Comments?
I'd love to hear
Mill's
Proof of Utility
Robert Bass
In the fourth chapter of Mill's Utilitarianism, he asks and seeks to answer the question "Of what sort of proof the principle of utility is susceptible?" In addition, of course, he seeks to supply at least a sketch of what he takes the proof to be:
Questions about
ends are ... questions what things are desirable. The utilitarian
doctrine is that happiness is desirable, and the only thing
desirable, as an end; all other things being only desirable as means
to that end. What ought to be required of this doctrine ... to make
good its claim to be believed?
The only proof capable of being
given that an object is visible is that people actually see it ....
and so of the other sources of our experience. In like manner, I
apprehend, the sole evidence it is possible to produce that anything
is desirable is that people actually do desire it. (Mill 1861, p. 61)
This has occasioned frequent criticism to the effect that Mill is fallaciously deriving an ‘ought' from an ‘is' or that he is committing what G.E. Moore called "the naturalistic fallacy," identifying good with some natural property.1, 2
Leaving aside questions about the interpretation of Mill's discussion of "higher pleasures",3 it seems that there are adequate grounds for accusing him of being a naturalist: The good or the desirable is identified with happiness, understood as a favorable balance of pleasures over pains, where pleasures and pains are, in Moore's sense, natural properties.
It is, however, less clear that the "naturalistic fallacy" is a fallacy. In any case, I do not think that Moore has shown that it is. In the first place, if Moore's "open question" argument worked, it would also show that scientific identifications, such as the identification of water with H2O, cannot be correct. That this is not a misinterpretation of the argument is shown by the fact that Moore himself accepted this sort of application of his argument, applying it to the question of whether a visual color, such as yellow, could be identified with electromagnetic vibrations. (Moore, p. 10) But, I submit, it is far more plausible that we are sometimes correct or justified in accepting scientific identifications than that Moore has succeeded in presenting a sound general argument that we cannot be.
Second, there seems to be a relatively straightforward mistake in Moore's argument. The core of his argument is the claim that if good were identical to some natural property -- pleasure, say -- then we would not be able to raise, or at least to be puzzled by, the question whether pleasure is good. For that question would be equivalent to the question, "Is pleasure pleasure?" But this confuses two different senses of "meaning." In one sense, the meaning of a term is something like a definition or a set of criteria for application in the mind of the term's user. In the other, the meaning of a term is its referent, what is meant by it. If "good" meant pleasure only in the first sense, Moore's argument would go through. But he has said nothing to show that "good" couldn't mean pleasure in the second sense, so that pleasure is what "good" refers to and hence is identical with. In that case, if pleasure were not part of the definition or set of criteria for application of the term, "good," then it could be entirely sensible to wonder if pleasure is identical with good. (Even if pleasure were part of the definition or set of criteria for application, we could still wonder if pleasure were identical with the good if there were some other element[s] within the definition or set of criteria. We might wonder, say, whether the good is identical to pleasure experienced by humans rather than to pleasure wherever and by whomever [or whatever] it is experienced.)
So, though it appears that Mill is a naturalist -- he does identify good or the desirable with a natural property, namely, the balance of pleasures over pains -- this by itself is not enough to show that his position or argument is flawed.
Are the complaints that Mill is trying (and failing) to derive an ‘ought' from an ‘is' well-founded? The case that this is what he is doing rests mostly on the passage quoted above. Plainly, as Moore and many others have pointed out (Moore, pp. 66f.), the sense in which an object may be visible has to do with whether it can be seen, but the sense in which something may be desirable is not a matter of whether it can be desired. Rather, it is a matter of whether it ought to be or is worthy to be desired. But Mill appears to be arguing that the fact that something is desired shows that it is desirable. If "desirable" meant "capable of being desired," this would be unproblematic; since it does not, Mill, so it is claimed, is illegitimately inferring an ought-statement, that happiness ought to be desired, from an is-statement, that happiness is desired.
