defending consequentialism handout

VBI 2011 Defending Consequentialism Nebel 1

I. What is consequentialism? A. What we ought to do is determined by the relative goodness of the way things

would be (i.e. the value of outcomes). B. Deontic facts are reducible to evaluative facts. C. Other disagreements are open questions within consequentialism.

II. Why is it important to learn how to defend consequentialism? A. Often unavoidable B. Strategic flexibility C. Refute or “open up” narrow standards D. Big impact scenarios E. Choose the position first, and the framework second.

III. Response strategies A. Delink B. Refute C. Bite the bullet

IV. Utilitarianism A. Classical utilitarianism B. Objection: Nozick’s experience machine1 C. Delink: preference utilitarianism

1) Objection: Hare’s fanaticism problem2 2) Response: malleable preferences

D. Objection: Parfit’s Repugnant Conclusion 1) Delink: average utilitarianism 2) Response: true genius (or true hell) 3) Bite the bullet vs. pluralism

E. Why “consequentialism” is not a standard V. Consequentialism and its Demands

A. Agent‐neutral, maximizing consequentialism 1) Objection: overdemandingness3 2) Response: begs the question 3) Revised objection: unrealistic 4) Response: not impossible, so no impact 5) Revised again: wrong to blame them 6) Response: blame is an act 7) Strongest version: alienation from agent‐relative values

B. Objection: agent‐relative reasons to care more about our friends4 or ourselves.5 1) Delink: Scheffler’s hybrid theory 2) Response: too permissive, too ad hoc 3) Delink again: satisficing consequentialism 4) Response: arbitrary 5) Delink yet again: indirect consequentialism

C. Rule Consequentialism6 (follow UA‐optimific principles) 1) A2 demandingness: fair shares 2) A2 alienation: UA‐optimific principles would include agent‐relative concerns7 3) Objection: exceptions. Impact: convergence with AC. 4) Response: compliance (following) vs. acceptance (believing)


5) Objection: universal acceptance is stupid (a) Cases where motivations, desires, and dispositions are orthogonal to moral

beliefs (b) Cases where evildoers are sensitive to moral beliefs8

VI. Two Kinds of Reasons A. A2 agent‐relative theories

1) Parfit: collectively self‐defeating9 2) Parfit’s better answer: incomplete relativity10

B. Delink: accommodate agent‐relative reasons 1) Restrict to government‐citizen relationship11 2) Incorporate agent‐relative values into good12

C. Back to agent‐neutrality 1) Esoteric morality13 2) Self‐effacing but not self‐defeating14

D. Other useful distinctions 1) Rightness vs. decision procedure 2) Objective (actual outcomes) vs. subjective (expected outcomes)

VII. Attack of the Kantians A. Universalizability

1) Universalizability: FUL (a) Wood: it’s useless15 (b) Specific maxims and the problem of relevance16 (c) Revise FUL to defend consequentialism?

(i) Hare: universal prescriptivism17 (ii) Pettit: universalizability and agent‐neutrality18 (iii) Parfit: FUL to Kantian Contractualism to Rule Consequentialism19

2) Humanity: FH (a) Conceptual clarification

(i) Humanity = rational nature, not the species (ii) Already existing end, not an end to be brought about (answers Pettit’s

arbitrariness argument) (iii) Dignity, so no price

(b) Kant’s warrant20 (i) Different flavors, e.g. Korsgaard21 (ii) Wood’s reconstruction22 (iii) Common theme: regress on conditions (iv) Committed to subjectivism about reasons23

(c) Key point: value of rational nature (i) Response: matter of degree24 so cannot be infinite worth25 (ii) Response: either implausible or circular26 (iii) Response: non‐moral activity is not good27 (iv) Response: can’t explain why we care about suffering28 (v) Response: biologically immature beings29 (vi) Response: Humanity ≠ rational nature30

(d) Can FH support consequentialism? (i) Cummiskey: Kantian consequentialism31


(ii) Assumption: agent‐neutral impartiality32 3) Freedom

(a) Why Kantianism requires freedom (i) Agency (ii) Value of rational nature (iii) Maxims have moral worth (iv) Wrongdoing merits blame (v) Kant’s “highest good”

(b) A ballsy strategy: hard determinism (i) Kant’s solution: the noumenal self (ii) Undermining free will may also undermine the Categorical Imperative33 (iii) Has specific links to applied areas, e.g. punishment (retributivism

assumes free will)34 VIII. Should the Numbers Count?

A. Nozick: separateness of persons35 1) Bad response: tradeoffs and paralysis 2) Better response: deny the separateness of persons36 3) Nozick himself: perhaps a threshold deontologist37 4) General problems with absolute side constraints

(a) Paradox of Deontological Constraints38 (b) Begs the question, so offer a positive argument for consequentialism first (c) Honoring vs. promoting39 (d) Non‐moral rationality40 (e) No act‐omission distinction41 (f) Uncertainty and infinity42 (g) Spectrum argument43 or scalar properties vs. binary judgments44

B. Taurek: more isn’t worse45 1) Response: violates intransitivity for indifference relation46 2) Response: no outcome can be better or worse?! 47 3) Response: conflates different senses of aggregation48 4) Response: prudential, so why not moral?49 5) What about a more radical stance?

C. Thomson: worlds can’t be good50 1) “Possible world” is not a goodness‐fixing kind51 2) Response: Consequentialism is about a different sense of “good” (goodness with

some motivational component52 or good in the reason‐implying sense that it is fitting to desire53)

D. Bostrom: infinitarian objection54 1) Delink: non‐fundamental principle55 2) Bite the bullet: fine, not the universe as a whole 3) Response: good‐ and bad‐making features only on Earth 4) Response: finite expansions56

IX. Conclusion


CARDS AND CITATIONS

1 Robert Nozick, Anarchy, State, and Utopia (1973), 42‐43.

Suppose there were an experience machine that would give you any experience you desired. Superduper neuropsychologists could stimulate your brain so that you would think and feel you were writing a great novel, or making a friend, or reading an interesting book. All the time you would be floating in a tank, with electrodes attached to your brain. Should you plug into this machine for life, preprogramming your life experiences? . . . Of course, while in the tank you won't know that you're there; you'll think that it's all actually happening. . . . Would you plug in? 2 Bernard Williams, Ethics and the Limits of Philosophy (1985), 87.

There is, first of all, the point that the sums have to come out right. It is naturally characteristic of utilitarianism that its results depend on calculations but, in connections such as this, that feature is particularly undesirable. If racist prejudice is directed toward a small minority by a majority that gets enough satisfaction from it, it could begin to touch and go whether racism might not be justified. The point is not how likely that is to arise, or in what circumstances, but that the whole question of how many racists are involved cannot begin to be an acceptable consideration on the question whether racism is acceptable (contrast Rawls’s treatment of slavery). Moreover, on the utilitarian argument it emerges as a consideration—though, if the sums come out right, not a decisive or winning consideration—that racists get some satisfaction out of the sufferings of the Jews; but this cannot be a consideration at all. This does not mean that the sufferings of racists never count. It means that the sufferings they experience solely because of their racist opinions do not count. Harsanyi, indeed, has built into his system a provision that aims to deal with this kind of problem, by rather briskly excluding “antisocial” preferences from the count. But he does not explain how they are to be defined; his rationale for the provision suggests that what is in question are antiutilitarian preferences. 3 Bernard Williams, Ethics and the Limits of Philosophy (1985), 77‐78. Any form of consequentialism locates ethical value ultimately in states of affairs. . . . [A]gency comes in only secondarily: our basic ethical relation to the world, as agents, is that of being the cause of desirable or undesirable states of affairs. . . . This means that there are states of affairs I can affect with respect to welfare which, because I can do so, turn out to be my own concern when, on nonutilitarian assumptions, they would be someone else’s concern. Moreover, because the class of beneficiaries is larger than that of agents, there are situations that turn out to be someone’s concern when on nonutilitarian assumptions they would have been no one’s concern. . . . The demands of utilitarianism for maximum welfare production are boundless. There is no limit to what a given person might be doing to improve the world, except the limits of time and strength. Moreover, because the relations of possible states of affairs to any given person’s actions are indeterminate, the demands are boundless in the further sense that there are often no clear boundaries between the demands on me and the demands on someone else.


