Utility and Rational Self-Interest

17
Utility and Rational Self-Interest Author(s): Michael McDermott Source: Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition, Vol. 46, No. 2 (Sep., 1984), pp. 199-214 Published by: Springer Stable URL: http://www.jstor.org/stable/4319701 . Accessed: 25/06/2014 08:02 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. . Springer is collaborating with JSTOR to digitize, preserve and extend access to Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition. http://www.jstor.org This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AM All use subject to JSTOR Terms and Conditions

Transcript of Utility and Rational Self-Interest

Page 1: Utility and Rational Self-Interest

Utility and Rational Self-InterestAuthor(s): Michael McDermottSource: Philosophical Studies: An International Journal for Philosophy in the AnalyticTradition, Vol. 46, No. 2 (Sep., 1984), pp. 199-214Published by: SpringerStable URL: http://www.jstor.org/stable/4319701 .

Accessed: 25/06/2014 08:02

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Philosophical Studies: AnInternational Journal for Philosophy in the Analytic Tradition.

http://www.jstor.org

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 2: Utility and Rational Self-Interest

MICHAEL McDERMOTT

UTILITY AND RATIONAL SELF-INTEREST*

(Received 26 July, 1983)

The aim of true morality, utilitarians agree, is the maximisation of utility. But if they mean different things by "utility" - pleasure, or happiness, or content- ment, or preference-satisfaction - their moral aims may be very different. What, then, do "utilitarian" moral theories have in common? Consequential- ism, certainly; but some consequentialist moral theories are more distinctively utilitarian than others. The special feature of what we would be most happy to call "utilitarian" theories is, I suggest, that they see the aim of morality as the maximisation for people in general of what it is in each person's interest to maximise for himself. Very often, when utilitarians disagree about whether it is morally better to maximise pleasure or happiness or whatever, it is just because they disagree about whether it is pleasure or happiness or whatever which is in an individual's best interests.

The aim of this paper is to provide a definition of "utility" which will allow the formulation and discussion of utilitarianism in abstraction from questions about the basis of rational self-interest. The idea is to define "utility" in terms of self-interest itself, rather than any of the various things which may be thought to constitute the basis of rational self-interest. In this way we might hope to formulate a "hard-core" utilitarianism which was neutral between competing accounts of rational self-interest - it would need to be combined with some specific theory of rational self-interest (plus empirical data) to yield moral judgements about specific cases.

As it turns out, my definition is not completely neutral between competing accounts of rationl self-interest. I don't need to enquire what kind of con- scious states are in a man's interest. I can even allow that things other than conscious states can be in a man's interest - e.g. the satisfaction of his preferences regarding things not within his experience. But I do need to assume that the satisfaction of his preferences is only in his interest if they are current preferences - a man is not benefited in virtue of the satisfaction of preferences he no longer has.

Philosophical Studies 46 (1984) 199-214. 0031-8116/84/0462-0199$01.6(0 C) 1984 by D. Reidel Publishing Company

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 3: Utility and Rational Self-Interest

200 MICHAEL McDERMOTT

My definition of utility in terms of rational self-interest is inspired by von Neumann and Morgenstern.' They, however, were operating with a very different conception of utility - the standard economist's concept of indi- vidual preference. Their aim was to provide measurement criteria for certain statements about individual preferences. (For instance, they argued that if a man prefers C to A, and A to B, then if he prefers A to a 50-50 gamble between B and C, "this provides a plausible base for the numerical estimate that his preference of A over B is in excess of his preference of C over A .,,2)

On my definition, however, statements about utility do not describe people's actual preferences, but are about what people should choose, from the point of view of rational self-interest, in certain hypothetical choice situations. They are thus value-judgements, of a kind, but not moral judgements.

On the basis of my definition, I put forward a brief argument for "hard- core" utilitarianism. This is inspired by Harsanyi's argument3 for what he calls "preference utilitarianism". His argument, using the economist's concept of utility, seems to me invalid for essentially the reasons given by Gauthier.4 I believe my argument captures what is attractive in Harsanyi's without repeating his errors.

