Canadian Journal of Philosophy

Transcription

1 Canadian Journal of Philosophy An Argument for Utilitarianism Author(s): Yew-Kwang Ng and Peter Singer Source: Canadian Journal of Philosophy, Vol. 11, No. 2 (Jun., 1981), pp Published by: Taylor & Francis, Ltd. Stable URL: Accessed: 02/02/ :50 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at. 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 support@jstor.org.. Taylor & Francis, Ltd. and Canadian Journal of Philosophy are collaborating with JSTOR to digitize, preserve and extend access to Canadian Journal of Philosophy.

2 CANADIAN JOURNAL OF PHILOSOPHY Volume XI Number 2, June 1981 An Argument for Utilitarianism YEW-KWANC NG and PETER SINGER Monash University Many utilitarians accept Bentham's view that to argue for the principle of utility is as 'impossible as it is needless'. They take utilitarianism as a first principle which one either accepts or does not. They do, of course, defend utilitarianism against objections, and make objections to other ethical positions; but the principle of utility itself, they hold, must stand on its own merits. In this article we use a different approach. We introduce a principle, which we call 'Weak Majority Preference', which we believe likely to be accepted by many who do not consider themselves utilitarians. We then show that from this principle it is possible to derive the general principle of utility. Preference and Welfare We begin with a preliminary distinction, between preference and welfare. Though the preferences of an individual have been used - especially by economists - as an indicator of his welfare, there are three reasons why they may not always be a good indicator. 229

3 Yew-Kwang Ng and Peter Singer First, the preference of an Individual may be affected not only by his own welfare but by his consideration for the welfare of other individuals. Thus it is possible for a person to be himself less happy in x than in y, and yet to prefer x to y, because he believes, e.g., that other people are happier in x than in y. It is true that the belief that other people are happy may make him happy. But this may not be strong enough to outweigh the loss he has to suffer from changing from y to x. For example a person may vote for party x, knowing that he himself will be better off with party y in government. The reason he votes for x is that he believes that the majority of people will be much better off with x. This in itself may make him feel better. However, this benefit may not be important enough to overbalance, in terms of his subjective happiness, his personal loss, say in income, under x. He may yet vote for x due to his moral concern for the majority. Secondly, preference may differ from welfare due to ignorance and imperfect foresight. While an individual may prefer x to y believing he will be better off in x than in y, it may turn out to be the other way round. This is the question of expectation versus realisation. While expectations are relevant for explaining behaviour, it is the realisation which is the actual welfare. Thirdly, an individual may have irrational preferences. The preference of an individual is defined here to be irrational if he prefers x over y despite the fact that his welfare is higher in y than in x, neither due to a consideration of the welfare of other individuals (any sentient creature can be an individual here), nor to ignorance or imperfect foresight. This definition of irrationality makes the three factors discussed here exhaustive causes of divergence between preference and welfare. There are at least two sources of irrational preference. First, an individual may stick rigidily to some habit, custom, 'principles', or the like, even if he knows that this is detrimental to his welfare and the welfare of others in the long run, taking account of all effects and repercussions. Customs, rules, moral principles, etc. may have a rational basis if they provide simple guides to behaviour which are, at least on the whole, conducive to social welfare. It will be too cumbersome and time-consuming if an individual is to weigh the gain and loss in terms of social welfare or his own welfare each time he has to make a decision. Thus he may stick to his routine, rules, principles, etc. without thinking about the gain and loss. If this results, occasionally, in decisions inconsistent with his welfare and the welfare of others, it may be regarded as a cost in pursuing generally good rules. If, say, there is a change in circumstances, the adherence to some rules may result in persistent net losses in welfare, taking everything into account. An individual may stick to these rules without knowing that they no longer are conducive to welfare. Then the divergence between preference and welfare can be 230