With regard to this charge, I think Mill is pretty plainly not guilty. The appearance is misleading. There is evidence that Mill did not believe that (substantive) ought-statements could properly be inferred from is-statements. Almost twenty years earlier, in his System of Logic, Mill provided his most extensive discussion of methods of justification as applied to ethics and other practical domains. There, he wrote:
Now, the
imperative mood is the characteristic of art, as distinguished from
science. Whatever speaks in rules or precepts, not in assertions
respecting matters of fact, is art; and ethics or morality is
properly a portion of the art corresponding to the sciences of human
nature and society.
The Method, therefore, of Ethics, can be
no other than that of Art, or Practice, in general; and the portion
yet uncompleted, of the task which we proposed to ourselves ... is to
characterize the general Method of Art, as distinguished from
Science. (Mill 1843, pp. 159f.)
Having distinguished art from science, he says
The relation in which the rules of art stand to doctrines of science may be thus characterized. The art proposes to itself an end to be attained, defines the end, and hands it over to the science. The science receives it, considers it as a phenomenon or effect to be studied and, having investigated its causes and conditions, sends it back to the art with a theorem of the combinations of circumstances by which it could be produced.... The only one of the premises, therefore, which Art supplies, is the original major premise, which asserts that the attainment of the given end is desirable. (Mill 1843, p. 161, emphasis added)
If this does not provide a sufficiently clear indication that Mill did not (at the time of the System of Logic) countenance is-ought inferences, he also writes, speaking of the desirable ends proposed by Art:
These are not propositions of science. Propositions of science assert a matter of fact: an existence, a coexistence, a succession, or a resemblance. The propositions now spoken of do not assert that anything is, but enjoin or recommend that something should be. They are a class by themselves. A proposition of which the predicate is expressed by the words ought or should be, is generically different from one which is expressed by is or will be. (Mill 1843, pp. 165f.)
If, in earlier work, Mill had indicated that he did not accept is-ought inferences, had he changed his views by the time that he wrote Utilitarianism? There is no reason to suppose that he did. It is quite clear within the text of Utilitarianism that he was not claiming to and did not mean to represent himself as deriving an ‘ought' from an ‘is.' In the first chapter, he asserts that
such proof as [the principle of utility] is susceptible of .... cannot be proof in the ordinary and popular meaning of the term. Questions of ultimate ends are not amenable to direct proof. Whatever can be proved to be good must be so by being shown to be a means to something admitted to be good without proof.... If ... it is asserted that there is a comprehensive formula, including all things that are in themselves good ... the formula may be accepted or rejected, but is not a subject of what is commonly understood by proof. (Mill 1861, p. 34)
At the beginning of the fourth chapter, he reiterates this and adds that "[to] be incapable of proof by reasoning is common to all first principles, to the first premises of our knowledge, as well as to the first principles of our conduct." (Mill 1861, p. 61)
From passages such as these, it may be inferred that, unless Mill is either disingenuous or simply not quite fully aware of what he is saying, he is not attempting to provide a proof "in the ordinary and popular meaning."
But if Mill wasn't trying to deduce an ‘ought' from an ‘is', what was he trying to do instead? I think the correct answer is suggested by noting what positions Utilitarianism is concerned to argue against. He begins by speaking of "the little progress which has been made in the decision of the controversy respecting the criterion of right and wrong," and refers to "the most gifted intellects .... divided ... into sects and schools carrying on a vigorous warfare against one another." (Mill 1861, p. 31) His antagonists are other moral thinkers -- intuitionists, believers in a "moral instinct," and a priori moralists such as Kant -- who differ from him with respect to the "criterion of right and wrong." All the arguments he offers against other positions are directed against different conceptions of the good or of right action than his own.
None of his arguments are directed against those who deny either that anything is objectively good or that we can know what it is.4 If the "proof of utility" had been directed to them, it would clearly fall flat, and if that had been the audience that Mill was seeking to convince, he could justly be charged with trying (and failing) to derive ought-statements from is-statements. However, given his actual intended audience, he could assume that they agreed that something (or other) was objectively good or right or desirable and also that we have some way of knowing what that is; the question was how to determine what that "something" is.
