Les ateliers de l'éthique The Ethics Forum Does Friendship Give Us Non-Derivative Partial Reasons?

pourra se régler au


ABSTRACT
One way to approach the question of whether there are non-derivative partial reasons of any kind is to give an account of what partial reasons are, and then to consider whether there are such reasons. If there are, then it is at least possible that there are partial reasons of friendship. It is this approach that will be taken here, and it produces several interesting results. The first is a point about the structure of partial reasons. It is at least a necessary condition of a reason's being partial that it has an explicit relational component. This component, technically, is a relatum in the reason relation that itself is a relation between the person to whom the reason applies and the person whom the action for which there is a reason concerns. The second conclusion of the paper is that this relational component is also required for a number of types of putatively impartial reasons. In order to avoid trivialising the distinction between partial and impartial reasons, some further sufficient condition must be applied. Finally, there is some prospect for a way of distinguishing between impartial reasons that contain a relational component and partial reasons, but that this approach suggests that the question of whether ethics is partial or impartial will be settled at the level of normative ethical discourse, or at least not at the level of discourse about the nature of reasons for action.
Most of us, no doubt, act partially towards our friends. In fact, we might be regarded as being poor friends if we did not; much of the time there are good reasons for us to do so. What is less clear is whether there are what I shall call 'non-derivative' reasons to be partial towards one's friends: whether the mere fact that one stands in the friend relation to another person provides a reason to be partial towards that person. One way to approach the question of whether there are nonderivative partial reasons of any kind is to give an account of what partial reasons are, and then to consider whether there are such reasons. If there are, then it is at least possible that there are partial reasons of friendship. It is this approach that will be taken here, and it produces several interesting results. The first is a point about the structure of partial reasons. It is at least a necessary condition of a reason's being partial that it has an explicit relational component. This component, technically, is a relatum in the reason relation that itself is a relation between the person to whom the reason applies and the person whom the action for which there is a reason concerns. The second conclusion of the paper is that this relational component is also required for a number of types of putatively impartial reasons. In order to avoid trivialising the distinction between partial and impartial reasons, some further sufficient condition must be applied. Finally, there is some prospect for a way of distinguishing between impartial reasons that contain a relational component and partial reasons, but that this approach suggests that the question of whether ethics is partial or impartial will be settled at the level of normative ethical discourse, or at least not at the level of discourse about the nature of reasons for action.

PARTIAL AND IMPARTIAL REASONS
Normally, when philosophers speak about reasons and partiality, they ask whether there are reasons to be partial in some way towards someone. In this paper, I speak of 'partial reasons'. It is worth saying something about why the discussion here concerns partial and impartial reasons, rather than reasons to be partial and reasons to be impartial.
There is an ambiguity in the formulation 'reason to be partial' and in its counterpart 'reason to be impartial' that requires some clarification. There are some situations in which it makes sense to describe and agent as acting partially or acting impartially. Consider a judge at a high-diving meet. A judge acts impartially with regards to each contestant if she both awards that contestant the score his performance merits and also awards to all other contestants the score that their performances merit. The judge acts partially if she adds or subtracts points on some basis apart from diving merit.
Here the sense of acting partially is quite clear. The judge may have reason to act partially (she's been bribed, or she knows that if the best competitor wins, he will use his success to promote an evil agenda, and so on). But, here the partiality does not pose a very interesting philosophical problem. Acting partially in this sense, choosing to favour (or disfavour) someone, could be done for many reasons, some of which might be best understood as impartial. The judge may have been trying to maximise total social wellbeing by skimming points from the wicked diver's score. Although that kind of action can be described as acting partially, it is not what we are after when we are thinking about impartiality and friendship.
What I do think we are after is an account of whether (and if so, how) friendship and other special relationships change the nonderivative reasons that we have. This is a question not of whether we should favour (or disfavour) our friends in certain contexts. Rather, it is a question about whether the fact that someone is my friend gives me a basic or non-derivative reason to act in certain ways that is different from the kind of reason I have towards people to whom I do not bear that relation. The target, then, is whether the reasons provided by friendship have a special character, and whether that character is partial.