I define "utility" by reference to choices or preferences among lotteries. The possible outcomes of these lotteries are to include the possibility of being "put in another man's shoes". To make this conception of outcome a little more precise, we may say that an outcome-description is a description of a stretch in a man's life, beginning at a specified time tn, and extending in- definitely into the future, but with a variable replacing the name of the man living that life. Thus "x has an apple for breakfast, enjoys it, gets indigestion, and ..." is an outcome-description. If Tom has an apple for breakfast, enjoys it, gets indigestion, and ..., and Dick has an apple for breakfast, enjoys it, gets indigestion, and..., then Tom and Dick get the same outcome. Outcome- descriptions may contain references to other individuals, who may or may not be specified independently of the person getting the outcome: for example "x takes x's wife to see the Queen." We allow that two descriptions can be of different outcomes even if a person getting those outcomes would not be aware of any difference: for example, "x's wife is faithful to him" describes a different outcome from "x's wife is unfaithful, but he never finds

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 4: Utility and Rational Self-Interest

UTILITY AND RATIONAL SELF-INTEREST 201

out". But outcomes cannot differ merely in what happens before tn, the time at which they begin: if tn is breakfast-time, then "x, who had previously requested an apple for breakfast, gets one" does not describe a different out- come from "x, who had not previously requested an apple for breakfast, gets one." 5

Next, let a, ...,ak, bI, ..., bm be outcomes, all beginning at the same time tn .

Let "U(ai)" stand for "the utility of ai". Let Pl, ***, Pk be positive real numbers summing to unity; and simrilarly q 1 , *--, q..

Now we can give the contextual definition of "utility":

Dfn (1)

k m

z Pi U(ai) > 1 qi U(bi)

means:

If any man has to choose between (i) a lottery in which he has, for each i(l S i < k), a probability

pi of getting outcome ai, and (ii) a lottery in which he has, for each i(l < i < m), a probability

qi of getting outcome bi, then, from the point of view of rational self-interest, he should choose (i).

Also,

k m

z piU(ai) = z qiU(bi) i=l i=l

means:

If any man has to choose between lotteries (i) and (ii) above, then, from the point of view of rational self-interest, it is indif- ferent which he chooses.

k We call , pi U(ai) the expected utility of the lottery (i).

i=1

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 5: Utility and Rational Self-Interest

202 MICHAEL McDERMOTT

I now give a few examples of the application of this definition.

1. The Average Utility of Twvo Societies

Let the outcomes ai and bi be the lives (beginning at tn) of the members of two actual or possible societies A and B, with k and m members respectively. Then the average utility of the lives of the members of A is I 1k/k U(a,), and of B m I1/m U(bi).

Dfn (1) says that

The average utility of the lives of people in A is greater than in B.

means:

If any man has to choose between (i) a lottery in which he has an equal chance of living any of the

lives of people in A, and (ii) a lottery in which he has an equal chance of living any of the

lives of people in B, then, from the point of view of rational self-interest, he should choose (i).

2. Outcome a Is of Greater Utility Than Outcome b

For any outcome a, U(a) = 1U 1= I1 U(a) -the utility of an outcome equals the expected utility of a "lottery" in which one is certain to experience that outcome. So

U(a) > U(b)

means:

If any man has to choose between (i) getting outcome a for certain, and (ii) getting outcome b for certain, then, from the point of view of rational self-interest, he should choose (i).

