V ARE THERE INCOMMENSURABLE VALUES? 1.On measuring well-being IN speaking of well-being, we all resort to quantitative language. We speak of "more' and "less'. At times we aim to "maximize' well-being. It is what all of us try to do in our own lives.1 And many of us, though by no means all, see it as one of the aims of morality. Yet how seriously does the quantitative talk have to be taken for this whole policy of maximizing to be feasible? What we want to know is not just, Is well-being measurable? There are many different scales of measurement, and it would be astonishing if well-being were not measurable on at least one of the less demanding ones. Nor do we want to know just, Is well-being measurable on a scale with origin and unit?, as if the crux were whether that sort of cardinal measurement were possible, in principle, in any circumstances. Perhaps, in special circumstances, when things fall out conveniently, it is. But that would not get us very far. Is it possible often enough in the sorts of messy everyday situations in which we want to maximize? But then that is not the crux either. Suppose it could not be. Do we really need a scale like that in order to maximize? To the extent that we can get by on pair-wise comparisons or rough cardinality, we do not. Our powers of measurement may be limited. But our demand on measure- ment may also be limited. In the end, what we need to know is, rather, Do our powers match our demands? There is one place they might not. If well-being were a simple state of mind that occurred in smoothly changing intensities, then we could at least hope to develop a powerful cardinal scale, and one that is without discontinuities or incommensurabilities. But well-being is nothing like that. It is the fulfilment of informed desire. So there may well turn out to 76 MEASUREMENT be cases in which, when informed, I want, say, a certain amount of one thing more than any amount of another, and not because the second thing cloys, and so adding to it merely produces diminishing marginal values. I may want it even though the second thing does not, with addition, lose its value; it may be that I think that no increase in that kind of value, even if constant and positive, can overtake a certain amount of this kind of value. We can make no assumptions about what the structure of informed desires will turn out to be like. There may well be incommensurable values in this fairly weak (or, worse, in some stronger) sense. But again, the question is not just the stark one, Are there incommensurable values? Incom- mensurabilities of the relatively weak form of "a certain amount of this is more valuable than any amount of that' are no threat to pair-wise comparisons, and it is clear that at least much maximizing deliberation uses only them. And if there are not many incommensurable values, or not many of a strong form, or not many at the centre of our maximizing deliberation, they may not be much of a worry. When we shift from prudential to moral deliberation, the demand on measurement goes up. Again, the question is not just the stark one, Are interpersonal comparisons of well-being possible? That is not a very live issue; some comparisons on some conceptions of well-being obviously are. The live issues are, rather, Are comparisons possible on the best conception for prudential or moral reasoning? And furthermore, Can we make enough of them, and will they be reliable enough, for the policy of maximizing to be a practical proposition? How difficult that last question is will depend upon what the best conception turns out to be. If it is the narrow conception of all-purpose means to a good life, our problem is, of course, much eased. But, unfortunately, morality has to use the broad conception out of prudential value theory, so our problem is greater. Still, in all of these cases, the important question comes down to, Will our measurement match our needs? That is the question for this Part of the book. I want to determine whether, and if so where, incommensurabilities appear and how they affect measurement; what sort of measurement is possible, and needed, in the one person case; and what 77 ARE THERE INCOMMENSURABLE VALUES? interpersonal comparisons are like and how feasible they are in the practical setting in which our deliberation has to take place. 2.Moral incommensurables and prudential incommensurables The word "value' covers a lot of ground. It is sometimes used just of prudential values---what makes a person's own life valuable to him, the "good' in contrast to the "right', as philosophers sometimes use the terms. But it is also quite properly used to include moral values---the "right' as well as the "good'---and indeed to include anything "normative' rather than merely "descriptive'. The word "incommensurable', as we have just seen, covers a fair amount of ground too. In a strong sense it can mean that two items cannot be compared quantitatively at all; the one is neither greater than, nor less than, and not equal to the other. Some writers regard this strong sense as the only proper sense;2 they use "incommensurable' to mean "cannot be fitted onto any scale of measurement', and one can see the point of the usage. But since there are many different scales, which can themselves be ranked from weak to strong,3 it is useful to consider whether values commensurable on a relatively un- demanding scale are incommensurable on a more demanding one. For one thing, although this sense can claim to be the strict sense of the term, we are interested in how incommensur- ability figures in the philosophical tradition, which has not always kept to the strict sense. Also, we have still to see how demanding a scale the policy of maximizing needs. So I shall use "incommensurable' in a way that allows different senses. In a weaker sense, then, "incommensurable' can mean, say, that no amount of one sort of item can equal, in respect of some quantity, a certain amount of another. And the term, as we shall shortly see, can mean several other things as well. Are there incommensurable values?4 Of course there are---if "value' is taken broadly enough and on some sense or other of "incommensurable'. For instance, the morally "right' trumps the prudentially "good'. At least, many moral theories, utilitarianism sometimes included, say so. If I could give my place in the lifeboat to two children whose prospects in life are 78 MEASUREMENT every bit as good as mine, then utilitarianism would say that I should. I have the strongest possible prudential reason against, but I have a moral reason for, and no amount of purely prudential weight, a utilitarian can say, outweighs any amount of moral weight. So when we ask whether there are incommen- surable values, we should keep in mind the domain of values and the kind of incommensurability that we are interested in. If the domain includes all practical reason, then it is highly likely that cross-category weightings (say, between moral and prudential reasons) will display, say, lexical ordering. If, however, the domain is confined to moral reasons, then we need to have settled what the structure of moral theory is before we can begin to decide about commensurabilities. It certainly looks, on the face of it, as if there are many, irreconcilably different, kinds of moral reasons: rights, promo- tion of well-being, obligations of special relations (such as promisor, parent, etc.), commitments to particular persons and institutions, etc.5 Two moral obligations can conflict, without, it seems, either's losing its standing as an obligation. In that event, it would help if we could find lexical orderings or relative moral weights. Perhaps we can even work out a scale of moral weight, formally analogous to the scale of prudential values.6 But perhaps not; perhaps we shall find some conflicts that none of these devices are available to resolve. What we cannot do, however, is to decide, independ- ently of a fairly well-developed substantive moral theory, what we shall indeed find. Otherwise, we should just be giving moral intuitions far more authority than they deserve. Are human rights "trumps'?7 It does not seem so; we think that a certain right can weigh heavily one time but not another, and that well-being can sometimes weigh more heavily on the same scale as rights. But we cannot really say, until we have a substantive theory of rights that tells us, for instance, what makes something a human right in the first place, and how the weight of rights is determined, and what governs trade-offs between one right and another and between rights and well-being.8 And anyone who finds it obvious, without the aid of a fairly full substantive theory, that rights, whatever else they are, exhibit this or that sort of incommensurability has missed the difficulties, both of subject-matter and of method, 79 ARE THERE INCOMMENSURABLE VALUES? in moral theory-building. The same stricture applies to other candidate moral reasons. Are all obligations of special relations moral obligations? It is, I believe, a conceptual truth that promises create obligations, but it is a matter for substantive moral theory to say which of the obligations generated by the internal workings of institutions such as promising, parent- hood, and property are moral and so even contend for a place in a moral conflict.9 And, yes, commitments to particular persons and institutions seem to set up moral claims on us. But even if they really do, we need to know at how deep a place in moral theory the claims appear.10 And perhaps some moral conflicts---perhaps some of what are spoken of as "tragic' moral dilemmas---arise from our having different moral points of view, different world views, none of which we are willing to drop, but not all of which we can reconcile within a single system of practical reason. Someone faced with the possibility of deliberately killing one innocent person to stop someone else from killing many more, might naturally, despairingly, think that there was simply no right thing to do.11 Here the two options may not only fail to fit on any single scale of moral weight but also, more radically, even fail to fit on a single scale of rationality or of theory-acceptability.12 This failure would raise questions not about the content of moral theory but about what a moral theory can, or should, be---what system or unity we can expect of it. I want for now to postpone the question, Are there incommensurable moral reasons? What I shall say later, in Part Three, will go a long way, as any substantive moral theory will, towards an answer. For now I want to confine myself to the domain of prudential values. Are there incommensurabilities there? Can the ends of life be got onto a single scale? That is a big enough question to get on with. 3.Forms of incommensurability: (a) Incomparability The very strongest sort of incommensurability arises if two values cannot even be ordered as regards value---that is, if, while A and B are both prudential values, A is neither more valuable than B, nor less, and not of equal value. 80 MEASUREMENT Are there values like that? In the moral domain, we might naturally first look for them among "tragic' cases.13 And, in the prudential domain, there are also some terrible choices. A talented but chronically depressed person may have to choose between suicide and accomplishment. And the two values, freedom from pain and accomplishment, are so far removed from one another that they are scarcely easy to compare. But, despite that, they come nowhere near strict incomparability. We can and do compare pain and accomplishment. If the pain is great enough and the accomplishment slight enough, we should not consider the accomplishment worth the pain. Of course, we might be pushed to know how to rank them if the pain got less or the accomplishment more. But true incomparability arises not when we cannot decide how to rank values but when we decide that they are unrankable. And this is not a case in which, since life forces choices upon us, we choose but do not prefer. On the contrary, what we think is, precisely, that it would not be worth it.14 We have to guard against exaggeration. We easily resort to hyperbole when we want to stress the difficulty of ranking certain values. For instance, we often despair of ranking artists of very diverse achievements: novelist A, say, is not better than, nor worse than, nor quite equal to, novelist B; they are good in such totally different ways that we cannot in any realistic way compare them. But though we find ourselves readily talking in these terms, they are overdrawn. A may be dry but full of insight, B unperceptive but hilarious. That they are good in such different ways of course makes it hard to compare them. But not impossible. If A were capable of only pretty minor insights, and B an absolute hoot, we could probably rank them. So, strictly speaking, their values to us---gaining insight, being amused---are not incomparable. Of course, when the values get close, we should probably have trouble with the ranking. We might wonder whether it is because it is hard to discriminate the differences in value that are really there or that there are no fine differences really there to discriminate. I think that it is not at all easy to determine which of those alternatives is true, but since the second one may well be, and is certainly the more awkward for prospects of measurement, let us simply assume that it is. Some values 81 ARE THERE INCOMMENSURABLE VALUES? are only roughly equal, and the roughness is not in our understanding but ineradicably in the values themselves. Still, rough equality is a long way from incomparability. Though rough equality does not create problems for commensurability, it does create problems, to which I shall return later, for finding any mathematical structure that will mirror the structure of prudential values. A feature of rough equality is that