Social choice with approximate interpersonal comparisons of well-being

Some social choice models assume that precise interpersonal comparisons of utility (either ordinal or cardinal) are possible, allowing a rich theory of distributive justice. Other models assume that absolutely no interpersonal comparisons are possible, or even meaningful; hence all Pareto-efficient outcomes are equally socially desirable. We compromise between these two extremes, by developing a model of `approximate' interpersonal comparisons of well-being, in terms of an incomplete preorder on the space of psychophysical states. We then define and characterize `approximate' versions of the classical egalitarian and utilitarian social welfare orderings. We show that even very weak assumptions about interpersonal comparability can yield preorders on the space of social alternatives which, while incomplete, are far more complete than the Pareto preorder (e.g. they select relatively small subsets of the Pareto frontier as being `socially optimal'). Along the way, we give sufficient conditions for an incomplete preorder to be representable using a collection of utility functions. We also develop a variant of Harsanyi's Social Aggregation Theorem.