4 An Argument for Utilitarianism traced to ignorance. If he knows this and yet sticks to the rules, he is irrational. Many readers may disagree with the definition of irrationality adopted here. For example, suppose a man sticks rigidly to the principle of honesty and would not tell a lie even if that would save his life and contribute to the welfare of others, taking everything into account. According to our definition here, he is acting irrationally. For those who are willing to accept honesty as an ultimate good in itself, he may not be considered irrational. Fortunately, one does not have to agree with the definition of irrationality given here to agree with the arguments of this paper. The second source of irrational preference is the excessive fear of danger or pain and the excessive temptation of pleasure. Consider a person with an aching tooth which will cause nagging pain for a long time but can be treated by a dentist with just a once-and-for-all agony. Fearing the agony, he may refuse to go to the dentist, thus subjecting himself to a greater amount of pain. His refusal to go to the dentist may be due to the ignorance of the greater amount of pain or to the ignorance of the fact that the agony of dental treatment is not really that bad. Then the cause of his preference that does not contribute to his welfare is classified under ignorance. But if he just lacks the determination to undergo a brief, severe pain, he is being irrational. It may be objected that an individual himself is the sole judge of his own welfare and irrationality cannot therefore arise. But some in- dividuals will admit that they have irrational preferences. One may, after a few days, decide to face up to the dentist and agree that one should have done so days earlier. After distinguishing preference and welfare, let us ask: which one is more interesting, preference or welfare? The answer depends partly on the problem on hand. Thus a statesman is more interested in the welfare of the people while a politician is more interested in their preference. In a market economy, preference is the relevant concept for descriptive theories of production and distribution because consumers pruchase goods according to their preferences. Welfare may be a more satisfactory criterion for assessing the performance of an economic, political or social system, for it seems a more credible ultimate end. However, welfare is difficult to measure. Preference is easier to capture operationally. Hence, one may wish to use the preference of an individual as an indicator of his welfare unless there are important specific reasons to suggest a definite divergence between his preference and his welfare. This is what we shall do in the remainder of this article. 231

5 Yew-Kwang Ng and Peter Singer Finite Sensibility There is one more preliminary point. If we are to use a person's preferences as an indication of his welfare, one might think it reasonable to require that the preference be transitive - in other words, that if he prefers x to y, and y to z, he will also prefer x to z. Similarly if he is indifferent between x and y, and prefers x to z, he should also prefer y to z, etc. Normally this will be the case; but the fact that human beings are not equipped with perfect powers of discrimination can lead to violations of transitivity. For example, suppose an individual prefers two spoons of sugar (x) to one (y) in his coffee. If we increase the amount of sugar continuously from one spoon, we will reach a point y' (say 1.8 spoons) for which the individual cannot tell the difference between x and y'. There may exist another point y" (say 1.6 spoons) for which he is indifferent to y' but he prefers x to y". Hence, a perfectly rational individual may have intransitive indifference.1 This intransitivity is well known in the literature of psychology on the concept of 'just noticeable difference'. The concept was touched on as far back as 1781 by Borda and in 1881 by Edgeworth. Edgeworth called it 'minimum sensibile' and took it as axiomatic, or, in his words 'a first principle incapable of proof, that the 'minimum sensible' or the just-perceivable increments of pleasure, of all pleasures for all persons, are equitable.2 Weak Majority Preference We can now introduce our value premise. It is as follows: For a community of n individuals choosing between two possibilities, x and y, if no individual prefers y to x, and at least - - individuals prefer x to y, then x increases social welfare and is preferable. 1 One of us has shown elsewhere that we may even have intransitivity of preference for choices involving multi-dimensions. See Yew-Kwang Ng, 'Sub-semiorder: A Model of Multi-dimensional Choice with Preference Intransitivity', Journal of Mathematical Psychology 16 (1977) J.C. de Borda, 'Memoire sur les Elections au Scrutin', Memoires de I'Academie Royale des Sciences, 1781, English translation by A. de Grazia, Isis, 1953; F.Y. Edgeworth Mathematical Psychics (London: Kegan Paul 1881) 7ff, 60ff. 232

