12.—
The Normative Thoughts of Friendship

Since the commitments of peer love add to the complexities of the integration problem for the agent of integrity, they must now be integrated with those of self-respect and self-esteem, with those of simple respect and esteem for others, and with those of parental love. Eventually, friendship and neighborly love must be integrated with each other. To accomplish this there must be a solution to the priorities problem. Thus the goal of this chapter is to provide a partial solution to the integration problem by giving a partial account of the priorities of an agent of integrity who has respect and esteem for self and others and who has the commitments of peer love as in friendship.

The solution is partial in the sense that I address here only the issue of priorities as they bear on the agent's other-restricting normative beliefs. The self-restricting beliefs of friendship are important in their own right, but I believe the continuity of the inquiry and the development of the conceptual scheme I wish to defend is best served at this point by providing an analysis of the other-restricting normative beliefs of friendship parallel to the other-restricting normative beliefs of parental love. Here, then, I consider the agent's priorities regarding what he or she ought to do in cases in which there are conflicts between the interests of a friend and the interests of others who are peers but not loved ones. The analysis focuses on what agent-centered restrictions might emerge in the normative thoughts of an agent who has simple respect and esteem for others and who loves his or her friends. The analysis includes both the central cases and their deviants.

The deviants are very important. If one looks at the work of the ancients on friendship, one will find a persistent concern for the influence of friendship on other matters and for what the limit of such influence should be.

This is true whether one is considering Plato, Aristotle, Cicero, or Seneca.^[1] Contrary to being a moral hobbyhorse, as Kant once called it,^[2] the concept of friendship was always discussed in the context of other goods among the ancients. It is no less important that we should so understand friendship in relationship to the goods of respect and other goods. How we understand this relationship turns on how far from the central case an instance of friendship is. Deviants involving friends with bad character do not yield the same normative dimensions as the central case. For friends with bad character enter the loving agent's deliberative field in a different way than those with good character. And, as we will see, the regulative functions of impartial respect are different for the deviant cases than for the central case, with the determinate function of friendship in the central case being much more extensive than in the deviant cases.

In this regard, some preliminary comments on current Kantian literature are in order. The treatment of friendship is more extensive among Kantian internalists than is their treatment of parental obligations. That said, however, the treatment of friendship is nonetheless very incomplete. Marcia Baron probably has written more about friendship than any of the other Kantians and has shown quite admirably that we do not accept a conception of friendship that is not regulated by impartial respect.^[3] But even in her work we do not get any developed treatment of how the CI procedure settles the priorities of friendship and impartial respect across a range of deliberative contexts. She merely alludes to the latitude allowed by imperfect duties. Baron also shows quite well that some treatments of Kant on friendship are caricatures.^[4] However, it is not enough to show that some conceptions of friendship are objectionable because they do not exhibit the regulative effects of impartial norms. Nor is it enough to show this and that some criticisms of Kant are based on caricatures. What the Kantian must show is that there is a clear rational procedure for setting the priorities of an agent who is a loving friend and who is respectful of others where that procedure employs impartial respect as an asymmetrical regulative norm. I cannot see

[1] . See Plato's Lysis , books 9 and 10 of Aristotle's Nicomachean Ethics , Cicero's De Amicitia , and Seneca's Epistle 60 and Epistle 63.

[2] . See Immanuel Kant, Lectures on Ethics , trans. Louis Infield (Indianapolis: Hackett, 1963), 209.

[3] . See Marcia Baron, Kantian Ethics Almost Without Apology (Ithaca: Cornell University Press, 1995), esp. chap. 4, 117–45.

[4] . See Marcia Baron, "Was Effi Briest a Victim of Kantian Morality?" in Friendship: A Philosophical Reader , ed. Neera Kapur Badhwar (Ithaca: Cornell University Press, 1993), 192–210.

that the Kantian literature even addresses this issue. To do so requires extensive analysis of different kinds of contexts and how the CI procedure works for those contexts. To be sure, it is open to the Kantian to abandon the purely procedural interpretation of the CI procedure and appeal to the substantive interpretation with its emphasis on the Formula of Humanity in accounting for latitude. But it is still a requirement of inquiry that the account of that latitude be shown to be restricted by a conception of respect for human dignity that is absolute, nonscalar, nonadditive, and immune to trade-offs in terms of other values. Currently we do not have such an analysis from the Kantian perspective, despite the recent work by Baron, Nancy Sherman, and Allen Wood.^[5] Consequently, in what follows I will provide what appears to me plausible Kantian possibilities in arguing for the Aristotelian conception. Central to the discussion will be the issue of agent-centered restrictions.

1.

The issue of agent-centered restrictions involved in friendship is an issue of other-restricting norms that prohibit the agent from acting on impartial norms in order to bring about the best available state of affairs from an impartial point of view. That is, they restrict what the agent can do for others in the name of simple respect and esteem and impartial sympathy. To the extent to which these norms are rational, I will argue, they exemplify the regulative function of our concept of friendship on our concept of impartiality. They do not reflect the rational result of treating human dignity as a value that is absolute, nonscalar, nonadditive, and immune to trade-offs. I will attempt to give an account of the scope and limits of such restrictions and how they emerge within an Aristotelian framework.

The important thing here is to understand the contexts in which these restrictions can have a place in the norms of a person who has simple respect and esteem for others among his or her categorical commitments. We can distinguish these contexts both by the kinds of interests at stake and by the kinds of deliberative patterns that emerge in these contexts, as we did in chapter 10. Broadly, the possibilities of contexts for conflicts between the

[5] . See Baron's Kantian Ethics Almost Without Apology ; Nancy Sherman's Making a Necessity of Virtue: Aristotle and Kant on Virtue (Cambridge: Cambridge University Press, 1997); and Allen Wood's "The Final Form of Kant's Practical Philosophy," in Kant's "Metaphysics of Morals," ed. Nelson Potter and Mark Timmons, Spindel Conference 1997, Southern Journal of Philosophy 36, supplement (1998): 1–20.

interests of a friend and the interests of others held in simple respect or esteem are the following: (i) where categorical interests of the friend conflict with noncategorical interests of others, (ii) where noncategorical interests of the friend conflict with categorical interests of others, (iii) where categorical interests of the friend conflict with categorical interests of others, and (iv) where noncategorical interests of the friend conflict with noncategorical interests of others.

Within these contexts, we can evaluate certain kinds of impartial deliberative patterns for their rationality. Here it is important to understand that what follows is not the sketch of a decision procedure so much as a description of a moral psychology. It assumes that the agent in these situations has a reasonably clear picture of what is at stake in these contexts. It then describes the deliberative pattern rational for agents with a certain character in these contexts. I say this to avoid the misunderstanding that I am attributing some mechanical decision procedure to an agent of this sort. The reason the agent deliberates in a certain way in the context is simply the fact that such a deliberation solves the integration problems the agent faces. This is consistent with the view of practical reason as an integrative function of consciousness that seeks an equilibrium of the goods valued in various ways by the agent. My thesis is that the rationality of these patterns is dependent on the character of the agent and the contexts in which the patterns emerge. I will consider several impartial patterns—some Kantian and some non-Kantian—and show how the impartial norms reflected in these patterns and the norms of friendship are symmetrical in their regulative functions. This could not be true on a Kantian view of any variety. But my purpose here is not simply either to create problems for Kantians or to restate the conclusions of chapter 2. Rather, the aim of my discussion of these patterns is to provide a developed account of the structure of the norms of simple respect and friendship and to show how they emerge in practical reason understood on the Aristotelian model as defined in the introduction.