In the context of trying to settle that question, he points out that "the first premises of our knowledge .... being matters of fact, may be the subject of a direct appeal to the faculties which judge of fact -- namely our senses and our internal consciousness." (Mill 1861, p. 61) Thus, though we do not prove those first premises (else, they would not be first), we do have or take ourselves to have cognitive capacities suitable for making judgments concerning them. The claim that we do have such cognitive capacities is supported (though not proven and not directly discussed here by Mill) by the extent to which we can reach intersubjective agreement on what we judge to be fact. Such support would not, of course, convince anyone who was comprehensively skeptical about the "faculties which judge of fact." But, since nothing would convince them, we are, in appealing to those faculties, giving "all the proof which the case admits of, [and] all which it is possible to require." (Mill 1861, p. 62)
Similarly, "the sole evidence it is possible to produce that anything is desirable is that people do actually desire it. If the end which the utilitarian doctrine proposes to itself were not, in theory and in practice, acknowledged to be an end, nothing could ever convince any person that it was so." (Mill 1861, pp. 61f.) That is, the fact that people do desire something is taken, for purposes of determining what is desirable, as the appropriate analogue to sensing (or "internal consciousness") in the judgment of facts.
Why is it an appropriate analogue? First, Mill is not addressing moral skeptics, but theorists who admit both that something is objectively good or desirable and that we have some way of knowing or recognizing what it is. Given that, the argument that nothing could commend an end to us if we did not in fact desire it is powerful. It's worth trying to set this argument out in a bit more detail.
Suppose that something, X, is never desired for its own sake. If people care about it at all, they care about it as a means to or as an obstacle to getting or bringing about what they want for other reasons. Could that be the good, the ultimate "criterion of right and wrong"? Perhaps, there could be an argument that proceeded from the ends that we do have and gave us reason to adopt X as an end. In one way or another, such an argument would have to say that adopting X as an end is a means to (at least some of) our other ends. Or, perhaps X is such that if we considered adopting it as an end -- which we so far have not -- we would do so. If either of these were the case, then it might be that the end that we would adopt could also be "the criterion of right and wrong." But neither of these really satisfies the initial condition, that X never be desired for its own sake. And, given that the present dispute is between moral theorists who admit that something (or other) is objectively good and that it is knowable to us what that something is, X could not be the good.
Second, if we do take desire to be crucial evidence as to desirability, though not something which deductively supports judgments of desirability, we find -- or so Mill would claim -- that there is a significant measure of intersubjective agreement as to what is desirable. What kind of intersubjective agreement could Mill be implicitly appealing to here?5 Plainly, it cannot be intersubjective agreement on moral theory, for then Mill would have no (or only very few) antagonists against whom to argue. His claim cannot be that everyone (or almost everyone) agrees that general happiness is the supreme standard of morality.