An intuitive, if imprecise, way to distinguish between partial reasons and impartial reasons is this: a partial reason is a reason to be partial towards someone (or towards his interests) in some way that goes beyond what is required by reasons that are impartial. 2 Thinking of partial reasons this way requires us to take impartial reasons as the primitive notion, and here we run into a difficulty in making much use of this way of making the distinction. Impartial reasons, so I shall argue, are sufficiently similar in structure to partial ones that it is not clear whether one can distinguish between impartial reasons with a relational component and partial reasons. The intuitive way of putting the distinction is no doubt correct as far as it goes, if there is a distinction to be made, but it leaves all the work to be done of specifying how to identify when there is a reason that is not impartial. So, we require a more precise definition. It is sometimes easier to get a grip on problems in practical reason by looking at issues in theoretical reason. So, I shall do some of the groundwork for an account of an impartial reason for action by first attempting an account of partiality and impartiality as regards friendship in the epistemic sphere. 3 Starting with epistemic reasons will help to highlight some of the general difficulties involved in spelling out what a partial reason is, and it will also make clearer some of the particular difficulties involved in giving an account of partial reasons for action.
The canonical notion of a normative reason for belief is that of an evidential reason. There are varying notions of evidence, but here is a simple one that will do for purposes here. For any proposition that is not a logical truth or a logical falsehood, p is evidence for q just if the conditional probability of q given p is higher than the conditional probability of q given not p. 4 This account of evidence provides a basis for an account of what an evidential reason for belief is. Fact f is an evidential reason for a to believe x just if f is evidence for x. It will not be hard to pick holes in this account of an evidential reason for belief, but the deficiencies do not affect the argument here.
Let us now assume that we have a theory of evidence e that gives us the genuine conditional probabilities for everything. With e in the background, we can now spell out what it means to be epistemically partial or impartial towards one's friends. There are, in fact, two general ways in which one might be epistemically partial or impartial towards one's friends: with regards to their testimony (broadly construed) and with regards to what you believe about them.
We can spell out the first way like this: you are epistemically impartial towards someone just if in light of her testimony and because of your relationship to her you adjust your conditional probabilities as required by e. You are epistemically partial if in light of her testimony and because of your relationship to her you adjust your conditional probabilities in a way different to that required by e. 5 Given the way that I have set up the first notion of epistemic partiality and impartiality, the natural way to set up the second one is like this: you are epistemically impartial towards someone just if you form beliefs about her in the way required by e, because of your relationship to her. And you are partial towards her just if you form beliefs about her because of your relationship to her in some way that does not correspond to that required by e, because of your relationship to her.
These definitions of being epistemically impartial and being epistemically partial may be problematic in at least one respect. Being impartial or partial involves adhering to or deviating from what is required by e because you are someone's friend. But, there is room for deviance with this because. It may be that people who are friends with a particular person are also systematically targeted by a mad scientist, who secretly uses an epistemic biasing ray to make sure that those people's beliefs do not conform to e. One's epistemic partiality may be because someone is your friend, but for indirect reasons. For sake of brevity, I shall not add non-deviance clauses throughout this paper, but, and this is especially pertinent in the case of partial reasons, they most likely belong.
So far I have just said something about you, namely under what conditions you would count as being epistemically partial towards your friends. Now I need to say something about reasons and partiality. As terms of art, I shall speak of partial and impartial reasons. Just to avoid confusion, it is important to remember that 'partial' here is used in the sense of 'partial towards' rather than in the sense of 'incomplete'. Here are the definitions of impartial and partial reasons; again, there will be two types: First: a testimony-given reason for belief is impartial if testimony t about x is a reason for a to update his belief about x in light of t according to e, because of a's relationship to x. A testimony-given reason for belief is partial if tes- timony t about x is a reason for a to update his belief about x in light of t not according to e, because of a's relationship to x.
