Note: I wrote this post before the FTX collapse resulted in lots of people questioning utilitarian principles, so it isn't a response to any of that, although I think it is relevant. It may be a bit more speculative than some people are comfortable with, and isn't a fully fleshed out philosophical argument. But, if you take the reality of consciousness experience seriously, I think these conclusions are unavoidable. Most ideas here are heavily inspired by work from @algekalipso and QRI


No conscious being can deny that suffering is real.

If somebody claims otherwise, I’ll bet they’re confusing it with pain. It’s true that pain usually leads to suffering, but it’s always possible to drug or meditate it away. This is because pain is just an intense feeling, while suffering is something else, a kind of discordance in consciousness. Consider emotional suffering, for example. A bad breakup or a loved one’s death doesn’t cause any physical pain, but it does cause suffering. It’s a real and unfortunate part of the universe, something that is undeniably bad.

The flip side of this though is that joy, happiness and love are real as well. They aren’t just a byproduct of physical pleasure or the evolutionary drive to reproduce, but things that really exist as part of reality. If anything is good, it’s these conscious experiences.

If we accept the reality of experience, it’s hard to deny that the universe has value built in. The good must derive from reducing suffering and promoting bliss, every coherent conception of it must ultimately point to this fact. In other words, if we buy the self-evident fact that conscious valence is real, we get an ought from an is.

 

An ethical theory of everything

Given this strong argument for moral realism, I find it hard to imagine an ultimate ethical theory that isn’t based on some form of utilitarianism. Deontological and virtue ethics may provide good tools for achieving good outcomes, but they don’t get to the heart of the matter. Any coherent ethical theory must aim to attain a world-state with less suffering. This ultimately reduces to a form of utilitarianism that is based on measuring the quality of conscious experience, often referred to as valence utilitarianism.1

Of course, things are more complicated than just this. It’s possible to be in both positive and negative valence states at the same time, or have to choose from many different types of positive experiences. Can we say that peace is always better than joy? Or that physical pain is worse than emotional pain? Deciding exactly how to formulate utilitarianism may get a bit tricky (as expertly explored by Andrés Gómez Emilsson on his blog) but that doesn’t mean that there isn’t a general direction along which we can move towards a better world. Creating more moments of joy or peace and fewer moments of suffering is always the right way to go.

In some cases, though, the details really do matter. As it is normally formalised, utilitarianism runs into a series of repugnant conclusions. But, if you look at them closely from the point of view of valence utilitarianism, some of them start to dissolve.

For example, take the the rogue surgeon thought experiment. If you only care about maximising the number of living people, it could make sense for surgeons to go around kidnapping healthy people and butchering them for their organs, which can then be transplanted into terminal patients, ultimately saving more people than are killed. However, this doesn’t take into account all the collateral effects caused by the fear and insecurity that this kind of practice would unleash on the general population, not to mention the violent deaths of the victims. If we are trying to maximise positive conscious valence, we need to take into account the fact that feeling secure has a very meaningful effect on the experience of our lives. The terror unleashed wouldn’t balance out the gains, so this scenario wouldn’t make sense under valence utilitarianism. Any modification to the thought experiment to get around these problems would require a fantastical level of secrecy, which wouldn’t play out in reality.

There are many other ways people try to point out holes in utilitarianism, with thought experiments about experience machines and fictitious utility monsters. But if you look at the details of a real world scenario from the perspective of valence utilitarianism, they can usually be explained away.2

Dissolving the ultimate critique

However, we are still left with the original repugnant conclusion, introduced by Derek Parfit. It states that with a naïve formulation of utilitarianism, you can always conceive of a world filled with many lives barely worth living that has higher utility than another world filled with a modest number of extremely happy people. This seems like an unacceptable possibility. We would all rather be one of the few in the world of extreme bliss than one of the just-surviving masses.

This problem can be dissolved by valence utilitarianism as well, but you have to be willing to accept two truths about the way the universe is configured. Firstly, that the degree of enjoyment or suffering one can experience is likely logarithmic — the best (or worse) experience is many thousands of times better (or worse) than a mildly pleasant (or unpleasant) experience. Something that has been argued for convincingly by the Qualia Research Institute.

The other, likely more controversial, truth is that closed individualism — the common-sense belief that we are all metaphysically separate observers — is almost certainly false. Believing that we have real separate and essential selves is a useful evolutionary tool, but doesn’t really stand up to scientific scrutiny or sustained introspection. This leaves us with open individualism, the belief that we are all one consciousness, or empty individualism, in which we are just a single moment of experience. Neither of the latter beliefs reify the existence of an individual self, and allow us to sum across the experiences of many different conscious beings, regardless of who they belong to.

Then, given logarithmic scales of valence and open or empty individualism, it’s always going to be easier to achieve a high utility world-state with a few beings enjoying spectacular experiences, rather than filling the universe with miserable people. This is especially true when you play out a realistic scenario, and take the resources available into account.

