Harsanyi’s rational choice utilitarianism
John Harsanyi’s (1953, 1955) articles are both leading contributions to modern utilitarianism. The 1953 paper offers a philosophical reconstruction of the notion of utilitarian impartiality.
Harsanyi imagines that an impartial spectator compares the utility of two populations. The fiction ought to help derive a function of social welfare from individual utilities, capturing the fact that social utility is based on the individuals’ personal assessment of their situation. A primary requirement is that the spectator should be impartial, that is to say, he should not favour the situation of one individual over another, whatever their differences. For instance, a poorer person should not be favoured over a richer one, nor the converse. Everyone is hence given a strictly equal weight. Harsanyi’s innovative idea was to extract this reasoning from the context of risk theory. Suppose you were to become one person in a given population: you have as much chance of becoming one person as any other - as much chance of being rich as poor. From outside, and not knowing who you end up eventually to be, you have no reason to favour one person over another in revealing your preference, so that you should consider being any person in the society with equal chance. You therefore want to maximize the sum of all individual utilities - for a given population, equivalently, the average utility. Impartiality is enforced by the fiction of a (thin) veil of ignorance.In his 1955 article, Harsanyi provides an axiomatic justification of the criterion of the sum. Based on the von Neumann and Morgenstern characterization of utility under risk, Harsanyi treats the individual reasoning under the veil of ignorance described above. For any weights αi, and for utility ui of all the individuals i of the society composed of N individuals, he obtains the following characterization of weighted utilitarianism:
W 5 Σi[Naiui (1)
This “construction” of what utility should be under impartiality entails that there exist two definitions of utility, whether actual or ethical.
Actual utility refers to subjective utility as it is, capturing the individual’s personal interest, including egoistic or altruistic views; it corresponds to a numerical indicator of preferences, which is revealed by actual choice behaviour. Such utilities are ordinal, and neither cardinal measures nor comparisons of utility in this sense are meaningful. By contrast, ethical utility corresponds to what utilities should be on the basis of impartial and impersonal considerations alone; they are cardinal and interpersonally comparable. The identification of ethical utility requires certain conditions to enforce impartiality. This is the task of the veil of ignorance. Furthermore, imagine that i is an altruistic person, whose utility is increasing with j,s utility. Then j,s utility is likely to be counted twice, in contravention of the impartiality criterion. Other-regarding preferences should therefore be excluded in order to avoid such double-counting (Dworkin 1978).Since their publication, however, the implications of Harsanyi’s results have spilled over into diverse issues: the interpretation of possible negative weights, problems of distribution, cardinality, difficulty with the introduction of subjective and heterogeneous utility, inconsistencies, and so on. The rich debate between Rawls and Harsanyi is one of its most famous outcomes (for example, Rawls 1974; Harsanyi 1975). The reasoning under uncertainty on which Harsanyi’s utilitarian social welfare function is based relies heavily on the specifics of his choices regarding risk aversion.
By contrast, John Rawls (1971) proposes a thicker veil of ignorance. As he thinks impartiality requires that less information is available to each individual, he conceives the following fiction. Individual i does not know any more the list of all possible situations: i then faces a situation of uncertainty rather than a risky situation. i can hardly weight all the possible situations equally, while i may fear falling into the worst situation. Any individual i under this veil is interested in improving the worst situation, such that the maximin criterion is more suitable. In the end, utilitarian impartiality implies the maximization of the sum of utilities, while the Rawlsian theory of justice as fairness entails devoting all attention to the most disfavoured group in society. As a consequence, the former ignores all distributive issues while the latter is fundamentally egalitarian.