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 6: Utility and Rational Self-Interest

UTILITY AND RATIONAL SELF-INTEREST 203

3. Comparison of differences in utility, for example

U(a) - U(b) r U(c) - U(d)

This is mathematically equivalent to

1 U(a)+ UrU(d) = U(b)+ rU( ) 1+ r 1+ r l+r l+ r

and hence means:

If any man has to choose between (i) a lottery in which he has a probability 1/1 + r of getting out-

come a, and a probability rnI + r of getting outcome d, and (ii) a lottery in which he has a probability 1/1 + r of getting out-

come b, and a probability rnI + r of getting outcome c, then, from the point of view of rational self-interest, it is indifferent which he chooses.6

4. Zero and Unit

There are, however, a variety of other statements about utility, apparently well-formed, to which Dfn (1) assigns no meaning. Some lack meaning just because we have defined no zero-point for the measurement of utility; for example:

Total utility is greater in society A than in society B.

U(a) - U(b) r U(c)

U(a) -U(b) > U(c)

To give meaning to these, it is necessary and sufficient that Dfn (1) be sup- plemented by an assignment of meaning to "U(a) = 0". One possible definition might be: "U(a) = 0" means that any man should, from the point of view of rational self-interest, be indifferent between the outcome a and being dead from tn on (where t,, is the time at which a begins). This would, for instance, make

Total utility is greater in society A than in society B.

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 7: Utility and Rational Self-Interest

204 MICHAEL McDERMOTT

mean:

If any rational man has to choose between (i) a lottery in which he has, for each ai, a probability 1 /k +m

of getting ai, and a probability m/k + m of being dead from tn on, and

(ii) a lottery in which he has, for each bi, a probability l/k +m of getting bi, and a probability k/k + m of being dead from tn on,

then, from the point of view of rational self-interest, he should choose (i).7

But I shall ignore this supplementation of Dfn (1) in what follows. I shall like- wise consider no suggestions as to how a unit of measurement for utility might be defined.

II

To establish the formal adequacy of Dfn (1), we need the following two principles of rational self-interest:

(P1) If any two men8 have to choose between any two given lotteries, then, from the point of view of rational self-interest, they should both choose the same lottery, or else both be indifferent between them.

(P2) If a man chooses among lotteries in accordance with rational self- interest, his preferences satisfy von Neumann and Morgenstern's axioms for rational choice.

Given these principles, an obvious adaption of von Neumann and Morgen- stem's fundamental theorem shows that there is a real-valued function U on outcomes satisfying Dfn (1).9

Now, as to (P2): certain of the vN-M axioms have been objected to, as requirements for rational choice. For instance, they imply that if living is rationally preferable to dying, living is rationally preferable to any lottery between living and dying; yet, it is argued, a man might enjoy taking part in such a lottery, and hence rationally choose it in preference to the certainty of life.

It seems, however, that all such objections assume the falsity of (P1). The

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 8: Utility and Rational Self-Interest

UTILITY AND RATIONAL SELF-INTEREST 205

plausibility of the particular objection above, for instance, depends on the assumption that whether a man enjoys taking part in life-and-death lotteries is relevant to the question whether he should do so (from the point of view of rational self-interest); it would not be claimed that everyone should prefer to life a certain life-and-death lottery. I shall, therefore, concentrate on defend- ing (P1).

II(a)

Let me first get out of the way a rather superficial objection to (P1). Suppose Tom likes apples but Dick prefers bananas, and they are both to choose whether to have an apple or a banana for breakfast. Then, it might be object- ed, Tom should choose the apple while Dick should choose the banana: they should choose differently among the same outcomes.

Reply: the obvious fact that a man's tastes determine how he should choose can be reconciled with a denial that his tastes determine how he should choose between given lotteries, and with a denial that men with different tastes should choose differently among the same lotteries. For the chooser's tastes are relevant to the question what the outcomes of his choices are likely to be. "Getting an apple for breakfast" does not specify a single outcome, but a lottery in which various outcomes are possible, for instance "getting an apple and liking it" and "getting an apple and not liking it". What are the probabilities of these outcomes? That will usually depend on whether he likes apples now. His present tastes partly detennine what are the lotteries he has to choose between. That is why this evidence is relevant to his choice between the apple and the banana.