though novelist A is roughly equal to novelist B, C's being clearly better than B does not allow us to conclude that C is better than A; the most that we might be able to say is that C is roughly equal to A. And rough equality may not be a rare phenomenon; we may find, upon close examination, that a large number of rankings are of this kind. So rough equality raises the problem of what sort of ordering prudential values exhibit. Strong orderings, partial orderings, weak orderings are all well understood; but values seem to present a case of a vague ordering, which is a sort that is not understood.15 Rough equality, therefore, leaves us with important problems about the sense in which well-being can be measured at all, which I shall come to in the next chapter. Where else might we expect to encounter genuine incom- parabilities? There is no denying that there are very sharp differences in kind between prudential values. Are they not sometimes sharp enough to constitute incomparability? For instance, Kant says that in the kingdom of ends things have either "price' or "dignity'---"price' being value in exchange (anything with an equivalent or a substitute therefore having a price) and "dignity' being above all price.16 It is a good distinction. Some values do indeed seem optional. I might go for a life of competition and achievement, or I might instead go for one of gentleness and contentment with my lot. We see these as alternative ways of living, one value to some degree excluding the other but also making up for it. If I am wise, I shall decide between them by deciding which will make my life better. But some values do not have alternatives; we do not think that we are similarly free to aim at them or not, as we choose. Some values constitute what we think of as the dignity of human existence, central among them being autonomy and liberty. If we surrender our human standing, we shall find no equivalent.17 Still, this point, though true, is not strong 82 MEASUREMENT enough. Non-equivalence is not incomparability. Nothing can take the place of human standing, but that is not to say that nothing can even be ranked with it as to value. In fact, in rare cases, other values can even outrank it. If for some unfortunate person autonomy brings extreme anxiety, it might be better to sacrifice some autonomy. The same is true of other examples. An individual human life has no equivalent. But that is not to say that nothing can be ranked with, let alone outrank, a human life. The French government knows that each year several drivers lose their lives because of the beautiful roadside avenues of trees, yet they do not cut them down. Even aesthetic pleasure is (rightly) allowed to outrank a certain number of human lives. It is easy here to move imperceptibly from Kant to cant. We want to express our sense of the enormous value of certain things---human standing, an indi- vidual human life---and once again we reach for strong expressions. But in real life, in concrete situations, we often do, and should, back down. We find that we did not mean literally what, in searching for strong expressions, we ended up saying.18 Perhaps, then, we have to look for genuine incomparabili- ties in more extreme clashes---the clash of different world views, of radically different metaphysics. In Part One of this book, I chose as examples of the ends of life such things as enjoyment, accomplishment, deep personal relationships, etc. They are all rather tamely secular. Someone else might see the ends of life as being self-denial, contemplation, solitude, etc.19 But when metaphysical views clash, we must try to choose between them. We might, of course, run up against a radical sort of incommensurability if these competing views could not themselves be ranked as to rationality, or even as to plausibil- ity, but that, besides being doubtful, is in danger of yielding altogether too strong a form of incomparability. We are interested in whether some of the values we use are incompar- able, and if two sets of values are themselves incomparable as to plausibility and we must just opt between them, then, though the sets are incomparable, we use only one. And if we do in fact wish to use both sets of values, as doubtless sometimes we do, then we shall be forced to organize them. But this involves ranking. For instance, self-denial, contemplation, and solitude 83 ARE THERE INCOMMENSURABLE VALUES? come out of a religious world-view, and they are not in the ordinary sense prudential values. They may be matters of vocation, and a calling from God introduces a change in category and so the live possibility of cross-category incommen- surabilities of the sort that we saw earlier. World views may clash in a way that allows us to choose only one. Or they may clash in a way that allows us to choose both, but then they ought themselves to show something about the the relative status of the values they propose. They should show us how to rank all of them. Some values may indeed outrank others. But this is not incomparability any longer, but some weaker form of incommensurability. Incomparability, I think, is too strong. We do not find values that are strictly incomparable. So we should look at weaker forms of incommensurability. 4.(b) Trumping The next strongest form of incommensurability allows compar- ability, but with one value outranking the others as strongly as possible. It takes the form: any amount of A, no matter how small, is more valuable than any amount of B, no matter how large. In short A trumps B; A is lexically prior to B.20 We just saw that the elements of what Kant calls "dignity' were not, strictly speaking, incomparable with the elements of "price'. Do they at least trump them? Even that, though, would be far too strong. How do we rank, say autonomy or liberty on the one hand, and prosperity or freedom from pain on the other? Nearly all of us would sacrifice some liberty to avert a catastrophe, or surrender some autonomy to avoid great pain. So people who would call certain values "trumps' or give them "lexical priority' probably do not mean these terms entirely seriously.21 What they may have in mind is some weaker form of incommensurability. 5.(c) Weighting It seems that we often find ourselves attaching greater weight, unit for unit, to one value than to another. For instance, we think that it takes a great deal of prosperity to outrank a 84 MEASUREMENT pretty slight liberty or that it takes a lot of pleasure or happiness to outrank a fairly small amount of pain or misery. These beliefs are entirely plausible. The second, for instance, is what gives "negative utilitarianism' much of its appeal.22 But the interpretation on which they are plausible is not the interpretation that yields any kind of incommensur- ability. What is plausible, for instance, is that it takes a relatively large amount of happiness, as happiness goes (that is, relative to other cases of happiness), to justify a relatively small amount of misery, as misery goes (that is, relative to other cases of misery). Thus, one is appealing, in a rough way, to two distinct scales. But neither scale is the ultimately important one, namely the single scale on which prudential values are ranked. And once we get on that scale, we shall find that the special weighting disappears. Since it is a scale of prudential value, a fairly small amount of misery will turn out to make life worse to a greater degree than a fairly large amount of happiness makes it better. Admittedly, there is an interpretation that would pre- serve special weighting and so would yield a kind of incom- mensurability. To state it, however, we should need ser- iously to appeal to two scales. We should have to be able to say that n amount of misery is not equal in prudential weight to n amount of happiness, that only n + m amount of happiness (where m > 0) has the same prudential weight as n amount of misery, and that only n + m + 1 amount of happiness justifies following a course of life involving n amount of misery. But is all that coherent? There is no scale on which we can fit all the various kinds of "happiness' and "misery', except the scale of prudential weight. Any scale that would have the scope necessary to let us talk about m or n amounts of "happiness' and "misery' would have to be built on a rich foundation of basic preference judgments---that is, judgments about relative prudential weight that themselves do not rest upon any other kinds of judgments about the amount of some quantity in the two options being ranked. This interpretation needs two scales, but there is only one coherent scale forthcoming, a scale of weight in choice. So this interpretation cannot even be stated.23 85 ARE THERE INCOMMENSURABLE VALUES? What I have called, "weighting', therefore, is not a promising source of incommensurability, so we have to look elsewhere. 6.(d) Discontinuity Incomparability is too strong; values can at least be ranked. Trumping is too strong; no value, no matter how little is at stake, outranks another, no matter how much is at stake. Weighting is a distraction. But that still leaves it possible that, so long as we have enough of B any amount of A outranks any further amount of B; or that enough of A outranks any amount of B. Both of these forms bring with them the suspension of addition; in both we have a positive value that, no matter how often a certain amount is added to itself, cannot become greater than another positive value, and cannot, not because with piling up we get diminishing value or even disvalue (though there are such cases), but because they are the sort of value that, even remaining constant, cannot add up to some other value. And here at last, it seems to me, we really do start encountering incommensurability. Not, however, in the case of the two values, liberty and prosperity. When we see that liberty is not, in the strict sense, lexically prior to prosperity, it is natural to retreat to the view that, if prosperity is assured to some minimal level, then the priority holds. But that is very doubtful. Liberty, I believe and shall later argue,24 may, from time to time, be at stake in only a small way. And on those occasions other values, such as the general prosperity of our lives, might well be at stake in a big way. So it might well then be better for me, all values properly considered, to surrender some small liberty to gain the conditions for bringing off some great achievement or to preserve deep, rewarding personal relations (for instance, staying in a country without certain political liberties because someone I love lives there). Liberty, as important as it is, is not such a heavyweight among prudential values that, even in the limited domain we have defined, it is bound to trump any other value capable of achieving fairly heavyweight status itself.25 The mistake here seems to be to think that certain values---liberty, for instance---as types outrank other values 86 MEASUREMENT ---prosperity, for instance---as types. Since values, as types, can vary greatly in weight from token to token, it would be surprising to find this kind of discontinuity at the type---or at least at a fairly abstract type---level. It is much more promising to look at less abstract types, or at tokens. It is far more plausible that an important sort of liberty---for instance, the right to live out the very central parts of one's life plan---can trump any amount of a trivial kind of disvalue---say, one's neighbours' getting huffy.26 Or, to take the purely prudential sort of case that we are concerned with now, it is more plausible that, say, fifty years of life at a very high level of well-being---say, the level which makes possible satisfying personal relations, some understand- ing of what makes life worth while, appreciation of great beauty, the chance to accomplish something with one's life---outranks any number of years at the level just barely worth living---say, the level at which none of the former values are possible and one is left with just enough surplus of simple pleasure over pain to go on with it. Could this be right? Why, if prudential values do not show discontinuity of the first form I have just mentioned (with enough of B, any amount of A outranks any further amount of B) would they start showing them here (enough of A outranks any amount of B)? I have already tried to give part of the answer: before we had quite abstract types of values, each of which could vary greatly in amount; now we have much more specific values tied as to amount at the opposite extremes of the scale. But this is only part of the answer. We should not expect to find sharp discontinuities at either extreme of the scale. Fifty years at a very high level---say, with enjoyment of a few of the very best Rembrandts, Vermeers, and de Hoochs---might be outranked by fifty-five years at a slightly lower level---no Rembrandts, Vermeers, or de Hoochs, but the rest of the Dutch School. And then fifty-five years at that level might be outranked by sixty years at a slightly lower level---no Dutch School but a lot more of the nineteenth-century revival of the Dutch School. And so on, step by step, until, it seems, we must eventually reach the point where the original fifty years would be outranked by a sufficiently large number of years of life just barely worth living. 