Among the impartial norms to be considered here are some that were considered in the discussion of parental love: a utility maximization pattern, a Kantian ordinal pattern, a Kantian maximization pattern, and an imperfect duties pattern. The point is not that all Kantians endorse these patterns but that an array of "Kantian" perspectives be evaluated for the sake of thoroughness. In addition to these "Kantian" patterns, I consider two other impartial patterns: one I call a catastrophe avoidance pattern; the other, a great benefits pattern.

A catastrophe avoidance pattern emerges only in contexts involving bad outcomes and requires avoiding bad outcomes only when they are extreme. It can attach to utilitarian or to Kantian considerations. But mere marginal losses of utility or of persons' autonomy are not to count as catastrophic outcomes. Such losses must be "significant" for this pattern to be at work. I will say more about the distinction between "marginal" and "significant" later.^[6]

A great benefits pattern emerges only in contexts in which relative social well-being is already good in some sense but can move to being much better by the agent's acting one way rather than another. Thus if things are in a good state of affairs from an impartial point of view and the agent can improve things significantly by doing x but does not damage things by doing y, then the agent ought to do x on this pattern. However, if doing x will only marginally improve things from an impartial point of view, then doing x is not required. A great benefits pattern, then, is not a simple maximizing pattern of either a Kantian or a utilitarian sort.

The scope and limits of agent-centered restrictions generated by friendship are measured by the degree to which impartial deliberative patterns are rational in these contexts. To the extent that the friend's interests require avoiding impartial deliberative patterns, considerations of the friend's interests generate agent-centered restrictions. To the extent that impartial deliberative patterns are rational in these contexts, the friend's interests do not generate agent-centered restrictions. A combined analysis of these contexts will give us an account of the scope and limits of the agent-centered restrictions generated by friendship as a partial norm. It will also confirm the Aristotelian model of both simple respect and friendship as symmetrically regulating norms.^[7] I begin with friendship in the central case.

[6] . Korsgaard seems to worry over related matters in "The Right to Lie: Kant on Dealing with Evil," in Creating the Kingdom of Ends (Cambridge: Cambridge University Press, 1996), 133–58.

[7] . Consequentialists will object that I have not included indirect impartial patterns. Cummiskey defers to Parfit in this regard but does not develop just what the indirect duties of special personal relations are. We are to assume, I suppose, that they will be just the right ones. As I note in chapter 2, Herman rejects indirect patterns in these matters, and I think she is correct. Moreover, I have nothing new to say to those who think that indirection accounts for our obligations along these lines. And I do not see what would count as convincing to those who are still unconvinced that if there are special duties to loved ones they are direct rather than indirect. Similarly, I do not know how to convince someone who insists that we have only indirect duties to animals that many of our duties to animals are direct. From my own point of view, that I have an obligation to animals not to torturethem for pleasure is clearer than the theory-laden claim that only rational nature has moral value. It is equally clear to me that my obligations to my children are to them rather than an indirect consequent of my concern for maximizing the respect for rational nature.

2.

Consider contexts of the first sort in which the friend's categorical interests conflict with the noncategorical interests of others. In these contexts, a loving friend never engages in the "moral mathematics" of utility maximization. Here numbers do not count , and it would be irrational for a loving friend to employ a utility maximization pattern in these contexts. No matter how many noncategorical interests of others are at stake, a loving friend's deliberations are at an end about numbers once it is perceived that the friend's categorical interests are at stake in these contexts . Here just imagine the person you believe to be your best friend even considering the possibility that if enough people would somewhat enjoy an evening at Disney World it might just be worth your life from his or her point of view. If you cannot imagine this, then you cannot imagine your friend doing the same thing where it is not your vital interests at stake but the categorical interests in terms of which your very life is meaningful to you. From this we can see that for these contexts a maximization pattern would order your priorities in a way that would not allow your friend a place as an end in a way of life that would capture the good he appears to be within your deliberative field. And it is this fact that makes the maximization pattern irrational in these contexts.

As far as what one ought to do is concerned, this resistance to maximizing is consistent with a Kantian understanding of respect for persons and a Kantian ordinal pattern. For Kantians, utility alone cannot add up to reasons for violating the personhood of others. No matter how many units of utility are generated by satisfying the noncategorical interests of others, these numbers cannot be a reason for violating the categorical interests of a person worthy of respect on a Kantian view. This would be to treat the person as a mere means and would violate the ordinal priority of categorical interests over noncategorical interests. It is important to note, however, that we can get the result that human dignity blocks consequentialist reasoning of the utilitarian sort without going on to say that the value of human dignity is absolute, nonscalar, and immune to trade-offs in terms of other values. All we need say is that human dignity is immune to trade-offs of this sort. But on a utilitarian view, at some point the numbers (however

small in individual cases) involved in adding up the minor satisfactions of others outweigh whatever utility is assigned to the satisfaction of a single person's categorical interests.^[8] What I am asserting here is that people who have simple respect and esteem for others and who love their friends do not and cannot think like utilitarians. The Kantian asserts that a person who has just simple respect for others (let alone friendship) cannot think like a utilitarian in this regard. To this extent, I agree with the Kantian.

There is, however, a difference between the Kantian view (on the ordinal pattern interpretation) and the Aristotelian view. The rationality of the Kantian view is traced through the employment of the impartial CI procedure and has nothing special to do with friendship. On the Aristotelian view, the irrationality of a maximization pattern for these contexts is that neither the agent's respect nor the agent's love for the friend will tolerate a utility maximization pattern because such a pattern will not order the agent's priorities in a way that makes a place for the goods that appear within the agent's deliberative field. This is seen at once when those priorities are placed within the context of how they would order a way of life and the goods valued by the agent. What the agent sees is that something good has either been distorted or is not within that way of life at all. And, given what that good is, the way of life is not worthy of choice. Clearly, these thoughts reflect the fact that the criteria of finality and self-sufficiency are at work in the perception of a way of life from the agent's point of view. Consequently, the agent's doing the right thing in these contexts is the nonaccidental result of his or her motives.

The Kantian can reply here that the normative thoughts of friendship do indeed have the structure I assert and that the criteria of self-sufficiency and finality are also reflected in these thoughts, but this does not undermine the claim that the simple respect that leads to the ordinal pattern is a regulative norm in these contexts. For if friendship were not involved, the simple respect that leads to the ordinal pattern would require the same priorities of friendship that make utility maximization irrational and wrong.

