Interpersonal comparability: the chasm

Interpersonal comparability of utility is a significant problem over several different domains. In welfare economics, for instance, there is a problem of getting to a social welfare function that respects each of our preferences, utilities, or goals. If we are to respect each of these, we must, to avoid Arrow's Impossibility Theorem, make discriminations of intensity between people's desires, preferences, and utilities. This is, at least, Amartya Sen's approach to Arrow's problem which may or may not get us out of that particular difficulty. Similarly, over the domain of politics, the same problem arises--how are we to judge between these preferences that respects the wishes of the voters? A simple voting system does not allow preference rankings over options--typically it is a binary choice between either 'yes' or 'no.' This greatly simplifies our social choice problem, but it fails to take account of just how strongly we may prefer 'yes,' to 'no,' and may result in an outcome were people who are mildly indifferent, but still choose 'no' may win over a group of people with an intense preference for 'yes.' Finally, over the domain of morality, if we are to assess what it is that is morally right to do with or to people, we should at least be sensitive to the preferences of those people. We may even need to make fine discriminations in cases where, and parents are put in this position all the time, we must determine whether we give the cake to our Betty or our Johnny. Our judgment about who would benefit most from this typically wins out, but it follows that we at least need to make some sort of comparison between Betty and Johnny and infer who would like it more, or would benefit from it more.

So we see that interpersonal comparisons of utility are greatly desired. Why is it seen as a problem? There are at least two reasons to suppose this to be a possibly insurmountable difficulty. The first is what philosophers have called the 'problem of other minds.' Just how do we access the minds of others? How can we come to understand the preferences, utilities, etc., that other people happen to have? Can we 'enter' the minds of others (so to speak)? Secondly, even supposing we could cross that hurdle, there is the problem of a single scale used for comparison. Are we comparing apples to apples, or are we comparing two states of mind, utility sets, preferences, that fall on different scales. To compare one set to the other requires a common denominator. If Betty benefits in terms of her health by my giving her the cake, and Johnny would benefit in terms of his spiritual preferences (Johnny being a follower of the cake religion), just how are we to square these two different scales of preferences?

Lionel Robbins, whose name dons the library I go to, reached the following pessimistic conclusion: "Every mind is inscrutable to every other mind, and no common denominator of feeling is possible." Strong words. What they imply is plain: interpersonal comparisons of utility fail to be possible not for lack of information, but in principle. The task is futile, no new information will get us anywhere. Forget about it.

I take it that this statement is far too strong. For one, we do make interpersonal comparisons of this sort, and I don't think we do so without justification. In fact, there are at least a few reasons to suspect that such comparisons, to paraphrase Friedman in a paper on the methodology of positive economics, can be and sometimes are possible.

Let's look at a few plausible candidates. I will rest on two pairs of shoulders to begin my argument: those of Amartya Sen and John Harsanyi. Eventually I will hop onto James Griffin's shoulders and see if the view gets any better from there.

For Harsanyi, interpersonal comparisons are possible. He points out that, in practice, we do make these sorts of comparisons all the time. Parents, as I've pointed out above, do so with children, for instance. They do it all the time and, I suspect, across all cultures. Are we doing so without any justifiability? Harsanyi suggests we do it on two grounds, the first is on the basis of revealed preference, and the second on the basis of expressed preference which implies either satisfaction or dissatisfaction. With revealed preference, we see the options others make, assume that those preferences remain constant when we witness it over and over, and infer that these preferences do certain things to us in terms of our satisfaction. Couple this with the expressions of satisfaction or dissatisfaction at the conclusion of some action, and we begin to create an informed account of the preference set of others.

Two objections arise, and Harsanyi deals with each in turn. The metaphysical objection is banal: "Well," says the philosopher, "we can never know whether what we choose is, in fact, a preference of ours, and neither can we be certain that people aren't telling us the truth about their preferences when we talk to them." Our preferences are buried deep inside of us, after all, so we might keep them hidden by choosing what we don't prefer, and expressing satisfaction with things that we don't like. The second objection is psychological: Can we ever know enough about people to figure out when they are making mistakes? People can, after all, be mistaken when they choose this or that, or when they express a preference for this or that. If we want to get at their preference scales, we would have to figure out when they are making mistakes, and when they are not.

Turning to the former, Harsanyi... to be continued (I have to look at other stuff just now...)

Major objection: Supposing Harsanyi is right in saying that we need to look to psychological laws to find out in what ways people respond to certain things, we still have the problem of character. James Griffin makes this plain when he compares the preferences of a professor, to the preferences of a mountaineer, and the ability of each to sit in a position of empathy or sympathy--like an impartial observer--with the other. The professor obviously prefers her life to the mountaineers (by hypothesis, I say 'obviously' but that need not be. She may be doing it out of a sense of expectation by others). The mountaineer, in turn, has the same preference (the caveat applies here as well). The one may be able to get into a situation where she 'understands' the other's choices, but can she ever do so outside of her own character?

This objection leans on the exposition of instrumental rationality versus procedural rationality found in Heap. Contrasting the two views (whose particulars are not that important for these purposes), Heap makes plain that the one relies on our own abstract individual preferences, and that these are motivational, while the second relies on us following norms (which are motivational). Part of the appeal of the latter is the fact that, in some cases, it is difficult to see how we might make non-norm-based decisions, since these infuse our character. Put differently, we make decisions 'in character,' as dramatists like to say, not outside of it. That character is shaped by something, that external shaping infuses it with certain predispositions.

To apply this, can we ever step out of our 'Professor' character to truly empathize or sympathize with the life of a mountaineer? Harsanyi and Sen (and, to a lesser extent, Griffin) assume that this is possible. But character is what makes any of our choices and opinions sensible in the first place. No evaluation can be made outside of character, only within character that is, at best, shorn of auxiliary features. Some 'core' character constraints will necessarily impact our assessments when we try to look at things from another's perspective. A rich interpersonal comparison is, and I put this strongly, not possible.

Just how rich does it need to be? This is, as I see it, the way out for those who would like to continue making interpersonal comparisons of utility (people like me). Just as Sen says, with a wide enough margin of error, we can probably strip ourselves of everything but the core (which, by hypothesis, we cannot relieve ourselves of), and come awfully close to a comparison. In cases where our core is similar to the core of others (assume similar backgrounds, cultural influences, family life, and so on), those comparisons will have much less of a margin of error.... TIME.