It might be suggested that he is appealing, first, to the wide consensus that exists on the moral rules of common sense, even among thinkers who disagree on what is their appropriate foundation, and second, to the even wider consensus on what are the morally relevant features of actions or situations. For example, almost everyone, though they may subscribe to different moral theories and disagree on some of the concrete applications to specific cases, agrees that saving life, preventing death, promoting health, preventing or curing illness and injury are both (a) morally relevant features of actions and situations and (b) count in favor of those actions.6 This suggests, at least, that our judgments are not merely arbitrary, but are somehow under the control of something real: we agree because, whether we realize it or not, we are attending to the same features of actions or at least to features highly correlated with one another.7
I think this kind of agreement is relevant to Mill's case. It is implicit in his remark to the effect that the "steadiness or consistency" that "moral beliefs have attained has been mainly due to the tacit influence of a principle not recognized." (Mill 1861, p. 33) It is also implicit in his answer to those who criticized utilitarianism for demanding too much of us in the way of calculation of consequences by appealing to the "rules of morality for the multitude" as already embodying utilitarian considerations. (Mill 1861, p. 52)
However, I do not think that that is the agreement Mill is appealing to here. Such considerations may strengthen the case for ethical objectivity of some sort, but they do not especially bear on the correctness of utilitarianism. Instead, the only agreement to which he directly appeals here is the agreement on happiness as an end. Immediately following his statement that if the utilitarian end "were not, in theory and practice, acknowledged to be an end, nothing could ever convince any person that it was so," he continues:
No reason can be given why the general happiness is desirable, except that each person, so far as he believes it to be attainable, desires his own happiness. This, however, being a fact, we have not only all the proof that the case admits of, but all which it is possible to require, that happiness is a good, that each person's happiness is a good to that person, and the general happiness, therefore, a good to the aggregate of all persons. (Mill 1861, p. 62)
On the surface, this appears to be an attempt to get to the utilitarian conclusion in a single step. Each person desires his own happiness; therefore -- because being desired is the only evidence available for desirability -- his own happiness is a good for each person, and the general happiness is a good to all persons. That has seemed to most critics to go too quickly. But, though there is something puzzling here, that is not how Mill takes the result of this stage of the argument. He says only that "[h]appiness has made out its title as one of the ends of conduct and, consequently, one of the criteria of morality." (Mill 1861, p. 62)
What is puzzling about the passage is the question as to what justifies the claim that the general happiness, even if Mill is not at this point arguing that it is the sole principle of morality, is a good to the aggregate of all persons. If Mill were saying that each person has a reason to care about or promote the general happiness because each treats his own happiness as an end, then, to say the least, argument is lacking. No reason has so far been given that one could not regard one's own happiness as a good without being concerned at all about others' happiness (except insofar as it promotes or inhibits one's own pursuit of happiness or other ends).
However, it appears that Mill did not mean to be making such an argument. In a letter written in 1868, he explained,
As to the sentence ... [where] I said that the general happiness is a good to the aggregate of all persons I did not mean that every human being's happiness is a good to every other human being, though I think that in a good state of society and education it would be so. I merely meant ... to argue that since A's happiness is a good, B's a good, C's a good, etc., the sum of all these goods must be a good. (quoted in Schneewind 1965, p. 339)
On this reading, the claim that the general happiness is a good to the aggregate of persons seems to say nothing more than the conjunction which combines statements about happiness as a good of each.
Yet, something still seems puzzling. Mill makes it clear that he takes something being good to be equivalent to its being desirable as an end. But ends are ends by virtue of actually being pursued. Why, then, if he is not saying that anyone does pursue or desire the general happiness, does he call it a good? What licenses the summing up of goods that are actually sought into an aggregate "good" that is not -- at least, he is not here claiming that it is -- actually sought?
I am not at all sure what the proper answer to this is, but I can offer a proposal. Mill may be relying upon his claim that desire is evidence of desirability. But it is consistent with that to hold that desirability is a property that the object of desire has apart from being desired. It is desired because it is desirable; its desirability is not constituted by desire and pursuit of it as an end. If this is correct, then Mill was not only claiming that happiness is objectively a good to or for those who experience it, but that its goodness is not agent-relative.
The argument, then, would be that just as we can infer from the facts that A, B, and C each have mass that the aggregate of A, B, and C has mass, we can also infer from happiness being a good for each of A, B, and C that the happiness of all of them is a good. There is still room to suspect that there may be a fallacy of composition here. (Compare "Subatomic particles are each invisible. Therefore, aggregates of subatomic particles are invisible.") But it isn't obviously mistaken.
Returning now to the main line of Mill's argument, he takes himself only to have established that happiness is a good and one of the ends of human conduct. Much of the remainder of the chapter is devoted to arguing that, appearances to the contrary, nothing else is an end of conduct. Clearly, he does not need to be concerned about his thesis being upset by those objects of pursuit which are only sought as means to something else. However, it does appear that some things other than happiness are sought for their own sake. In fact, he wants to claim that, on utilitarian grounds, there is reason that some of them, such as virtue, should be sought for their own sake.
