Originally published on Medium on Mar 9, 2018.
In ethical arguments, doing good and preventing harm are often treated symmetrically. All else being equal, doing some amount of good is considered equivalent to preventing the same amount of harm. Although this is not exclusively a utilitarian idea, it follows naturally from utilitarian thinking. Whether it came about via preventing harm or causing good, the end utility in the world is the same, so the end states are equally good. In symbols, let world W1 have utility U1 and let world W2 have utility U2. If U1 is greater than U2, then W1 is better than W2, regardless of how W1 and W2 were reached.
I propose an alternate formulation where utility is not a property of the world but is instead a property of the transition between worlds. Note that this is still a utilitarian model and suffers from both the practical and fundamental issues that utilitarianism has. Let’s ignore that for the moment. The intuition behind this formulation is that for two worlds to be comparable, there must be a common predecessor world from which both worlds transitioned. Without a common predecessor, we are comparing two fundamentally different things, apples to oranges to use the proverbial example.
One way to think about this is that instead of comparing two worlds to each other directly, we always compare two worlds relative to the common predecessor. Symbolically, instead of comparing the utility of W1 to that of W2, we compare them each to the utility of a common predecessor W0. Note that in every case where W0 exists, this will give the same ranking in utility as comparing the worlds directly: (U1-U0)-(U2-U0) = U1-U2 . However, it excludes many pairs of worlds from comparison.
We can look at this in terms of total utility but knowing total utility is not necessary. The comparative aspect of this view is better modeled by associating utility with transitions between worlds. Because we are always comparing relative to a common predecessor, only the change in utility between the common world and its successors matters. The utility difference between two worlds separated by many transitions is the sum of the utility changes for each transition in the chain. This ends up equivalent to the difference in utility between the starting and ending worlds while explicitly ignoring everything that doesn’t matter for the comparison.
Let’s see how this solves the problem of doing good vs preventing harm. When we require utility comparisons to have a common predecessor, we can articulate the issue with this comparison. This comparison asks us to compare two worlds with equal utility, one where the current state resulted from doing good and another where the current state resulted from preventing harm. Can we compare these? What is the common predecessor?
The immediate predecessor could not have been common because in one case, preventing harm, the predecessor was a world with equal utility to the current world and in the other predecessor, doing good, it was a world with less utility than the current world. Perhaps we can go further back to find the common ancestor. At this point, we are no longer comparing doing good and preventing harm, but a more complicated sequence of events such as comparing a world where harm was prevented to one where harm was done then good done which balanced it (whatever that means — another challenge with utilitarianism is that the trade-off between utility types in fundamentally non-obvious).
When we make the cases comparable by starting from a common predecessor, things look rather different. From this common world, if one fails to prevent harm, the utility of the world decreases. If one prevents harm, the utility of the world does not change. If one does not do good, the utility of the world stays the same. If one does good, the utility of the world increases. From the common predecessor, doing good and preventing harm are distinct. They do not result in the same world.
Given the above, it is important to ask why preventing harm seems like a stronger moral imperative than doing good. It seems more important to decrease someone’s suffering than to increase their pleasure. But doing good is the only way to increase utility. Preventing harm only avoids decreasing utility. If anything, one might think, doing good should seem more important then than preventing harm. This is where the comparative model, when combined with a little psychology, becomes more than just a mathematical trick to make some worlds incomparable.
In the comparative model, it is not the total utility that matters. It is the utility of the transition itself. We compare our state not to some ideal but to the state that came before. And here we get into that bit of human psychology. Although more recent research indicates that loss aversion may not be as strong as an effect as previously thought (or so I’ve heard), there does seem to be a fundamental tendency to perceive changes framed as losses more negatively than those framed as gains. In this sense, the comparison between worlds is more than just the sum of the individual transitions from the common predecessor. The shape of the path between the worlds also matters. I don’t have an idea yet for how to capture that precisely, although I do feel the comparative utility model is one step closer than the total utility model.
(Note: it seems someone must have developed this sort of idea before now. It is probably obvious enough to be well known among those who think about ethics. I’m intentionally not looking up related ideas yet to allow myself more time to think about it, but I would appreciate pointers to look at later.)