4 David McNaughton and Piers Rawling, “Deontology,” Ethics in Practice, ed. Hugh LaFollette, 3rd edition (2007), 39.

Loyalty is essential to friendship. Loyalty involves the recognition of an underivative agent‐relative obligation to one’s friends. Consequentialism has no place for underivative agent‐relative obligations; thus it has no room for friendship. But friendship, as is generally acknowledged by consequentialists, is an important intrinsic good. Consequentialism holds that the good is to be promoted; but here is a good that it apparently cannot accommodate. . . . Even if we abandon the thought that we are required to favor friends, surely we must be permitted to favor them if our bonds to them are to be special. That is, we must be morally permitted to favor our friends even when we could do more good overall by not doing so. But act consequentialism denies us this permission: an act is wrong if it fails to maximize the good. 5 David McNaughton and Piers Rawling, “Deontology,” Ethics in Practice, ed. Hugh LaFollette, 3rd edition (2007), 40. Does someone have special personal reason to pursue her own benefit, just in virtue of it being hers? We think she does. These agent‐relative personal reasons arise because each of us has our own point of view. I have personal reason to care about my pain that I cannot have to care about yours, namely that it is mine. This does not mean that I have no reason to care about your pain, nor does it commit me to denying that pain is equally bad whoever has it. Options can be justified by appeal to the personal perspective. Each agent has moral permission not to maximize the good when the cost to her would be significant. An agent is allowed, in determining what she is morally required to do, to accord greater weight to the cost borne by her than is warranted by its impersonal badness.

How can this be? Since I am a creature with a personal point of view, a morality that required me to transcend that point of view and think of the world as if I had no particular place in it would not merely be unreasonably demanding, it would deny all moral significance to the fact that my life is, in a sense, all I have. There has, therefore, to be some balance between the demands that the needs of others put on us and our right to live our own lives. Determining where that balance lies is notoriously difficult. But this does not entail that there is no balance to be struck. 6 Brad Hooker, Ideal Code, Real World: A Rule‐Consequentialist Theory of Morality (2000). 7 Derek Parfit, On What Matters, vol. 1 (2011), 404‐405. As Sidgwick argued, the answer is No. If everyone always tried to do whatever would make things go best, these attempts would often fail. When predicting the effects of possible acts, people would often make mistakes, or deceive themselves in self‐benefiting ways. It would be easy, for example, to believe that we were justified in stealing or lying, because we falsely believed that the benefits to us would outweigh the burdens that our acts would impose on others. If we were all Act Consequentialists, that would also undermine or weaken some valuable practices or institutions, such as the practice of trust‐requiring promises. If everyone had the motives of an Act Consequentialist, that would be bad in other ways. For it to be true that everyone nearly always tried to make things go best, most


of us would have to lose too many of the strong loves, loyalties, personal aims, and other motives in which much of our happiness consists, and that also make our lives in other ways worth living. For these and other such reasons, we can claim that (A) if everyone accepted AC, things would go worse than they would go if everyone accepted certain other principles.

These other, UA‐optimific principles would partly overlap with the principles of common sense morality. These principles would often require us, for example, not to steal, lie, or break our promises, even when such acts would predictably make things go best. These principles would permit us to give some kinds of strong priority to our own well‐ being. And they would often permit us, and often require us, to give some kinds of strong priority to the well‐being of certain other people, such as our close relatives and friends, and those to whom we may be related in various other ways, such as our pupils, patients, clients, colleagues, customers, neighbours, and those whom we represent. Since AC is not the principle whose universal acceptance would make things go best, the Kantian Formula does not require us to be Act Consequentialists. 8 Gideon Rosen [Princeton], “The Supreme Principle of Morality,” Ratio 2009, 86‐87.

The examples can be adapted to undermine even this weak claim. Suppose that the gremlin will do his worst if unanimity is achieved, unless the consensus embraces the moral system of (say) the Roman aristocracy under Caligula. It is then a non‐trivial fact about this world that this repugnant moral system is the system whose universal acceptance would make things go best. (UARC) entails that if the world contains this sort of gremlin, it is morally permissible to beat one’s slaves, even if the gremlin will never make his presence felt. But that’s absurd.

[ . . . ] Suppose that Sade will retaliate against Juliette and her family unless she accepts his own corrupt moral system. Then it may be a hard question whether any decent moral system that includes the prohibition against recreational torture amounts to an optimific system. The benefits of universal acceptance would have to be weighed against the costs to Juliette and her family of such acceptance. But while this might be a hard question, it would not be a hard question whether ordinary acts of torture are morally permissible. This shows conclusively, in my view, that it is one thing to ask whether an act is morally permissible, and quite another to ask whether it is permitted by principles whose universal acceptance would make things go best. In a gremlin world these two questions may have very different answers. 9 Derek Parfit, Reasons and Persons (1984), 98. Two‐Person Parent’s Dilemmas are unlikely to occur. But we often face Many‐Person Versions. It is often true that, if all rather than none give priority to our own children, this will either be worse for all our children, or will enable each to benefit his own children less. Thus there are many public goods: outcomes that would benefit our children whether or not we help to produce them. It can be true of each parent that, if he does not help, this will


be better for his own children. What he saves—whether in money, time, or energy—he can spend to benefit only his own children. But, if no parent helps to produce the public good, this will be worse for all our children than if all do. In another common case, such as the Fisherman’s Dilemma, each could either (1) add to his own earnings or (2) (by self‐restraint) add more to the earnings of others. It will here be true of each that, if he does (1) rather than (2), he can benefit his children more. This is so whatever others do. But if all do (1) rather than (2) each can benefit his children less. These are only two of the ways in which such cases occur. There are many others. . . . The problem comes from the giving of priority. It makes no difference whether this is given to oneself or others. 10 Derek Parfit, Reasons and Persons (1984), 141.

Consider any pair of claims that are related in the following way: The first claim contains the word ‘I,’ but does not contain the word ‘now.’ The second claim is just like the first, except that it does contain the word ‘now.’ Call such a pair of claims analogous. If the first claim conflicts with the second, it is incompatible with full relativity, and should therefore be rejected. If the first claim does not conflict with the second, it is an open question whether we should accept the first claim. But, if we accept the first claim, we should also accept the second. This is because, if we accept the first but reject the second, our view is incompatible with full relativity. And we are not, as we ought to be, giving to ‘I’ and ‘now’ the same treatment. 11 Cass Sunstein and Adrian Vermuele, “Is Capital Punishment Morally Required? The Relevance of Life‐Life Tradeoffs,” Chicago Public Law & Legal Theory Working Paper No. 85 (March 2005), p. 17. The most fundamental point is that unlike individuals, governments always and necessarily face a choice between or among possible policies for regulating third parties. The distinction between acts and omissions may not be intelligible in this context, and even if it is, the distinction does not make a morally relevant difference. Most generally, government is in the business of creating permissions and prohibitions. When it explicitly or implicitly authorizes private action, it is not omitting to do anything or refusing to act. Moreover, the distinction between authorized and unauthorized private action – for example, private killing – becomes obscure when government formally forbids private action but chooses a set of policy instruments that do not adequately or fully discourage it. 12 Douglas Portmore, “Can an Act‐Consequentialist Theory be Agent Relative?” American Philosophical Quarterly 38 (2001). 13 Peter Singer and Katarzyna De Lazari‐Radek, “Secrecy in Consequentialism: A Defense of Esoteric Morality,” Ratio (2010). 14 Derek Parfit, Reasons and Persons (1984), section 17. 15 Allan Wood, ‘The Supreme Principle of Morality’, in The Cambridge Companion to Kant and Modern Philosophy, ed. Paul Guyer (Cambridge, 2006), p. 345. 16 Onora O’Neill, Constructions of Reason (Cambridge University Press, 1989), p. 85


17 Peter Singer [Ira W. DeCamp Professor of Bioethics, Princeton], “The Groundwork of Utilitarian Morals: Reconsidering Hare’s Argument for Utilitarianism,” draft prepared for the Conference on Issues in Modern Philosophy: “The Foundations of Morality,” NYU Philosophy Department, November 7, 2009, 3‐4.