Mill's general strategy for reconciling this apparent incompatibility is to claim that everything other than happiness that is sought as an end is something that has come to be, through a process of habit-formation in the person who seeks it, experienced as a part of happiness. "Those who desire virtue for its own sake desire it either because the consciousness of it is a pleasure, or because the consciousness of being without it is a pain, or for both reasons united...." (Mill 1861, p. 64)
On the whole, I do not think this reconciliation works. It can be questioned on a number of grounds, but for the present, I will mention only one. In discussing the case of the person whose "virtuous will is still feeble, conquerable by temptation, and not to be fully relied on," Mill says that the will to be virtuous can only be "implanted or awakened .... by making him think of [virtue] in a pleasurable light, or of its absence in a painful one. It is by [such means] that it is possible to call forth that will to be virtuous which, when confirmed, acts without any thought of either pleasure or pain." (Mill 1861, p. 66) This may be true, though I think Mill has not done nearly enough to show that there is no other way in which a desire for virtue for its own sake can be inculcated, but even if it is universally true, it does not show what he needs. For if it is, and if the case can be generalized, it shows that the acquisition of a desire for something other than happiness as an end involves seeing that other end as a means to happiness. It does not show that, once the new end is acquired it is, at the later time, desired as a part of happiness. To put the point more shortly: A desire for happiness may be part of the cause of my coming to desire something else as an end; it does not follow that when I do desire it as an end, that I am "really" desiring happiness.
If this is correct, then Mill was mistaken to hold that we pursue nothing except as a means to or a part of happiness. Accordingly, he should -- in common with most more recent utilitarians -- have reconstrued the utilitarian ideal in terms of preference satisfaction.
However, something, I think, can be salvaged from Mill's approach. If the utilitarian end is preference satisfaction, and if preference satisfaction is construed in terms of von Neumann-Morgenstern utility functions (perhaps the only reasonably rigorous way to construe it), then there is a significant problem with interpersonal comparisons. Since a utility function is defined in terms of gambles over pairs of options offered to a single agent and picks out an arbitrary zero-point, there seems to be no way to say that my preference satisfaction is greater than, less than or equal to yours.
However, interpersonal comparisons do not seem to be a problem in principle with respect to happiness. Though there are numerous difficulties in practice in comparing the happiness of different persons, there is little doubt that, at least in cases in which there is substantial divergence, we can reliably judge that one person is happier than another. But, given this and the fact that each person seeks his own happiness as an end, we can use the standard apparatus of defining utility functions to say what the satisfaction of other preferences is worth in terms of happiness. Even though most or all persons will have preferences for other things in addition to happiness, it will be possible (in principle) to assign a value to the satisfaction of their other preferences expressed in happiness. (This is no stranger than the fact that we can calculate a person's net worth in terms of dollars without implying that he possesses nothing but dollars; it only matters that the other things he possesses be exchangeable at approximately knowable ratios for dollars.) If we can say how much the satisfaction of other preferences is worth in terms of happiness, then, since happiness can be compared between persons, we have the rudiments necessary to compare preference satisfactions between persons in general. Even if we can't say that happiness is the only thing sought as an end, we can use it as a metric for the kinds of comparisons that utilitarians need to make in order to derive any determinate and distinctive prescriptions from their principles. So, in a different form, utilitarians would still be able to speak of and employ a "greatest happiness principle."
Nonetheless, Mill accepted the argument that we sought nothing but happiness, while undogmatically leaving it to the consideration of others as "a question of fact and experience, dependent, like all similar questions, upon evidence." (Mill 1861, p. 65) Having established this point to his own satisfaction, it is at this point (and not before) that he claims that
[since] human nature is so constituted as to desire nothing which is not either a part of happiness or a means of happiness . . . we can have no other proof, and we require no other, that these are the only things desirable. If so, happiness is the sole end of human action, and the promotion of it the test by which to judge of all human conduct; from whence it necessarily follows that it must be the criterion of morality, since a part is included in the whole. (Mill 1861, p. 65)
Is this enough?8 I think Mill believed that it was, at least for the purpose of identifying "the criterion of right and wrong." He has presented an argument which, though questionable in places, does not seem to fall into any of the more obvious pitfalls that unsympathetic critics have accused him of straying into. It also seems that if, for the sake of argument, we grant that the questionable steps are correct, it is an argument that actually does support its conclusion. It can be set out in this way:
1) The good or
the desirable is both objective and knowable (as Mill's critics
agree).
