Introduction

Harsanyi (1953) proposed a veil-of-ignorance argument for concluding that a rational soul, behind the veil of ignorance, would behave like a utilitarian – more precisely, that it would maximize a weighted sum of von Neumann–Morgenstern utilities of individuals. The argument is justly famous, as the first attempt to formalize the idea of the veil of ignorance, using the then recently developed tool of von Neumann–Morgenstern utility, that is, of decision theory under uncertainty. Indeed, Harsanyi used the terminology of the impartial observer (IO), rather than the veil of ignorance, but I shall assume these two metaphors are attempts at capturing the same, ethically correct stance. Weymark (1991) calls the argument Harsanyi's impartial observer theorem. I shall argue that Harsanyi's conclusion is incorrect: It does not follow from his argument that the IO is a utilitarian. The essential point is that utilitarianism requires, for its coherence, a conception of interpersonal comparability of welfare, and no such conception adheres to the concept of von Neumann–Morgenstern utility that Harsanyi invokes.

Let X be the set of social alternatives, or states of the world, and let H be the set of types. Think of X, for instance, as a set of possible income distributions among persons. Define the set of extended prospects as Y = H × X, whose generic member is (h, x). Behind Harsanyi's veil of ignorance, the IO faces the set of extended prospects, where (h, x) is interpreted as meaning “I shall become a type h person in state of the world x.”