This claim, that the chooser's tastes are relevant only because they are evidence for his tastes during the relevant outcomes, gains support if we imagine a case in which the chooser knows for certain what his tastes are going to be at breakfast time, and they are going to be quite different to his tastes at the time of the choice. In such a case we would be happy to admit that it was the latter tastes which were relevant to the question what he should choose, not those at the time of the choice.

II(b)

Now a more difficult case. Tom is to read, at some future time, a certain

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 9: Utility and Rational Self-Interest

206 MICHAEL McDERMOTT

pornographic book; his choice is whether to be lewd or virtuous at that time - that is the only difference between the two outcomes. Dick faces the same choice. Because Tom is now virtuous, he chooses to remain so, even though reading the book in that state will be quite upsetting. Dick, on the other hand, is now lewd, and chooses to be lewd when he reads the book. Tom and Dick have different preferences between the very same outcomes. Here the apparent relevance of their divergent preferences cannot be accounted for by reference to what they show about the likelihood of various outcomes on a given alternative: for each alternative, the outcome (fully described) is certain.

To get a counterexample to (P1), however, we need more than a case in which two men do choose differently between the same outcomes; we need a case in which they should choose differently, from the point of view of rational self-interest. Is it true, in the case described, that Tom should choose the virtuous outcome, while Dick should choose the lewd outcome?

Where a choice is between two certain (i.e., sure) outcomes, the one Tom should choose, from the point of view of rational self-interest, is the one on which he will be better off. He should choose the virtuous outcome iff he will be better off on the virtuous outcome than on the lewd outcome. Dick should choose the virtuous outcome iff he will be better off on the virtuous outcome. To make the objection work, then, it would have to be shown that Tom will be better off on the virtuous outcome, while Dick will be better off on the lewd outcome, because of the difference in their present preferences.

It seems obvious, however, that how well off a man is over a given period (beginning at t,n say) is completely determined by what happens, or what is true of him, over that period. If Tom and Dick both get the virtuous out- come, they will both have exactly the same preferences, experiences, memories, reactions and so on over that period. They will have different histories, to be sure - different things will have been true of them before t,. But if these past differences result in no differences whatsoever over the period beginning at tn, they must be equally well off over that period.

So Tom would be exactly as well off on the virtuous outcome as Dick would be. Similarly, Tom would be exactly as well off on the lewd outcome as Dick would be. Hence Tom would be better off on the virtuous outcome than on the lewd outcome iff Dick would be, too. If Tom does choose the virtuous outcome, and Dick does choose the lewd outcome, one of them is not choosing what he should choose, from the point of view of rational self- interest. There is nothing surprising in this: people often fail to see where

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 10: Utility and Rational Self-Interest

UTILITY AND RATIONAL SELF-INTEREST 207

their own best interests lie; moreover, people often act, and act rationally, out of motives other than self-interest ("rational" does not always mean "rational, from the point of view of self-interest" 10).

I conclude, then, that men with different tastes should not choose differently between given outcomes, because how well off a man will be on a given outcome is not affected by his present tastes.

H1(c)

I have already described one misunderstanding which might lead people to think that a man's present desires can be relevant to the question how he should choose between given outcomes - a failure to specify outcomes fully. Another is the ambiguity of hypothetical statements of rational self-interest.

Consider a choice between posthumous outcomes: Smith has decided to bequeath his fortune to the Art Gallery; he wonders whether to make the bequest anonymous; the following seems plausible:

(S) If Smith wants to be posthumously famous, he should make the bequest in his name; if he wants to be posthumously obscure, he should make it anonymous.

It is tempting to think that we have here a counterexample to (P1). For compare two cases: in one case Smith wants posthumous fame, and faces this choice between posthumous fame and posthumous obscurity; in the other case Smith faces the same choice, but wants posthumous obscurity. (S) seems to imply that the Smith of the first case should choose one thing, the Smith of the second case another thing, contrary to (P1).