2) If desire were not evidence of desirability, then
nothing could be.
3) Therefore, desire is evidence of
desirability.
4) But all people desire happiness as an
end.
5) Therefore, happiness is desirable for each person.
6)
And because the good is objective, so is the general happiness for
all persons.
7) Moreover, people seek nothing except as means
to or part of happiness.
8) Therefore, the general happiness
is the proper end or criterion of all action.
References
Mill, John
Stuart [1843]. "Of the Logic of Practice, or Art; Including
Morality and Policy," reprinted in Schneewind, J. B. (ed.)
[1965]. Mill's Ethical Writings. New York: Collier.
Mill,
John Stuart [1861]. Utilitarianism in Smith, James M. and
Ernest Sosa (eds.) [1969]. Mill's Utilitarianism: Text and
Criticism. Belmont: Wadsworth.
Moore, G. E. [1976 (1st
ed., 1903)]. Principia Ethica. Cambridge: Cambridge University
Press.
Schneewind, J. B. (ed.) [1965]. Mill's Ethical
Writings. New York: Collier.
1 These are often conflated so that
anyone who tries to deduce an 'ought' from an 'is' is said to be
committing the naturalistic fallacy. I think they should be kept
distinct. Though it is true that if the good is identical with some
natural property, then an ought-statement could be deduced from an
is-statement (and an is-from an ought-statement, for that matter), it
is not necessarily the case that a deduction of an ought-statement
from an is-statement involves the identification of the good with
some natural property.
2 Understanding
what Moore means by this is problematic since it is less than
completely clear what he means by a natural property: he defines it
differently in different places. Consequently, it is less than
entirely clear what he is denying in denying that good is a natural
property. For present purposes, I shall just assume that some clear
sense can be made of that denial (see Moore, especially pp. 13-14,
38-41).
3 In "Mill's
Higher Pleasures and the Choice of Character," Roderick Long has
argued that there is an interpretation that reconciles the references
to higher pleasures with an undiluted commitment to hedonism. (Draft,
1991. A shorter version of his argument was later published in
Utilitas, November 1992, pp. 279-297.)
4 I
will use "moral skeptics" to refer to those who would deny
either or both of these disjuncts. Calling them moral skeptics does
not seem quite appropriate, but I have not found a better term.
5
That is, what kind of intersubjective
agreement could Mill be appealing to if I am correct in thinking that
his proof is meant to highlight an analogue with the knowledge of
fact derived through perception and "internal consciousness"?
6 Similarly, there are a vast
array of features of actions and situations on which there is
widespread consensus that they are morally irrelevant, for example,
that (without special explanation) an act should occur at a certain
time of day. There is more controversial ground on the question of
whether moral theories allow discrimination on the basis of features
such as skin color or sex, for many apparently have. In large part,
however, and whatever the motivations of the proponents, there have
been attempts to represent discriminatory treatment as based on other
differences. (The interests of women are alleged to be better
protected if they are denied certain legal rights, slaves are really
not competent to run their own lives, etc.)
7 Of
course, Mill could press the further question why moral thinkers have
tended to converge on features that correlate highly with one another
if they are not tacitly attending to the same thing.
8
What may appear to be missing from this is
any account of why (motivationally) the general happiness should be
pursued. But I think this is not really an omission. Mill has
discussed motivations bearing on the pursuit of the general happiness
elsewhere -- largely in the earlier chapter, "Of the ultimate
sanction of the principle of utility," and had clearly stated
elsewhere that one should not "confound the rule of action with
the motive of it." (Mill 1861, p. 46)