6 An Argument for Utilitarianism We call this the principle of Weak Majority Preference, or WMP.3 We believe that WMP is a modest and reasonable premise, and should be acceptable to many who do not consider themselves utilitarians. It could roughly be described as a combination of majority rule and the Pareto Criterion. The Pareto Criterion - which states that in a choice between x and y, x increases social welfare and ought to be adopted if at least one individual prefers x to y and no individual prefers y to x - is very widely accepted in the literature of welfare economics. It is taken to be a less controversial principle than the utilitarian principle, since it does not state that we ought to override one individual's preference in order to satisfy the preferences of a larger number of individuals, or the more intense preferences of an equal number. (It is important to remember that the Pareto principle does not say that overriding an individual's preference would be wrong; it simply makes no comment on the situation. It is a sufficient condition of correct choice, not a necessary one.) It is easy to see that those who accept the Pareto principle must also accept WMP, and conversely those who reject WMP must reject Pareto. Common to both principles is the requirement that no-one should prefer y to x, if the choice of x over y is to be recommended. In addition to this, the Pareto principle requires merely that at least one individual should prefer x to y, while the rest remain indifferent. WMP requires that at least half the individuals of the community prefer x. Hence whenever WMP is satisfied, the Pareto principle will be satisfied too. Of course, there are some who reject the Pareto principle - and would also reject WMP in its present form - because of the absence of any reference to the preservation of individual rights, equality, justice, or other non-consequentialist moral principles. We do not ourselves regard as ethically acceptable any principles which are supposed to be intrinsic values independently of their contribution to any or all individual welfares; but since we cannot argue this point here, those who insist that rights, equality, justice etc. are intrinsic values independently of their effect on the welfare of individuals may wish to add a ceteris paribus clause holding these alleged values unaffected. In this case WMP yields only a qualified form of utilitarianism in which the maximization of the sum of individual utilities is subject to the nonviolation of rights, equality, justice, etc. 3 Strictly, this formulation covers only cases in which n, the number of individuals, is even. If n is odd, we would require that, instead of -J- individuals preferring x to y, at least - ^- individuals prefer x to y, and at least another individual's utility level has not decreased. 233

7 Yew-Kwang Ng and Peter Singer We shall now explain why WMP leads to utilitarianism. WMP stipulates that if the consequences of an alternative are preferred by half a population, this suffices to make that alternative the one that ought to be chosen, so long as it is not contrary to the preferences of the remainder. When we couple this idea with the fact of finite sensibility, it indicates that alterations in people's situations marked enough to affect their preferences are more significant than alterations which do not affect their preferences. This allows - indeed requires - us to redistribute goods from one half of the population to the other, if the amount we take from the first half is insufficient to affect their preferences, while the amount we bestow upon the second half is noticeable to, and preferred by, the second half. Since any group of individuals comprising 50% of the population is an acceptable half, this has the effect of making a just-perceivable increment of pleasure of any individual an inter-personally comparable unit of utility. An example may help. Suppose that in a given society everyone prefers 2 teaspoons of sugar in their coffee to any noticeably larger or smaller amount. Suppose, too, that people in this society can tell the difference between 1.0 and 1.2 teaspoons of sugar, and they prefer the greater amount; they cannot, however, discriminate between 2.0 and 1.8 teaspoons. (There is, we may assume, a law of diminishing marginal discriminability, which means that the more you have, the less likely you are to be able to tell if a set amount is added or subtracted.) Well then, if it so happened that in this society half of the population had two teaspoons of sugar per cup of coffee, while the other half had only one, WMP would require us to redistribute 0.2 teaspoons of sugar from the first group to the second, provided this can be done without other costs. Finite sensibility has already led us beyond what most discussions of the Pareto principle assume, namely that Pareto does not allow us to subtract from any individual, or trade the gains of some against the losses of others. Of course, there is a sense in which we have not subtracted from anyone, or caused anyone to suffer a loss: we have not done anything contrary to anyone's preferences. On the other hand, we could, by interpolating between a person's explicit preferences in cases where he fails to discriminate, construct a graph representing his 'underlying' preferences. In terms of these underlying preferences, we would be trading the gains of some against the losses of others. Let us now focus on utility in a sense which includes these underlying preferences. If we use a particular number, say one, to represent the utility of a just unnoticeable difference for each individual, the acceptance of WMP will effectively lead us to the utilitarian principle of maximizing the sum of individual utilities. This result can be illustrated more vividly in the following diagrammatical argument. 234