This response is partially successful, because it is true that simple respect would require the ordinal pattern rather than the utility maximization pattern in these contexts were friendship not involved. The beauty of the Aristotelian view is that it covers both norms. The reason that the ordinal pattern would be rational in these contexts were friendship not involved is that

[8] . This is at least true on total sum utilitarian view. Again see Derek Parfit's discussion of the repugnant conclusion in Reasons and Persons (London: Clarendon Press, 1984), 381–90.

it is the only pattern that orders priorities in a way that structures a way of life in which the goods of that way of life are self-sufficient and final from the agent's own point of view. Thus there is a simplicity to the Aristotelian model that is lacking on the Kantian model: The kind of explanation for the rationality of friendship as a norm is the same as for the rationality of respect as a norm.

I believe that a similar analysis applies to the first deviant case of friendship, where the friend has lost R-qualities but has the possibility of redemption. Of course, a great deal depends on how bad the change in character of the friend is, how likely redemption is, and how the change in character is related to the interests of those involved. For present purposes, I will assume contexts in which the change is a significant one involving a serious flaw, the likelihood of redemption is reasonably probable, and the change forces a conflict involving the friend's categorical interests. To fix intuitions, imagine that a friend's ambitions have grown to the point that they have eclipsed his sense of fairness in dealing with the important but non-categorical interests of others. His relationship with you is such that it is the only really personal relationship he has. He loves you, and without you not even his ambitions will carry enough meaning to motivate his life. Yet the mere threat of withdrawal will not move him to fairness regarding the important noncategorical interests of others. If his ambitions are curtailed through his collapse as a result of your actual withdrawal from the relationship, other respectable people will receive what they fairly deserve regarding some of their important noncategorical interests. What would the priorities of a loving friend and respectful person be in such a context?

Even disregarding friendship, most of us believe that any reasonable probability of redemption from even some fairly significant character flaws is worth some significant noncategorical sacrifices on the part of respectable people. This is simply to say that retrieving lost character is worth more than minimal costs. Here there are two issues, and both involve the concept of hope. First, is it consistent with having simple respect for others to abandon hope for a person who has lost R-qualities where redemption seems reasonably probable? Second, does this hope generate normative thoughts about giving priority to the categorical interests of a person who stands a reasonable chance at redemption over the noncategorical interests of others ? The answer to the first question, I believe, is no. At least it is where the failings are not overwhelmingly short of tolerable and where the probabilities of redemption are reasonable. As to the second question, the answer, I believe, is yes, given the same assumptions about the failings and the probabilities of redemption.

But if this is true of simple respect and esteem, there can be no doubt about friendship: Under these conditions the answer to the first question is no and to the second yes. If the agent is not disposed to hope in these contexts, the agent simply is not a loving agent. Of course, simple respect and esteem for others will require that this hope has a foundation in fact if it is to be normative for the agent. But if the hope does have a foundation in fact, the hope will be normative. It will set the agent's priorities regarding the friend's categorical interests far above any number of noncategorical interests of others. This is enough to foreclose, from the loving and respectful agent's point of view, the consideration of a utility maximization procedure in these contexts.

The second deviant of the central case of friendship is different, however. This is where, due to a change in the friend's way of life, the friend has lost R-qualities without the possibility (or at least the significant probability) of redemption. Although it is possible to love "friends" of this sort and to have simple respect and esteem for others, it does not seem that such love is as normative as in the previous cases of friendship. Can one have simple respect and esteem for others and consider it an obligation to sacrifice their important noncategorical interests for the sake of the categorical interests of an incorrigibly contemptible loved one?

Perhaps this depends on how awful the loved one is.^[9] It does not seem that our public conceptions of what it is to respect, esteem, and love people in these contexts are clear enough to be determinate and to provide guidance except in the extreme cases. These are cases in which the loved one has become thoroughly rotten and the noncategorical interests are important ones of estimable persons. But one thing is certain: There is no room for the normative thoughts of hope in these contexts.

If what I have said here is correct, we can conclude some things about the scope of agent-centered restrictions from observations about the first kind of context. First, friendship generates agent-centered restrictions against utility maximization patterns in the central case of friendship and its first deviant. The second deviant is unclear. Second, there are no agent-centered restrictions against giving priority to simple respect and esteem for others in these contexts. That is, there are no agent-centered restrictions against

[9] . In NE 1165b:31–36, Aristotle says, "Should, then, a former friend be treated just as if he had never been a friend at all? No; we should remember our past familiarity with him, and just as we feel more obliged to do favors for friends than for strangers, we must show some consideration to him for old friendship's sake, provided that it was not excessive wickedness on his part that broke the friendship."

a Kantian ordinal pattern in these contexts. This is because Kantian respect, like friendship, would give priority to the categorical interests independent of the numbers. Kantian maximization patterns do not emerge in these contexts, because there are no conflicts of similar kinds of interests. Thus there are no agent-centered restrictions against either a Kantian maximization pattern or a Kantian ordinal pattern in these contexts, though for different reasons. Nevertheless, consideration of the role of personal love in these contexts explains something about the motivational structure of the agent's deliberations on the Aristotelian view that the Kantian analysis does not, namely, that the structure of the agent's reflective endorsement of a set of priorities for these contexts involves the criteria of finality and self-sufficiency for both simple respect and friendship in exactly the same way.

3.

Now consider contexts of the second kind in which the noncategorical interests of the friend conflict with the categorical interests of others for whom the agent has simple respect and esteem. We need only consider friendship in the central case because friendship in the central case does not generate agent-centered restrictions in these contexts. If this is true, then the weaker deviants of friendship with their diminished normative structures certainly do not.

The priorities of the agent of integrity who is respectful of others and who is a loving friend are reversed in these contexts as to what they were in the previous contexts. Nothing about friendship in the central case requires giving priority to the friend's noncategorical interests, and simple respect for others will not allow it. If this is true, then a Kantian ordinal pattern is rational relative to these contexts. Also, if this is true, the Kantian ordinal pattern restricts the rationality of both utility maximization patterns and the agent-centered concerns of friendship.

As to the latter point, consider the case of Diane. Diane has a great love for opera and wants to share this with her friend, Karen. This is very important but not categorically important to Diane, and being very important to her it is very important to Karen. Besides, Karen thinks exposing herself to opera is an important thing on its own. When Diane discovers that one of her favorite operas is scheduled for the upcoming weekend, she invites Karen to attend. Unfortunately, Karen cannot attend due to a previous commitment that involves developing a reading skills program for functionally illiterate adults who are trying to become productive members of society. (If Karen does not keep her commitment, the program will fold, not to be

taken up by others.) Diane becomes angry and alienated from Karen because of the declined invitation.

Now, it seems that we can conclude two things about Diane, one about respect and the other about friendship. Clearly she does not have the disposition of respect for others that either the Kantian or any mature person would want us to have. But, equally clear, she knows very little about friendship, because she does not recognize the nature of her friend's commitments.