Moral judgments, in other words, cannot be based on just any reason. In particular, we cannot make exceptions for ourselves. If I claim that you ought to give to the poor, I must either agree that I too ought to give to the poor, or I must find a reason why it is not the case that I ought to give to the poor, and this reason must be specifiable without reference to individuals. That reason might be that you are a billionaire and I am struggling to keep up with my mortgage payments. But it cannot be that I am the one who will benefit if I keep all my money, whereas you will benefit if you keep all your money. Although Hare’s words sometimes gave the impression that he understood universalizability to be a requirement of reason (for example: “to universalize is to give the reason”) his considered position leaves room for amoralism, and the amoralist, in his view, need not be inconsistent or irrational. Moral language excludes non‐universalizable judgments, but there is no logical requirement that one use moral language, or guide one’s life by universalizable principles. If that is right, the significance of any conclusions that we may be able to reach by drawing out the implications of universalizability will be weakened. One can always escape those conclusions by refusing to make moral judgments. Nevertheless, since many of us do want to use terms like “ought” and “right,” the possibility of rational amoralism does not render pointless the exercise of exploring the implications of universalizability, and seeing what normative conclusions may follow from its application.

When I prescribe something, using moral language, my prescription commits me to a substantive moral judgment about all relevantly similar cases. This includes hypothetical cases in which I am in a different position from my actual one. So to make a moral judgment, I must put myself in the position of the other person affected by my proposed action – or to be more precise, in the position of all those affected by my action. Whether I can accept the judgment – that is, whether I can prescribe it universally – will then depend on whether I could accept it if I had to live the lives of all those affected by the action. 18 Philip Pettit, “Non‐Consequentialism and Universalizability,” Philosophical Quarterly 50 (2000). The upshot is that if as a non‐consequentialist theorist I straightforwardly universalize the prescription that in a certain situation I should instantiate a favored pattern, P, then the prescription to which I thereby commit myself — that in that situation any X ought to instantiate pattern, P — may force me to revise my original self‐prescription. I have equal reason to prefer both that I instantiate P and that any agent instantiate P — this reason is expressed by the use of the word ‘right’ or ‘ought’ in each case — and the spirit of universalisability blocks me from treating myself as in any way special. Thus, if the preferences are inconsistent in a certain situation — if the choice is between my instantiating P alone, for example, or my acting so that many others instantiate P instead —


then I will have reason not to instantiate P myself. As a would‐be non‐consequentialist thinker, my initial claim must have been that the point is to instantiate P in my own life, not promote it generally. But I countenance the general claims of the P‐pattern when I universalize in the straightforward way: I prescribe general conformity to that pattern, not just conformity in my own case. Thus it now seems that what I must think is that this general conformity is to be promoted, even if that means not myself instantiating the pattern in my own behavior or psychology or relationships. It seems that what I must embrace, in effect, is a consequentialism in which conformity to pattern P is the ultimate value to be promoted. 19 Derek Parfit, On What Matters (2011). Quoted in Michael Otsuka, “The Kantian Argument for Consequentialism,” Essays on Derek Parfit’s On What Matters, p. 62. A. Everyone ought to follow the principles whose universal acceptance everyone could rationally will, or choose.

B. Everyone could rationally choose whatever they would have sufficient reason to choose. C. There are some UA‐optimific principles.

D. These are the principles that everyone would have the strongest impartial reasons to choose. E. No one’s impartial reasons to choose these principles would be decisively outweighed by any relevant conflicting reasons. Therefore,

F. Everyone would have sufficient reasons to choose these optimific principles.

G. There are no other significantly non‐optimific principles that everyone would have sufficient reasons to choose.

Therefore, H. It is only these optimific principles that everyone would have sufficient reasons to choose, and could therefore rationally choose.

Therefore, Everyone ought to follow these principles. 20 Kant, Groundwork (4:429).

The human being necessarily represents his own existence in this way; so far it is thus a subjective principle of human actions. But every other rational being also represents his existence in this way consequent on just the same rational ground that also holds for me; thus it is at the same time an objective principle from which, as a supreme practical ground, it must be possible to derive all laws of the will. The practical imperative will therefore be the following: So act that you use humanity, whether in your own person or in the person of any other, always at the same time as an end, never merely as a means.


21 Christine M. Korsgaard—Working Paper as of May 2010. [Professor of Philosophy at Harvard. PhD in Philosophy From Harvard]. Valuing Our Humanity.

This question brings us back to the issues with which I began. In “Kant’s Formula of Humanity” and even more in The Sources of Normativity, I argued that there is a sense in which we must value our moral nature, that is, our practical identity as moral beings, because it is the condition of the value of anything else. In The Sources of Normativity, in particular, I argued that we must value our moral nature as what I call a form of practical identity, a description under which we value ourselves and find our lives worth living. I claimed that our moral identity, which I took to be equivalent to our human identity, is the ultimate source of reasons, because the moral law is the ultimate source of justification – a consideration must be capable of being embodied in a universalizable maxim if it is to count as a reason at all. So if we are to have any reasons, or to see anything as being valuable, we must value and identify with our nature as moral beings. But how can we value our moral nature if we can think neither that it is praiseworthy nor that is prizeworthy to have it? 22 Allen W. Wood, Kant’s Ethical Thought (New York: Cambridge University Press, 1999), pp. 124‐125. (1) This is how the human being necessarily represents his own existence; to this extent it is a subjective principle of action. (2) But every other rational being also represents its existence in this way, based on the same rational ground which is valid for me.

(3) Therefore it is the rational ground of an objective principle which can serve as the foundation for all moral laws.

(4) The practical imperative will therefore be: act so that you use humanity in your own person, as well as the person of every other, always at the same time as end, and never merely as a means. 23 Derek Parfit [Oxford], On What Matters (July 10, 2010 draft), 67. Korsgaard’s remarks provide one example. To illustrate her claim that something can be good ‘because it is desired for its own sake’, Korsgaard writes: ‘chocolate gets its value from the way it affects us. We confer value on it by liking it’. Such examples do not, I believe, show that our desires can create or confer value, or disvalue, by making what we want to have, or to avoid, good or bad. Our future pleasures or pains are not made to be good or bad by our present desires to have these pleasures, and to avoid these pains. And when we are in great pain, by having some sensation that we intensely dislike, what makes our conscious state bad is our intense dislike, not our present desire not to be having the sensation that we dislike. Since our meta‐hedonic desires do not make their objects good or bad, the examples of pleasure and pain do not decisively, or even, I believe, strongly support the view that our other desires have such value‐creating power. Though it is good to have sensations that we like, nothing is good merely because we want this thing. 24 Jeff McMahan [Rutgers], “Challenges to Human Equality,” The Journal of Ethics (2008).


Each of us began life below the threshold of psychological capacity that we had to reach in order to become the moral equal of others. We are now above it. But as I have noted, equality is all‐or‐nothing, and it is implausible to suppose that there was a point when each of us was suddenly transformed into the moral equal of others. For the development and maturation of our psychological capacities is gradual and continuous, without abrupt discontinuities. Given that the difference between being and not being within the scope of liberal egalitarian principles is profoundly significant, it is impossible to make the transition from one type of status to the other through a merely incremental increase in psychological capacity. Because the relevant capacities are matters of degree and lack sharp boundaries, the difference between an individual just above the threshold and an individual just below it cannot be the difference between the possession of some capacity and the failure to possess that capacity. [Even if] Assuming that problem 2 can be overcome and we can identify a sharp threshold dividing those who are our moral equals from those who are not, where does the threshold lie? If it is sufficiently low that very young children are above it, then some nonhuman animals, such as chimpanzees, should be above it as well. If it is high enough to exclude all animals, it will also exclude young children and adults whose psychological capacities are permanently arrested at or below the level of young children. 25 Richard Arneson [University of California, San Diego], “Joel Feinberg and the Justification of Hard Paternalism,” Legal Theory (2005).