Although this is a tempting line of thought, it must be wrong. It would hardly be suggested that a man's choice of one posthumous outcome rather than another could make him any better off during the relevant period, the period beginning at his death: from the point of view of rational self-interest, no posthumous outcome is better than any other. Whence, then, the plausibil- ity of (S)?

(I am assuming that a bequest is just a choice between posthumous out- comes. Actually, though, what Smith chooses to do will affect his state of mind in the period preceding death, and his present desires are relevant to his choice to the extent that they bear on the likelihood of various pre-death outcomes, on a given altemative. (If he chooses fame, will he enjoy that

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 11: Utility and Rational Self-Interest

208 MICHAEL McDERMOTT

prospect in his remaining life?) This may be one source of the plausibility of (S). But I shall explain the plausibility of (S) even on the assumption that the bequest has no significant pre-death effects.)

A sentence of the form

If agent A wants goal G he should do act x.

is generally capable of two quite different interpretations. (I assume the "should" is the "should" of rational self-interest, rather than of morality, say.) First, as a genuine conditional, to which modus ponens applies. Second, as merely a recommendation of x as a means to G. Here are two sentences for which the most natural interpretations are respectively the first and the second of these:

If Jones wants to kill himself, he should ring Lifeline. If Jones wants to kill himself, he should take an overdose.

A single speaker may accept both of these, in their different senses, if he believes that the best way to kill yourself with an overdose, but that it is better to try to get talked out of a suicidal intention than to act on it. Given the further information that Jones does want to kill himself, he will apply modus ponens to the first only, drawing as conclusion the categorical "should"- statement

Jones should ring Lifeline.

He will not conclude that Jones should, categorically, take an overdose. He would only conclude that if he thought that Jones should kill himself. In general, the difference between the two interpretations of "If A wants G he should do x" seems to be that on the first interpretation the hypothesis is merely that A does want G, whereas on the second interpretation the hypoth- esis is that this is a rational desire, from the point of view of the agent's self-interest - the real hypothesis is "A should (try to) get G".1

Returning now to (S), its plausibility comes from the fact that it is prob- ably true if interpreted in the second of the two ways we have distinguished - and hence this becomes the natural interpretation. A bequest in his own name would be the best way to achieve posthumous fame, an anonymous bequest the best way to preserve his obscurity; we can accept this, and yet not be committed to different "should"judgements about the case where Smith wants posthumous fame and the case where he wants obscurity. (So

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 12: Utility and Rational Self-Interest

UTILITY AND RATIONAL SELF-INTEREST 209

(S) provides no counterexample to (P1).) We can accept (S), on the natural interpretation, without being committed to any categorical "should"judge- ments about either case. We need not say of the Smith who wants posthumous fame that he should (categorically) make the bequest in his name; we need not say of the Smith who wants obscurity that he should make it anonymous- ly.

Where a man chooses between two purely posthumous outcomes, neither choice is dictated by rational self-interest - neither outcome makes the chooser any better off. But this implies that neither choice is contrary to rational self-interest: a man facing such a choice can do what he likes, as far as rational self-interest is concerned. So if Smith does want to be posthu- mously famous, we can reasonably advise him "Make the bequest in your own name". That, however, is quite different from saying that he should do it - it does not suggest that such a choice is required by rational self-interest.

(I should perhaps make it clear that I have not claimed that there is never anything to choose between posthumous outcomes from a moral point of view. Nor does my position imply that there is no moral obligation to respect the wishes of the dead; that would only follow on the crudest kind of act- utilitarianism. 12)

II(d)

People do choose differently between the same lotteries. Partly this is because they place different values on the various outcomes which each lottery might result in. But it is also partly because they have different attitudes to risk. So far I have been considering the claim that the first factor is relevant to how people should choose, and to simplify that discussion I have used examples of "lotteries" with no element of risk - straight choices between certain, fully- described, outcomes. I now turn to the second factor.