Worries that friendship does not allow significant room for simple respect and esteem for others are simply misplaced. Diane's negligence and indifference to others is not due to her love for Karen and the opera but to a lack of concern for others. After all, we can assume that Karen has an equal concern for sharing the opera with Diane as Diane has with sharing it with her. But Karen has the concern for others, which affects her normative thoughts in these contexts. By forgoing the opera she is not being a bad friend, but if she neglected these admirable people she would be showing a lack of respect. Indeed, it is Diane who is lacking not only in respect for others but in friendship as well. Karen's conception of friendship is regulated by simple respect and esteem for others; Diane's is not. To the extent to which Diane is a friend, she illustrates a deviant case, because if she were a friend in the central case she would understand Karen's priorities. Unfortunately, friends like Diane are familiar to us all.

Consider also how this might be thought to apply to the addiction to smoking. Suppose your friend's addiction to smoking conflicts with the vital and categorical interests of others for whom you have simple respect and esteem. I am thinking here primarily of the dangers of secondhand smoke. In this context, it will be important to consider a special feature of a responsible agent's sense of obligation to be a reliable critic. If you are a respectful and sympathetic agent, you will have a sense of obligation to be a reliable critic on behalf of those worthy of respect and sympathy. To be unwilling to criticize others for acting disrespectfully and unsympathetically toward others when the circumstances indicate that effective criticism is in order is simply to be lacking in respect and sympathy. In this sense, respect and sympathy are intolerant dispositions. Moreover, as a respectful person you will, on reflection, think of it as a good thing that others can be relied on in this regard, even if you at times come up for criticism. One of the goods of simple respect, then, is that there are others who are reliable critics of your behavior, and one of the roles of being a respectful person is being a reliable critic of others. How do these thoughts bear on the present context?

One of the categorical interests of friendship is the interest in the friend's

character. Remember that friendship is a peer relationship and that one feature of equality in this regard is that your friend is your moral equal. Thus the loving friend in the central case will be especially concerned to be a reliable moral critic for his or her friends. As a loving friend and as a respectful and sympathetic agent, then, you will have two sources of your concern to be a reliable critic: One attaches to your concern as expressed in your respect and sympathy for others; the other, to your concern for your friend's character. Now, in cases like the one involving smoking and the dangers of your friend's secondhand smoke harming others, it is important to keep in mind both roles regarding criticism.

One might think that the kind of smoking issue described is a straight-forward case of a deliberative context of the second sort in which noncategorical interests of the friend conflict with categorical interests of others. But this would be a mistake. True, there are noncategorical interests at stake for the friend and vital and therefore categorical interests at stake for others. The fact that the addiction to smoking is in a way compelling to your friend does not make that interest a categorical one. It is not that your friend thinks that smoking is more important than the health and well-being of others. Rather, the addiction simply blinds your friend to the values at stake. Apparently, this is often one of the features of addiction. But the context is complicated by your interest in a continued relationship with the friend and in the friend's character. Given the influence and possibly alienating consequences of criticism, you have a contrary inclination to being critical and to giving priority tothe interests of others. Nevertheless, if the context is one in which it is plausible that criticism might prove effective (even at the cost of some alienation) the agent of integrity in these contexts will give priority to the categorical interests of others. But note that this involves categorical risk—risk required for the sake of others. The reason I say that you will give priority to the categorical interests of others is because there is no sense to be made of saying that your doing so in these contexts is a sign of being a bad friend. Thus if anyone is being a bad friend, it is the addicted person who puts the relationship at risk for the sake of an addiction. Of course, if your friend is one in the central case, he or she will take criticism that is thoughtfully given for what it is, and to the extent that you understand your friend to be a friend in this sense you will have less anxiety about the criticism causing an end to the relationship. The problem is that such addictions are simply difficult to maintain and retain the kind of character to be a friend in the central case, which is one of the major reasons for avoiding addictions.

Of course, this analysis can apply to many, many contexts involving addiction and other forms of irrationality, not the least of which are those including racism and sexism. Still the categorical interests of those affected by these attitudes are so important to the respectful and esteeming agent of integrity that not even the categorical risk involved in putting one's friendships in jeopardy is enough to justify (from the agent's point of view) the tolerance of such behavior. To be sure, we often fail to live up to our own norms here, and we understand why this is sometimes so. But admitting our failures only reinforces the claims here about what the priorities of simple respect and esteem are relative to friendship in these contexts. Without thoughts of self-reproach for such failures, an agent is simply unrecognizable as being respectful of others. Friendship in the central case, however, does not have this result, because friends in the central case are open to the reliable criticism of their friends, as well as that of others who are held in respect and esteem.

Yet let me add that when we do act in accordance with our commitments in these contexts we act not only for the sake of others; we also act for the sake of our friends and in accordance with our own good (if not for the sake of it). This is because as loving friends we act in the role of reliable critics for them, and we expect from them that they act as reliable critics for us. In this way, our partial and impartial norms coordinate in these contexts. Without such coordination we cannot make sense of our concern in these contexts both for our friends and for others. Therefore, although friendship can generate conflicts here, it is also part of the solution: Good friends do not stand by idly while friends are failing in serious ways. They stand their ground as reliable critics, acting not only for the sake of the categorical good of others but also for the sake of the categorical good of the friend's character. Thus these contexts are not of the second sort. Rather, they are contexts that involve multiple categorical conflicts. This is why they are so difficult to manage.

I conclude, then, that contexts of the second sort are not ones in which agent-centered restrictions are even permissible from the perspective of a respectful and loving agent of integrity. This is due to the fact that simple respect and esteem is reflected in the rationality of a Kantian ordinal pattern in these contexts. Since this is true of the central case of friendship, there is no need to consider the deviant forms. Finally, it is also important to point out that there is no need for a disagreement between a Kantian account and my account of the priorities of an agent of integrity in these contexts. Once again, the Kantian ordinal pattern is consistent with the priori-

ties of friendship. However, the structure of the agent's deliberations can be described more accurately on the Aristotelian view. The right set of priorities in these contexts structures a way of life in which both other respectable people and one's friends are ends in their respectively relevant senses. The choice of that way of life with its priorities reflects the criteria of finality and self-sufficiency: The way of life includes both one's friends and others as ends and is chosen for itself. So, once again, the rightness of the Aristotelian agent's actions is the nonaccidental result of their motives, as Herman and Kant require, but without the need for an impartial decision procedure.

4.