Second: fact f is an impartial reason for a to believe x about person p just if f is a reason for a to update his belief about x according to e, because of a's relationship to x. Fact f is a partial reason for a to believe x about person p just if f is a reason for a to update his belief about x not according to e, because of a's relationship to x.
Epistemic partiality towards one's friends, to put things less technically, involves giving excessive weight towards one's friends' testimony because those individuals are one's friends, or being excessively trusting about one's friends' behaviours and motivations because they are one's friends. 6 Let me note for now, as will become clear in the discussion of partial reasons for action, that this account of the structure of partial reasons for belief is incomplete in a way that will prove problematic to resolve.

DERIVATIVE AND NON-DERIVATIVE REASONS
Having set up a first attempt at an account of partial and impartial reasons, so far only in the epistemic sphere, I now must say something about derivative and non-derivative reasons. It will be best to treat a non-derivative reason as the primitive notion, and then to define a derivative reason as any reason that is not non-derivative.
Some theories tell us about the source of reasons. For example, a teleological theory of reasons will tell us that the source, or basis, of reasons for action is goodness. One may imagine a simple teleological theory that asserts the following about reasons: TTR: Fact f is a reason for a to phi just if f makes a's phiing good This simple theory is monistic in the sense that it only allows for one source of reasons: goodness. Strict evidentialism about reasons for belief is another theory of this kind. It says the only reason to believe something is that the fact that is the reason is evidence for the contents of the belief for which it is a reason.
Theories about the sources of reasons need not be monistic. Moderate evidentialists might accept that both evidence and more general considerations in making an inference to the best explanation (IBE) are sources of reasons for belief. Likewise, moderate deontologists believe that there are many sources of reasons for action, not just goodness. Moderate evidentialism and moderate deontology are both pluralistic theories of the sources of reasons.
Sometimes, it is difficult to tell whether you adhere to a monistic or pluralistic theory of reasons. Many philosophers think that, in general, there is good reason to believe the proposition arrived at by IBE. But, they can disagree as to why. Our moderate evidentialist is committed to saying that something's being the result of making the inference to the best explanation is a non-derivative reason for believing it. A strict evidentialist might say that there is usually a reason to believe the conclusion of an inference to the best explanation, but that reason is a derivative of the tendency of IBE to agree with, or point towards, what the evidence suggests.
Putting things this way allows for a reasonably intuitive picture of the difference between derivative and non-derivative reasons; now, we can be moved towards a precise account. Remember our simple reasons schema: Fact f is a reason for agent a to phi: Non-derivative reasons: Fact f is a non-derivative reason for a to phi just if the most direct explanation of f's being a reason for a to phi appeals to a source of normativity. Derivative reasons: Fact f is a derivative reason for a to phi just if the most direct explanation of f's being a reason for a to phi does not appeal to a source of normativity.
When asking about partiality, it is the non-derivative reasons that are of interest. On a very great number of normative ethical theories, there will sometimes be reasons to be partial towards one's friends. The interesting question is whether the mere fact that one is friends with someone gives her partial reasons towards her friend.

PARTIAL AND IMPARTIAL REASONS FOR ACTION
I have now offered a rough account of derivative and non-derivative reasons (in general). I have also offered a tentative account of partial and impartial reasons for belief. In this section, I want to discuss partial and impartial reasons for action. I shall discuss two proposals for partial and impartial reasons, the first paralleling the model used in belief, and the second working in a slightly different way. I shall argue that neither proposal about partial reasons for action is correct, but that the logical components of partial reasons described in those proposals are necessary parts of partial reasons for action. Showing this much does not, of course, settle the debate about whether there are partial reasons for action from friendship, or any other source.