Moreover, part of what has gone wrong is the Kantian extravagance that Feinberg pinpoints. Rational‐agency capacity is a complex of abilities, each of which varies by degree. Its value is great, no doubt, but finite, not “unsurpassable.” But as the separate components of rational‐agency capacity dwindle and overall capacity shrinks, the overall value also shrinks. Moreover, our reverence for rational‐agency capacity is, to a large degree, reverence for the potential that rational‐agency capacity gives the bearer in most normal circumstances of human life. This is potential to develop one’s individuality in particular ways, to make something worthwhile of one’s life for oneself and others, to achieve any of an enormously wide range of great goods according to our choices and the luck of circumstances. In abnormal, unfortunate circumstances when this potential disappears or almost entirely disappears, what is left can be of very modest value indeed, and it can be per fectly reasonable to trade off the good of continued existence for some extra period of time as a (somewhat) rational agent against other goods that one can secure by cutting short one’s existence. The idea that these trade‐offs, which are sometimes difficult and complex and sometimes obviously favor cutting short one’s life, are always automatically wrong to contemplate because they offend against the unsurpassable value of rational agency capacity itself is simply not credible. But even rational‐agency capacity in full flowering, possessed by a capable, intelligent human in fortunate circumstances, is reasonably assessed at some large but finite value. The basic idea of the no‐exchange constraint is misguided. 26 Derek Parfit, On What Matters vol. II (2011), p. 161.


If the dignity of persons were a kind of supreme goodness, and Hitler and Stalin had this kind of goodness, that would imply that Hitler and Stalin were supremely good. Since that is clearly false, as Kant would have agreed, we should conclude that, at least when had by persons, dignity is not a kind of goodness. As Wood, Hill, and others claim, the dignity of persons is a kind of ‘moral status’, or a ‘value to be respected’. Though Hitler and Stalin were not good, they had dignity in the sense that, as rational beings, they had the moral status of being entities who ought always to be treated only in certain ways.

Return now to Wood’s Foundational Thesis. If we take ‘rational nature’ to refer to rational beings, or persons, this thesis implies that (1) our reason to treat all persons only in certain ways is provided by the fact that persons have supreme value. This supreme value, as we have just seen, is not a kind of goodness but a kind of moral status. So we can restate (1) as (2) our reason to treat all persons only in certain ways is provided by the fact that persons have the moral status of being entities who ought to be treated only in these ways. In this restatement, Wood’s Thesis becomes less appealing. Nor could (2) be claimed to ground morality’s requirements in what Herman calls ‘a correct analysis of the Good’. (2) claims only that our reason to follow these requirements is provided by the fact that morality requires these acts. This claim does not give morality what Herman calls a value that could be its end or point, showing how morality’s demands make sense. 27 Derek Parfit, On What Matters vol. II (2011), 165. Kant uses ‘rational nature’ to refer both to rational beings and to the rationality of these beings. The value of rational nature therefore consists in part in the dignity of all rational beings, or persons. As we have seen, this dignity is not a kind of goodness, but is the moral status of being entities who ought to be treated only in certain ways. The claim that persons have this status does not help us to decide how persons ought to be treated. When Wood refers to the supreme value of rational nature, he is more often referring to the value of non‐moral rationality, such as prudential rationality. Though Wood no longer claims that our capacity to set an end confers goodness on what we choose, he still takes Kant to be claiming truly that ‘the correct exercise of one’s rational capacities . . . must be esteemed as unconditionally good’. On Kant’s view, Herman similarly writes, ‘the domain of ‘‘the Good’’ is rational activity and agency: that is willing’. These claims are not, I believe, justified. Some kinds of rational activity may have great intrinsic value as achievements, and this would support Kant’s claim that we ought to develop and use our various rational abilities. But unlike good wills, non‐moral rationality cannot be claimed to be supremely good. The rational agency of Hitler and Stalin was not good. Nor, I believe, would Kant have made this claim. On Kant’s view, as Herman notes, what is good is only good willing. 28 Derek Parfit, On What Matters vol. II (2011), p. 161.

Wood also claims that concern for alleviating human suffering is ‘grounded’ in the ‘fundamental value’ of non‐moral rationality. That is not, I believe, true. Our concern to relieve people’s suffering should be grounded, not in the value of these people’s rationality, but in the ways in which suffering is bad for these people, by being a state that they have strong reasons to want not to be in. We have similar reasons to relieve the suffering of


those abnormal human beings who have no rational abilities, and the suffering of non‐rational animals. As Bentham said, our question should not be ‘Can they reason?’ but ‘Can they suffer?’ 29 Jeff McMahan [Rutgers], “Challenges to Human Equality,” The Journal of Ethics (2008).

Most mature human beings have a rational nature and most immature human beings do seem to be inherently or internally directed toward the development of the capacity for rationality. Suppose we grant – what I believe to be plausible – that the actual possession of a rational nature is sufficient to bring an individual within the scope of liberal egalitarian principles.10 There remains the familiar question about the potential for rationality as a basis for this moral status: namely, why should an immature human being’s internal directedness toward the development of a rational nature affect how we ought to treat that individual now? There is, of course, a good answer to that question in cases in which how we treat the individual now will affect her for better or worse later, when she actually has a rational nature. But there are certain types of act the effects of which are limited entirely to the present, including – crucially – acts of killing. Why should the morality of an act of killing be governed by the kind of respect that is appropriate for a nature that the individual killed does not have now but may have later, though only if it is not killed? Why should an act of killing not be governed instead by due consideration for the nature of the individual at the time of action (or of the death, if it occurs later)? 30 Jeff McMahan [Rutgers], “Challenges to Human Equality,” The Journal of Ethics (2008).

I believe, as I argued in The Ethics of Killing, that there are no good answers to these questions. I will not rehearse those arguments here but will instead raise what I think is a more difficult question. What reason is there to suppose that all human beings are in fact internally directed or programmed toward the development of a rational nature? There are some human beings – those who are congenitally and radically cognitively impaired (henceforth the ‘‘radically impaired’’) – who in at least one obvious sense lack the potential for the development of a rational nature. While most immature human beings (embryos, fetuses, newborn infants) will, given a favorable environment, develop a capacity for rationality, those immature human beings that are radically cognitively impaired cannot develop this category, even with the most extensive forms of assistance that we are currently able to provide. 31 David Cummiskey, Kantian Consequentialism (Oxford, 1996). In such a situation, what would a conscientious Kantian agent, an agent motivated by the unconditional value of rational being choose? We have a duty to promote the conditions necessary for the existence of rational beings, but both choosing to act and choosing not to act will cost the life of a rational being. Since the basis of Kant’s principle is rational nature exists as an end‐in‐itself, the reasonable solution to such a dilemma involves promoting, insofar as one can, the conditions necessary for rational beings. If I sacrifice some for the sake of other rational beings, I do not use them arbitrarily and I do not deny the unconditional value of rational beings. Persons may have dignity, an unconditional and incomparable value that transcends any market value, but, as rational beings, persons also


have a fundamental equality, which dictates that some must sometimes give way for the sake of others. The formula of the end‐in‐itself thus does not support the view that we may never force another to bear some cost in order to benefit others. If one focuses on the equal value of all rational beings, then equal consideration dictates that one may sacrifice some to save many. 32 Paul Hurley [Professor of Philosophy, Claremont McKenna College], “ The Hidden Consequentialist Assumption,” Analysis 1992, 244.

