I prefer ethical systems that make sense in all situations and all circumstances that can occur with reasonable probability.

The following scenario made me think that classical utilitarianism has some intriguing limitations.

Consider that David Pearce's Abolitionist Project one day will finally be successful and all sentient beings will have been modified so that they cannot experience suffering anymore - without impairing any of their functions critically. As negative utilitarian you could declare final victory (modulo freeing the rest of the universe from suffering). But then what? Time to look for another job! Let's switch to classical utilitarianism and try to maximize happiness - while not creating any suffering in the process.

Further imagine that an indefinite time later people has found a way to maximize happiness (or generally all pleasant emotions) quite easily while leaving everyone being fully functional (no "wireheading"). Perhaps that's as "easy" as equipping every mind with some kind of hedonic device that is able to maximize every desired emotion at will. As classical utilitarian you could declare final victory (modulo bringing joy and happiness to the rest of the universe). But then what? Time to look for another job! Let's switch to preference utilitarianism and try to fulfill all remaining preferences - while staying in the state of maximum happiness and freedom from suffering.

For the rest of your time you try to fulfill as many preferences as possible throughout the whole universe. If you are actually very successful at that and reach what I call a "non-plus-ultra state" you may declare victory again. From now on, you just try to keep up the good work / the NPU state and strive for stability. But anyway utilitarianism does prescribe to keep up maximum utility anyway, so there's no extra level beyond negative + classical + preference utilitarianism. In fact, if there was something else you wanted to maximize after reaching an NPU state you would have the preference to maximize that extra thing, so it wouldn't be anything new. So, it's really the maximal demand to approximate a NPU state.

This thought experiment shows that there's a relatively natural hierarchy of utilitarian priorities:
1. Negative utilitarianism / minimize suffering
-> 2. classical utilitarianism / also maximize happiness
-> 3. preference utilitarianism / also fulfill preferences

But you might not want to wait until you really got rid of all suffering to increase happiness, or wait until also happiness is maximized to fulfill all kinds of independent preferences. It would be legitimate to increase happiness if it doesn't cause any suffering, or to fulfill independent preferences if it doesn't make anyone less happy or cause any suffering. For those who want to be strict you can replace if with iff (if and only if). In the case that you want to be completely strict, you can argue that the Abolitionist Project has priority above all else and people should devote all their resources to that project first. And you can argue that the project of maximizing happiness is much more important than fulfilling all kinds of emotionally not very relevant preferences.

These considerations give rise to (at least) three different types of hierarchical utilitarianism:
Radical hierarchical utilitarianism: First focus completely on abolishing all suffering, then focus completely on also maximizing happiness, then also fulfill preferences.
Strict hierarchical utilitarianism: You try to maximize freedom from suffering, happiness, and preference congruence all at once, but only if no higher priority is harmed - do not create extra suffering in any case, and do not diminish happiness once suffering has been abolished.
Lenient hierarchical utilitarianism: There are many ways in which you can be lenient regarding those priorities:
(i) Causing suffering may be acceptable iff the gain in happiness clearly outweighs that cost.
(ii) Causing suffering may be acceptable iff the gain in preference fulfillment clearly outweighs that cost.
Obviously (i) requires that suffering and happiness are commensurable, while (ii) requires that suffering and preference fulfillment are commensurable. The latter seems to me quite counter-intuitive, but perhaps there's a good way to make sense of it. If you even regard suffering, happiness, and preference fulfillment as commensurable you can formulate
(iii) Causing suffering may be acceptable iff the gain in happiness + preference fulfillment clearly outweighs that cost.
And there's also
(iv) Diminishing happiness may be acceptable iff the gain in preference fulfillment clearly outweighs that cost.

In the end, it all just boils down to the question which qualities are commensurable and how you weight them. You can say that you weight everything the same, but the point of hierarchical utilitarianism is that preventing suffering is still more important than everything else, and that increasing happiness is still more important than preference fulfillment (in the sense that you invest more resources into the more important projects, if everything else is equal). In very short form this could be codified as:
dislike > like > want

It may be possible to chose an alternative hierarchy, but that doesn't feel natural.

Thoughts?

My first thought is that you could post this on the Wiki Felicifia site (enter the proposed title in the search box and it will prompt you to create the page -- capitalize the first word only (unless you include a proper noun).

I might answer the question differently if I apply it to my personal ethic as opposed to a legislative ethic. My current preference for my personal ethic is relatively consistent with that hierarchy: remove suffering first, add some degree of happiness, but I'm not sure of anything beyond that. I still perceive focusing on satisfying every preference is essentially irrational and that it will result in the misdirection of time, money, technology, energy, brain power, etc. The Buddhist approach would be to seek contentment independently of great thrills.

A parallel from science is a recent study that shows that people who drink diet soft drinks have larger waist-lines than those who drink regular soft drinks (and they aren't exactly skinny and they are at risk for diabetes). Something similar was also shown with rats: add a zero-calorie artificial sweetener to food and the rats gain lots of weight. The human study suggests that people condition themselves to expect gustatory delights, thereby leading them to eat additional calories that exceed the calories that they would have gotten in a regular soft drink (and another theory is that the process caloric regulation is disrupted by the sweeteners). My point is that people who endlessly pursue their every whim are conditioning themselves to non-productive (or possibly self-destructive) behavior that might be better addressed by a Buddhist ethic of seeking contentment in the way things are. From a utilitarian perspective, this equates to the hypothesis that pursuing every whim will actually reduce my true utility in the long run.

As negative utilitarian you could declare final victory (modulo freeing the rest of the universe from suffering).

No you can't. You have to make sure there really is no suffering and you didn't just mess up on your calculations. You have to make sure suffering never happens again. If you follow evidential decision theory, you have to minimize the suffering in the past and in alternate Everett branches. If you follow timeless, updateless, or ambient decision theory, you have to minimize the suffering in those as well as alternate possible universes.

Hi Mike,
A completely fantastic post. As Rehoot says, it belongs [url=felicifia.com/index.php?title=Hierarchical_consequentialism]on the Wiki[/url].

I think DanielLC's point is a good one. It reinforces your distinction between radical hierarchies and others: In theory, a radical hierarchical utilitarian would proceed to maximising happiness only after maximising suffering perfectly, and in practise, this would never occur.

You may notice that I linked you to the Wiki page for Hierarchical consequentialism rather than Hierarchical utilitarianism. This raises question of whether your theory is consequentialist, rather than utilitarian. The building blocks in this theory are forms of utilitarianism. But once you put them together, I think you've created a consequentialist system. My line of thinking is that utilitarianism is that it permits only one utility. Once you go pluralistic, you go consequentialist. I'd be interested to hear your thoughts.

maximising suffering

You mean minimizing?

My line of thinking is that utilitarianism is that it permits only one utility.

This works with utility if you allow infinitesimal utility. Also, this is grammatically incorrect.

I'd call it consequentialism because there's no reason to limit this to utilitarian stuff. You could very well have a layer that's egoist or prioritarian or some such. Then again, you could have a layer that's deontological.

DanielLC wrote:No you can't. You have to make sure there really is no suffering and you didn't just mess up on your calculations. You have to make sure suffering never happens again.

I think this is a key criticism of all the eventualities in the OP. Once you've decided on your definition and attitude towards utility, your task is infinite. You can always create one more person with positive utility, or look for one more way to prevent suffering. It doesn't make sense within the parameters you set yourself to suddenly switch algorithms just because you're doing well at the moment. I guess it might work for satisficing utilitarians, but does anyone seriously accept SU?