What it does show is that making a distinction between partial reasons for action and a common type of impartial reason for action is problematic. Because it turns out that partial reasons and impartial reasons of a certain type share a distinctive structural feature, excluding the possibility of partial reasons for friendship on the grounds of that feature comes at a high cost.
Reasons for action take roughly the same shape as reasons for belief. Being a reason for action is a property of a fact, and it is the property of standing in the reason relation to an agent and an action. Like reasons for belief, reasons for action can have varying degrees of strength, or, if you prefer, varying weights. I might have strong reason to go to the store, but at the same time also a weaker reason not to go to the store.
Unlike for reasons for belief, it is not clear that there is a canonical type of reason for action. The nearest parallel to evidence may, in the case of action, be goodness, although I take it that saying goodness provides the basis for reasons for action is controversial on a large number of views. Nonetheless, it is worth seeing how far we can get with an account of a partial reason for action that uses goodness in the same way that the putative account of a partial reason for belief used evidence.
Let us assume we have a complete theory of goodness, g. g will play a part in the following discussion. I shall make reference to it periodically. It is supposed to function in the same way that e does in the case of reasons for belief. Now we can make a first attempt at defining partial and impartial reasons for action. Impartial reasons for action: Fact f is an impartial reason for action just if f is a reason for a to phi regarding person p, and the strength of that reason is as g requires, because of a's relationship to p. Partial reasons for action: Fact f is a partial reason for action just if f is a reason for a to phi regarding person p, and the strength of the reason is not as g requires, because of a's relationship to person p. 7 I have added 'regarding person p' to allow a place in the relation for mentioning the other relevant party, while leaving it open as to how the other relevant party fits in.
This analysis may be problematic, although not obviously false, for two reasons. The possible worries are significant in a way that will be discussed later in the paper. For the moment, I shall just observe what those worries are.
First, we can imagine, in the case of belief, two different factors that affect how much reason we have to believe something. One factor might be evidence, another might be betterness of explanation. In the case of action, we might have the same kind of pluralism about sources of reasons. For example if we are moderate deontologists, we can imagine two or more factors that would provide reasons: that something promotes virtue and that something promotes knowledge, for example. What is apparently odd in this analysis is that it presumes that goodness simpliciter is not the only factor determining what reasons we have. If goodness did all the work, then it would be impossible to have a reason in which some factor (here a special relationship) brings the strength of the reason out of harmony with the amount of goodness that the action produces. There is nothing that rules out a pluralistic theory for the sources of reasons for actions that treats goodness as the primary source of reasons for action, but allows that other factors are also sources of reasons for action. But, one might worry that once goodness is a standard part of the picture, there may be a temptation to assimilate all considerations in favour of action into the theory's axiology.
The second reason why this analysis of partiality may be worrying comes out when we think about what work 'regarding p' is doing. The partial formulation invites a reading that suggests that who p is matters. That p is one's friend is the explanation of why the reason is of a different strength to what g requires. Again, mixed views are not by any means impossible, and it seems to me that a teleologist who is tempted by partiality in general will just say that the value is different when p is one's friend from when p is not. A parallel move is not available in the belief case with e. That there is partiality in value is a live possibility. It might be better, ceteris paribus, to save one's friend from drowning than to save a stranger. And, if we accept the teleological theory of reasons, then we will have a stronger non-derivative reason to save a friend than to save a non-friend. That sounds like partiality, but it is a partiality in reasons only because it is a partiality in axiology. We can give a definition of a putative partial reason for action that is partial due to a partialist axiology like this: Partial reason: Fact f is a partial reason for a to phi regarding p just if a's reason to phi regarding p is of a different strength than his reason to phi regarding q, and that difference in strength is because a's respective relationships to p and q makes, according to g, the value of a's saving p different from that of a's saving q.
This latter definition is only of interest if we are willing to invoke g. But, this account provides a template that can be amended slightly to work with non-teleological theories: Partial reason: Fact f is a partial reason for a to phi regarding p just if a's reason to phi regarding p is of a different strength than his reason to phi regarding q, and that difference in strength is because of a's respective relationships to p and q.