If the agent‐neutral restriction on principles is abandoned, and certain agent‐centred principles or standards are allowed to be relevant to the determination of the best available state of affairs, then appeal to the value of states of affairs can result in agent‐ centred restrictions, i.e. in its being wrong to act, in some cases, so as to bring about the best state of affairs as determined from an impersonal point of view. Such value‐based restrictions on bringing about the best state of affairs from an impersonal point of view can result if the best state of affairs available to any agent A is not the best state of affairs available to A as determined from an impersonal point of view. Such a failure of coincidence can occur whenever the principles or standards relevant to the determination of the best state of affairs available to A are not all agent‐neutral principles or standards. 33 Allen Wood, “Kant’s Compatibilism,” Self and Nature in Kant’s Philosophy (1984) 77. The free will problem arises for Kant because he believes that practical freedom requires transcendental freedom and that there is no room in the causal mechanism of nature for a transcendentally free being. All causes in the natural world are sensible or empirical and all necessitate their effects. Consequently, if our actions were determined by them, then they would be necessitated by something sensuous (so that we would lack practical freedom in the negative sense) and there would be no room for a nonsensuous motive for our actions (so that we would lack practical freedom in the positive sense). Practical freedom requires that we be able to determine our actions entirely from within ourselves, through our own legislative reason. Natural causes, however, belong to an endless regressive chain in which there is no spontaneous or first cause. We can think of ourselves as practically free, therefore, only by thinking of our actions as subject to a transcendentally free cause lying outside nature.

Because practical freedom, the freedom required for moral responsibility, must involve the capacity to do what morality demands of us, it is obvious that any account of practical freedom must depend on some account of the demands of morality. In Kant’s philosophy morality has an especially close association with practical freedom. For Kant, morality is autonomy, it is the conformity of the will to a self‐given law of reason. Acting morally, then, is not merely one thing among others that a practically free being can do, but it is the peculiar function of morality to actualize practical freedom. 34 Joshua Greene and Jonathan Cohen, “For the law, neuroscience changes nothing and everything,” Phil. Trans. R. Soc. Lond. B (2004).


The forward‐looking–consequentialist approach to punishment works with all three responses to the problem of free will, including hard determinism. This is because consequentialists are not concerned with whether anyone is really innocent or guilty in some ultimate sense that might depend on people’s having free will, but only with the likely effects of punishment. (Of course, one might wonder what it means for a hard determinist to justify any sort of choice. We will return to this issue in x 8.) The retributivist approach, by contrast, is plausibly regarded as requiring free will and the rejection of hard determinism. Retributivists want to know whether the defendant truly deserves to be punished. Assuming one can deserve to be punished only for actions that are freely willed, hard determinism implies that no one really deserves to be punished. Thus, hard determinism combined with retributivism requires the elimination of all punishment, which does not seem reasonable. This leaves retributivists with two options: compatibilism and libertarianism. Libertarianism, for reasons given above, and despite its intuitive appeal, is scientifically suspect. At the very least, the law should not depend on it. It seems, then, that retributivism requires compatibilism. Accordingly, the standard legal account of punishment is compatibilist. 35 Robert Nozick. “Moral Constraints and Moral Goals.” Anarchy, State and Utopia, New York: Basic Books, 1974. p 31‐33

Individuals are ends and not merely means; they may not be sacrificed or used for the achieving of other ends without their consent. Side constraints express the inviolability of other persons. But why may not one violate persons for the greatest social good? Individually, we each sometimes choose to undergo some pain or sacrifice for a greater benefit or to avoid a greater harm: we go to the dentist to avoid worse suffering later; we do some unpleasant work for its results; some persons diet to improve their health or looks; some save money to sup‐port themselves when they are older. In each case, some cost is borne for the sake of the general overall good. Why not, similarly, hold that some persons have to bear some costs that benefit other persons more, for the sake of the overall social good? But there is no social entity with a good that undergoes some sacrifice for its own good. They are only individual people, different individual people, with their all individual lives. Using one of these people for the benefit of others uses him and benefits the others. Nothing more. What happens is that something is done to him for the sake of others. Talk of an overall social good covers this up. To use a person in this way does not sufficiently respect and take account of the fact that he is a separate person, that his is the only life he has. He does not get some overbalancing good from his sacrifice, and no one is entitled to force this upon him – least of all a state or government that claims his allegiance (as other individuals do not) and that therefore scrupulously must be neutral between its citizens. The moral side constraints upon what we may do, I claim, reflect the fact of our separate existences. They reflect the fact that no moral balancing act can take place among us; there is no moral outweighing of one of our lives by others so as to lead to a greater overall social good. There is no justified sacrifice of some of us for others. 36 Derek Parfit, Reasons and Persons (1984), 338‐339.


These claims can be explained in a different way. Even those who object to balancing think that it can be justified to impose burdens on a child for his own greater benefit later in his life. Their claim is that a person’s burden, while it can be morally outweighed by benefits to him, cannot ever be outweighed by mere benefits to others. This is held to be so even if the benefits are far greater than the burden. The claim thus gives to the boundaries between lives—or to the fact of non‐identity—overwhelming significance. It allows within the same life what, over different lives, it totally forbids.

This claim would be more plausible on the Non‐Reductionist View. Since the fact of identity is, here, thought to be deeper, the fact of non‐identity could more plausibly seem to have such importance. On this view, it is a deep truth that all of a person’s life is as much his life. If we are impressed by this truth—by the unity of each life—the boundaries between lives will seem to be deeper. This supports the claim that, in the moral calculus, these boundaries cannot be crossed. On the Reductionist View, we are less impressed by this truth. We regard the unity of each life as, in its nature, less deep, and as a matter of degree. We may therefore think the boundaries between lives to be less like those between, say, the squares on a chess‐board—dividing what is all pure white from what is all jet black—and more like the boundaries between different countries. They may then seem less morally important. 37 Robert Nozick, Anarchy, State, and Utopia (Basic Books Press, 1974), pp. 29‐30 at fn. 6. Unfortunately, too few models of the structure of moral views have been specified heretofore, though there are surely other interesting structures. Hence an argument for a side‐constraint structure that consists largely in arguing against an end‐state maximization structure is inconclusive, for these alternatives are not exhaustive. (On page 46 we describe a view which fits neither structure happily.) An array of structures must be precisely formulated and investigated; perhaps some novel structure then will seem most appropriate. The issue of whether a side‐constraint view can be put in the form of the goal‐without‐side‐constraint view is a tricky one. One might think, for example, that each person could distinguish in his goal between his violating rights and someone else’s doing it. Give the former infinite (NC) weight in his goal, and no amount of stopping others from violating rights can outweigh his violating someone’s rights. In addition to a component of a goal receiving infinite weight, indexical expressions also appear, for example, “my doing something.” A careful statement delimiting “constraint views” would exclude these gimmicky ways of transforming side constraints into the form of an end‐state view as sufficient to constitute a view as end state. … The question of whether these side constraints are absolute, or whether they may be violated in order to avoid catastrophic moral horror, and if the latter, what the resulting structure might look like, is one I hope largely to avoid. 38 Samuel Scheffler 2003 [Professor of Philosophy at University of California at Berkeley], “The Rejection of Consequentialism,” in Stephen Darwall (ed.) Consequentialism (Maldin, MA: Blackwell, 2003), p. 127


[…] The [Nozick’s] passage suggests that the rationality of agent‐centered restrictions like R as a response to the separateness of persons derives from the fact that the violation of such a restriction harms some victim without compensating him for this harm. But the relevance of this fact is unclear. Why does this consideration make rational the view that it is wrong to violate such a restriction even in order to prevent a still greater number of equally weighty violations of the very same restriction? Whether A1 harms P1, in our schematic example, or A2 . . . A6 harm P2 . . . P6, someone’s separate and distinct life will be violated without compensation. There is no question here of avoiding uncompensated violations altogether. The question instead is: why isn’t it at least permissible to disrupt one distinct individual life, without compensation, in order to prevent the uncompensated disruption of five equally distinct lives? It is obviously no answer to this question to simply reiterate that people are distinct, that each has only one life to lead, and that P1 will not be compensated for the harm he suffers if A1 violates R. This restatement provides no explanation whatsoever of the impermissibility of A1’s inflicting one uncompensated violation in order to prevent A2 . . . A6 from inflicting five similarly uncompensated violations. 39 Phillip Pettit [Professor of Politics and Philosophy, Princeton University], “Consequentialism” in A Companion to Ethics, ed. Peter Singer (1993), 238.