8 An Argument for Utilitarianism According to WMP, as long as 50% of the population are made noticeably better off and the other 50% un noticeably worse off, the society should be regarded as being better off. Conversely, as long as 50% are made noticeably worse off and the other 50% un noticeably better off, the society is worse off. Thus, taking the limit, social welfare stays unchanged as 50% of the population are made just unnoticeably better off and the other 50% just unnoticeably worse off.4 Figure I. 4 Strictly speaking, we have to assume that social welfare is a continuous function of individual utilities. This is however a very reasonable assumption, stating merely that social welfare does not jump as the utility of an individual changes by an infinitesimal amount. 235

9 Yew-Kwang Ng and Peter Singer Now consider Figure 1 where all the n individuals are arranged along the horizontal axis and individual utilities are represented by the vertical axis. Starting from an initial situation x, let the curve x represent the utilities of the various individuals. For example, if each individual occupies a distance of half a centimeter (along the horizontal axis), the utility of the first individual in situation x is the shaded area. The curve x need not be continuous. This does not affect our argument. Consider another situation y which, in comparison to x, involves making the first -^-individuals just unnoticeably better off and the rest just unnoticeably worse off. From WMP and the argument of the preceding paragraph, social welfare remains unchanged. Now, starting from situation y (see the curve y in Figure 1), consider another situation z. In comparison to y, z involves making another 50% (the first individual, and the second half of all individuals except the last one) unnoticeably better off and the other 50% (the last individual and the first half of all individuals except the first) unnoticeably worse off. Hence, social welfare again remains unchanged. Since there is no change between x and y, or between y and z, social welfare must be the same at x as at z. Comparing x and z, we see that all have exactly the same levels of utilities except the first and the last individuals. The first individual's utility has increased by two units and the last decreased by two units. This demonstration does not depend on the choice of these two particular individuals, since WMP applies to any 50% of individuals. Thus, we may conclude that, irrespective of the initial situation, if we increase the utility of any one individual by two units and decrease that of any other by two units, holding those of all others unchanged, social welfare must remain unchanged. Effectively, this means that the utility of one individual can be traded off against that of another, and the best state for the community is that in which the total utility (the unweighted sum of the individual utilites) is as high as possible.5 Objections One likely reaction to this argument is that it involves a trick, an intellectual sleight of hand. We innocently accept WMP in the belief that it cannot lead us to any action which makes anyone noticeably worse 5 This conclusion is established formally in Yew-Kwang Ng, 'Bentham or Bergson? Finite Sensibility, Utility Functions and Social Welfare Functions', Review of Economic Studies (1 975)

10 An Argument for Utilitarianism off - only to find that, by repeated applications of the principle, we reach a situation in which people are noticeably worse off than they originally were. In other words, although WMP does not enable us to go directly from x to z, it does allow us to go from x to y and from y to z - and this, surely, is trickery. We readily concede that at first glance WMP does not seem to have the implications we have shown it to have; but we do not think that the derivation of these implications can be rejected on this ground. The argument withstands closer scrutiny. There is nothing up our sleeve. Nor is the conclusion strictly paradoxical. One might think it to be paradoxical if one misconstrues WMP as a necessary condition of a preferable state of affairs - that is, if one take WMP as stating that it is wrong to choose an alternative which makes some individuals noticeably worse off. If this were correct we would have a situation in which WMP directs us, in two steps, to a situation which it actually prohibits us reaching in a single step. That would be paradoxical. But WMP, like the Pareto principle, does not prohibit transfers which make some people noticeably worse off: it does not apply to such cases, and so says nothing explicitly about them. Nor does WMP in itself imply utilitariansim. It is only when WMP is coupled with a social ordering, including the transitivity of social preferences (so that if x is socially preferred to y, and y is socially preferred to z, x is socially preferred to z) that we can get to utilitarianism. Another possible reaction poses a more serious threat to our strategy. One could say: I accepted WMP because I thought it could not lead us to make people noticeably worse off; now I see that it can; therefore I withdraw my initial acceptance of WMP. If WMP is rejected, our argument cannot get started. Nor can we properly try to hold readers to their original acceptance of WMP, irrespective of its implications. What we can do is invite readers to look again at WMP and consider if they really can deny what it says. For WMP is, as we have said, a very modest value premise. Suppose that we have a community of 100 individuals, none of whom prefers y to x. WMP directs us to prefer x only when no fewer than 50 individuals do positively prefer it. (For Pareto, of course, a single positive preference would suffice, if all others are indifferent.) If we are not prepared to accept this, under what conditions would we be prepared to act in accordance with community preferences? Are we to require unanimity of positive preference? That would, in practice, be impossible to achieve and would effectively block virtually all schemes for improving social welfare. WMP substitutes majority rule for unanimity, while retaining the requirement that the minority be, at worst, not opposed to the 237