 

Repugnancy signals an error in reasoning

I actually think that the reason we find Parfit’s repugnant conclusion so problematic is that our intuitions already get the above argument. We intuitively know that the best day of our lives was exponentially better than an average drizzly November afternoon. We also take for granted that the suffering of any being is equally bad, it doesn’t matter who. It’s just that we’re not used to thinking in logarithmic scales or outside of closed individualism. So, when we start trying to carry out rough utility calculations, we don’t take these points into account.

In fact, it seems like all repugnant conclusions point to a problem with a particular formulation of utilitarianism applied to an unrealistic scenario, not utilitarianism itself. From the standpoint of valence utilitarianism, if a solution brings us to a feeling of repugnancy, it itself is inducing a negative conscious experience. Given that the people living in an imagined world-state are probably still going to have fairly similar sensibilities to us, if that world state is unappealing to us, it would be unappealing to them. We can therefore never end up in a repugnant conclusion if we apply valence utilitarianism correctly.

Now, I realise it is always possible to push back against this with specialist thought experiments. But, applying it in realistic scenarios actually has pretty reasonable outcomes. This is helped by the fact that the human nervous system is set up in a way that actually keeps us in quite a narrow range of conscious valence. This is especially true for people free from the struggle to survive, with enough resources to keep them comfortable. We may be able to enjoy a few peak experiences in our life, but even the richest people in the world are firmly stuck in the human condition most of the time.

The practical application of utilitarianism in the real world is therefore probably best achieved with a sustainable and careful application of humanist values. Slowly increasing the number of people living meaningful and secure lives. For this purpose, I don’t disagree with many of the criticisms of misconceived versions of utilitarianism that don’t pay enough attention to conscious experience. In most cases, the application of deontological and virtue ethics can work well.

The other reason we have to be cautious when following valence utilitarianism is that there’s no way to measure conscious experience. You know it when you have it, but that’s it. We should therefore be incredibly careful when designing a future utopia, given that we have no idea the suffering we could unwittingly inflict. These kind of considerations will likely come to a head as AI gets more powerful. Not really ever being sure if they are actually conscious should give us pause before we aim for a future dominated by AGIs.

 

Consciousness must come first

So, even though utilitarianism is probably true, we have to be very thoughtful about how we formulate and apply it. If we are too ideological and insensitive, we will end up in repugnancy. But, if we are willing to take the reality of our conscious experience seriously, there should be a way to make it work. After all, it is the only thing we are sure of and the only thing that could possibly bring value. Many of the bad outcomes of history came from forgetting of this fact and forcefully applying artificial and inhuman values.

These considerations are now more important than ever. As it’s looking like the future is going to get quite crazy quite quickly, our moral intuitions will likely start breaking down. We therefore need a concrete ethical framework to know how should we respond to transhuman technologies, or the possibility of producing artificial consciousnesses. It is therefore now incredibly important to get a strong understanding of the real nature of consciousness. But, we also need a clear way of thinking about value. Valence utilitarianism is the only system that stands up to this task. If we don’t want to lose ourselves in the coming future, we have to always bear one thing in mind —consciousness comes first.


For further reading or listening:

New Comment
7 comments, sorted by Click to highlight new comments since:

This doesn't address any of the strong objections to Utilitarianism (around whether and how individual values get aggregated).

No conscious being can deny that suffering is real.

I deny that "real" is a well-defined word in this sentence.  I experience suffering (and joy and other psychological states), but I can't measure them very well, and I can't compare those experiences to what (if anything) any other cluster of spacetime experiences.  I'm willing to stipulate that such things are, in fact, common.  But I don't stipulate that they aggregate in any useful way, nor that they're important to anything except themselves.

Any coherent ethical theory must aim to attain a world-state with less suffering.

I think that's a misunderstanding of the word "coherent". A coherent ethical theory is one that aims to attain a world state that is logically consistent with itself. Maybe that means less suffering. Maybe that means more suffering. Maybe that means extreme suffering for some and very little suffering for others. All of these world-states are logically consistent, and, thus it's possible to create coherent ethical theories that justify any of them.

Any modification to the thought experiment to get around these problems would require a fantastical level of secrecy, which wouldn’t play out in reality.

That sounds like a hope which reality may or may not be benign enough to fulfil. There are philosophers who argue that the surgeon should kill the one to harvest their organs to save five, and who do not hastily back away from the conclusion, but say yes, yes he should, and not only keep the act secret, but keep secret the doctrine of true consequentialism, which is not for the public. See "Secrecy in Consequentialism: A Defence of Esoteric Morality" by Katarzyna de Lazari-Radek and Peter Singer. The surgeon well-placed to save five lives by cutting up one in secret is morally obliged to do so. The best action is always compulsory, and this is the best action.

