Since the moral values of society have to be rationally pursued, our analysis of the different welfare theories has demonstrated that they cannot be classified according to a single moral criterion, and a single kind of instrumental rationality.
Benthamism, NWE, TSC and RU are based on an ethics of motive, which considers only subjective values represented with a utility function; while Keynes's macroeconomic theory is grounded in an ethics of end, which also admits the rational choice of moral values, which are assumed to be not always measurable.
As for right conduct, when subjective probability is assumed to be equal to objective probability, welfare theories based on utilitarianism and the neo-Humean view are cases of substantive rationality because they describe agents' behavior in a situation of complete information about consequences and probability. When these theories admit bounded rationality, subjective probabilities are assumed to be different from objective probabilities in learning maximization models; on the other hand, when there is insufficient information to establish subjective probabilities, the principle of indifference should be applied. Keynes's macroeconomics, in contrast, also admits other situations of bounded rationality by recognizing that agents may act under conditions of fundamental uncertainty. In this case, bounded rationality focuses on non-maximization procedures (such as those of intermediate objectives and conventional judgment), not only because moral goods may be non-measurable but also because of the awareness that, when a numerical probability cannot be identified, the principle of indifference leads to absurdities.
Since no rational choice can be made between competing moral systems, a GTW is needed. GTW admits all the possible values that a society espouses. As for instrumental rationality, since agents may have inadequate information and insufficient computing capability for applying the maximization procedure, it must be based on rational dualism. Maximization models are applied when situations are simple enough to be fully understood and first-best solutions are found; on the other hand, models of bounded or procedural rationality are appropriate for addressing more complex problems which are only partially understood, as is the case with most practical applications and under conditions of fundamental uncertainty.