Sharp Blue: Utility principles

In discusions of utilitarianism, one key point is often elided: what is meant by maximising utility. Instead, arguments seem to focus on what utility function we should use, with some advocating happiness, others lack of mutual harm, still others the satisfying of individual prefences. The default assumption about the maximisation of utility appears to be what I shall call differential utilitarianism. In essence, this says that when making a choice between possible actions, we should choose the one that results in the greatest increase in happiness (or whatever) between the situation immediately before the act and the situation immediately after it.

Differential utilitarianism is prone to several types of undesirable outcome. The most obvious is the "lotus eater catastrophe", in which everybody chooses to takes drugs that induce a permanent state of euphoria. Clearly at each stage of the progressive intoxication, everyone is happier than they were a moment before. However, in the end the machinery of modern civilisation would presumably grind to a halt and everybody would die, simply because to take actions to preserve civilisation would mean temporarily giving up wallowing in pure pleasure, and under differential utilitarianism nobody would make such a choice. A less extreme but still undesirable outcome is the "Brave New World" scenario, in which social engineering is used to modify everybody’s preferences so that they become willing to happily accept a state of affairs that most of us would consider a dystopian nightmare.

However, differential utilitarianism is not the only approach to the theory of utility. A much more sensible formulation is that of integral utilitarianism. This says that when making a decision, we should consider the future histories produced by each possible choice and then pick the one that results in the highest possible total utility. In this formulation, we must clearly think of utility in terms of something like "happiness hours" and then compare future histories in terms of the total number of hours during which people are happy (weighted by some measure of just how happy they are during each interval). This, then, avoids the lotus eater catastrophe because the future histories in which the entire population descend into drugged apathy are much shorter than those in which people act in other more sensible ways. Indeed, it also allows choices that make things worse in the short term if they also make things better in the long term, thus avoiding not just the lotus eater catastrophe but a whole spectrum of locally optimal but globally suboptimal outcomes.

At first glance, integral utilitarianism might seem prone to its own spectrum of pathological outcomes: "Year Zero" scenarios in which an ever longer period of ever worse conditions is accepted now in the hope of ever greater rewards in the future. Unlike the lotus eater catastrophes, these outcomes are not inherent in the principle of integral utilitarianism itself but rather are defects in our ability to assess the ways in which our choices here and now will affect the future. These defects can be reduced but never entirely eliminated by the progress of the sciences.

An obvious development of the integral utility principle is the cosmological utility principle (a conflation of the cosmological principle and the integral utility principle). This says that the contribution to utility of all people (or people-equivalent entitites) everywhere in the universe and at all times are equivalently weighted. In other words, we should not discount the contributions to the utility integral for a given history of the circumstances of people far in the future relative to the contributions of people in the near future. From this viewpoint, we should act in ways that tend to promote the survival of our species, our society and our selves arbitrarily far into the future. It’s fortunate, of course, that these ways tend to be consilient with the sorts of actions permitted by the moral side-constraints imposed by our current society, and more generally with the idea of the advance of science, the promotion of open societies, the establishment of free markets and so forth.

Furthermore, as described in my earlier article, "Living at the Fulcrum", on the largest scales there are two possible classes of history: those in which civilisations continue for an infinite amount of subjective time and those which are merely finite. Clearly, for many sensible choices of utility, the total utility in an infinite history will be infinite. This means that we should very strongly prefer choices that will tend to lead to our civilisation expanding out into the universe and existing forever. Even if the universe does not cooperate in allowing us an infinite future, the total utility of histories in which we fully colonise and exploit the solar system will utterly dwarf the total utility of histories in which we remain confined to the Earth. In turn, the utility of histories in which we are satisfied with just the solar system will be miniscule in comparison to histories in which we colonise our entire galaxy, or beyond. And this is true pretty much regardless of what choice of utility one makes so long as one uses an integral utility principle. This, then, is the lesson of the cosmological utility principle: that we should make choices now that maximise the probability that in the future humanity will colonise and industrialise space as fully as possible.