87 ARE THERE INCOMMENSURABLE VALUES? This step-by-step approach seems irresistible. Yet it presents us with two embarrassments: a Sorites Paradox and a slippery slope. If we take enough pebbles from a heap, it ceases being a heap. But then, since one pebble more or less could never make the difference between its being a heap or not, if we remove them singly it can never cease being a heap. Similarly, one might argue, with "appreciation of beauty' or "deep loving relationships'; with slow, step-by-step changes we can never lose the appreciation of beauty or deep loving relation- ships either. But we obviously can---I want to put the Sorites problem to the side. We can indeed reach a point, with the series of subtractions that we just saw (first Rembrandt, Vermeer, and de Hooch, then the whole seventeenth-century Dutch School ...), at which we lost the appreciation of beauty. We may instead have the kicks of kitsch but they are different. Even with the loss of one important value, such as appreciation of beauty, it must be possible for enough more life with the substantial values that remain---deep loving relations, accom- plishment, and the rest---still to be more valuable. That seems right, but then, by parity of reasoning, it would seem that, since kicks of kitsch are not nothing, enough of them must be more valuable than a certain amount of the appreciation of beauty, so we are on our way down the slippery slope. If we also lose deep relations, accomplishment, and all the rest of the substantial values of life, then, by parity of reasoning, since what is left is, though slight, not nothing, enough life with this residue must be more valuable. But there is one impediment to this slide down the slope. We do seem, when informed, to rank a certain amount of life at a very high level above any amount of life at a very low level. And if we do, it is likely to be a basic preference; that is, it could not be based upon other judgments about the amount of value in the objects being ranked. Nor need there be any extreme perfectionism at work in this preference; those who held this preference could also be willing to sacrifice the best Rembrandt, Vermeer, or de Hooch, for all the various, though lesser, achievements of the Dutch School. To call this preference "basic' is simply to say that it is basic among quantitative judgments. It is not to say that it is 88 MEASUREMENT ground-floor in every way. We can offer arguments---the sort of deliberation that I talked about at some length in Part One---for seeing the objects as valuable, which may also lead to seeing them as of varying value. Basic preferences can certainly be argued over. And we can dispel other doubts. The doubts about this kind of basic preference, I think, often come down in the end to an assumption of the totting-up conception of measuring well-being. They all focus on a series of small changes, as if the relation of one level of well-being to another had to be a matter of additions and subtractions of a small unit. But the assumption that there is a single additive measure of well-being continuously across its range is just what is in question. That additive scales of well-being are possible here and there in its range is a long way from one additive scale's being possible across its whole range. The assumption of changes in small units is plausible enough in some cases. When we move from portraits by Rembrandt to portraits by Hals, we lose something. We should be likely to attach less value to appreciating the beauty of one portrait by Hals than one portrait by Rembrandt. But if we have all of the rest of the Dutch school except Rembrandt then we might well attach more value to the number and variety of paintings that we should then have. This looks like a loss' being made up by the sum of a larger number of smaller values. However, when we get to the point where beauty totally disappears and only kitsch is left, it looks as if no totting-up could be done and only a basic preference is possible. We have simply to rank a life with beauty against a life with only lots of kitsch. And then, from this perspective, when we look back at the earlier cases where totting-up seemed to be going on, we can see that it was not what was going on after all. What goes on in comparing a few Rembrandts with all the rest of the Dutch School is not arithmetical addition of a larger num- ber of slightly smaller values to get a greater overall sum. We have to decide how we value greater number and more variety against a few supreme, less varied examples. And that is itself a basic preference. Any additive scale that we might introduce in this domain will have to be defined on basic preferences such as this. Therefore, one cannot har- bour doubts about the appearance of discontinuities in the 89 ARE THERE INCOMMENSURABLE VALUES? addition of well-being by assuming the continuity of the addition.27 So there are, I wish to propose, incommensurabilities of the form: enough of A outranks any amount of B. That leaves us with the job of looking closely at each judgment of this form to separate the ones that are literally true from the exaggera- tions,28 but some, I think, prove literally true. 7.(e) Pluralism Pluralism says that there are irreducibly many prudential values. That seems to me right.29 But how is pluralism connected with incommensurability? Is the form of reduction- ism that pluralism rightly denies a form that commensur- ability needs? If the denial of reduction is just the denial that there is a single mental state running through all the things that we rank in terms of which we rank them (the denial of a crude mental state account, or of hedonism), then, it can be agreed by everyone, utilitarians included. In one sense, different kinds of pleasure, or pleasure on the one hand and pain on the other, are incomparable: namely, there is no deeper unitary mental state in terms of which they can be compared.30 But the failure of crude mental state accounts, or crude hedonisms, does not mean that there is no single scale for ranking prudential values.31 If, on the other hand, the denial of reduction is the denial that there is a single substantive value running through all the things that we rank (the denial of substantive monism), then it seems to me entirely correct. There is no single substantive super-value.32 But we do not need a super-value to have a scale.33 It is enough to have the quantitative attribute "value'. If the denial of reduction is the denial of a single scale of origin and units continuous across the domain of prudential values (the denial of complete, strong cardinality), then again it seems to me right. Even if we abandon appeal to a super-value and use the concept of "value' itself as our quantitative attribute, we do not get such a scale. But that does not show much. It does not show, for instance, that there is no single scale of well-being, nor that not everything can be 90 MEASUREMENT ranked on it. And it certainly does not show that our powers of measurement are not up to the policy of maximizing. So we might take the denial of reduction to be the denial of complete ranking. We might think that though some values are quantifiable others are not, or that though all are they are quantifiable only on different scales. The absence of a complete ranking would indeed upset the policy of maximizing (though upset and incompleteness come in degrees). Still, it is that strong form of incommensurability---namely, incomparability ---that we found difficult to produce convincing examples of. The lure of the apparent examples is to be traced, I think, to a mistake about what the scale of well-being is. We can sometimes rank pleasures as to intensity, duration, and number. We can also measure degrees of one component of happiness (say, contentment with one's lot) and separately measure degrees of another, radically different component (say, achievement of one's ambitions). We seem therefore to have several partial scales---scales for certain dimensions of pleasure or happiness---but no super-scale. But it does not follow from there being no super-value that there is no super-scale. To think so would be to misunderstand how the notion of "quantity' of well-being enters. It enters through ranking; quantitative differences are defined on qualitative ones. The quantity we are talking about is "prudential value', defined on informed rankings. All that we need for the all-encompassing scale is the possibility of ranking items on the basis of their nature. And we can, in fact, rank them in that way. We can work out trade-offs between different dimensions of pleasure or happiness. And when we do, we rank in a strong sense: not just choose one rather than the other, but regard it as worth more. That is the ultimate scale here: worth to one's life.34 Commensurability does not require monism. At least on what seems to me the most common interpretation of the terms, the monism-pluralism issue is not especially central to the issue of incommensurability.35 The quantity used to define the scale of measurement of prudential value is not itself a substantive prudential value. The real threat to prudential values is the threat that their diversity, their incompatibility, and, above all, the great value that each at times constitutes 91 ARE THERE INCOMMENSURABLE VALUES? will be ignored. But that threat comes not from believing in commensurability, but from being arrogant and intolerant. It can come as easily from the pluralist who believes in incommensurability---from someone, say, who fervently be- lieves that liberty is "incommensurably higher' than prosperity, or even than life itself. If we assembled all the deplorable fanatics that history has ever seen and asked them to divide into two lobbies labelled "monist' and "pluralist', my money would be on the pluralists' winning hands down. It is hard to think that a happy, productive life counts for nothing but unfortunately it seems terribly easy to think that it counts for nothing up against what is seen as the "incommensurably higher'.36 What needs defending is not the mere plurality of values but a certain important picture of how they are related---that they clash, that they all matter, that they all have their day, that there are no permanent orderings or rankings among them, that life depressingly often ties gain in one value to terrible loss in another, that persons may go in very different directions and still lead equally valuable lives---call this picture "liberalism'. But one does not defend liberalism by attacking commensurability or computation. One does not defend it by denying the possibility of trade-offs between values, even of trade-offs between liberty itself and other values, or by denying that values can be ranked on a single scale. Trade-offs are possible; values can be ranked on a single scale. So if we want a defence of liberalism, it will have to take a different form. What is needed is to show just how valuable our being able autonomously to form and freely to carry out a life plan is, how central it is to our living a human existence, and just how greatly we value our humanity, even at the cost of our happiness. All of that has to be put as emphatically as we can. But it is easy to slide from emphasis to hyperbole. We reach for the powerful language of incomparability and trumping. But it does liberalism no service, it offers it in the end no real defence, to tie it to the indefensible. If liberty is to be defended, it has to be taken down from the shaky perch of "incomparability' and placed in the same domain as other prudential values, where they will conflict, where none is above the fray, where some, however, are weighty enough 92 MEASUREMENT usually to win a conflict, and where even, occasionally, a weighty value can assume in a particular case such weight that no amount of some lightweight value can overcome it. It is in those terms that liberalism has to be---and can be---defended. VI THE CASE OF ONE PERSON 1.Is well-being the sort of thing that can be measured at all? THIS question, to which confident but contradictory answers are regularly given,1 is itself far from clear. There are several forms of measurement. Are we asking, Is well-being measur- able on any scale, no matter how weak? We tend to treat one scale, usually a fairly strong one such as a scale for measuring length, as the standard that we expect all measurement to live up to, which is a mistake. As a corrective, we might keep in mind S. S. Steven's well-known definition, "Measurement [is] the assignment of numerals to objects or events according to rule---any rule'.2 On that deliberately capacious definition, the following all count as scales of measurement. There are ordinal scales, where the rule assigns numbers in order to the objects or events we observe occurring in an order. For example, we can rank minerals by a scratch test for hardness and assign them numbers, either ascending or descending, in order of their hardness. Ordinal scales are unique up to a monotone transformation. Then there are more powerful interval scales, where the rule requires, more strictly, assigning numbers so that equal differences between them correspond to equal intervals in the attribute being measured. Here a unit of measurement and an origin appear, both arbitrary. Fahren- heit and Centigrade scales of temperature, for example, are interval scales. So interval scales are unique up to a linear transformation of the form y = ax + b. Then there are the still stricter ratio scales, in which the rule makes the further demand that the ratios of the numbers (doubles, trebles, etc.) be the same as the ratios of the amounts of the attribute to which they are assigned. Here the unit is still arbitrary but the origin no longer is. After the discovery of an absolute zero, temperature could be measured on a ratio scale, which, unlike the Fahrenheit and Centigrade scales, allows us to say that 94 94 MEASUREMENT 20 is twice as hot as 10. Therefore ratio scales are unique up to a similarity transformation of the form y = ax. I shall stop there, though that list is just a start. Since a rule of assignment defines a scale, and other rules are possible, other scales are possible. There is, in principle, a non-denumerable infinity of types of scale,3 although in practice only a few are of empirical interest. None the less, more than I have listed have in fact proved of interest. For instance, the area between ordinal and interval is a