There are at least three respects in which consequentialism scores on simplicity. The first is that whereas consequentialists endorse only one way of responding to values, non‐consequentialists endorse two. Non‐consequentialists all commit themselves to the view that certain values should be honoured rather than promoted: say, values like those associated with loyalty and respect. But they all agree, whether or not in their role as moral theorists, that certain other values should be promoted: values as various as economic prosperity, personal hygiene, and the safety of nuclear installations. Thus where consequentialists introduce a single axiom on how values justify choices, non‐consequentialists must introduce two. But not only is non‐consequentialism less simple for losing the numbers game. It is also less simple for playing the game in an ad hoc way. Non‐consequentialists all identify certain values as suitable for honouring rather than promoting. But they do not generally explain what it is about the values identified which means that justification comes from their being honoured rather than promoted. And indeed it is not clear what satisfactory explanation can be provided. It is one thing to make a list of the values which allegedly require honouring: values, say, like personal loyalty, respect for others, and punishment for wrongdoing. It is another to say why these values are so very different from the ordinary run of desirable properties. There may be features that mark them off from other values, but why do those features matter so much? That question typically goes unconsidered by non‐consequentialists. Not only do they have a duality then where consequentialists have a unity; they also have an unexplained duality.

The third respect in which consequentialism scores on the simplicity count is that it fits nicely with our standard views of what rationality requires, whereas non‐consequentialism is in tension with such views. The agent concerned with a value is in a parallel position to


that of an agent concerned with some personal good: say, health or income or status. In thinking about how an agent should act on the concern for a personal good, we unhesitatingly say that of course the rational thing to do, the rationally justified action, is to act so that the good is promoted. That means then that whereas the consequentialist line on how values justify choices is continuous with the standard line on rationality in the pursuit of personal goods, the non‐consequentialist line is not. The non‐consequentialist has the embarrassment of having to defend a position on what certain values require which is without analogue in the non‐moral area of practical rationality. 40 Michael Smith [Professor of Philosophy, Princeton], “Two Kinds of Consequentialism,” Philosophical Issues 19 (2009).

The first part is a reminder of why some people think that the arguments in favour of some form of consequentialism are overwhelming. As Joseph Raz points out, one of the dominant views theorists have held about reasons for action—he calls it the "classical view"—is that an agent has a reason to act in a certain way just in case something good will result from his acting in that way (Raz 1999, especially Chapters 11‐13). Raz traces this view back to Plato and Aristotle, but it is also extremely widely held by contemporary theorists, both consequentialists (see, for example, Parfit forthcoming), and non‐consequentialists alike (see, for example, Anscombe 1957, and of course Raz himself). But it is difficult to see how those who hold this view of reasons for action could give anything other than a consequentialist explanation of what agents ought to do. In order to see that this is so, consider side‐constraints. If those who accept the classical view are to resist a consequentialist explanation of side‐constraints of the kind given above, such theorists must suppose, on the one hand, that we each have an obligation not to violate people's rights ourselves, and yet also hold, on the other, either that we have no reason to act in that way, which is inconsistent with the normativity of obligation, or that we have a reason that doesn't entail the good of what will thereby come about, which is a denial of the classical view. On the face of it, this combination of views looks to be flat‐out inconsistent. Those who hold the classical view thus seem to be committed to some form of consequentialism. Of course, many who hold the classical view don't see things this way themselves. But we are now in a position to hazard a guess as to why not. 41 Ingmar Persson, “The Two Act‐Omission Paradox,” 2004. Jul. 2004.

There are two ways in which the act‐omission doctrine, which implies that it may be permissible to let people die or be killed when it is wrong to kill them, gives rise to a paradox. First, it may be that when you let a victim be killed, you let yourself kill this victim. On the assumption that, if it would be wrong of you to act in a certain fashion, it would be wrong of you let yourself act in this fashion, this yields the paradox that it is both permissible and impermissible to let yourself act in this fashion. Second, you may let yourself kill somebody by letting an action you have already initiated cause death, e.g., by not lending a helping hand to somebody you have pushed. This, too, yields the paradox that it is both permissible and impermissible to let yourself kill if you are in a situation in which killing is impermissible but letting be killed permissible.


42 Frank Jackson and Michael Smith [Professors of Philosophy, Australian National University and Princeton University], “Absolutist Moral Theories and Uncertainty,” The Journal of Philosophy CIII (2006). A skier is heading in a direction you know for sure will trigger an avalanche that will kill ten people. You know the only way to save the ten people is for you to shoot him. The probability that the skier intends to trigger the avalanche and kill the ten people is 1–p. We can agree that our target absolutist theory says it is right for you to shoot if it is certain that the skier intends to kill the ten, that is, if p = 0, for in that case you would not be killing someone innocent—you would be protecting the ten in the only way possible from an unprovoked attack. We can agree that our target theory says that it would be wrong for you to shoot if you are certain he simply happens to be skiing in that direction, that is, if p 5 = 1, for then you would be intentionally killing someone innocent, and that is never right no matter how many you will be allowing to die by your failure to shoot. The number of lives that would be saved in the example as described is ten, but of course the distinctive position of absolutism is that the number does not matter: it is never right intentionally to kill the innocent no matter how many lives would be saved by doing so. Our question is, What should the theory say for other values of p? […]

The trouble with this response is that there is nearly always some greater than zero chance that someone is innocent. All the evidence may be against them, but induction from the past record of overturned verdicts in cases that looked watertight at the time tells us that there is nearly always some chance that someone who looks clearly to be guilty is in fact innocent. We will get the result that it is never, or hardly ever, right to shoot the skier. Indeed, it will be hard to find any cases where it is right intentionally to kill someone as there is always some chance that the someone is innocent, and a small chance times an infinite disvalue equals an infinite disvalue. We will have a quick (too quick) argument from absolutism against intentionally killing the innocent to an extreme kind of personal pacifism.

[…]

The problem noted in section iii with giving intentionally killing the innocent an infinite disvalue is that whenever p is nonzero, that is, nearly always, (A) will come out worse than (B), which means that it is never right for you to shoot in the skier case—clearly the wrong answer by all lights except extreme personal pacifism, as we noted. The problem noted in section iv with giving intentionally killing the innocent a finite but very large disvalue is that it threatens to make it the case that increasing the number allowed to die in (B) can make (B) worse than (A) even when p, that is, the probability that shooting is indeed the killing of the innocent, is high (in fact, as high as you like). This would mean that we lose the distinctive feature of absolutism, as we noted. 43 Tim Mulgan, “Critical Notice of Jeff McMahon, The Ethics of Killing: Problems at the Margin of Life,” Canadian Journal of Philosophy 34 (2004).