Suppose Tom has to choose between two lotteries. One presents the pos- sibility of a great gain, but also the possibility of a great loss. The other is less risky - the possible loss is small, but so is the possible gain. What Tom actually chooses may depend on whether he is a risk-lover, or risk-averse. But I shall argue that his attitude to risk is not relevant to the question what he should choose, from the point of view of rational self-interest.

I appeal to the concept of benevolence. To act benevolently is to be guided solely by the interests of the object of your benevolence. If I have to

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 13: Utility and Rational Self-Interest

210 MICHAEL McDERMOTT

choose between lotteries where the various outcomes will be got by Tom, say, I act benevolently if I choose the lottery which is in his best interests, i.e., the lottery which he should choose, from the point of view of his own best interests, if he were choosing between the same lotteries.

Suppose, then, that (i) I have to choose between the risky lottery and the less-risky lottery for Tom, that (ii) I am motivated solely by benevolence for Tom, and that (iii) Tom knows nothing of this - he does not realise that his future is in my hands. (The set-up is that I will choose one alternative, then further factors will bring about one outcome or another; this will be an out- come for Tom, but all Tom ever knows about is the outcome he eventually gets.) In this situation, I suggest, there is no plausibility in the claim that Tom's attitude to risk is relevant to the rationality of my choice. For Tom takes no risk - only I do. But the lottery I should choose in this situation must be the same as the one Tom should choose, when it is he who does the choosing. In both cases, the lottery which should be chosen is the one which is in Tom's best interests. Tom's attitude to risk does not affect which lottery I should choose when acting out of benevolence for Tom; so it does not affect which he should choose either, from the point of view of his own self-interest.

The plausibility of the claim that what a man should choose depends on his attitude to risk comes, I think, from a failure to include enough in the outcome of a choice. It is tempting to think that he makes his choice, and then waits - anxiously, perhaps - for the outcome; if he had chosen a safer alternative, he might have had less anxiety about the outcome. But this can't be right: if the level of anxiety depends on the alternative chosen, it must be included as part of the outcome of that choice. The outcome includes every- thing that depends on the choice. If two men differ in their liking for the anxiety which follows their choice, they do not experience quite the same outcomes. For the man who is certain to enjoy the anxiety, the possible outcomes of the risky lottery, say, are "a period of enjoyable anxiety fol- lowed by a great gain" and "a period of enjoyable anxiety followed by a great loss". For the risk-averse man, the possible outcomes of the risky lottery are very different. It is therefore not really the same lottery. A man's attitude to risk, then, only helps determine how he should choose because it bears on the question what are the lotteries he has to choose between. There is no reason to think that men with different attitudes to risk should choose differently among the same lotteries.

This completes my defence of (P1).

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 14: Utility and Rational Self-Interest

UTILITY AND RATIONAL SELF-INTEREST 211

III

I want to comment on the bearing of my definition of "utility" on the tradi- tional problem of the logical status of interpersonal comparisons of utility. On my account, interpersonal comparisons, and such other statements about utility as have been assigned meaning, are about what people should choose, from the point of view of rational self-interest, in certain hypothetical choice situations. The question is, what is the logical status of such judgements of rational self-interest?

They are, I believe, value judgements. It is not just a coincidence that a whole battery of terns which play a key role in value judgements of other kinds occur injudgements of rational self-interest also. "Should", for instance, can also be used to make moral judgements, or aesthetic judgements (e.g. "You shouldn't put a magenta cushion on a scarlet sofa" - Hare's example). We can speak of a good choice, or the best choice, from the point of view of rational self-interest, as from the moral or aesthetic points of view. We can say he did the right thing or the wrong thing, either as a judgement of rational self-interest or as a moral or aesthetic judgement. And so on.

Like other value judgements, judgements of rational self-interest commend the actions mentioned. They indicate that the speaker is in favour of that action.