rich one for the development of new scales, especially important in measuring psychological attributes.4 Another defect of the question with which I started is that it is not clear whether it is asking a theoretical or a practical question. The theoretical question is, What quantitative statements about well-being can we give sense to? Can we, for instance, give sense to ordinal talk about "more' and "less' and "maximum' well-being? Can we go further and give sense to such cardinal talk as "the gap between these values is greater than the gap between those' (interval scale) or even that "this is twice as valuable as that' (a ratio scale)? But we might satisfactorily answer the theoretical question and it still be unclear whether, in the messy everyday situations in which our deliberation actually takes place, we can know enough for measurement, at least for reliable measurement, to succeed. It is not much help to practical deliberation if while certain forms of measurement of well-being can be given sense no form can really figure in our actual thought. Questions about measurability are often ambiguous as between the theoretical question and the practical question, and both are important. And is the question asked of the one-person case, the many-person case, or both? It is easy to exaggerate or misconceive the difference between one-person and many- person cases. We are inclined to think that all one-person cases are relatively uncomplicated because we always have a single preference order to fall back on, while in many-person cases we clearly do not. But often we do not in one-person cases either. If I am faced with two very different courses of life, which will eventually bring changes to my tastes, attitudes, and concerns so that, whichever I choose, I shall end up happy that I chose as I did and not the other way, there is no 95 THE CASE OF ONE PERSON single preference to fall back on either. Intertemporal intra- personal comparison can get very close to interpersonal. What I want to do, to isolate issues, is to concentrate in this chapter on the simplest case---one person with stable preferences---and take up the more complicated cases in the following chapter. 2.An ordinal scale of well-being Let us start with the theoretical question. In general, we can prove measurability by showing that a system of empirical relations and operations is isomorphic with a certain system of numerical operations and relations, and by showing what in a numerical system is to be taken as having meaning when applied to the empirical system. This can be thought of as our having to prove two theorems: a representation theorem (that, by being isomorphic, certain aspects of the arithmetic of numbers represent certain aspects of the world) and a uniqueness theorem (that the scale of measurement that results has a certain sort of uniqueness; in short, that we can determine the sort of scale it is---ordinal or interval or ratio or ...).5 It is almost universally accepted by economists that in simple one-person cases well-being is measurable ordinally. Unfortunately, incommensurabilities create problems even for that relatively modest claim. "Well-being' is certainly a quantitative attribute, in the sense that we can sometimes say that one thing makes us better off, or at least as well off, or exactly as well off, as another.6 Let us concentrate on the relation "at least as well off', in terms of which the others can be defined, and symbolise it "w'. Now we can show that well-being is measurable ordinally if we can show that "w' has the properties that constitute one or other logical ordering ---for example, those of a weak ordering, namely: reflexivity: (@Ax) xRx transitivity: (@Ax) (@Ay) (@Az) (xRy & yRz ~ xRz) and completeness: (@Ax) (@Ay) (xRy |v yRx |v xy) (To put this into English, a relation is reflexive if each individual in a certain domain stands in that relation to itself; it is transitive if, for any trio of individuals, the fact that the first stands in the relation to the second and the second to the 96 MEASUREMENT third implies that the first does to the third; it is complete if, for any pair, one stands in the relation to the other or the two are equal.) Does well-being behave like this? Over much of its range it does. But there are places where it seems distinctly not to. The trouble, therefore, comes with the requirement of completeness. As we have seen,7 sometimes the best we can say is that one thing roughly equals another, where the roughness may not be in our understanding of what is before us but in the objects of understanding themselves. To say that AwB is to say that either AwB or AwB is true. But rough equality does not behave like strict equality. If A roughly equals B, then we can add something to A (say, by improving it in some clear but slight way), getting A+, but it may still not be the case that AwB nor of course that A+wB or that A+wB. A+ may still be only roughly equal to B. We could then retreat and ask whether "w' has the properties of a partial ordering, namely those of a weak ordering without completeness. But this raises the question of how extensive the phenomenon of rough equality is. If, as is likely, we find it here and there right across the domain of well-being, then the retreat to a partial ordering is not much help. The trouble with rough equality is that it makes the strict ranking statements that it infects neither definitely true nor definitely false. For instance, one might be able to decide that a life of great accomplishment even at the cost of day-to-day pleasure is better than a life of slight day-to-day pleasures. And one might be able to decide that a life rich in day-to-day pleasures is better than a life of only slight accomplishment. But when the balance between these two values gets more level, we enter the area of indistinctness. At a certain point in this borderland none of the claims---not AwB, AwB, nor BwA---is either definitely true or definitely false. We seem, therefore, to have not so much a partial ordering as something new and little understood, namely a vague ordering.8 What effect does this have on the measurability of well-being? There is no obvious mathematical model that mirrors the behaviour of "w' given its vagueness, so we may wonder whether we shall be able to prove a representation theorem. But I do not think that the vagueness has such 97 THE CASE OF ONE PERSON devastating results. Rough equality (along with that other form of incommensurability, discontinuity) does certainly create problems for a cardinal scale with units that can be added across the domain of values. If A, A+, and A++ can all be roughly equal to B, small variations in value will get lost if we simply treat A and B as equal. But this will not matter if we can keep to ordinal judgements. And with ordinal judge- ments, it does not matter either that we have only a vague ordering. It is uncommon for rough equality to matter to prudential deliberation. Where we have rough equality, we treat the items, when it comes to choice, simply as equals. We are indifferent between them, even though our indifference in this case has an uneasiness to it absent in cases of strict equality. There can be, it is true, quite complex cases of rough equality, though they are fairly rare. For instance, we might have to rank not A and B, which are roughly equal to one another, but A++ and B{-. But so long as we can rank A++ and B{- against one another, even if only by "roughly equal to', we can proceed. It is a mistake to conclude from the fact that rough equality crops up fairly often that the difference between strict and rough equality matters fairly often. But unfortunately we cannot just conclude that for purposes of deliberation we may treat "w' as a weak ordering. The indistinctness that gives rise to "rough equality' also undermines the transitivity of "w' (which means that we cannot treat "w' for purposes of deliberation as a partial ordering either). "w' is transitive only if "w' is transitive, and "rough equality' is only non-transitive. That is, if A is roughly equal to B and B to C and C to D, we might sometimes decide, in a pair-wise comparison, that A is roughly equal to D but also sometimes decide that it is not. Though we are assuming that the roughness of "rough equality' is in the world rather than in our understanding of it, "rough equality' will, as regards transitivity, behave like such epistemological relations as "is indistinguishable from'. However, the break- down in the transitivity of "w' would probably be very rare---far rarer, for instance, than the appearance of rough equality itself. And the transitivity of "w' and of "w' would remain undisturbed. 98 MEASUREMENT So "w' gives us what might be called a partially transitive weak ordering. The vagueness that sometimes disturbs the transitivity and completeness of the relation "w' is important and little understood, and I have only scratched the surface of the problems it creates. But it need not change our minds on the theoretical question now before us: well-being is ordinally measurable. The attribute well-being is itself quantitative in a sufficiently strong sense for the ordering it generates to constitute an ordinal scale. There is one implication of the discussion of this section that is worth a comment. What is not at issue over the measurability of well-being is whether there is a single substantive super-value to serve as the common-denominator of all the other values. The dispute between monists and pluralists is irrelevant here. The measurement of values needs, not a super-value (fortunately, since there is none) but an attribute that is quantitative. And well-being and prudential value are themselves quantitative attributes. 3.Pockets of cardinality Let us stay with the theoretical question for a little longer. We want to know in how strong a sense well-being is measurable, and we may be able to give sense to a much stronger scale of measurement for well-being than we find ourselves often able to use in practice. Surely, well-being is sometimes measurable on a cardinal scale. But in now arguing that, we have to keep a firm grip on the quantity that we want to measure. It is very easy, in the course of arguing for the measurability of some quantity, to twist the quantity to fit the needs of the argument, so ending up with an impeccable proof of measurability but not of the quantity that we started with.9 Our quantity is the strength of informed desire. "Strength' and "informed' have to be understood in the way that I explained earlier;10 unfortunately neither is simply observable in action. The notion "informed' introduces an ideal element: what one desires under certain conditions, which may or may not obtain. Let us, therefore, to improve conditions for measurement, restrict our attention to the not 99 THE CASE OF ONE PERSON uncommon cases of desires which are in fact informed. And, as we have seen, the sense of "strength' that is relevant here is not a felt intensity or a motivational force; it is, rather, a place in an informed preference order. So we cannot conclude from the fact that a reformed drunk's will cracks and he takes a drink, that his desire to take a drink was "strongest' in the sense that we are out to measure. Nor can we conclude from the fact that a dedicated teetotaller never in his life takes a drink that he has never had a desire to do so. The link between desire and action is obviously nothing so simple. We have many desires, not all compatible; we have beliefs, which are themselves linked to desires, that certain desires ought not to be satisfied; we also have beliefs about which desires are within our capacity to fulfil, and these beliefs clearly affect action. But there is at least this link between desire and action: if a person believes that he can fulfil a certain desire, and he has no desire incompatible with it, then if he has the desire, he will act to fulfil it. This link is in part empirical and part what Wittgenstein calls criterial. We should have no understanding of what desire is unless desires had some connection with action. Our learning how to use the word "want' involves learning what makes people act, so that if, when there are no obstacles to a person's doing what he sincerely says he wants to do, no competing desires or inhibiting beliefs, he does not do it, then we should (and he would too) give up the belief that he does actually want to do it. And the whole set of hypotheses that we operate with about the connection between beliefs, desires, and actions, has a large empirical element that must turn out to be descriptively adequate. So let us now further restrict our attention to the simple, but by no means out-of-the-way, cases of desires that have no competing desires or inhibiting beliefs. An informed desire will now issue in action, and desires of different strengths will, in certain settings, issue in different action. Suppose, for example, that the County Council decides to change the layout of a road I use, threatening various things along the way that I find very beautiful. I decide that I am willing to give up five hours of my leisure time campaigning to save the chestnut tree, ten hours to save the stone wall, and twenty hours to save the copse. Of course, the marginal value to me of an hour of 100 MEASUREMENT leisure can change as the amount that I have changes, but let us restrict attention to a range within which the marginal value does not change---say that I should not have to give up more than two hours of leisure in a week so that any one hour is the same sacrifice as any other. Then, too, the marginal value of beauty changes as the amount there is changes, but again let us just restrict attention to a range in which it has not yet begun to change---say that there is so little beauty along the road that the value of each of these beautiful things is the same whether