Contexts of the third sort, however, are ones in which the categorical interests of friends come into conflict with the categorical interests of others for whom the agent has simple respect and esteem. Here, I believe, an adequate solution to the priorities problem clearly conflicts with three of the impartial deliberative patterns mentioned above: a utility maximization pattern, a Kantian ordinal pattern, and a Kantian maximization pattern. Since it is unclear what an imperfect duties pattern would be in these contexts, it is difficult to assess its relevance. Whatever that pattern is, however, it must be consistent with the value of human dignity as absolute, nonscalar, nonadditive, and immune to trade-offs in terms of other values. I will argue that our thinking about these contexts reveals that we do not endorse that conception of the value of human dignity, and hence we have reasons for rejecting a Kantian analysis of latitude, even where we recognize a role for latitude. The role for latitude can best be made sense of on the Aristotelian model that does not treat human dignity as having the kind of value Kantians claim it has. Rather, it treats human dignity as it appears within one's deliberative field as a very important intrinsic value but one the overall value of which is affected by the need to make a place for other intrinsic values. Great benefits patterns are not contextually relevant because these contexts involve bad outcomes. Some of these contexts, however, involve not only bad outcomes but outcomes that are catastrophic. In some such cases, I will argue, impartial deliberative patterns are rational for the personally loving friend who also has simple respect and esteem for others. Thus, in some of these contexts, agent-centered restrictions are not generated by friendship but antithetical to the commitments of respect and esteem. The sum of these observations defines the symmetrical regulative relations between the norms of friendship and the agent's impartial norms.

In some sense, all contexts of the third sort are catastrophic, because someone worthy of respect loses categorically. But there is another sense in which some are catastrophic in ways that others are not. This sense involves the sheer numbers of similarly worthy people involved in the conflict situation. In some situations, the friend's categorical interests conflict with those of someone else equally worthy of respect and esteem. Given the influence of love and the greater interest in the friend, the agents norms will prioritize the friend's interests over those of the other person. If this is true, then a Kantian ordinal pattern is irrational in these contexts, for it allows but does not require giving priority to either the friend's interests or the interests of the other person. In other contexts, there will be marginally more equally worthy persons with categorical interests at stake than the friend. Still, given the priorities of deep personal love, the friend's categorical interests will hold first in the agent's normative thoughts. If this is true, then neither a Kantian nor a utilitarian maximization pattern is rational in these contexts. In still other contexts, there will be extremely large numbers of equally worthy persons with categorical interests at stake. It is these that I am calling catastrophic contexts, and it is in these that friendship does not generate agent-centered restrictions.

Friendship prevents treating the categorical interests of a friend on an equal par with similar interests of another equally worthy person, everything else being equal.^[10] This is true even for a person who undeniably has simple respect and esteem for others. Of course, having to do something with the foreseen consequence of destroying the categorical structure of another person will fill the respectful and esteeming person with sadness.^[11] Indeed, without this sadness, we would be hesitant about the person's respect and esteem for others. But if the agent did not give priority to the friend's interests, we would be clear about the absence of friendship. Still,

[10] . It might seem that there are clear counterexamples to this. For instance, it seems that in official contexts agents often treat the interests of friends and others equally. Reflection reveals, however, that we do not usually allow an agent to fill an official role in contexts in which the categorical interests of his or her loved ones are likely to come into conflict with the interests of others in a way that requires the agent to adjudicate the conflict.

[11] . Isaiah Berlin was probably right that for any way of life to succeed in its own terms for any substantial number of people other ways of life will be ruled out for other people. This is an unpleasant fact, but it should be kept in mind when criticizing ways of life. One should not assume that as long as it can be pointed out that a way of life has a price for other people that this in itself counts as a criticism. For this can probably be pointed out for most any meaningful way of life whatever. Otherworldly criticisms of this sort have their home in utopian instincts, instinctsthat have probably introduced as much misery into the world as anything else. See Berlin's The Crooked Timber of Humanity (Princeton: Princeton University Press, 1990).

where there is both sadness at the loss of others and the disposition to give priority to the friend's interests, we have no hesitation at all in these contexts regarding the agent's simple respect and esteem for others. No one would say that a person choosing to save a beloved friend from a fire instead of an equally worthy stranger was being disrespectful to the stranger.

Nor can I see that anyone would be inclined to say that a person choosing in this kind of context to save a beloved friend rather than some marginally greater number of equally worthy strangers was being disrespectful to the strangers. Yet there would be serious doubts about friendship if the love for the friend operated only as a tie-breaker criterion. The priorities of friendship and other forms of love give significantly greater weight to the categorical interests of the loved one than to others worthy of respect and esteem. Of course, the dispositions of the loving friend in these contexts are coupled with simple respect and esteem for others. This means that having to act in a way that results in these categorical losses for others means even more sadness for the agent than in the previous case. Without this sadness, we are at a loss to ascribe the dispositions of respect and esteem for others to the agent, even in the face of deep personal friendship. Such sadness is one regulatory effect of simple respect on friendship, but it is also evidence of friendship as a determinate and dominant partial norm. I do not see how the Kantian can account for this, even on an imperfect duties pattern with its emphasis on latitude and appeal to the Formula of Humanity.

Clearly, loving friends do not have the latitude to employ either a tie-breaker criterion or a maximizing pattern by appealing to the value of human dignity in these contexts. To do so would violate the criterion of self-sufficiency in regard to how friends appear within the loving agent's deliberative field, even where others appear there as worthy of respect. What explains the latitude is not the value of human dignity as absolute, nonscalar, nonadditive, and immune to trade-offs. If human dignity were absolute, how could we explain a person allowing friendship to provide the latitude for deliberation on marginal and significant differences in the way that makes sense of catastrophe avoidance patterns at some point but not others? If human dignity is immune to trade-offs in terms of other values, then how do we explain the fact that the loving friend in such contexts does make such trade-offs without denying his or her respect for the value of

human dignity? "Just so" stories that allude to latitude regarding the actions of friendship will not do; rigorous analysis is called for, analysis that shows how the latitude employed is restricted by the value of human dignity as absolute, nonscalar, nonadditive, and immune to trade-offs. Too of-ten, the Kantian literature makes valid points about the regulatory effect of impartial respect on partial norms but then leaves it to intuition to fill in the details of the rest of the theory. But it is just those intuitions that I am challenging here. If I am wrong, there should be some clear specification of the CI procedure in its imperfect duties deployment that accounts for these thoughts. What is it, and how is it compatible with the Kantian account of the value of human dignity?

Imperfect duties are said to allow for latitude in what the agent does in ways that perfect duties do not. Baron's work makes a great deal of this.^[12] In addition to latitude about just what one does, there is also latitude in many contexts about how much we are to do, even whether we are required to do anything at all. No doubt there are contexts in which these kinds of latitude would make sense of our judgments. However, it does not follow from this that the CI procedure explains the rationality of the agent's actions in these contexts. There are many distinctions in Kant's ethics, especially in The Metaphysics of Morals , but there is less argument than one would desire that shows that it is the CI procedure when applied to human beings that generates these distinctions. Rather, it often seems that these distinctions are generated by our intuitive judgments and the CI procedure is assumed to make a place for them. What is needed is the kind of careful casuistry that links the distinctions to the CI procedure in the way that Korsgaard does regarding perfect duties. I cannot see that Baron ever does this regarding friendship and the kind of context in question. Instead, she focuses on contexts in which respect regulates friendship. At any rate, the context in question leaves little latitude for choice about how much to do or whether to do anything at all. Yet the latitude makes perfect sense on the Aristotelian view: It is required to make sense of the place of both friends and other respectable people within a life that is self-sufficient and final. Any other set of priorities would distort the way that friends and others appear within the loving and respectful agent's deliberative field. When friends appear within one's deliberative field as more important than oth-

[12] . I will address Baron's views on imperfect duties in much more detail in chapter 18. For her views in this regard, see Kantian Ethics Almost Without Apology , esp. 88–110.

ers with equal dignity, it can hardly be the case that the notion of respect that generates that field is one that conceives of human dignity as absolute, nonscalar, nonadditive, and immune to trade-offs.