44 Larry Alexander, “Scalar Properties, Binary Judgments,” Journal of Applied Philosophy 25 (2008).

In the moral realm, our deontic judgments are usually (always?) binary. An act (or omission) is either morally forbidden or morally permissible. Yet the determination of an act’s deontic status frequently turns on the existence of properties that are matters of degree. In what follows I shall give several examples of binary moral judgments that turn on scalar properties, and I shall claim that these examples should puzzle us. How can the existence of a property to a specific degree demarcate a boundary between an act’s being morally forbidden and its not being morally forbidden? Why aren’t our moral judgments of acts scalar in the way that the properties on which those judgments are based are scalar, so that acts, like states of affairs, can be morally better or worse rather than right or wrong? I conceive of this inquiry as operating primarily within the realm of normative theory. Presumably it will give aid and comfort to consequentialists, who have no trouble mapping their binary categories onto scalar properties. For example, a straightforward act utilitarian, for whom one act out of all possible acts is morally required (and hence permissible) and all others morally forbidden, can, in theory at least, provide an answer to every one of the puzzles I raise. And, in theory, so can all other types of act and rule consequentialists. They will find nothing of interest here beyond embarrassment for their deontological adversaries. The deontologists, however, must meet the challenges of these puzzles. And for them, the puzzles may raise not just normative questions, but questions of moral epistemology and moral ontology. Just how do we know that the act consequentialist’s way of, say, trading off lives against lives is wrong? For example, do we merely intuit that taking one innocent, uninvolved person’s life to save two others is wrong? Can our method of reflective equilibrium work if we have no theory to rationalize our intuitions? And what things in the world make it true, if it is true, that one may not make the act consequentialist’s tradeoff? I do not provide any answers to these questions any more than I provide answers to the normative ones. But they surely lurk in the background. 45 John Taurek, “Should the Numbers Count?” Philosophy and Public Affairs 6 (1977).

I cannot see how or why the mere addition of numbers should change anything. It seems to me that those who, in situations of the kind in question, would have me count the relative numbers of people involved as something in itself of significance, would have me attach importance to human beings and what happens to them in merely the way I would to objects which I valued. If six objects are threatened by fire and I am in a position to retrieve the five in this room or the one in that room, but unable to get out all six, I would decide what to do in just the way I am told I should when it is human beings who are threatened. Each object will have a certain value in my eyes. If it happens that all six are of equal value, I will naturally preserve the many rather than the one. Why? Because the five objects are together five times more valuable in my eyes than the one.

And so it is in the original situation. I cannot but think of the situation in this way. For each of these six persons it is no doubt a terrible thing to die. Each faces the loss of something


among the things he values most. His loss means something to me only, or chiefly, because of what it means to him. It is the loss to the individual that matters to me, not the loss of the individual. But should any one of these five lose his life, his loss is no greater a loss to him because, as it happens, four others (or forty‐nine others) lose theirs as well. And neither he nor anyone else loses anything of greater value to him than does David, should David lose his life. Five individuals each losing his life does not add up to anyone's experiencing a loss five times greater than the loss suffered by any one of the five.

Now, adding still others to our number, not one of whom will suffer as much as you are asked to bear, will not change things for me. It ought not to change things for any of us. If not one of us can give you a good reason why you should be willing to undergo a greater suffering so that he might be spared a lesser one, then there is simply no good reason why you should be asked to suffer so that the group may be spared. Suffering is not additive in this way. The discomfort of each of a large number of individuals experiencing a minor headache does not add up to anyone's experiencing a migraine. In such a trade‐off situation as this we are to compare your pain or your loss, not to our collective or total pain, whatever exactly that is supposed to be, but to what will be suffered or lost by any given single one of us. 46 Gregory Kavka [Professor of Philosophy, UCLA], “The Numbers Should Count,” Philosophical Studies 36 (1979). Let us name the six strangers, calling the one who needs all the drug Sam, and the others Mel, Tim, Art, Cal, and Len. Let A 1 be the act of giving the whole dose to Sam, and A2 be the act of giving one‐fifth of it to each of the others. According to Taurek, we should be indifferent between A1 and A2. But note that there are other acts the owner could perform. He could, for example, give one‐fifth of the drug to each of Mel, Tim, Art, and Cal, and pour the rest on the ground. Call this act A3.6 Compare A3 and A 1‐According to Taurek's analysis, one should be indifferent between them, between saving the four and saving the one. But it is a general presupposition of rational choice that indifference (i.e. the absence of preference) is a transitive relation. It follows, therefore, that we should also be indifferent between A2 and A3. But this is to be indifferent between saving Mel, Tim, Art, and Cal, on the one hand, and saving all four of them plus Len, on the other. None of us (especially Len) is likely to find this implication of Taurek's view acceptable. 47 Derek Parfit, “Innumerate Ethics,” Philosophy and Public Affairs 7 (1978). At certain points, Taurek suggests that no outcome can be worse than its alternative. One of the two outcomes may be worse for particular people, but it cannot be simply “worse.” On this suggestion, Taurek’s first premise becomes implausible. It implies that, in the absence of special obligations, we have no moral reason to prevent anything. We have no reason even to prevent those outcomes which are worse for everyone. Even Taurek would not accept this. If he keeps the suggestion that no outcome can be “worse” than its alternative, he must therefore abandon his first premise. 48 Derek Parfit, “Innumerate Ethics,” Philosophy and Public Affairs 7 (1978).


What does Taurek not understand? A puzzling passage reads: “Suffering is not additive in this way. The discomfort of each of a large number of individuals experiencing a minor headache does not add up to anyone’s experiencing a migraine.” If “add up to” meant “be the same as,” this would be true. But it would not be relevant. Those who believe that suffering is “additive” do not believe that many lesser pains might be the same thing as one greater pain. What they believe is that the lesser pains might together be as bad. 49 Gregory Kavka [Professor of Philosophy, UCLA], “The Numbers Should Count,” Philosophical Studies 36 (1979).

There are also more general reasons for doubting Taurek's contention that the numbers should not count at all in cases of this kind, reasons based on a plausible analogy between morality and prudence. Consider Taurek's claim that one party should not be forced to undergo a large pain (on the grounds that the total amount of pain suffered would be less) to save a large number of persons from each suffering smaller pains. 8 This seems correct in many cases. But the same principle applies to a rational person's choices between his own pains. Some pains are so awful that it may well be rational for a person to prefer to suffer several somewhat lesser pains than to suffer such a pain once, even if this increases the total amount of pain he will suffer. 9 Yet one can admit this without jumping to the absurd conclusion that a rational person will be indifferent between suffering one pain of a given intensity and many pains of that same intensity. The fact that pains should not be summed in the neat way utilitarians suggest does not mean their numbers do not count at all ‐ for rational prudence or for morality. 50 Judith Jarvis Thomson [MIT], Normativity (2008). 51 (1) “Being a good K is being good qua K,” in contrast to being good and a K. (The latter can’t be true because a good K1 who is also a good K2 need not be a good K2.) So goodness is attached to the kind in question.

(2) “There is such a property as being good qua K if and only if K is a goodness‐fixing kind.” If goodness is attached to the kind in question, there must be some standards for members of some kind to be good.

(3) “Possible world” is not a goodness‐fixing kind. Goodness‐fixing kinds include functional kinds (e.g., knives), kinds designed for certain purposes (e.g., organs), biological kinds (e.g., dogs), kinds with capacities that members of the kind might or might not exercise (e.g., human beings with moral capacities). But “possible world” is not one of these kinds. 52 Michael Smith [Princeton], “On Normativity,” Analysis (2010).

Certain uses of 'good', Hare would say, like the one used in the statement of Consequentialism, accord with the doctrine known as Judgement Internalism. The scope and strength of this doctrine are controversial, but in Hare's view it is a doctrine about the judgements that are in play when we deliberate about what to do. What Judgement Internalism says is that, first, such judgements are about the values of the possible worlds that are available to agents through the pursuit of their options, and second, that when someone judges that a certain available possible world is better than alternatives, he has a