or not the others exist. With all of these restrictions in place, our understanding of the relation of desire and action would lead us to treat these figures as representing how much I value each of the things at risk. We could say that I value the copse about twice as much as the stone wall, and the stone wall twice as much as the chestnut. And we could say that the combined value to me of the wall and the chestnut is still well short of the value of the copse. We might hesitate attaching much significance to the precise figures that I announce (but then we might also with my estimates of the length of the stone wall and the height of the chestnut).11 Still, if we are satisfied that I have estimated carefully, we should be willing to treat them as accurate to within a tolerably small margin of error. Have we got the makings of a cardinal scale here? The ingredients we now have are part of an economist's stock- in-trade. We have concentrated on the marginal rate of substitution between two goods. We can define a unit of well-being on a unit of leisure time. With it, we can measure the size of the gap between things we value. So we have at least an interval scale. Whether or not we have a strong form of ratio scale depends upon the nature of the zero point we use to construct the scale.12 However, I want to leave aside difficult conceptual issues about "zero' well-being; both scales are cardinal and allow addition in a fairly powerful sense. The important question is not which of these two cardinal scales we have got here, but what the scope of the scale that we have got is. An economist would stress---what is right---that this sort of approach will produce, at best, only cardinal measures of local significance, the locality of which (defined, for instance, in the case I have discussed by the range in which the marginal 101 THE CASE OF ONE PERSON value of time and beauty is the same) may turn out to be very small. But one of the things that we want to know is to what statements about the measurability of well-being we can give sense. The existence even of pockets of cardinality shows that at least sometimes we can make sense of well-being as a cardinal notion. These pockets of cardinality also show that, contrary to the claims of some advocates of a probabilistic notion of utility, we can make sense of well-being, and of its cardinal measurement, independently of probability. There are economists and deci- sion theorists who take the tough line that it is only by defining utility in connection with probability that one gets a notion of utility worth bothering with. It is true that if utility numbers do nothing more than indicate place in an ordinal ranking they are not doing much. I prefer saving the copse to saving the stone wall, and saving the stone wall to saving the chestnut tree. If this ordinal information is all that we have, we can still attach numbers to each of the options: say, 1 to the least preferred, 2 to the next, 3 to the most preferred. We can even use these numbers to describe my behaviour; we can say that I behave so as to maximize the index number. But all that this talk in terms of utility numbers means is that I prefer saving the copse to saving the wall, and saving the wall to saving the chestnut. The numbers are unique only up to an increasing monotonic transformation; any three numbers in ascending order would do, say, {-14, 206{.3, and 3,000 (even three numbers in descending order would do if we were willing to give up the anyway not very informative talk about maximizing the index number). It is also true that, once probabilities are added to the picture in the following way, we need more than ordinal rankings to explain what is going on. I am sure, with some lobbying, to save the wall; or, by directing my lobbying efforts elsewhere, I have a two-thirds chance of saving the copse and one-third chance of saving the chestnut. An adequate explanation in this case requires giving weight not only to what probability I attach to the alternatives but also to how much I want each. As for the latter, it is not enough to know my ordinal ranking, because whether I should prefer the certainty of the wall to the gamble between the copse and the chestnut turns on where the middle item in the ranking comes 102 MEASUREMENT in relation to the two extremes. We could then call how much I desire something its "subjective value' or "utility', and utility numbers now would have some explanatory role. So long as a person's choices satisfy certain axioms, say the von Neumann-- Morgenstern axioms,13 his behaviour can be seen as an attempt to maximize the mathematical expectation of these numbers. And these numbers will be unique up to a positive linear transformation; that is, they will constitute an interval scale. But this is not really enough to justify the very tough line that the only interesting notion of utility must be defined in terms of probabilities, or that it is even "mystical' to speak of utility apart from probability.14 That tough line rests on the belief that the only way to cardinalize utility is in connection with probabilities. But it is not. And it would be surprising if it were. "Utility' figures in a very complex theoretical setting; it connects with more than just probability, and these other connections, as we have seen, are rich enough to allow other forms of cardinalization. There is no denying that the pocket of cardinality in my example is not large. We have no theory that yields correlations of belief, desire, and action that permit scales as broad in application as do the theoretical correlations of temperature, pressure, and volume. And it is not just a matter of hoping eventually to find richer correlations between belief, desire, and action; once outside a fairly narrow domain, incommensurabilities start arising. But, despite all this, the correlations of belief, desire, and action that we are able to make are rich enough in themselves, without probability, to allow the introduction of a cardinal notion of well-being. None of this is to deny that one of the advantages of taking well-being (even defined non-probabilistically) in conjunction with probability is that it allows us to expand the pockets of cardinality.15 But that brings us to the practical question: how are we going to measure well-being in the messy everyday situations in which we have to apply the policy of maximizing? Can we devise a scale with sufficient scope and reliability for a prudential policy of maximizing well-being to be possible? I think that, even there, probabilities do not figure importantly. But let us turn now to these practical matters. 103 THE CASE OF ONE PERSON 4.What powers of measurement do we actually need? Suppose that I wonder, Should I go for a demanding life of accomplishment, or for a more tranquil life of day-to-day pleasures? It might look as if I could make my decision more rational by introducing measurement of, say, the amount of pleasure that would come my way in the life of day-to-day pleasures. But that would not help much. What I need is a scale, not of pleasures, but of prudential value, a scale on which I can fit both the lives that I have to compare. What seems most important is to get an imaginative grasp on each of these lives as wholes. When that is full enough, I can then just place each on the scale. It is not that I have first to measure the two options in respect of some quantity, "prudential value', in order to be able then to rank them. The ranking is the decision about their value. Judgments of preference are often---as they are likely to be in this case---quantitatively basic: that is, they are judgments that do not depend upon other judgments about the amount of some quantity each option has. It is with basic preferences that the construction of a scale of measurement of well-being begins. The scale is generated out of these raw materials, so no scale of well-being could be used at this earlier stage to generate the materials. And although, as we have seen, in certain restricted domains, quite powerful scales can be generated, we do not need them in this case of prudential deliberation. It can get by on less---often on pair-wise comparisons. The same seems true of many preferences which need not be basic. Suppose that I am facing an academic's dilemma, Should I take a job with fewer hours of teaching, but in a less beautiful town, and with more pay? Well, how much weight do I attach to hours of teaching, beauty of surroundings, and amount of pay? Perhaps here I could introduce an additive measure. But such an additive scale could be introduced only on a fairly rich base of information about my preferences on certain closely related matters. In most cases, therefore, it is as easy simply to form the preference between the options in front of me, as I would have to if they were basic, by grasping as fully as I can what the two options involve. 104 MEASUREMENT There may be a few cases---cases, say, in which there are too many pros and cons for me to be able to hold them all in my mind simultaneously---when deliberation in terms of cardinal measurement will seem necessary. Many economists have taken it for granted that only ordinal scales are needed in one person cases (strictly, in one person cases where preferences are stable). It is true that preference ordering is all that is needed, in theory, to find the maximum. But we are now concerned with what is needed in practice, and it may be that we cannot form any reliable non-basic preference between two many-sided options without resorting to some cardinal measure- ment. But it would be rare if rough cardinality would not do, and rough cardinality is much less restricted in application than the sort of full interval or ratio scale that appeared in the pockets of cardinality that we discussed earlier. Those scales met the most demanding requirements of isomorphism with arithmetical addition: cardinality with origin and unit. But we can still argue, though only impressionistically, for isomorphism between a system of combining rough estimates of the size of the gaps between items that are ordinally ranked and arithmet- ical addition. And it would be little upset to a prudential policy of maximizing that from time to time we met cases where our capacity for measurement was not up to producing a clear answer. There is, though, the threat of incommensurabilities. But they too, if confined to rough equality and discontinuities, are not much of a worry for prudential deliberation. Prudential deliberation can get by largely on pair-wise comparisons, so it can tolerate those forms of incommensurabilities. Additive scales have assumed great importance in the past because, whether we have been for or against utilitarianism, a rather crude mental state account of well-being has dominated our thinking. Even if we have dropped the mental state for the desire account, we have not always managed to drop its views about measurement. If the goodness of a life consists in a lot of short-term pleasures or experiences, then to rank two courses of life we should clearly need to do a lot of totting-up. Therefore we should need the equipment for totting-up, namely, additive scales. But such a crude mental state account is the wrong account. So prudential judgment does not 105 THE CASE OF ONE PERSON depend upon totting-up. So the possibility of additive scales is not a central issue. What, for the most part, we need for measurement is knowledge not of a person's present, indi- vidual, perhaps idiosyncratic, tastes and preferences (they may not be for what is in his own best interests) but of what in general makes life good. And if the person deviates from the norm, then we shall still need general causal knowledge about how persons of this sort work. For the most part, what we need is the sort of knowledge of informed global preferences that is derivable from a general theory of prudential good and a causal theory of human nature. This general theory of prudential good and this causal theory are also, as we shall see in the next chapter, central to understanding the character of interpersonal comparisons. So far as prudential deliberation goes, the demand for additive scales is often a hangover from an account of well-being that we should anyway give up. VII THE CASE OF MANY PERSONS 1.The link between conceptions of well-being and problems of comparability THE problem of interpersonal comparisons is itself problematic- ally protean. It changes shape with every change in conception of well-being. And we cannot simply decide on what is the best conception of well-being before broaching the subject of comparability, because the discussion of comparability is one of the things that helps us to decide on the best conception. On a mental state conception, the problem about compar- ability is largely, though by no means entirely, a problem about knowledge of other minds. We saw earlier1 some of the troubles that mental state accounts run into, the main one being lack of scope; they have a hard time accommodating all that it seems right to regard as part of well-being. We might therefore use instead an actual-desire conception of well-being, say the conception of utility as the subjective value that a person attaches to a gamble. If we do, the problem about comparability is still largely one of knowledge of other minds, though with a different focus. One trouble with the actual-desire account, as we saw earlier,2 is that, though it may suit empirical parts of decision theory and of economic theory, it is much less suitable for moral theory (and for welfare economics, and the normative parts of decision theory, and large parts of social choice theory).3 A person may in fact want what will be bad for him, and the notion of well-being that we, as moral philosophers, are after must be centred on real, not subjectively perceived, benefit. So if we manage to give an adequate account of the comparability of utilities defined on actual preferences over gambles, it is not clear that we should yet have shown how well-being, in the sense that matters to normative theories, is to be made comparable. 