There comes a point, however, at which the categorical losses for others become considerably more than marginal and therefore catastrophic in the strong sense. There are, I believe, two kinds of cases here.

First, in some of these contexts simple respect and esteem for others and deep personal friendship come incommensurably into conflict. At some point—and I cannot see that there is any algorithm for this—the losses on both sides are stultifying to the agent. Yet there cannot be any thought of flipping a coin: No arbitrary tie-breaker criterion is conceivable to the agent. Try to imagine in this regard the features of a context in which flipping a coin would be appropriate to resolve a conflict between the categorical interests of a friend and those of others. Either the numbers are decisive, or stultifying, or they just do not add up. When they are stultifying, the agent cannot resolve the conflict, priorities cannot be established, and there is a deliberative impasse. When this occurs, the agent breaks down in ways that are indicative of a loss of integrity—a loss of the will to live, hysteria, self-deception, denial, debilitating depression, and so on.

Breakdowns of this sort teach us three very important lessons about stultifying catastrophic contexts. First, they give us a barometer for measuring the adequacy of a theory of human rationality. Any theory that does not recognize stultifying decision contexts by requiring the equivalent of flipping a coin to resolve the conflict is an inadequate theory of human rationality. It fails to recognize the psychological features of agent rationality as it applies to human organisms. Breakdowns in these contexts are a function of the presence rather than the absence of human rationality. Second, these contexts do not generate either agent-centered restrictions or impartial catastrophe avoidance patterns. Rather, they result in agent breakdown and reflect the limits of regulative norms, whatever their type. Finally, they show that there is no value, either of human dignity or of friendship, that is absolute, nonscalar, nonadditive, and immune to trade-offs. If there were such a value appearing within the agent's deliberative field, there would be no such stultifying contexts.

Yet there is a straightforward explanation for this on the Aristotelian model, as it is interpreted here. In such contexts, there is no set of priorities that will allow the agent to project a vision of a way of life in which all the categorically important parts are within it in a way that makes that way of life even minimally worthy of choice from the agent's point of view. The

criteria of self-sufficiency and finality applied to categorical goods yield the desired explanation. I do not see how the Kantian has an explanation for such phenomena other than attributing them to our nonmoral, pathological nature. If what I have said is correct, however, the phenomena are due to what is best about us.^[13] It is the failure to see how the criteria of finality and self-sufficiency function in practical reason that leads some to think of breakdowns in these contexts as irrational human failure.

The second kind of response to catastrophic contexts involves commensurability. The agent is simply overwhelmed by the numbers of those who will suffer categorically if the friend's categorical interests are secured. Think, for example, of facing the choice of either saving one's friend or preventing the Holocaust. In this case, the agent is able and feels compelled to deliberate to the normative thought of sacrificing the friend's categorical interests. Indeed, it is a case of the agent deliberating to the normative thought that his or her categorical interests are to be sacrificed, and we have already seen how this is possible. Just as friendship can give one reasons for dying for the sake of another, simple respect and esteem can give one reasons for allowing the death of one's dearest friend for the sake of others. Yet engaging in such deliberations and acting on them come with categorical consequences to the agent. Whether they are survivable is dependent in part on what other commitments there are in an agent's complex ground project. But it should not be surprising that, more often than not, being forced by the contingencies of life into making catastrophic decisions of this sort simply destroys an agent.^[14] In these contexts, the agent can reflectively endorse a set of priorities that will structure a life for the moment but will not project into the future as a way of life in which the agent has a clear role. It is more important that the way of life projected on the agent's priorities include a place for these respected people as ends than that either the agent or the friend has a clear place there. The most choice-worthy life available is not self-sufficient, and the goods that are missing are the agent and the friend. And it is important to note that these contexts reveal the plausibility of objective over subjective eudaimonism: What one projects into the

[13] . I argue this more extensively in Dignity and Vulnerability: Strength and Quality of Character (Berkeley: University of California Press, 1997).

[14] . In these contexts, the breakdown comes as a result of the deliberations and actions that are possible for the agent. In the previous case, breakdown is a function of deliberative impasse and incommensurability. That is, in some cases, a decision must be made, but, due to incommensurability, it cannot be made. In other cases, a decision can be made, but it cannot be lived with.

future is not one's virtuous activities or one's happiness but others and their well-being.

The conclusion is that in the central case of friendship and in contexts involving categorical conflicts between the interests of friends and equally estimable strangers, simple respect and esteem sometimes generate catastrophe avoidance patterns that are inconsistent with agent-centered restrictions. When the numbers are marginal rather than catastrophic, however, agent-centered restrictions are generated by friendship, which establishes that the relationship between simple respect and friendship is that of symmetrically regulating norms within practical reason. This, of course, confirms the Aristotelian view, and explains the latitude in these contexts in a way that appeals to human dignity as a value that is absolute, nonscalar, nonadditive, and immune to trade-offs in terms of other values cannot. As for the deviant forms of friendship, it seems that the instances of incommensurability become less as the deviation from the central case increases. In the second deviation (where redemption is unlikely), it is unclear that stultifying contexts apply here for persons with categorical commitments of simple respect and esteem for others. It is, after all, one thing to know what one's commitments are and another to have a competing contrary inclination to keeping them. This can happen with people we love but for whom we have no hope of respect.

5.

Finally, I must say something about contexts of the fourth kind in which the noncategorical interests of the friend conflict with the noncategorical interests of others the agent respects and esteems. The distinction between important and minor noncategorical interests is relevant, as are distinctions regarding different contexts. I have in mind those contexts in which the friend's noncategorical interests are important and others in which they are minor.

We must be aware, however, that the distinction between important and minor noncategorical interests is clear only at the extremes, which makes an account of an agent's normative thoughts difficult in terms of precision. Where the distinction is difficult to perceive, the agent's normative thoughts will be less clear, from the agent's own point of view. This, I believe, is a strength of my account rather than a weakness. Our normative thoughts simply are not always clear, nor can they always achieve clarity with reflection, except arbitrarily. It is a fault of some moral theories that

they fail to recognize this complexity and why it exists. Yet I think that in these contexts there is more room under conditions of unclarity for arbitrary tie-breaker criteria than in the contexts involving categorical conflicts. Still some of these contexts are ones in which friendship generates agent-centered restrictions against both utility maximization patterns and in others against certain kinds of Kantian reasoning.

