Rationalizing Rationality

Julian and Will have a discussion going on the meaning of rationality and its relationship to ethics. The discussion was stoked by Matthew Yglesias’s observation that libertarians have a marked tendency to characterize their own views as more “rational” than others’.

This is a useful context in which to emphasize the narrow definition of the word “rationality” as employed by economists – a definition I would like to see adopted more widely. To an economist, rationality is almost entirely a matter of consistency between ends and means; an action is rational to the extent that it increases the perceived likelihood of achieving one’s own goals, whatever they might be. Rationality is not a normative stance, as it does not pass judgment on the correctness or desirability of others’ goals. It isn’t anti-normative, either: it doesn’t say you can’t pass judgment on others’ goals, only that such judgments are not strictly a part of rationality.

I said rationality is “almost entirely” just a matter of consistency because there are two more specific requirements that economists build into their notion of rationality. The first is completeness, meaning that a person is actually capable of comparing alternatives at all. The second is transitivity, meaning that if you prefer A to B and B to C, you also prefer A to C. Strictly speaking, these are both assumptions about the kind of goals people have, not merely the consistency of their choices with their goals. These assumptions are justified, though, by the fact that “consistency” doesn’t have much meaning without them. If there are people with incomplete or intransitive preferences (and there may well be such people), then to that extent the term rationality just doesn’t apply. For instance, the same action could be simultaneously consistent and inconsistent with the same set of intransitive preferences. If I had my druthers, economists would stop saying that completeness and intransitivity are part of the definition of rationality, and instead say that they are necessary conditions for rationality to be applicable as a concept.

Julian argues that the economists’ notion of rationality may be excessively narrow. As an instance of behavior that is seemingly consistent with the maximization of the satisfaction of one’s goals, but that many people would nonetheless consider rational, he offers the notion of satisficing, which means accepting any outcome that meets a certain minimum standard. For instance, a satisficing dater might stick with a girlfriend who is not ideal but who meets certain minimum requirements for looks, personality, and so on. However, satisficing turns out to be an instance of maximizing behavior in the context of a dynamic search. When it is costly to search for something – say, a new mate, or a lower price on a product, or a job with a higher wage – you have to weigh the cost of searching further against the expected gains from finding something better than the status quo (as well as the distinct possibility of ending up with something worse than the status quo). It’s a standard result from search theory that a rational person facing these conditions will adopt a reservation level of satisfaction (quality of mate, price of product, wage of job) and stick with any outcome at least that good. In other words, he will engage in satisficing behavior.

Julian correctly observes (in somewhat different words) that one can always cook up a utility function that produces the behavior one is trying to cast as rational, and I agree there’s something suspect about that. But the question we should ask is whether the behavior in question can *reasonably* be construed as maximizing, and in the case of satisficing, I think it clearly can. Find a person who is engaged in satisficing behavior, such as buying a product once they find a price below a certain threshold level (as I did today, when I bought a Playstation 2 for $180; my threshold was about $185). Then ask that person why they are acting like that. I think the answer will almost always be that it’s too time-consuming or annoying or inconvenient to keep looking longer when you’ve already found something good enough. The reference to the search cost is a dead giveaway that the satisficing behavior is really maximization in a dynamic context.

As another example of behavior that is seemingly not maximizing, and yet might be construed as rational, Julian offers Nozick’s idea of moral side constraints on behavior. This is example is stronger, I think, because it raises a big fat question mark about what a “constraint” is. Economic models typically assume that an agent maximizes utility subject to a set of objective constraints such as income, prices, gravity, etc. Morals could be incorporated in the model as constraints like any other, but they are notably subjective constraints. I can choose to follow a different moral code; I cannot choose to change the laws of physics. And once we realize that constraints may be subjective, all kinds of stuff is up for grabs – maybe those prices aren’t really objective constraints, because I could take a five-finger discount.

But I think the narrow definition of rationality is still helpful here. Indeed, it may be the only thing that allows us to address the conundrum while retaining a non-vacuous notion of rationality. While philosophers can talk about binding moral side constraints, in reality we all know they are not generally so binding. We would like to act morally, but we also face temptations that occasionally prove too strong to resist. In econo-jargon, the marginal rate of substitution of moral behavior for other desirable things is, for most people, not infinite. Thus, morality may be treated as one more preference among many, and rationality need say nothing more about it than about one’s preferences about apples and oranges.

This is not to say that morality is undebatable, or that rationality has nothing to say about morality. It simply means that we shouldn’t try to stuff all moral arguments into the rationality sack. (The attempt to do so is precisely what is so irritating about Rand: she tries to characterize all her arguments as following directly from rationality and pure logic, when in fact she imports value premises from the beginning.) Broadening the definition of rationality to include more than testing consistency of means and ends rarely illuminates moral discourse, and often just confuses matters. Rationality is more helpful as a concept if we use it to make a certain kind of directed and internal critique; e.g., we might say that particular moral rules are irrational if they systematically thwart achievement of whatever goals the rules are supposed to advance. That job becomes much more difficult when the definition of rationality is so expansive that any charge of “irrationality” could be just another way of expressing disagreement.