107 THE CASE OF MANY PERSONS Were we, therefore, to abandon conceptions of well-being out of the utilitarian tradition and use, say, an objective conception such as John Rawls' index of primary goods,4 the problem of comparability would obviously be much eased. We should now compare persons in respect of such objective and relatively accessible things as their income and the social institutions in which they live. Still, as we have seen,5 these narrow objective conceptions of well-being have their troubles too. They impose a cut-off on considerations available to moral theory that it may not be able to accept and still answer questions at the centre of its interest. We ease problems about comparability almost as much, if we adopt what Derek Parfit has called an "objective list' account of well-being.6 How well off a person is would then turn on the extent to which he realizes the objective prudential values on the list---say, such things as autonomy, accomplish- ment, deep personal relationships, etc. But again, as we have seen,7 objective-list accounts have their troubles. They are, at least on simple interpretations of them, too insensitive to variation between individuals to provide a plausible account of well-being, at least if "well-being' is understood as an all-encompassing assessment of the quality of a particular person's life. That brings us to informed-desire conceptions---say, the sort that I developed in Part One. Comparability would now, as it was on the actual-desire conception, partly be a matter of knowing how strongly a person wants something (when he is informed, or would want it if he were informed). But, since the informed-desire conception can be developed in ways that bring it at least in the vicinity of an objective-list conception,8 many interpersonal comparisons might often be made simply in terms of the items on the list. Still, informed-desire accounts have their troubles, chief among them being whether, in the end, they really can explain comparability.9 Can they even explain intrapersonal intertemporal comparisons, when the latter involve radical change in preferences? If not, the informed-desire account will not do even as a theory of prudence. What these links between comparability and conceptions of well-being show is that we need three things to come together 108 MEASUREMENT for us at the same time: first, we need the account of well-being that we adopt, whether it is of a broad utilitarian sort or a narrower objective sort, to be a plausible account of the domain of prudential value that it tries to cover; second, it must be what we want to use, for purposes of moral judgment, as the basis for comparison between different persons; and, third, it has to lend itself to the sorts of measurement that moral deliberation needs. That these three things must all come together for us enlarges the problem of interpersonal comparisons. I think, however, that we ought not to let the problem get too large. Comparing different persons' well-being can be seen as, in effect, moving from the interests of several different persons, often in conflict, to some sort of common interest, and so, in effect, to the fair or moral decision. But then the choice of a particular method of making interpersonal comparisons be- comes, for all practical purposes, the choice of a whole morality, or at least of the principles of fair distribution. But that is letting the issue get far too large; it is taking too many steps at once. We can decide about the comparability of well-being, on a certain conception, without begging questions about how this information is to be used in moral judgment. We can decide about it, leaving open which conception of well-being is the one that morality needs. And what we decide will be one, but only one, consideration in deciding which conception of well-being morality does indeed need. There is the question, prominent since Robbins' famous article,10 of what value judgments one makes simply in making an interpersonal comparison of well-being, but that question has to be left until we have a much clearer view of the nature of these comparisons. So, for the time being, I want simply to ask, How do we make interpersonal comparisons of well-being, on the informed-desire conception of what well-being is? 2.A natural proposal for comparability and a problem with it Let us, in order to get at what seem to me to be the most live and difficult issues about comparability, make some assump- tions. Let us put complications about informed desires aside for the moment, by supposing that all the desires that we shall be 109 THE CASE OF MANY PERSONS talking about are in fact informed. And let us put the problem of knowledge of other minds aside, too, by making what most persons would accept as the safe enough assumption that we can to some extent know what experiences other persons are having, including the sorts of experiences relevant to their well-being. To make this second assumption may seem to be helping ourselves to far too much, in effect to most of what comparability needs. But it has a practical pay-off; it allows us to focus our attention on the remaining moves that compar- ability also needs. What exactly is the nature of an interpersonal comparison? Are comparisons possible on some conceptions of well-being but not others? Are they possible on the scale that moral and political thinking needs? The first is the important question. And it is a difficult question even with the assumption of knowledge of other minds. Suppose I know a lot about your experiences. I can correctly, fully, even vividly, represent them to myself. But my being able to represent to myself the feel of your experience is, in a way, too much of a good thing. It leaves me with one perception of the feel of my own experience and a second perception of the feel of yours. There is still a gap. How do I get the two experiences on to one scale? Knowledge of other minds does not take us far enough. It is natural, and nowadays common,11 to answer along these lines. Preference bridges the gap. If I can represent two personal states to myself---say, in my shoes with my outlook on things and in your shoes with your outlook---I can rank my being in the one state against my being in the other. I can then take my indifference between the states as showing them to be equal, and my preference for one as showing it to be higher. This answer thus reduces interpersonal comparisons to intrapersonal comparisons by appeal to the judger's own preference as to possible states of himself.12 But there is a problem. Could the crucial judger's preference really be a preference of his at all? I am supposed to introduce interpersonal measurement for my own use by forming a preference between possible states of myself. I prefer, I decide, taking on a life like that quiet scholar's, with his risk-aversion and security and contentment, to taking on a life like that mountaineer's, with his taste for adventure and 110 MEASUREMENT the perils and challenges he faces. Now, usually when I decide whether I should prefer one of these lives to the other, I appeal to my own values or tastes or attitudes. I might, for instance, prefer the scholar's life because I morally disapprove of risking one's life on mountains if one has, as that mountaineer has, children to support. But that would not be a preference that gets us at the well-being of the two lives. To get at that I must strip away my own moral views. But then, stripped of my moral views, I might still prefer living the scholar's life simply because I am, like him, risk-averse. But that would not get us at an interpersonal comparison of well-being, either; it merely shows what I, as I am now, enjoy. So I must strip away my own tastes and inclinations as well. Now, the problem is not whether one can ever learn enough about the scholar's and the mountaineer's lives to form a preference between them, or whether one can ever decide that one of them is better off than the other. It seems to me that one can. The problem is, rather, how I, stripped as I have to become in order not to distort the comparison, can form any preference of my own between what are supposed to be in some sense states of myself, me- as-the-scholar and me-as-the-mountaineer. By reference to what can I now form the preference? Disendowed as I have to be, I have only what anyone else has at his disposal, namely general knowledge of human nature and particular knowledge about that scholar and that mountaineer. I can appeal to the human sciences; I can appeal to information about the extent to which that scholar's and that mountaineer's interests are satisfied. But then the reference to my preference, to which states I should choose to enter, is superfluous. It is not, at least in the first instance, a matter of my personal choice or preference at all, but rather of judgment on grounds available to anyone about two lives. I enter only as making the judgment that the one person is better off than the other. My judgment in certain cases may be wrong; it may differ from yours. But these are possibilities with any sort of judgment. And if you and I disagree over the scholar and the mountaineer, our disagree- ment is not like a disagreement in our personal preferences, which typically arise from our differing tastes and attitudes. The purified preference that this proposal for comparability needs, preference purged of any particular point of view, 111 THE CASE OF MANY PERSONS seems to leave too little; it looks like preference purged of what is needed to make sense of preference. Could there be preference in these circumstances? So could preference, after all, be what bridges the gap? 3.Can the problem be solved? We could admit that my comparison of the scholar-life and the mountaineer-life is a factual judgment that I make based on causal generalities about human nature. But we could go on regarding it, nonetheless, as still a preference of mine in a special sense---the sort of preference that I have when I look at things in a certain detached way. This is a solution offered by John Harsanyi.13 If we lived in a simple world where everyone's preferences were the same, Harsanyi begins, then everyone would have the same utility function, and interpersonal comparisons could be made simply by reference to it and would, in this way, reduce to intrapersonal ones. Of course, in reality, people have different preferences. But one person's preferences, Harsanyi says, are formed by the same general causal variables that affect everyone else's. Thus, differences in preferences can be predicted, in principle, from differences in these variables. If two persons have the same biological inheritance and the same life history, then since they are subject to the same general psychological laws governing the formation of preferences, they will end up with the same preferences. Now, the form of an interpersonal comparison of utility is that the judger would prefer to be in m's objective position with m's outlook than to be in n's position with n's outlook. But since the utility that one judger assigns to this m state is based on general causal knowledge of what anyone with m's biological inheritance and life history prefers, the utility that all judgers assign will, if knowledge is full enough, be identical. If we call the utility that a judger would assign to his entering m's state or n's state his "extended' utility, then everyone has the same extended utility function. Therefore although two persons' ordinary utility functions are likely to be different, their extended utility function will prove to be the same. This restores us to that favourable situation where interpersonal comparisons are 112 MEASUREMENT reducible to intrapersonal ones. It is because interpersonal comparisons are a certain sort of factual proposition that they can also be expressions of personal preference. Most of what Harsanyi says seems to me right. It is plausible that everyone's preferences are determined by the same general causal variables. And a loose way of putting this would be to say, as he does, that if I were like you in biological make-up and life history, then I should have the preferences you have. What strictly is meant, however, is a perfectly general causal regularity: anyone with a certain biological make-up and life history will have certain preferences. There- fore, with enough information about the scholar and the mountaineer and enough general causal knowledge, I can come to understand the preferences of each. But there is still the problem of the gap. How do I bring my understanding of their states together in an interpersonal comparison? Harsanyi says that I, the judger, supply the bridge. Yet, having banned my preferences, in the ordinary sense, by accepting that interpersonal comparisons are factual judgments, Harsanyi seeks to restore them by arguing that, since these factual judgments are the same for everyone, they justify the introduc- tion of an extended utility function that is also the same for everyone. But all that the claim that there is an extended utility function in this case means is that preferences are subject to general causal regularity and that, for that reason, everyone is constrained to make the same causal judgments about preferences. The extended utility function is not my utility function. In ranking the utilities of the scholar's state and the mountaineer's state, I do not rank my utilities. It is not a question of which I prefer, me-like-this or me-like-that. Harsanyi moves from the causal explanation of preferences to the existence of a common extended utility function and, finally, to the reduction