Consider first those contexts in which the friend's noncategorical interests are important.

In one kind of context, the friend's important interests conflict with the minor interests of others who are held in simple respect and esteem by the agent. Here friendship restricts utility calculations, as would a Kantian ordinal pattern. Neither a Kantian nor a good friend would think that calculations regarding the minor interests of any number of other people would permit the frustration of a clearly important interest of someone, whether it is one's friend or not. Reconsider here actions that would prevent Diane and Karen from attending the aforementioned opera. What if these actions had the sole benefit of preventing an infinite number of people from experiencing the normal unpleasantness of a very temporary itch? The stranger who would prevent the itch, it seems, would lack respect (and probably sympathy as well) for Diane and Karen. If Karen canceled plans for the opera on these grounds, Diane would be more than justified in her anger at Karen's priorities, even doubtful about the status of the friendship. At the very least, Karen would be a bad friend to have such a sense of priorities, for these priorities would not project a way of life that is both self-sufficient and final because her friend would not have her proper place there as a beloved end. Yet the alternative that does give her friend the proper place in a way of life is not one that makes inadequate room for respect for others.

Now consider the reverse situation: the important noncategorical interests of others conflict with the minor interests of the friend. Unless the priorities are reversed here, it is difficult to make sense of the place of simple respect and esteem in the life of an agent. A person who values relieving the normal unpleasantness of a friend's temporary itch over the clearly important (even if noncategorical) interests of others is simply puzzling as a respectful agent. How could a respectful person reflectively endorse as self-sufficient and final a way of life that allowed such treatment of others, and how could a loving friend see that friendship required such priorities? Indeed, such a person could be described as insensitive to or negligent of the interests of others, rather than respectful of them. Yet a set of priorities that

reflects sensitivity and mindfulness in these contexts is not at all evidence of a lack of friendship. Thus preoccupation with Karen's minor interests might dominate Diane's sense of caring to the exclusion of concern for the important interests of others. But if it does, it is testimony neither to a respectful character nor to mature friendship. Therefore, friendship in such contexts does not conflict with simple respect and esteem for others but is regulated by them.

But what about contexts in which the conflicts are between equally important noncategorical interests of a friend and others held in respect and esteem? Does an analysis of the dispositions of simple respect and esteem and the personal love of friendship yield a clear solution to the priorities problem here?

One might think that these contexts should be analyzed in a way parallel to the contexts in which the categorical interests of friends and the categorical interests of other respectable people come into conflict. But this would be a mistake. First, these contexts are never catastrophic, simply because there is not enough at stake given the kinds of interests involved, which means that they never give rise to catastrophe avoidance patterns, and second, because there is not enough at stake, they are never stultifying contexts in the way that contexts involving categorical interests can be. If this is true, then the agent of integrity is always able to deliberate to a solution and resolve the conflict. The relevant questions, then, are, do these deliberations yield agent-centered restrictions and, if so, in what contexts?

First, consider a case in which a friend and one other equally respectable and esteemed person have equally important noncategorical interests in conflict. Surely, there can be no doubt here that favoring one's friend is not a sign of disrespect to the nonfriend. Yet, everything else being equal, failing to favor the friend would be contrary to friendship. Otherwise, we are unable to account for the status of the friend within the loving agent's deliberative field, given what we have said about the general features of love.

Imagine, for example, that Karen decides to share the opera with Paul instead of Diane and justifies this to Diane with the response that Paul wanted to go to the opera just as much as Diane. Everything else being equal, this would be a sign either (i) that Karen considers Paul at least an equally good friend as she does Diane or (ii) that Karen does not love Diane as a friend. If Diane could rule out the first alternative, it would seem that if she were not hurt and offended by Karen's priorities, we could only conclude that she does not care for Karen as a friend. Moreover, if Paul understands friendship and does not see himself as a friend to either Karen or Diane, he also

will be puzzled by Karen's priorities rather than see them as a sign of respect. Thus even Paul will see Karen as appropriately having the otherrestricting belief that Diane's interest ought to come first in these contexts.

This other-restricting belief is an agent-centered restriction, despite the fact that the states of affairs resulting from either alternative are equal from an impartial point of view alone. The fact remains that Karen's personal commitments restrict her impersonal commitments. She is not free from her point of view to employ an arbitrary tie-breaker criterion. To do so would violate her commitment to Diane, her friend. Her other-restricting norm, then, is an agent-centered restriction against giving priority to the interests of others on the basis of equal respect and esteem. Where this is contrary to at least some ways of reading Kant is that a Kantian ordinal pattern permits, but does not require, the woman to favor the friend as a tiebreaker criterion.^[15] But I fail to see how thinking of Karen's priorities as settled by appeal to friendship as a tie-breaker criterion, as a kind of moral permit to favor Diane, is a reflection of Karen's love. Love, as we have seen, is not like that. What matters to one's loved one is not just as important but more important than what matters to some other person, even if the other is an equally worthy person. This is a fact about how one's friend appears within the deliberative field, which does not deny at all that others appear there as ends as well. Thus, in these contexts, a Kantian ordinal pattern is not rational for an agent who is a loving friend and who is respectful of others, nor is a prerogative pattern. This leaves open the possibility of some other Kantian pattern, but it is unclear what it is. The concept of imperfect duties is so amorphous, despite the extensive Kantian taxonomy, that it is difficult to see how one gets from the CI procedure on either the purely procedural or the substantive interpretation to a clear enough concept of imperfect duties to set priorities in these contexts. Moreover, this problem is compounded by the fact that the employment of the CI procedure must avoid filtering respect through considerations of friendship, as indicated in chapter 2. The burden is on the Kantian to provide a clear account here, and I do not see that such an account has even been attempted. Given the fact that there is a clear account on the Aristotelian view vis-à-vis the criteria

[15] . See Marcia Baron, "The Alleged Repugnance of Acting from Duty," Journal of Philosophy 81 (April 1984): 197–220; Barbara Herman, "On the Value of Acting from the Motive of Duty," Philosophical Review 66, no. 2 (July 1981): 233–50; and Charles Fried, An Anatomy of Values (Cambridge, Mass.: Harvard University Press, 1970), 27.

of self-sufficiency and finality and the categorical/noncategorical interest distinction, we are justified in accepting the Aristotelian view.

But what of the contexts in which there are marginally more equally worthy persons with equally important noncategorical interests at stake as the friend and those in which there are substantially more equally worthy persons involved? Is either of these contexts one in which friendship generates agent-centered restrictions? The answer, I believe, is that in the first kind of context it does, but in the second it does not. Yet I do not believe that there is any bright-line and still nonarbitrary criterion for determining when numbers are marginal and when they are substantial.