Here we have two putative accounts of a partial reason for action, and the difference between them is that one absorbs partiality into axiology, and the other does not. Crucially, they have in common an approach to thinking about partiality: that reasons may vary in strength because of the relation of the agent to the person whom the relevant action concerns.
To provide a parallel account of an impartial reason, we need to make it explicit that the relationship of a to p does not affect the strength of the reason: Impartial reason: Fact f is an impartial reason for a to phi regarding p just if a's reason to phi regarding p is of the same strength as his reason to phi regarding any person x, regardless of the relationship that a bears to x.
Unfortunately, this version of impartiality rules out various self-identified impartial ethical views, for reasons that will be discussed in the next section.

PROBLEMS WITH IDENTIFYING PARTIAL REASONS
Earlier in the paper, I said that, while these putative accounts of partial reasons for action provided some necessary structural features required for partial reasons, they prove incomplete. To see why this is so, it will be helpful to start by thinking about the account just given of an impartial reason.
This account of an impartial reason for action makes explicit mention of the following relata: a fact (that has the property of being the reason), an agent (for whom there is a reason), an action (for which there is a reason), an object (the person whom the action concerns), and a relationship (between the agent and the object). Nothing yet has been said about the nature of the fact that is the reason, but the nature of the fact matters.
There has been a tacit assumption in the construction of both partial and impartial reasons for action that the fact that is the reason is a consideration that counts in favour of the agent performing the action regarding the object. One possible consideration that might be taken to count in favour of certain actions is that I am someone's friend. That so-and-so is my friend is a reason for me to phi regarding him. If we are permissive about what kind of facts count as reasons, then we will admit facts about my relation to the object of my action as being reasons themselves.
In order to draw a structural distinction between partial and impartial reasons for action, we cannot be quite so permissive. Otherwise partial and impartial reasons will have the same structure: fact f is a reason of some strength for agent a to phi regarding p. Suppressing the relation of a to p as a relatum in the reason, and putting the relational information into the fact itself would push the focus of an account of partial reasons towards the nature of the fact. Doing so would complicate the analysis, and there are good independent reasons for making the relation between a and p one of the relata. What is surprising is that it is apparently impartial theories of ethics that provide the strongest case for including a relation relatum in just exactly the way it occurs in the partial reason relation.
If we must include the relation relatum for impartial reasons in the same way that we must for partial reasons, then we lose our ability to distinguish between partial and impartial reasons based on whether the reasons include a place for the relation between the agent and the object. The remainder of this section is devoted to showing why apparently impartial ethical theories generate reasons that require a place for this relation. In the next section, I shall discuss some unsuccessful proposals for trying to rescue a clear distinction between partial and impartial reasons. Let us consider one ethical theory that might reasonably be considered impartial: telic egalitarianism. Telic egalitarianism takes various forms, but a common one says that it is good to add wellbeing to a society, and it is good to decrease inequality. Likewise, it is bad to decrease well-being in a society, and it is bad to increase inequality. We can represent a simple form of telic egalitarianism for a two person society like this: V is the total value or goodness in a society. W 1 is the wellbeing of one person in that society, and W 2 is the well-being of the other person in that society. In this version of TE, each additional point of well-being added increases the total goodness by 1/2 of a point, and of course each point subtracted reduces the total goodness by the same amount. Each one point increase in inequality decreases the total value in society by 1/4 of a point, while each one point decrease in inequality increases it by the same amount. As you can see, adding one point of value to the person who is better off in society will increase V by 1/4 of a point, whereas adding that same point of well-being instead to the worse off person increases V by 3/4 of a point.