preference for that possible world over the alternatives (Hare 1981 pp.20‐24). Note, however, that if Judgement Internalism, so understood, is indeed a constraint on the legitimate use of 'good' as it is used when we deliberate about what to do, then this suggests one way in which we might mark off the sense of 'good' that is in play in the statement of Consequentialism from the sense of 'good' that Thomson is concerned to analyze. Imagine someone who judges that a particular K is good qua K. It plainly does not follow from this that he prefers that particular K to alternatives. A good virus is presumably one that replicates in a whole variety of hosts. However it does not follow that someone who judges a particular virus to be a good one—imagine a scientist who discovers that some virus he is studying in the lab replicates in nearly every host—prefers that virus to alternatives. He might consistently despise that virus, precisely because it is so good qua virus. Judgement Internalism is thus not a constraint on the meaning of good qua K. Nor is it a constraint on the meaning of good‐modified. Someone who judges that a certain brand of toothpaste would (say) be good for use in infecting whole populations with a certain virus need have no preference for that brand of toothpaste over other brands. But if the sense of 'good' as it is used in the statement of Consequentialism is different from the senses that Thomson is concerned to analyze—if the former is constrained by Judgement Internalism, whereas the latter are not—then her objection to Consequentialism lapses. 53 Michael Smith [Princeton], “On Normativity,” Analysis (2010). If this is right, however, then it seems to me that Thomson's earlier argument against Consequentialism was based on a miscommunication of ideas. That argument, you will recall, assumed that Consequentialists ascribe the property of being good to possible worlds, where 'good' in 'good possible world' means the same as 'good' in 'good K' where K is a goodness‐fixing kind. This is why Thomson thought that Consequentialism makes sense only if possible world is itself a goodness‐fixing kind, which it isn't. Against this, I suggested that Consequentialists might insist that their theory should be stated in terms of a use of 'good' as 'good' is used in deliberative contexts, and I pointed out that Scanlon's suggestion that something is good just in case it has the higher‐order property of having some property that provides a reason to desire it seems the perfect candidate for what's picked out by 'good' in such contexts. But it now turns out that this line of reply is one that Thomson herself could have given on the Consequentialist's behalf. For she believes that the higher‐order property that Scanlon identifies with the property of being good exists, and she agrees that it is an evaluative property, she just doesn't think that it is the property of being good: instead, she thinks that it is the property of being desirable. Thomson might therefore have restated Consequentialism in more plausible terms before attacking it. So stated, Consequentialism is the view that that we can analyze the directives that are true of agents in terms of the desirability of the possible worlds in which those agents perform the actions that are available to them. As far as I can tell, Thomson is in no position to think that Consequentialism, so understood, is incoherent. Indeed, it seems that her own account of what it is for a possible world to be desirable tells us not just that the view is coherent, but also what its attractions are. According to Consequentialism, so understood, someone's φ‐ing is impermissible just in case, when you compare the possible world in which he φs with the possible worlds in which he performs each of the alternative acts


available to him, what you discover is that there is a possible world in which he performs an alternative act that has a feature that makes it deserve to be desired more than the possible world in which he φs. The question to which the Consequentialists quite reasonably demand an answer is what could possibly justify the agent's φ‐ing in these circumstances. How could it be permissible for an agent to act in a certain way if his acting in that way did not deserve being desired? 54 Nick Bostrom, “The Infinitarian Challenge to Aggregative Ethics,” Future of Humanity Institute (2008).

Recent cosmological evidence suggests that the world is probably infinite. Moreover, if the totality of physical existence is indeed infinite, in the kind of way that modern cosmology suggests it is, then it contains an infinite number of galaxies, stars, and planets. If there are an infinite number of planets then there is, with probability one, an infinite number of people. Infinitely many of these people are happy, infinitely many are unhappy. Likewise for other local properties that are plausible candidates for having value, pertaining to person‐states, lives, or entire societies, ecosystems, or civilizations—there are infinitely many democratic states, and infinitely many that are ruled by despots, etc. . . .

Suppose the world contains an infinite number of people and a corresponding infinity of joys and sorrows, preference satisfactions and frustrations, instances of virtue and depravation, and other such local phenomena at least some of which have positive or negative value. More precisely, suppose that there is some finite value ε such that there exists an infinite number of local phenomena (this could be a subset of e.g. persons, experiences, characters, virtuous acts, lives, relationships, civilizations, or ecosystems) each of which has a value ≥ ε and also an infinite number of local phenomena each of which has a value ≤ (‒ ε). Call such a world canonically infinite. Ethical theories that hold that value is aggregative imply that a canonically infinite world contains an infinite quantity of positive value and an infinite quantity of negative value. This gives rise to a peculiar predicament. We can do only a finite amount of good or bad. Yet in cardinal arithmetic, adding or subtracting a finite quantity does not change an infinite quantity. Every possible act of ours therefore has the same net effect on the total amount of good and bad in a canonically infinite world: none whatsoever. Aggregative consequentialist theories are threatened by infinitarian paralysis: they seem to imply that if the world is canonically infinite then it is always ethically indifferent what we do. In particular, they would imply that it is ethically indifferent whether we cause another holocaust or prevent one from occurring. If any non‐contradictory normative implication is a reductio ad absurdum, this one is.” 55 Nick Bostrom, “The Infinitarian Challenge to Aggregative Ethics,” Future of Humanity Institute (2008). Were one to reject aggregative consequentialism as a fundamental moral theory, one may still find an important place for its core idea as a lower‐level moral principle—a principle of limited validity that would be embedded in a more encompassing normative framework and that would come into play only in particular circumscribed contexts. For example, one might hold that certain institutions ought to take an aggregative maximizing


stance with regard to the interests of their constituencies; or, more weakly, that such a stance reflects one type of consideration that some institutions have a duty to incorporate into their decision making. This would lead to none of the difficulties described in this paper, so long as the constituencies are necessarily finite and there is an upper bound on the amount of harm or benefit that can be imposed on any member. To the extent that utilitarian and other aggregationist ideas make their way into the world outside philosophy departments, it is usually in such a circumscribed capacity. Social choice theory often finds it convenient to proceed on the assumption that policies and social institutions exist to serve finite populations of individuals whose interests are defined in terms of their preference structures in such a way as to avoid problematic infinities. We also find echoes of aggregationist consequentialism in such policy tools as cost‐effectiveness analysis, impact statements, and QALY‐based evaluations of health care policies. Properly circumscribed and qualified, these real‐world applications are immune to infinitarian paralysis. 46 56 Peter Vallentyne and Shelly Kagan, “Infinite Value and Finitely Additive Value Theory,” The Journal of Philosophy 94 (1997). The point here is that it is not necessary for judgments of betterness that one world be better than a second relative to all finite sets of locations. It is sufficient for this sort of case, it seems, that any finite set can be expanded sufficiently so that, relative to all further finite expansions, the first world is better. The rough idea is that, if, no matter what finite set of locations you start with, you can expand enough so that, relative to all further expansions, one world is better than another, then the former world is better tout court than the latter world. (In example 2, no matter what finite set one considers, if one expands it to include at least two locations, wl is better than w2, relative to all further expansions.) Finitely additive theories of value rank worlds with a finite number of locations (peoples, times, and so on) on the basis of the total goodness they contain. It is commonly supposed that the only way that such theories can apply to worlds with unbounded infinite numbers of locations is to rank them on the basis of the total goodness they contain. But this has the crazy result that where time (for example) is unbounded, a world with 2 units of goodness at each time is not better than a world with the same locations but only 1 unit at each location. Fortunately, this way of dealing with the infinite case is not necessary for finitely additive theories. And given the implausible rankings it generates, it is not plausible. We have argued in favor of a different approach. At the most basic level we have defended BI, which says that, if two worlds have the same locations, and, relative to every finite set of locations (and a fortiori, every location), one world is better, then that world is better tout court. Somewhat more tentatively, we have defended the strengthening of BI to GM. GM is a strengthening in that, as long as the two worlds are location isometric, it applies even when the locations are not the same in the two worlds. Furthermore, it does not require that one world be better than the other, relative to all finite sets of locations, but only to all bounded uniform expansions of some bounded uniform expansion of any bounded region of locations. Although we can imagine that, due to a faulty understanding of the issues, we may be mistaken about the plausibility of these strengthenings, we are very confident that some strengthenings of BI are plausible. And it should be remembered that we are only


claiming that GM is plausible as far as it goes. We certainly do not think that GM is the strongest plausible metaprinciple for dealing with unbounded locations.

defending consequentialism handout

Documents

sufferingisnotadditivein thisway

good

theinfinitarianchallengetoaggregativeethics

inparticular

philosophicalstudies36

ucla

onnormativity

asrationalbeings