If we return to the opera with Karen and Diane, I think we can see this. When the interest is an important one, friendship in the central case does not allow Karen to say to Diane, "I can't go to the opera with you because I must go with Paul and Don who want to go just as much as you, even if they are not my friends. You see, they are fine people from the homeless shelter who need a sponsor to get in. I know that you are also in need of a break from your heavy schedule and have been very much looking forward to this, but after all, there are two of them and that makes the difference." Clearly, these are not the thoughts of friendship. If this is correct, then friendship renders a Kantian maximization pattern irrational in these contexts. Can a Kantian imperfect duties pattern yield the desired result? Perhaps, but two points are crucial here. First, the account must be given in a way that does not simply allude to the latitude that comes with imperfect duties. It must show how to get from the CI procedure to the notion of an imperfect duty that covers the case in question without filtering and while employing a conception of the value of human dignity in a way that is absolute, nonscalar, nonadditive, and immune to trade-offs. That is, the account must be context-sensitive rather than a hand-wave in the direction of imperfect duties. Second, since we already have an account in the Aristotelian view that shows how the agent's actions in these contexts are the nonaccidental result of their motives, what is the need for the CI procedure, even assuming that some interpretation of the imperfect duties pattern can produce the desired result? What prevents the numbers from counting in the way imagined is simply a function of the fact that such a set of priorities would not structure a way of life in which the beloved friend has a place as the good she appears to be within the loving agent's deliberative field. Moreover, the alternative set of priorities does structure a way of life in which both the friend and the others have their perceived value. Practical reasoning in terms of the Aristotelian criteria of finality and self-sufficiency provides the rational choice.

On the other hand, imagine that you are faced with a choice involving the allocation of scarce time: Either you can devote a certain amount of time to a project that brings an important but not categorical benefit to a friend or you can devote the same amount of time to a project that will benefit in a similar way a large number of people with whom you do not have a loving relationship but for whom you have both sympathy and respect. Suppose, for example, that your regular job has been demanding a great deal of your time but now things have changed to the point that you have more time to devote to other things. One alternative is that you can spend more time on the weekend with a friend whose life would be importantly though not categorically enriched by sharing more time with you. It is not that your friend is depressed but that life would be significantly enriched by a more regular sharing of life with you. A second alternative is that you can spend your weekends working with an organization that teaches a large number of elderly people skills that will add to their lives to the same degree that your friend's life will be improved by sharing his weekends with you.

I am sure that many of those immersed in moral theory will think that the answer is obvious: You would, if you are respectful and esteeming of others, take the second alternative. Others, perhaps, will think that your friendship will require the first alternative. But is either of these clear? What if you take the first alternative and feel little joy in it because of what you could have done for all these other fine people? You do not feel guilty, or shameful, but simply bad that these fine people could benefit from your help. What would such feelings show: that you do not fully love your friend? Or what if you take the second alternative with similar feelings about what you could have done with your friend? Does this show that you do not fully respect and esteem these other worthy people?

It seems that you are simply faced with a somewhat unpleasant choice, despite the fact that on either alternative no one is going to be worse off than before and someone is going to benefit. I cannot see that either your love for your friend or your respect and esteem for others would necessarily settle your priorities. You might find that you simply cannot see that there is anything that you ought to do. Nonetheless, thoughts about permissibility might not be liberating. Flipping a coin would also seem odd as a deliberative alternative, at least as a means of relieving the stress. It is possible, however, that your respect and esteem would require that you take the second alternative. I do not think there is anything about the concept of friendship that is inconsistent with such a set of priorities. It is just that I do not see that the concept of respect and esteem is sufficiently clear across

persons for us to predict this set of priorities. What is clear to me, however, is that there is nothing about the concept of friendship in the central case that requires you to deliberate to the first alternative. The good friend who has simple respect and esteem for others will understand another friend having a set of priorities that either allows or requires the second alternative. Without understanding of this sort, it is difficult to see how friendship could exist as a peer relationship between persons who value both their friends and other worthy people in these contexts.

If what I have said about this case is true, then deliberative contexts of this sort do not generate agent-centered restrictions. For if friendship can be integrated with simple respect and esteem in either of these two ways—as allowing or requiring the second alternative—then friendship does not generate agent-centered restrictions in these contexts. But note that the absence of agent-centered restrictions here is not accounted for on either a utilitarian or a Kantian basis. Nor is it accounted for on the basis of catastrophe avoidance considerations, for there will be no catastrophe either way. Rather, what is at work here are considerations of flourishing, not at the borders of marginal increases in utility, but at the rapidly expanding horizons of the well-being of many people. Thus what is at work here is a great benefits pattern.

Sometimes we are struck with the differences in alternatives because one alternative would produce a great deal of good by improving many already good lives in significant ways. Yet there are other alternatives we could choose that would not result in as much good but would not result in catastrophe. This is the kind of context we are considering. When the sacrifice of overall flourishing would be neither noncatastrophic nor merely marginal but significant, this sometimes affects what we think it is rational to do. In such a context, neither mere maximization nor catastrophe avoidance patterns are at work. It is the sheer greatness of the benefit, not just that it is better, that moves us. Why does it move us? Because it allows us to project a way of life that accommodates all that we see as good within our deliberative field where that way of life is worthy of choice for itself. This is the Aristotelian view.

What would make such a pattern rational over a mere maximization pattern that would require always selecting for any marginal increase in utility no matter what the context? One answer is this: It would be rational to employ a great benefits pattern in a context in which the costs in terms of flourishing for other good people weighs too heavily on an agent in terms of the important interests of his or her loved ones. I am sure that many moral theorists will cringe at this because of its messiness. But I cannot

see what the philosophical argument is that there is a nonarbitrary way of achieving neatness here. Human rationality simply is not as neat as some people want to think, even under the most ideal cognitive conditions. My claim here is that the rationality of great benefits patterns is context dependent, as is the rationality of Kantian, utilitarian maximization, and catastrophe avoidance patterns. In contexts in which great benefit patterns are rational, one of the limits of agent-centered restrictions is reached. This I have argued applies in the current kind of context.

6.

A summary here reveals that friendship both limits and is limited by our concern for self and others. Where agent-centered restrictions are generated within the dispositions of a person who has simple respect and esteem for others but who is also a loving friend, friendship limits the concern for others in setting the priorities of the agent. Where impartial deliberative patterns generate a sense of obligation to others or even simply allow the agent to give priority to others, friendships restrictions are limited by the concern for others. I have tried to argue that, in different contexts, depending on the kinds of interests at stake and the kinds of relationships that exist between persons, there is a significant place for agent-centered restrictions, on the one hand, and impartial patterns, on the other. In other contexts, friendship and impartial patterns converge. Within these boundaries lie the priorities of a well-integrated agent who is personally loving and respectful and esteeming of others. But it should be clear that this agent is neither a Kantian nor a utilitarian, because there is no one kind of concern—even the concern for human dignity—that is lexically privileged across deliberative contexts that can solve for the integration problems. Nor is there any one kind of concern that either directly or indirectly sets the priorities of an agent of integrity without regard to context. Moreover, it should be clear that the priorities expressed here are decidedly not those of an overly conservative conception of morality, nor is the Aristotelian conception of practical reason they reflect.