And, like other value judgements, they commend on the basis of standards. (Equivalently, they are universalisable; goodness from the point of view of rational self-interest, like other kinds of goodness, is supervenient.) If Tom should do x, and Dick should not do x, there must be some difference between Tom and Dick, or their circumstances, which is the reason for the difference in what they should do; this follows whether the "should" is that or morals, aesthetics, or rational self-interest.

As with morals and aesthetics, again, standards of rational self-interest can vary from person to person. This is not just a matter of using some key words with different meanings. People can have different moral or aesthetic stan- dards, even though they both speak ordinary English, and use the terms of moral or aesthetic appraisal in their ordinary English sense. People can have different moral or aesthetic standards even if they speak no language. So it is with rational self-interest: we can fundamentally disagree about what counts as a good reason, from the point of view of rational self-interest, without our disagreement being merely verbal.

This last point has been denied, in connection with both rational self-

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 15: Utility and Rational Self-Interest

212 MICHAEL McDERMOTT

interest and morality. Just as some have thought, for example, that "(moral- ly) right" means "productive of maximum happiness for people in general", so some have thought that "right (from the point of view of rational self- interest)" means "productive of maximum happiness for the agent". Others have defined rational self-interest in terms of the agent's pleasure or satisfac- tion. My own view is that this is just as much a fallacy (the "naturalistic fallacy") in the field of rational self-interest as it is in ethics. For instance, it may be that anyone who denies that rational self-interest lies in the maximisa- tion of satisfaction is wrong, but it is surely not self-contradictory to main- tain that it is better (from the point of view of rational self-interest) to be Socrates dissatisfied that a fool satisfied.

What I have said about he possibility of different standards of rational self-interest may seem to be incompatible with my defence of (P1) above. For if people have different standards of rational self-interest, they will some- times choose differently among the same outcomes. What (P1) denies, how- ever, is the possibility of a case in which two people should (from the point of view of rational self-interest) choose differently among the same out- comes: that they do choose differently does not mean they should. The objection cannot succeed unless we assume that people should (from the point of view of rational self-interest) choose according to their own stan- dards of rational self-interest. But there is no reason to think that people's choices should be determined by their standards of rational self-interest (at the time of the choice), any more than by their tastes. We do not accept that it is morally right for a man to do something because it is right by his stan- dards; why should we accept that it is in a man's interests to do something because it is in accordance with his standards of rational self-interest? (This is something on which we can agree even if we ourselves differ about rational self-interest: (P1) is compatible with a wide range of standards of rational self-interest. Similarly, people with a wide range of moral views can agree that a man does not do the morally correct thing just because he acts on his own moral principles.)

On my account, then, interpersonal comparisons or utility are value judge- ments - as, indeed, are intrapersonal comparisons of utility. But they are not moral judgements. (It has sometimes been maintaned that interpersonal com- parisons of utility are really moral judgements.) That morality and rational self-interest are distinct seems obvious enough. Although many philosophical analyses of value judgements have failed to distinguish between moral judge-

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 16: Utility and Rational Self-Interest

UTILITY AND RATIONAL SELF-INTEREST 213

ments and judgements of rational self-interest, or even between those and aesthetic judgements, I take it for granted that these apparently distinct kinds of value-judgements are really distinct. Exactly what is the distinction between them, however, remains an open question.'3

IV

Now here is my argument for utilitarianism. Its premise is the following moral principle:

The morally best society is that which any man should choose (from the point of view of rational self-interest), if he has an equal chance of living the life of any person in the chosen society.

This principle is perhaps not entirely self-evident. But it fares well, I think, against its more faithfully Rawlsian competitors. (Letting our chooser know he has an equal chance of getting any position, rather than withholding knowledge of probabilities, or saying his position will be assigned by his enemy, seems the best way of ensuring that he gives equal weight to every- one's interests.) Moreover, as I argue elsewhere,14 the usual kind of cases put forward as counterexamples to utilitarianism on the score of distributive justice have little force against this principle. Yet the principle is, by Dfn (1), equivalent to a form of utilitarianism:

The morally best society is that in which the average utility of people's lives is greatest.