11 Yew-Kwang Ng and Peter Singer scheme. If we give up complete unanimity, WMP is a logical next stopping place. It might be said that the problem with WMP is that it is stated in terms of preferences, rather than underlying utility. It is this which allows us to use the fact of finite sensibility to derive utilitarian implications from it. So it might be suggested that we should reformulate WMP along the following lines: For a community of n individuals choosing between two possibilities, x and y, if no individual's utility is reduced by the choice of y over x, and the utility of at least - - individuals is increased by the choice of x over y, then x increases social welfare and is preferable. We could call this Weak Majority Utility, or WMU. Provided that 'utility' is understood as including underlying utility as well as noticeable differences of utility, WMU does not imply utilitariansim, as WMP does. However, WMU is not of much help in actual choices, where some individuals ususally must be made worse off if the majority are to be made better off. This is true even if we are prepared to go a little beyond WMU and accept the Pareto principle defined in terms of underlying utility instead of explicit preference. (In other words, - - individuals' in the specification of WMU is replaced by 'some individual'). If we insist on not trading-off the loss in utility of one individual against the gain of others, no decision can be made except in special cases where everyone is made better off in terms of his underlying utility, or at least no worse off. Due to finite sensibility, we may not know that this is the case even if no individual objects to the decision. This is a serious practical difficulty. There is a further argument against the 'no trade-off' principle. Let us classify all changes that affect utilities of individuals as (a) those due to our conscious actions, and (b) all others which we shall call natural forces. It is fair to assume that, in all probability, if we do not take any conscious action, the effects of natural forces will tend to reduce our wellbeing. This is so since we have already built a fairly advanced civilization and need all sorts of actions just to maintain it, e.g. to meet the energy shortage, etc. If we insist on the no trade-off principle, very little scope will be left for conscious actions. With natural changes occurring, it is most probable that, after a period of time, most of us will be made worse off. On the other hand, if we trade-off a small loss for a larger gain (in utility) irrespective of who is affected, then in the long term our conscious actions will serve to increase the utilities of most of us and may then balance or over-balance the unfavourable effects of natural forces. By adopting the utilitarian principle of trading-off gain 238

12 An Argument for Utilitarianism and loss, in the long-run, we may be able to make no one or very few people worse-off. This contrasts sharply with the adoption of the no trade-off principle which in the long term is likely to make most people worse off. It may be thought that, if we apply our value principle to the longrun outcome, then the above argument against the no trade-off principle does not hold. This is not so. Given the working of natural forces, it is quite impossible to ensure that no one be left worse off in the long term. It may appear that this would be possible if we compensate the losers. When we take into account the practical problems of doing so, however, especially the problem of false claims of damage, clearly it is quite impossible to ensure that no one is left worse off. The difficulty of applying the no trade-off principle to the long run is compounded by the fact that we are uncertain about the effects of both natural forces and our conscious actions. Pursued to its logical conclusion, the notrade-off principle will preclude any action, even one that has only a remote probability of making one person slightly worse off. Realizing the inadequacy of the no trade-off principle, aren't we compelled to go a little further? Should we not agree that noticeable differences are more important than unnoticeable ones, and that the fact that choosing x will make a noticeable difference to the utility of half the community should outweigh the fact that it will make the other half worse off by an amount that they cannot notice? These rhetorical questions cannot force the reader to accept WMP rather than WMU, although they might persuade some of the reasonableness of so doing. If our readers remain unconvinced and choose WMU but refuse to accept WMP we will have failed in our stated purpose of producing an argument for utilitarianism. We will, however, have achieved something else: an argument against principles, like WMP and the Pareto Criterion, stated in terms of preferences. Such principles may seem quite distinct from principles, like the principle of utility, which seek to maximise welfare even when this requires making some worse off in order to make others better off; when we take finite sensibility into account, however, this difference may disappear.6 December Though this paper was jointly written, credit (or blame?) for its central idea must gotong. -P.S. 239