Under TE, there is always a reason to improve someone's wellbeing. And, ceteris paribus, there is always more reason to distribute a fixed amount of well-being to the worse off person rather than to the better off person. Here we now have an instance where we cannot know the strength of the reason without know knowing what relation a bears to p. If a bears the relation of being better off than to p, then a will have a stronger reason to distribute an available point of well-being to p than to distribute it to herself. If a bears the relation to p of being worse of than, then a will have more reason to distribute the resource to herself than to p.
In the case of TE, in order to know how strong a reason is, we must know the relation that a bears to p. The strength of the reason will vary depending on, and because of, the relation that a bears to p. So this impartial ethical theory evidently has need of the basic reason structure required for a theory of partial reasons.
In fact, TE is not the only impartial ethical theory that requires knowledge of the relation between a and p to determine the strength of the reason; various forms of cooperative utilitarianism will require a relational relatum to explain the strength of the reasons one has, and indeed whether one has any reason at all. That there are impartial ethical theories that require structures of the kind given here as proposals for partial reasons shows that these proposals are inadequate as accounts of partial reasons. 8 Accepting the proposed accounts of a partial reason given so far as correct would turn impartial ethical theories into partial ones, blurring the intelligibility of the distinction.

CONCLUDING WITH THREE PROPOSALS
This last section of the paper will conclude with three proposals for distinguishing between partial and impartial reasons for action with the upshot that the debate about whether ethics is fundamentally partial or fundamentally impartial may well need to be settled at the level of normative ethics, rather than by the theory of reasons. It is not clear what further amendments could be made to the reason relation that would allow for a clear distinction between partial reasons and impartial reasons with important relational features. The alternative to amending the structure of reasons further is to mark out the feature(s) of the relation between an agent and an object that identifies that relation as the sort that occurs in partiality. Three proposals for how to do so are presented here. Each raises significant concerns.
The first proposal is that we can distinguish between the kinds of relations occurring in partial reasons from and those occurring in impartial, but relational, reasons by trying to capture the distinctive features of impartial, but relational, reasons. This proposal is problematic because it is not clear that there is a distinctive class of relations that appear only in impartial, but relational, reasons.
Consider again the case of telic egalitarianism. Here, the proposal to be rejected is that positionality is the distinctive feature of the kind of relation that appears in the impartial but relational rea- son. The relation in telic egalitarianism is positional with respect to well-being. Having more well-being gives you a stronger reason to distribute resources to someone with less well-being than that person would have to distribute well-being to you. But, positional goods are not unique to impartial ethical theories. Rank, for example, is a positional good. And, it is plausible on a partialist account of ethics that those with a higher rank within an organisation have obligations to those of lower ranks that differ from the obligations of those with lower ranks towards those with higher ranks. For example, one might argue that a private in the army has an obligation to obey a reasonable order from his sergeant, but that the sergeant has no obligation to obey orders from the private. Likewise, the sergeant may have a strong, special duty of care towards a private that the private lacks, or has only more weakly, towards the sergeant. A second proposal is that partial reasons are distinguished by the importance of the relation between the two particular individuals involved, rather than only the general features of the relation. This may be a tempting view with friendship. My relations with each friend of mine are slightly different from all the others. We have unique ways of interacting with each other, and we may feel that each of our individual friendships brings a different set of particular commitments and obligations. To know what my reasons are towards a friend, I would need to know who that friend is and what our friendship is like.
To put this proposal a different way, the strength of the reason depends not only on my having the formal relation of friendship with someone, but also on who each of us is. It is a relation in which the identity of the people involved matters fundamentally. Put this way, it is less clear just how to interpret this claim. Whether identity matters in ethical theories is a matter of some controversy. To say that it does denies a very plausible supervenience claim that the ethical features of one's life are determined solely by the way one is and one's circumstances. A complete description of an individual's properties and circumstances, save for her identity, would seem to suffice to identify all the morally relevant information about her. If Jim and Joe are identical in all respects, save numerically, and they are in identical circumstances, it is not clear how the mere fact that Jim is Jim and Joe is Joe would change the obligations that they bear to those around them and that those in their surroundings bear to them.
