a p p e n d i x
(iv)views about the advisability of greater equality for social stability
(v)views about general advisability of reducing envy and depression
(vi)views about insuring against possible losses in the future to oneself: risk aversion to greater inequalities and so a wish to establish public policy of less inequality
(c)The last three are compatible with teleological constraints; while the first three, with the possible exception of the first, (i), are not.
From the standpoint of moral theory, what we want to know is not on what basis people actually do take distribution into account (presumably all six above reasons and more influence this or that individual), but on what basis they think they should take them into account. And what their moral conceptions are in this respect. Thus, in particular, what allowance does classical utilitarianism make for distribution?
§5. Some Points Regarding Greatest Numbers and Happiness and Maximizing Total vs. Average Utility
(1)The phrase “the greatest happiness for the greatest numbers” occurs first, it seems, in Francis Hutcheson, An Inquiry Concerning Moral Good and Evil (1725), Sec. III, §8. This phrase has occasionally given rise to the confusion as to whether the principle of utility directs us to maximize total pleasure, or to maximize the number of persons, or some weighted mix of both.
The classical doctrine is clear: we are to maximize total pleasure (net balance of pleasure over pain). In the long run numbers (or the size of society) is to be adjusted accordingly. This is what the classical view says. To maximize a weighted mix of net pleasure (total) and numbers represents an intuitionist view and is no part of the classical doctrine. Perhaps it should also be said that the absurdity attributed to Bentham’s utilitarianism by von Neumann and Morgenstern (in the 2nd edition of Theory of Games, 1947), namely, that Bentham wanted to maximize two things at once, aggregate happiness and numbers, is incorrect. Neither Bentham, Edgeworth, nor Sidgwick would fall into this foolishness.
(2)Second, there is the question whether we are to maximize total utility or average utility (well-being per capita over the members of society).
[ 408 ]
Copyright © The President and Fellows of Harvard College
Four Lectures on Henry Sidgwick
Here again the classical view is perfectly clear: we are to maximize total, not average, utility. Of course, the two are the same in the short run with n fixed (n = population of society). But in the long run n is not fixed, and under certain conditions the total and average views give different results for population policy or indeed for any social policies insofar as they influence the size of population (e.g., via effects on birth rate, death rate, etc.). Crucial among the conditions in determining population size is the relative rate at which average utility decreases if population increases. If average utility falls slowly enough, the loss in the average may always be less than the gain in the total from greater numbers, so one can theoretically be led to a very large population of very low (for each individual, positive > 0) utility; rather than a far smaller population that would maximize average utility. Perhaps, however, utilitarianism is not a feasible principle for population policy in any case. If so, then the question is, how is the classical utilitarian to get a complete view?
It is to be observed that Sidgwick is clear on both these points. (See ME, pp. 415f.) This is, among other reasons, why I say Sidgwick is the best representative of the classical view: he is aware of all these sorts of points and resolves them each time in ways consistent with that view. Note again that if we say average rather than total utility is what should be maximized, we seem to be introducing a new or distinct first principle. For if pleasure alone is the sole good, it seems clear that we are to maximize the sum of it. Why in this case does the average matter at all?
§6. Concluding Remarks
(1) It is important to emphasize the point of the preceding remarks. The introductory historical remarks are to highlight the long continuity of the utilitarian tradition; its striking dominance (in at least three respects) over the course of English (speaking) moral philosophy since the first quarter (at least) of the 18th century; and its close tie with social theory, especially political theory and economics: there is nothing else anywhere that approaches this pattern of features.
Also, we must realize that, with some but not much qualification, the leading utilitarians attempted to work out, in reaction to Hobbes, an acceptable moral theory for a secular society with the features of the modern
[ 409 ]
Copyright © The President and Fellows of Harvard College
a p p e n d i x
world. Their reaction to Hobbes (in contrast with the orthodox Christian response, e.g. Cudworth) brings out this aspect of their work: they represent the first modern moral and political theory.
(2)Also, second, I have discussed Sidgwick as the last of the leading strict classical utilitarians: the Bentham-Edgeworth-Sidgwick trio. Sidgwick’s presentation is the most detailed; he is fully aware of most (at least) of the problems (as not even all contemporary utilitarians are). One needs to recognize the many complications in formulating what seems at first sight a rather simple principle (conception). Actually, the initial simplicity makes the complexity easier to see.
