Plausibly there are actually two problems of interpersonal utility comparison- the epistemic problem and the ontological problem. The epistemic problem is the problem of how we could know that a given comparison is correct. The ontological problem is the problem of why we should think such comparisons are meaningful- anymore than measuring temperature against length would be. The ontological problem is the more fundamental problem, in that establishing a method for knowing a given comparison is correct only makes sense if comparisons are meaningful. Assuming a solution to the ontological problem, the epistemic problem is, at least arguably, well addressed as as a practical concern by Lerner’s (1944) equal ignorance theorem.
What I want to argue is that if the logic that we are bound to endorse policies that we would endorse behind a veil of ignorance is correct, then there may be an argument for the sort of “utility-function egalitarianism” indicated by the equal ignorance theorem where the scale factor of everyone’s utility is treated the same, regardless of the actual meaningfulness of interpersonal utility comparisons. That is to say this is a normative solution to the problem of interpersonal utility comparison- it aims to show that we are obliged to act as if interpersonal welfare comparison is possible, and do it at a certain egalitarian ‘rate’.
Happyville is a society run by utilitarian monks. You are about to join it.
(Assume there is no difficult with constructing an interval scale of utility over income. Perhaps we extrapolate from preferences under risk, or from time preferences, both of which give sufficient information to construct such a scale, or perhaps we use Fisher’s method.)
On the day before you make your journey to Happyville you receive an email. While the monks know the shape of the curve of everyone’s utility on income, the data concerning the scaling of the curves has been lost. This includes your own curve, which they have added to the database in preparation for your arrival. The monks have decided to delegate you the task of setting the scale of the curves relative to each other. The magnitude of the curves will determine how they distribute a finite pool of income.
(Assume you can’t recognise your own curve of utility over income- even probabilistically-, or even narrow it down. Assume you aim to maximise your own EU and your preferences are entirely selfish.)
I do not yet have a proof that all rational agents would converge on a single answer to this question, but I would contend that their answers are likely to be similar- they are unlikely to decide, for example, that one of the agents cares 100,000x more about their 10,000th dollar than another, for if they are the person whose preference is made to matter 100,000* less they will lose out greatly, and given the concavity of income, the possibility of being the person who matters 100,000* more isn’t adequate compensation. They will at least approximate something like a Lernerian equal ignorance approach.
Of course our society doesn’t have utilitarian monks, but many societies including our own do operate on an ‘each according to their needs’ basis, at least to some degree. Arguably this argument gives an ethical reason to treat the scale factor of different interpersonal utilities as similar.
If this were an argument for the possibility of interpersonal comparisons, it would be a rather bad one, in essence, simply amounting to a statement that people make relatively similar judgements about these sorts of things which is, at best, relatively weak evidence offering little or nothing on top of the standard extended sympathy treatment. The point is subtly different though. It is instead that we are ethically obliged to engage in these kinds of comparisons, at this sort of egalitarian rate, simply because this is what we would do to protect our own interests were it to come to that, and hence by veil of ignorance type logic, we are obliged to treat others similarly, lest we arbitrarily prefer ourselves. As mentioned above, this is a normative solution to the problem, we are left agnostic on whether such comparisons are “really” meaningful, yet we are obliged to treat each other as if they were.
A.P. Lerner (1944), The Economics of Control. London: Macmillan
J.C. Harsanyi (1955), “Cardinal Welfare, Individualistic Ethics, and Interpersonal Comparisons of Utility,” Journal of Political Economy, 63: 309–321