NOTES

* I am grateful to J. B. Griffin, R. M. Hare and Nicholas Rescher for helpful comments and suggestions. 1 J. von Neumann and 0. Morgenstern, Theory of Games and Economic Behaviour (New York, 1964). 2 Op. cit., p. 18.

The key papers are collected in J. C. Harsanyi, Essays on Ethics, Social Behaviour and Scientific Explanation (Dordrecht, 1976). See also his Rational Behaviour and Bargain- ing Equilibrium in Games and Social Situations (Cambridge, 1977), Ch. 4. 4 David Gauthier, Critical Notice of Harsanyi's Essays, Dialogue 17 (1978), pp. 696- 706. 5 Whether this outcome (an apple for breakfast) benefits a man may depend on his previous request, as far as this account of outcome goes. (But I argue against such a dependence below.)

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions

Page 17: Utility and Rational Self-Interest

214 MICHAEL McDERMOTT

6 This works as long as r is positive. If r is negative, use

U(b) - U(a) = -r U(c) - U(d)

For, if U(death) = 0, then k m

T, U(ai) > 2: U(bi) i=1 i=1

iff k+ U(al) + ... + m U(ak) + k + m U(death)

>1 k lU(b) + + U(b + k U(death). k+m k+m k eam 8 Including the one man at different times: (P1) is meant to assert that how a man should choose between given lotteries does not depend on when he chooses between them. 9 If (P1) were not a true principle of rational self-interest, Dfn (1) would have the un- desirable consequence that some expected utilities are incomparable. For suppose there were two men who should choose differently (from the point of view of rational self- interest) between a certain two lotteries: then it would not be true to say that any man should choose the first, so it would not be true to say its expected utility is greater than the second's; and similarly, it would not be true to say that its expected utility is equal to the second's, nor that it is less.

If (P2) were not true, Dfn (1) would be incompatible with the standard interpreta- tions of the contexts in which "utility" is defined. For instance, the vN-M axioms say that the relation among lotteries "is chosen in preference to" is transitive. Suppose that certain rational preferences were not transitive, and it were rational to choose outcome a in preference to b, b in preference to c, and c in preference to a. Then, by Dfn (1), U(a) > U(b) > U(c) > U(a). This is impossible on the standard interpretation of " > ". 10 Similarly for "prudent": one can act prudently or imprudently, in ordinary usage, where one's own interests are not affected. There seems, however, to be a philosopher's use of "prudence" in which it does imply self-interest. " This account agrees in essential respects with that of Hare's 'Wanting: some pitfalls' (reprinted in R. M. Hare, Practical Inferences (London and Basingstoke, 1971), pp. 44-58). 12 Similarly, consider this case (R. B. Brandt, A Theory of the Good and the Right (Ox- ford, 1979), p. 250 - the case is due to Parfit and Griffin): "a convinced sceptic who has rebelled against a religious background wants, most of his life, no priest to be called when he is about to die. But he weakens on his deathbed, and asks for a priest. Do we maximise his welfare by summoning a priest?" There is certainly some plausibility in the claim that we should (morally) deny his request, in accordance with his past desires. But that is consistent with the view that he will be better off, because happier, in his last moments, if we get the priest. Getting the priest might be an act of benevolence, and yet wrong, because it represents a kind of betrayal. 13 Some tentative suggestions, which I still believe to be on the right track, are made in my 'How to preach', Canadian Journal of Philosophy 8 (1978), pp. 633-652. 14 'Utility and distribution',Mind XCI (1982), pp. 572-578.

Department of Philosophy,

University of Sydney, N.S. W. 2006,

Australia

This content downloaded from 195.34.79.254 on Wed, 25 Jun 2014 08:02:00 AMAll use subject to JSTOR Terms and Conditions