(3)Finally, we went over quickly some of the matters that arise in establishing interpersonal comparisons of well-being. This was merely to highlight the problems, to illustrate the difficulties. For example, the 0–1 rule is not, I think, even approximately satisfactory over all sentient beings, though it’s better over human beings. But it shows where many deep troubles lie. We shan’t pursue these questions further now. Yet one must be aware of the problems.
Appendix to Lecture III:
Regarding Cardinal Interpersonal Comparisons
Points: The main points are these:
(1)We need two things to apply utilitarian principle:
(a)A cardinal measure for each individual.
(b)Meaningful correspondence rules to correlate these measures: at least unit comparability.
(2)Cardinal vs. Ordinal Measures:
(a)Ordinal simply defines a complete ordering of better and worse, but not how much worse or better.
(b)Cardinal measure defines a zero and a unit and says how many units between levels.
(c)There is no unique scale but all scales in cardinal measures related by positive linear transformation (e.g. temperature scales). In ordinal, positive monotonic transformation.
(3)In classical cardinal individual measures, e.g. Sidgwick: ME, Bk. III,
Ch. II:
[ 410 ]
Copyright © The President and Fellows of Harvard College
Four Lectures on Henry Sidgwick
(a)Individuals can rank levels of well-being by introspection (complete ordering).
(b)Individuals can rank differences between levels by introspection (complete ordering of differences between levels). These ((a) and (b)) give a cardinal individual measure.
(c)In making these rankings, no choices or decision involving risk or uncertainty are involved.
(4)Re Correspondence Rules: These are needed to match up the measures of distinct individuals:25
(a)Level comparability is a matching of levels.
(b)Unit comparability is a matching of units (how many units of individual A = 1 unit of individual B).
(c)Full comparability = Level plus unit. For application of utilitarianism we need only unit comparability (utilitarianism interested in totals, not levels).
(5)Zero-One Rule as Illustration:
(a)If we work out the consequences of this rule for a society of people and cats, we get what may be odd results. But if we reject the rule for people-cat correspondence, why do we do so? Is it just that we don’t want to be slaves to cats? And what is the correct ratio, not exactly, but even roughly? Very roughly?
(b)If the 0–1 rule fails to hold for people and cats, does it hold between all people? Why can’t the reasons that lead us to reject it for the people-cat correspondence lead us to reject it between different kinds of people? And can we, or should we, accept this implication? Again, how consistent are the reasons for rejection with the classical utilitarian view? Do they involve a doctrine of the different qualities of pleasure in a way utilitarianism can’t allow? (e.g., J. S. Mill, although his view is unclear).
(c)Does the 0–1 rule imply that we should cultivate simple and easily satisfiable pleasures as a matter of social policy? Or even further: simple kinds of people who are easily made happy in ways that require few social resources?
(d)What moral principle is implicit in the 0–1 rule: is it a claim to
25. Here cf. Amartya Sen, Collective Choice and Social Welfare (San Francisco: HoldenDay, 1970), Chs. 7 and 7*.
[ 411 ]
Copyright © The President and Fellows of Harvard College
a p p e n d i x
equal or to maximal satisfaction on the part of all beings with a capacity for satisfaction?
The 0–1 rule is a useful way simply to illustrate the problem of cardinal interpersonal comparisons. And the conjecture is that any scheme of correspondence rules seems to embody deep ethical assumptions that are hard to get out of utilitarianism. It is here that the complexity of utilitarianism starts to appear.
l e c t u re iv
Summary of Utilitarianism (1976)
(1) In the past three classes we have been discussing the classical utilitarian doctrine as formulated by Sidgwick in Methods of Ethics.26 (Think of this as B-E-S utilitarianism, although there are variations between them. Sidgwick’s statement is the most complete and consistent. He pushes the classical view to its philosophical limits.)
Recall that the classical utilitarian view says to maximize:
∑ ∑ uij = u11 + u21 + . . . + unm; where i = individuals and j = periods of time.
Where for each uij:
(a)The real number represents an interpersonal cardinal measure of net well-being for individual i in period j; and where this net well-being is interpreted hedonistically as the desirable feeling aspect of consciousness (cf. ME, I: ix; II: ii; III: x).
(b)This cardinal measure is based on introspection: assessment by the individual (who it is assumed can order levels of well-being and also order differences between these levels: and all this by judgments that do not involve risk and uncertainty (contra von Neumann–Morgenstern cardinal measure of utility)).
26. [The following remarks, entitled “Transition from Utilitarianism,” constitute the first section of a 1976 lecture on “Locke and Social Contract Doctrine.” —Ed.]