of interpersonal to intrapersonal comparisons. But what seems the best interpretation of the existence of a common extended utility function gives no adequate motivation for reintroducing preference. We might, therefore, try solving the problem another way. Perhaps we ought to conceive of the differing utility functions of different persons as a rather superficial phenomenon. Our tastes and attitudes do indeed differ, but perhaps they 113 THE CASE OF MANY PERSONS represent merely different approaches to, in effect, the same substantial end. Perhaps, deep down we are all after the same thing---enjoyment in a broad sense would seem to be the most plausible candidate. This is the assumption that we all have the same deep utility function.14 Therefore, when I think about the prospect of my entering a scholar-state or a mountaineer-state, I should put aside my personal tastes and attitudes, which constitute my superficial utility function, and work only with my deep utility function. Then the preference I should form would be the same as the preference anyone else would form from this ultimately authoritative, basic, perspective. But this solution is even less promising. It falls back on the notion of a single, substantive super-value, and it is highly doubtful that there is one.15 There are many different, substantive prudential values, and different persons may, and often do, differ over what they are. I may aim at realizing certain values in my life; you may aim at realizing others. Since there is no substantive super-value that our different aims may be seen as merely means to or parts of, there need not be any one thing that, over all, we are both in fact aiming at. So there is no deep utility function to ensure that our particularly basic preferences, on which interpersonal compar- isons are to be defined, will coincide.16 Neither an extended nor a deep utility function solves the problem. Neither really manages to dispel the mystery of where, once his personal tastes, attitudes, and concerns are banned, the judger's preference will come from. It is hard to find any coherent explanation of comparability in terms of the judger's preference. This would be bad news for desire accounts if this were the only sort of explanation open to them. But I now want to suggest that it is not. 4.Interpersonal comparisons of well-being There is another possibility. Interpersonal comparisons of well-being, we might say, are judgments of the following sort. I, informed as I am, want this thing very much. You, or virtually anyone, would want it, if informed, the same amount. It is, for most persons, roughly that desirable. He, on 114 MEASUREMENT the other hand, wants it less; he lacks, let us say, certain normal capacities (for instance, he is depressed so wants nothing very much). The quantitative phrases "very much', "same amount', and "less' that appear in these judgments come from the same scale. We can make judgments, based on causal knowledge about human nature and information about partic- ular persons, of how much persons want things, and these judgments place the desires of different persons on the same scale. Perhaps the mistake made by accounts of comparability in terms of judger's preference is to misunderstand the forms that knowledge of other minds can take. If one thinks of it as limited, say, to representations of the texture of experiences, then of course there is a gap that needs bridging, and preference looks a likely bridge. But if we know that you want this only a little and I want that a lot, and these terms are not relative to others things that each of us, in his own case, wants but relative to each other, then there is no gap, so no need for a bridge. This seems to me the best solution to the comparison problem, and I want now to suggest a way of developing it. We have a picture of normal human desires: virtually all persons, when informed, want to live autonomously, to have deep personal relations, to accomplish something with their lives, to enjoy themselves. With experience, we build up such a profile of the components of a valuable life, including their relative importance---a chart to the various peaks that human life can reach.17 These values, if our profile is complete, cover the whole domain of prudential value. They are valuable in any life; individual differences matter not to what appears in this profile of general prudential values, but to how, or how much, a particular person can realize one or other particular value.18 Then, we also build up understanding of how individuals deviate from the norm. For instance, one person may find autonomy anxiety-making, so his life is more complex than the normal one: he faces, as a normal person does not, a hard choice between competing values. Or you may enjoy things more than most persons, while he is depressed and enjoys nothing very much. Also, there will be differences in the form that a value takes in different lives: what you can accomplish, or enjoy, in your life may well be different from what I can in mine. But all this reasoning about 115 THE CASE OF MANY PERSONS individual differences takes place within the framework of a set of values that apply to everyone. And these three elements, a list of universal prudential values, general causal knowledge of human nature, and the information about particular persons relevant to these causal generalities, make up our grounds for judging how well off persons are. Take this case. Smith's great ambition in life is to become a millionaire. He sees it as life's crowning achievement. Whether or not it is, that is how he sees things. Perhaps he lacks imagination; perhaps his horizons are limited; but he does not even entertain the possibility of another goal in life. In a way the very limits of his vision contribute to the intensity of his desire: being a millionaire gets invested with all the attraction of being what it is to have a valuable life. Jones, on the other hand, attaches no intrinsic value to money; his ends in life are to live autonomously, to love and be loved, to accomplish something important in his life, to enjoy himself. Suppose Smith and Jones both reach their goals. How do I compare their well-being? On the model of the judger's preference, my knowing Smith's situation and aims would allow me to form a desire of a certain strength about my landing in that situation with those aims---and similarly for Jones. But "strength' in what sense? No doubt, with a million in the bank and seeing things Smith's way, I should want it a lot. And, with Jones's success seen through Jones's eyes, I should want that a lot too. But there is still a gap. How do I get these two strong desires on to one scale? We cannot be after "strength' in the sense of felt intensity, because just how strongly we feel our desires is largely a matter of upbringing (Smith easily gets emotional; Jones has a stiff upper lip) and has no secure correlation with how well off we end up. Nor could it be "strength' in the sense of motivational force; a person can succumb to desires that are not in his best interest. What we are after must be "strength' in some such sense as "place in an informed preference ordering'. The relevant desires here must be desires formed by at least some appreciation of the nature of the objects of the desire.19 Maybe Smith is sufficiently cushioned by his lack of imagina- tion that he is not at all disillusioned when he gets his million; maybe he simply enjoys the fact of his success. This is obviously important to how well off he is, and I should need 116 MEASUREMENT information about him as an individual to know whether this is so. But perhaps also, if his horizons were not so limited, he would want some of the things Jones is after even more. Perhaps some things just do make life more valuable than others; some things may just be more desirable, when we are informed, than others. It seems very likely. But then judgments to that effect have little to do with the judger's own preference about entering one state or another. It is less a matter of what the judger in fact wants, even of his desires formed on the spot,20 than of how desirable certain things are. The judger cannot form a preference between entering the total Smith- state or the total Jones-state until he knows the strength of Smith's and Jones' desires, in the relevant sense of "strength'. The model of the judger's preference gives no clear answer to the question, What is the relevant sense of "strength'? If the answer is, "strength' in the sense of "place in an informed preference ordering', then part of the ground for interpersonal comparisons, besides the things that are peculiar to him as an individual, are the things that are desirable for persons generally. This case is reminiscent of Mill's interpersonal comparison of Socrates and the Fool. The Fool attaches no value to Socrates' life. Socrates attaches none to the Fool's life. How would each decide how relatively well off they are? This is a case where there is no obvious overlap in the values, tastes, and attitudes of the judger and his subject---at least, in the values, tastes, and attitudes relevant to the judgment that has to be made here. Socrates can, of course, decide: if I had the Fool's values, tastes, and attitudes, I should find the Fool's life valuable; I should actually want it a lot. But finding it valuable and actually wanting it in those circumstances cannot be what matter. In any case, would this decision of Socrates be a personal preference of his? And is this the sort of decision that leads to comparability? The answer in both cases seems to be, No. What Socrates needs to make is a judgment of a very different sort from what we ordinarily understand by a personal preference. He needs to know how much persons generally, when informed, would want each life, how desirable they are. This judgment can be expressed as a personal preference, but the nature of the judgment is very special: it is 117 THE CASE OF MANY PERSONS a judgment about prudential values that is independent of what any particular individual's desires or preferences happen to be. That is, Socrates should need to know, primarily, what made life valuable.21 He should have to appeal to his understanding of what humans, or sometimes humans of a certain type, are capable of, and of the various peaks that human life can reach. Then he should have to decide how close he and the Fool came to some peak. What he should not particularly need to consult is the phenomenological "feel' of their experience, nor their personal tastes and attitudes, nor his own preferences about landing in the one sort of life or the other. It is true, of course, that Mill's own discussion of Socrates and the Fool gives preference considerable prominence, but I do not think that the prominence he gives it is at odds with what I am saying. It was not at all implausible of Mill to take as the authoritative comparison of Socrates' life and the Fool's the preference of persons who have experience of both. Still, there is an important sense in which the preferences that appear in this comparison are not personal ones; they are not expressions of each individual's tastes or attitudes or concerns. They are desires formed by the perception of the nature of the two lives. The preferences relevant to this comparison are not the ones formed by anything peculiar to the judger.22 They are formed, in a way, both from scratch and from no particular point of view, simply from an understanding of the objects before us. So reference to the judger's personal preference between himself-as-this or himself-as-that drop away as irrelevant. And the judgments relevant to comparisons do not even need to take the form of pair-wise comparisons. Each object can be placed singly on the general profile of values: the values in a Socratic life matter a lot; the Fool's gratifications do not matter much. Take one more case, a case in which individual differences matter much more. One prudential value is enjoyment, and it is the plainest of facts that different persons enjoy different things, or the same thing to very different degrees. So we cannot here make comparisons by appeal to the common profile. We have to appeal to the other grounds---general causal knowledge and information about particular persons. 