Although it seems doubtful that it can, if the claim that identity does matter for ethics could be defended, then perhaps that a rea-son contains a relational relatum in which identity matters is a sufficient condition for identifying it as a partial reason for action. It is not clear, however, that it is a necessary condition. While perhaps friendship and parenthood would meet this condition, partial reasons deriving from the obligations of rank will clearly not meet this condition.
Finally, we may consider the proposal that what formally makes a reason a partial reason for action rather than an impartial, but relational, reason for action is that the relation involved is one of the partial ones. This definition is uninformative, but it may be formally correct. It may also be useful for understanding the debate about whether ethics are partial or impartial. To see why, it is worth briefly reviewing the project here.
One difficulty in making substantive claims in normative ethics is that there is always the danger debates will reduce to the trading of intuitions. This is by no means the rule, but it is always a danger. Sometimes, therefore, it can be helpful to see if there are any structural tools available for analysing the competing positions to see if they can be ruled in or ruled out on some basis that is more reliable than the disputants' surface-level moral intuitions. The initial project in this paper was to provide an account of impartial and partial reasons for action to see if either contained features that would make the existence of such reasons implausible.
What became apparent as the analysis progressed is that it is difficult to distinguish partial reasons for action from an important class of impartial reasons for action, namely those that have fundamentally relational features. This dampened the prospect for using arguments about the structure of reasons to rule in or rule out partial or impartial reasons.
The last proposal provides a formal account of a partial reason that distinguishes it from an impartial reason based on the particular relation involved, not on some general features of partial relations that is lacking in impartial ones. While this makes an analysis of partial and impartial reasons much less helpful, perhaps not helpful at all, for resolving questions about whether ethics is partial or impartial, it does tell us what work there is to be done. There must be substantive work done at the level of normative ethics to explain clearly which relations are partial and which are impartial. This suggests, too, that any resolution to the debate about whether ethics is fundamentally partial or impartial will be adjudicated at the level of normative ethical discourse, or at least not through an analysis of reasons. This paper has benefited greatly from the input of a number of philosophers. I wish to thank Dirk Baltzy, John Bigelow, Monima Chadha, Karen Green, Jessica Pepp, Sarah Stroud, and Christine Tappolet. In addition to those named in particular, I received a number of valuable comments from others who attended presentations of this paper at the Monash University Philosophy Department and at the Séminaire international: Friendship and Partiality in Ethics, hosted by the Centre de recherche en ethique de l'Université de Montréal. 2 Another way of understanding partial reasons would involve reasons that we only have when certain relationships exist between ourselves and other people. Although I shall not argue the point in this paper, I believe such reasons can be treated as special cases of the kinds of partial reasons I consider here. 3 For a comprehensive study of the relationship between epistemic partiality and friendship, see Stroud, Sarah, "Epistemic Partiality in Friendship",Ethics,vol. 116, There are particular difficulties with providing evidence for propositions that already have a probability of one or zero, as on standard models the probabilities of such propositions are the same under all conditions. If, for example, a logician provides testimony that a particular logical truth is true, that will not change the probability that the logical truth is true, and yet one might want to count such testimony as evidence. 5 We may want a stronger claim here with the upshot that the deviance from e is effected because the person is your friend. 6 Christine Tappolet has made a number of useful suggestions, which I have taken up, in giving the formulation for partial and impartial epistemic reasons. 7 I thank Christine Tappolet for stressing the necessity of the 'because of' clauses. Without them, the possibility remains open that the bias in favour or against the person whom the action is regarding might be due to inherent features of that person, rather than her relation to the agent. 8 John Bigelow, Dirk Baltzy, and Karen Green offered examples to me of further difficulties with the proposed accounts of partial reasons. They observed that there are classes of impartial ethical theories that factor in agent responsibility and an agent's causal relations, and that these, too, would require a relational place in the reason relation.