[ 412 ]
Copyright © The President and Fellows of Harvard College
Four Lectures on Henry Sidgwick
(c) The characterization of the uij’s is teleological: no notions of the right are required for the definition of utilities.
(2) Now the intuitive idea of classical utilitarianism has a number of attractive features. Indeed, it can seem self-evident that what we should do is maximize the good; or that we should always do that action which, in the circumstances, will most likely have the best consequences, all things considered. Classical utilitarianism seems a clear way of formulating this view.
So formulated, it has many admirable features:
(a)It is a single principle maximizing conception.
(b)Thus, it needs no priority rules, in theory; all working rules are approximations, rules of thumb, etc.
(c)It is a completely general conception, applying uniformly to all sub-
jects.
(d)It has but one basic notion—the good; and then introduces others (right and moral worth) via the maximizing idea.
(e)It has often seemed to be readily expressible in mathematical form suitable for reasoning using calculus, and it has been so used in economics.
One should keep these features in mind. Throughout these lectures I have tried to stress the simple, underlying intuitive ideas that seem to guide the development and formulation of ethical and political conceptions. So understood:
Classical utilitarianism evolves from the notion of maximizing the good; it readily fits with the idea of using all means for the promotion of good in the best way (the most rational husbanding of social instruments and resources).
Such is the structure of the classical view.
(3) In discussing this doctrine I have suggested that its simplicity may be deceptive:
(a)That in defining the good itself, what the uij’s are a measure of, we must be careful to abide by the constraints of a teleological theory: the uij’s cannot include an adjustment for feelings of resentment, or disinterested aversions to unreason (Sidgwick); for attitudes regarding distribution (Gini-coef- ficient), and so on.
(b)The procedures of interpersonal comparisons may themselves embed principles of right, which need to be then made explicit and require an explanation: e.g. zero-one rule; other standard assumptions.
[ 413 ]
Copyright © The President and Fellows of Harvard College
a p p e n d i x
(c)That the uij’s are to be summed rather than, e.g., multiplied is itself an ethical assumption; e.g. only summing is indifferent to distribution.
(d)Moreover, all the various constraints (a), (b), (c) above may naturally suggest, or even force, a certain conception of the person: e.g. that of a con- tainer-person (as found in Sidgwick).
Thus the thought is this: if we pay attention to what is going on in the classical doctrine we see that it is not so simple as it at first appears. Of course, this is no objection. But it alerts us to the fact that any reasonable political conception is bound to have a complex structure, even if it does develop from some simple intuitive idea. Presumably, social contract theory will show the same features.
(4)A final comment on the use of the notion of a utility function. This term is often used (in economics and elsewhere) as a mathematical representation of someone’s preferences, choices, decisions, etc. For example, one could use a utility function to represent the decisions, or judgments, of an intuitionist (TJ, §7). Or these functions could be used to represent the collective social decisions made by a society; or by its members as they make choices via their constitution.
In my view, it is very unfortunate that the notion of utility function is used in this broad way: it would be much better to do the following:
(a)Adopt some other term as appropriate in each case: such as, (multiple) objective function; social decision function; or constitutional choice function. Avoid terms like “utility”- or “welfare-functions” that have special and restrictive connotations.
(b)Realize that such objectiveor decisionor judgment-functions merely represent or describe for purposes of some theory what some agent’s choices or decisions are (the agent may be a person, firm, association, society, etc.). The function may not take any account at all of how this agent decides, of what complex of principles they actually employ. E.g., take the case of an intuitionist judgment-function.
(c)Understand that the problem from the standpoint of moral theory is not representation in the thin sense, but grasping the complex of principles that enter into and regulate the judgments actually made, or that would be made in reflective equilibrium.
(d)Also, mathematically speaking, the representation-function may be such that there is no natural sense in which it describes the agent as maximizing anything. E.g., there may be multiple objectives; or lexical orderings (no continuous representative function).
[ 414 ]
Copyright © The President and Fellows of Harvard College
Four Lectures on Henry Sidgwick
(e) Finally, we will not have a classical utilitarian theory unless the rep- resentation-function observes all the restrictions of the appropriate teleological theory.
To sum up the main point overall:
Anyone’s moral or political judgments can, let’s assume, be represented by some mathematical function. In terms of this function, one can say: they judge as if they think that in each case society should maximize this function, promote the best consequences (as defined by this function).
But this way of speaking implies no specific political conception. The question then is: what is the shape, or what are the special features of this function; and what conceptions and principles stand behind it in the thought and judgments of agents (individuals and society)?
[ 415 ]
Copyright © The President and Fellows of Harvard College