118 MEASUREMENT When I go to decide whether you enjoy wine more than I do, I need to know what your powers of discrimination are and whether your capacities for enjoyment are more or less normal. If I decide that nothing has dulled your enjoyment, that your powers of discrimination are much greater than mine, and that for someone like you greater discrimination goes along with greater enjoyment, then I have my answer. But even enjoyment is misunderstood if the large common element in it is overlooked. There are natural human enjoy- ments. When I consider ways in which my tastes and interests might develop, I look at general issues such as what normal human capacities are and how I am placed to develop them. For example, you like fine clarets; I like only plonk. My palate, no doubt, could be trained too. And faced with that possibility, I should reason something like this. Is it more enjoyable to have these powers of discrimination? Well, persons who have them do not, in general, lose their capacity to enjoy plonk, and most persons find that they have more to enjoy. So I, and most other persons, would be better off developing these powers of discrimination. And we decide this not by deciding how strong this or that particular person's desire is but by deciding how strong this kind of desire is compared to that kind. We enter them on the general profile of human desires. To decide how much someone enjoys life, one does not usually need to get inside his skin; one needs to know both what makes life enjoyable and how he, with his individual differences, is placed to exploit its possibilities. So it is not that the general profile of prudential values is the whole ground of interpersonal comparisons. My point about the general profile is that the ground of interpersonal comparisons is not full without it, not that it is the full ground. Individual differences of the sort that I have just been discussing are obviously an important type of argument in any plausible utility function that aspires to completeness. The profile of prudential values gives the general framework for comparisons. It forms much of the ground of comparison when we choose between different ways of life. But often we are interested in how good various options are within a single way of life, and often with individual ways of reacting and responding. My stressing the role of the profile of prudential 119 THE CASE OF MANY PERSONS values may make my account sound too objectivist and so defective in just the way that I said at the start objectivist accounts tend to be. But one defect of objectivist accounts is that they have no place for individual differences, and the account of prudential values that I developed in Part One does. Another defect is that it is hard to see where an objectivist's values are coming from, but there is nothing to prevent our saying that all values in the general profile must find their place inside informed desires. If a solution to the problem of comparability is to be found anywhere, it has to be found in the context of a theory of prudential values---one that makes it clear both the extent to which values are not personal and the way in which individual differences affect values, one that gets the mix between the personal and the impersonal right. Not surprisingly, using the wrong conception of well-being distorts the problem of interpersonal comparisons. If well- being consists in mental states, then interpersonal comparisons present the daunting task, first, of learning about the texture of individual experiences and, then, of finding a way of ranking them. If well-being is fulfilment of desires, but desires are seen largely as a product of tastes that are personal and varying, then interpersonal comparisons present the equally daunting task of learning each individual's desires and their intensity.23 But tastes do not take up so much of the ground for comparisons. Basic prudential values provide us with an important standard for judging many (ordinary) human lives. They let us say, though only roughly, how good the life is, how it could be better, and how it compares to other lives. They considerably ease the burden of comparison. The deepest and most decisive issues about comparability are ones about the nature of well-being. Does this mean that comparisons are value judgments? That is too vague a question to answer outright. What some writers mean in saying that interpersonal comparisons are, or involve, evaluations is often not clear, perhaps even to themselves. It is true that Robbins and many subsequent economists seem to have meant that there is no way to compare information about individual utilities without making what are, in effect, very strong assumptions about what would 120 MEASUREMENT be fair for the group. For instance, the zero-one rule is the assumption that everybody's best state (give it the value 1) and worst state (give it the value 0) is the same. Our being able to measure each individual's own utility, even on a fairly powerful scale unique up to a linear transformation, is not enough for comparability. But the further assumption of the zero-one rule would give it to us. But the zero-one rule is just false. It is not the case that we all reach the same peaks and valleys. What is needed for comparability is something less than such strong assumptions about fairness but something more than simple matters of fact. Some comparisons are judgments about, say, the factual matter of whether you want to drink a certain glass of wine more than I do. But in focusing attention on informed desires in this way, we are already accepting one particular account of the nature of well-being. And on that account many comparisons involve appeal to a general profile of prudential values. So interpersonal compar- isons are value judgments in this sense: they are part and parcel of a complex normative exercise. 5.Intrapersonal intertemporal comparisons This account of interpersonal comparisons also helps with intrapersonal intertemporal comparisons. When I think at all ambitiously about what will enhance my own well-being, I do not consult my own present tastes and desires, as if which desires I have at any moment, and how they change, is something that just happens to me. In fact, when I am thinking radically, I do not consult any of my tastes and desires, seen as mine. Instead I consult the profile of prudential values. Would I be better off giving up my fool-like gratifica- tions for a more demanding Socratic life? I should answer that question by deciding what sorts of life are valuable and what I am capable of. I should reason in much the same way even over rather trivial matters. Should I learn to like oysters? Well, if I am full of food fads, then I am missing a lot in life, and it would probably be worth changing. But if it is only oysters that I hate, and if I think that my dislike is pretty stubborn, then it would probably not be worth changing. I do not consult my own particular tastes, attitudes, and concerns; 121 THE CASE OF MANY PERSONS I appeal to the same mix of the personal and the impersonal as before---what is valuable and how I am placed. Part of the insight in the wish to reduce interpersonal to intrapersonal comparisons is that the reasoning in the two is virtually the same. Is Socrates dissatisfied better off than the Fool satisfied? Am I better off going down the Socratic path or the Fool's? I often need the same materials to answer either the interpersonal or the intrapersonal question. The problems thought to be connected with our forming desires about these two sorts of life are avoided if we realize that what we need is, not a personal preference of the judger (which, anyway, seems not to be available), but desires---yours, mine, and the other person's---shaped by our understanding of the two options (which is available). In first forming a preference between the options, I give expression to a value. I do not consult a value that is already built into me in the form of a utility function of one sort or other; on the contrary, I create and give shape to part of my utility function. Neither intrapersonal nor inter- personal comparisons are the more fundamental; both rely on the same sort of reasoning. 6.Comparability on a social scale There are serious worries about comparability at the other end of the scale---on the large social scale. Could a government ever carry on its deliberation using a broad conception of well-being, such as one from the utilitarian tradition? Could it realistically expect to collect the enormous amount of informa- tion that it would then need? Could it hope to get its citizens to accept what would sometimes be bound to be deeply damaging results of interpersonal comparisons based on value judgments that they themselves might hotly dispute?24 This worry misunderstands desire accounts. To give a friend a present I should indeed want to know his tastes and interests. But governments are not in the business of giving presents. Their chief business is enhancing possibilities for co-operation. That involves setting up social institutions, defending and amending them, and remedying their unwanted consequences. It is not easy to say how much control and correction governments should go in for, but in general a 122 MEASUREMENT government has to restrict its attention to the general frame- work of its citizens' lives. Any realistic government would accept the limitations imposed by the large scale of its operations: it would adjust to its own lack of knowledge of individual lives and to the intrusiveness needed to overcome it by keeping its hands off a lot of our business; it would meet the need for citizens to accept the outcome of its interpersonal comparisons by prudently steering clear of contentious issues as much as it can. In any case, governments---just as you and I in our interpersonal comparisons---would use, for the most part, the list of prudential values, supplemented by their necessary conditions. So we should need to know not each person's individual desires with the intensities peculiar to his nature, but what is in general desirable in life. And we should need to know what the unavoidable means to those ends are: usually, the healthier, wealthier, freer, and more knowledgeable we are the better we can realize our life plans. So governments would in general compare different citizens' well-being by appeal to things that it is assumed they all value. They would do that in general, but not in every detail. There are items on the combined list of prudential values and their unavoidable means that not everyone aims at, but so long as governments do not become too contentious they can go beyond items that are universally agreed. As political experience shows, stability does not require total neutrality. Anyway, the basic list of prudential values and unavoidable means provides an "objec- tive' measure (in the sense that the measure does not depend upon persons' own individual desires),25 which greatly eases the burden of interpersonal comparisons on the social scale. In practice, governments face a wide variety of kinds of decision, and the basis of their interpersonal measurement should change from kind to kind. Any interpersonal measure meant for use on a social scale involves simplification, but the degree of simplification can, and no doubt should, vary. In any case, the difference between the way we operate at opposite ends of the scale of size---parents with their children, a government with its citizens---should not be exaggerated. Parents' judgments about their children are always made within the framework of their theoretical understanding of 123 THE CASE OF MANY PERSONS human nature. Think of how parents' views about the stress that their children feel have changed since Freud. And legislators appeal to the same sort of theoretical understanding of human nature. In simple cases, a legislator can just take an increase of something on the list of prudential values and unavoidable means as an increase in general well-being, as when the government can promote an increase in GNP that can be used to promote any of the items on the list. In more complex cases, a legislator has to decide between promoting one item on the list at the cost of another, for which purpose he will need a schedule of the trade-offs that virtually all of us are willing to make. For instance, most people rank health and freedom from anxiety high. Should a government promote polio vaccine though it will leave less money for the arts and education? It is not straining the normal trade-off profile that one would build up to say that, given the costs of a polio vaccination programme, the loss in promotion of the arts is less than the gain in freedom from disease and anxiety. Sometimes, however, our general theory of human nature will run out of answers. Would the residents of old people's homes prefer more privacy or more physical comfort? Perhaps a govern- ment ought to find out. Sometimes, of course, research or polls would be too costly or too unreliable to bother with. Some- times, for lack of anything better, a legislator would just have to think himself imaginatively into the position of those affected by the legislation and guess the outcome. And, of course, one legislator's guess may differ from another's. When there is disagreement, we need a procedure, such as majority vote, to resolve it---not any longer to tell us what maximizes well-being but to do the different job of giving us a fair resolution. These disagreements do not, in fact, produce social instability. On the contrary, our legislators now reason far less satisfactorily than this, and if we can stomach what we get now, why should we be upset by better? Some political philosophers argue that governments should restrict their attention to the all-purpose means to a good life, and stay neutral between differing substantive conceptions of what a good life is.26 They argue this, not primarily because such neutrality would ease the burden of interpersonal compar- isons (though it would), but because injustices would otherwise 124 MEASUREMENT arise (some persons would, just because of the strength of their desires, come off better, even much better, in the distribution of society's resources than others, in violation of a deep requirement of equality). I find it doubtful that justice requires such strict neutrality.27 But, in any case, a govern- ment would not survive if it restricted itself so severely. As we have seen,28 if it did, it would not be able to make decisions that are central to its function. The appreciation of art is not part of everyone's conception of a good life; health, in contrast, is an all-purpose means. Yet there must be a point where the moral demands of promoting health are fully met, and also a point at which, though not fully met, they are so little affected that they matter less than some important but non-neutral value such as art. There seems to be no way to decide where these points are without deciding what else is valuable---say, art---and how valuable it is, and then ranking all of these competing claims on some scale of overall importance to the quality of life. Governments need a standard of well-being that, like a dictionary, comes in many sizes. Usually they need consult only the "pocket' standard; but sometimes they will need the "desk' standard, and so on. The informed-desire conception allows this, while a neutral standard restricted to all-purpose means does not.