someone and enjoy their company, and therefore regard them as a friend. But if one's views of them importantly diverge from who they truly are, this in itself undermines the friendship.
However, that friendship is in some sense based on the friend's character or merits need not lead to the thought that friendship requires an overall positive evaluation of the other. Nor need it imply that the best friendships will be friendships with those in whom one perceives the most good qualities, or even the best qualities. It is presumably something like one of these thoughts that would be necessary to bridge the gap from the thought that friendship is in some sense based on the friend's character and merits, to the thought that epistemic partiality can strengthen friendship. It is plausibly essential for friendship that one sees something good in one's friends, but this can coexist with awareness of negative qualities in them. A friend might be kind and caring, which could form the basis of a good friendship, whilst nonetheless having a terrible lack of humour and a tendency to complain. Seeing a good quality in one's friend does not preclude also seeing their negative qualities, even overall negative qualities. 35 Moreover, that friendship is based on a friend's character or merits need not suggest that the more good qualities one perceives, or the better those qualities, the better the friendship. One can recognise that someone possesses some positive qualities without particularly appreciating those qualities, in which case such qualities would not be a good basis for friendship. 36 I can recognise, for example, that someone has a lively sense of humour, but unless I care about or share their sense of humour, this cannot be a basis for friendship. Friendship thus seems to be based on something more specific than simple belief that another person has positive qualities, something like apprecia- 34 Aristotle seems to have held this view (see Cooper 1977). Elder (2014) defends the related view that bad people cannot be good friends. 35 There are plausibly reasons not to express one's beliefs about one's friends' flaws (or at least, not too often). And we would also be sceptical of the claim that someone who attended closely and critically to all potential flaws and failings in another person was a good friend. But these are consistent with this point. 36 Aristotle seems to make a related point when he notes: Goodwill, then, seems to be the first principle of friendship, as pleasure is the first principle of love. For no one loves another is he has not been pleased beforehand by his appearance. But finding enjoyment in the form of the other does not mean that one loves him; this happens only when one longs for him in his absence and wants him to be there. So too, though people cannot be friends if they have not already developed goodwill for each other, those who have goodwill are nevertheless not friends; they merely wish what is good on those towards whom they have goodwill, and would not cooperate with them in any action or go to any trouble on their behalf (NE 1167a4-10). tive awareness of another's positive qualities. Importantly, such appreciativeness need not correspond to either the number or the extent of the good qualities that one perceives. 37 Rejecting epistemic partiality for the Murdochian reasons I have offered is thus compatible with accepting the idea that a friend's character and merits do matter for friendship, and indeed that friendship is in some respect based on such features.
To conclude, the Murdochian conception of friendship that I have sketched entails a rejection of an assumption about friendship that has gone unquestioned in most of the literature on epistemic partiality. The Murdochian conception suggests that the norms of epistemology and friendship are not independent of one another, since friendship involves knowledge of the other person. As such, on the Murdochian conception the norms of friendship are informed by those of epistemology, so there could not be constitutive conflict between the two. I have not offered a full defence of this conception of friendship, but I have suggested that it makes sense of our intuitions regarding the impact of importantly mistaken beliefs on friendships, and that it also makes sense of the ways in which we would expect the good friend to behave. There is thus good reason to think that friendship is an epistemically rich state, and that discussions of epistemic partiality go awry in ignoring one of its core features.