The neo-Humean view
In the Humean view no single summum bonum emerges. According to Hume (1910 ), considering that a thing is good means that people prefer it. Feeling and reason are equally involved in morals: feeling, intended as preference, is the fundamental motive of conduct, while reason is an instrument in the service of our preferences.
The welfare theories based on the Humean moral view assume an impartiality among the ends: morality requires every one of us to be impartial as between the ends of different persons. Welfare is identified with preferences, but preferences are not specifically identified with pleasure or happiness. Simply, a good is useful if it is preferred, and a utility function is considered a very convenient way to describe decision makers' preferences.
Ordinal utility: new welfare economics and the theory of social choice
New welfare economics (NWE) (Kaldor, 1939; Hicks, 1939) does not admit interpersonal comparisons of utility, so preferences can only be ordered. The welfare function is intended as a function that produces an ordering of social preferences (Bergson, 1938). Thus, the problem of aggregating individual preferences arises. A logical-mathematical criterion is Pareto optimality (Pareto, 1964 [1896-7]), which has been more successful as an ethical principle than Benthamism because its moral content is less controversial. The first value judgement admitted is the individualistic foundation: each agent is considered the sole judge of his or her own welfare. Given two allocations A and B, allocation B is better than A if B is preferred to A. The second value judgement establishes that allocation B is optimal compared to A if and only if at least one agent prefers B to A and nobody else prefers A to B. According to Hicks
(1939, pp. 700-1), changes 'which benefit some people without damaging others ... represent an increase in economic welfare - or better, an increase in the efficiency of the system as a means of satisfying wants, that is to say, in the efficiency of the system tout court'. A non- Pareto- optimal situation becomes optimal if the distributional constraint that a sacrifice (measured in terms of utility) requires a compensation (SRC) is satisfied, at least potentially (Brink, 1993, pp. 252-8). Therefore, uncompensated sacrifices are morally unacceptable. Thus a distribution is morally acceptable only if it does not impose an uncompensated or a net loss of welfare on one or more persons in order to provide benefits to others.
The compensation principle can apply to people of the same generation, but when present and future generations are considered, such as in sustainable development situations, some difficulty arises. The Pareto-optimal interpretation of SRC not only raises the issue of compensating future generations, if actions of the present generation impose losses on future generations, but also raises the philosophical issue of compensating the losses of the present generation when its behaviour favours future generations. This issue is still open, since future generations may not exist: they are only 'possible' individuals. Therefore, when a policymaker (who represents the present generation) applies a take-care-of-tomorrow policy, such a policy satisfies the moral value of intergenerational justice but is incompatible with the Pareto criterion of efficiency, because in this case nobody can compensate the present generation (Parfit, 1982; Temkin, 1993).
The Pareto criterion is also admitted by the theory of social choice (TSC), according to which voting is a way of aggregating individual preferences. Nevertheless, a well-known result of this theory is the Arrow impossibility theorem (Arrow, 1951), which claims that, given a minimum number of desirable value judgements or conditions (unrestricted domain, non-dictatorship, Pareto efficiency, independence of irrelevant alternatives), there is no social welfare function which produces a transitive order of preferences and satisfies those conditions.