They characterise the "esoteric morality" of their title by the following tenets:

• There are acts which are right only if no one – or virtually no one – will get to know about them.

• Some people know better, or can learn better, than others what it is right to do in certain circumstances.

• There are at least two different sets of instruction, or moral codes, suitable for the different categories of people.

• Though the consequentialist believes that acts are right only if they have consequences at least as good as anything else the agent could have done, the consequentialist may need to discourage others from embracing consequentialism.

• Paradoxically, it may be the case that philosophers who support esoteric morality should not do so openly, because as Sidgwick said: ‘it seems expedient that the doctrine that esoteric morality is expedient should itself be kept esoteric’

They go on to say that despite various philosophers arguing against it, "Esoteric morality is a necessary part of a consequentialist theory, and all of the points above can be defended." They proceed to defend them.

The reference to Sidgwick is to his book, "The Methods of Ethics", whose thesis is summarised (and agreed with) by the authors:

Sidgwick famously divided society into ‘enlightened utilitarians’ who may be able to live by ‘refined and complicated’ rules that admit exceptions, and the rest of the community to whom such sophisticated rules ‘would be dangerous.’

I've quoted all this just to point out that there are consequentialists, notably Peter Singer, inspiration for EA, who take consequentialism to be absolutely axiomatic and firmly bite every bullet. Although not to the extent of not publishing their esoteric morality.

Eliezer has written, "Go three-quarters of the way from deontology to utilitarianism and then stop. You are now in the right place. Stay there at least until you have become a god." de Lazari-Radek and Singer say: We are sufficiently enlightened to be able to be total utilitarians, and therefore we must. Deontology is a second-best that is all that the less mentally able can handle.

Knowing your own suffering is on a pretty solid footing. But in taking into account how we impact others we do not have direct perception. Essentially I deploy a theory-of-mind that blob over there probably corresponds to the same kind of activity that I am. But this does not raise anywhere near to self-evident bar. Openness or closedness has no import here. Even if I am that guy over there, if I don't know whether they are a masochist or not I don't know whether causing them to experience pain is a good action or not.

The other reason we have to be cautious when following valence utilitarianism is that there’s no way to measure conscious experience. You know it when you have it, but that’s it.

Does this take imply that if you are employing numbers in your application of utilitarianism that you are doing a misapplication? How can we analyse that a utility monster does not happen if we are not allowed to compare experiences?

The repugnancy avoidance has an issue of representation levels. If you have a repugnant analysis, agreeing with its assumptions is inconsistent to disagreeing with its conclusions. That is when you write down a number (which I know was systematically distanced from) to represent suffering, the symbol manipulations do not ask permission to pass a "intuition filter". Sure you can say after reflecting a long time on a particular formula that its incongruent and "not the true formula". But in order to get the analysis started you have to take some stance (even if it uses some unusual and fancy maths or whatever). And the basedness of that particular stance is not saved by it having been possible that we could have chosen another. "If what I said is wrong, then I didn't mean it" is a way to be "always right" but forfeits meaning anything. If you just use your intuitive feelings on whether a repugnant conclusion should be accepted or not and do not refer at all to the analysis itself, the analysis is not a gear in your decision procedure.

Open individualism bypassing population size problem I could not really follow. We still phase a problem of generating different experience viewpoints. Would it not still follow that it is better to have a world like Game of Thrones with lots of characters in constanly struggling conditions than a book where the one single protagonist is the only character. Sure both being "books" gives a ground to compare them on but if comparability keeps addition it would seem that more points of view leads to more experience. That is if we have some world state with some humans etc and an area of flat space and then consider it contrasting to a state where instead of being flat there is some kind of experiencer there (say a human). Even if we disregard borders it seems this is a strict improvement in experience. Is it better to be one unified brain or equal amount of neurons split into separate "mini-experiencers"? Do persons with multiple personality conditions contribute more experience weight to the world? Do unconcious persons contribute less weight? Does each ant contribute as much as a human? Do artists count more? The repugnant steps can still be taken.

[-]ZT510

Then, given logarithmic scales of valence and open or empty individualism, it’s always going to be easier to achieve a high utility world-state with a few beings enjoying spectacular experiences, rather than filling the universe with miserable people. This is especially true when you play out a realistic scenario, and take the resources available into account.

In the least convenient possible world where this isnt't the case, do you accept the repugnant conclusion?

I think that preference utilitarism dissolves a lot of the problems with utilitarism, including the one of the "repugnant conclusion". Turns out, we do care about other things than merely a really big Happiness Number. Our values are a lot more complicated than that. Even though, ultimately, all existence is about some sort of conscious experience, otherwise, what would be the point?

P. S. Thanks for the post and the links! I think this is an important topic to address.

There is a common Idea that there is only one soul experiencing reality. So like Feynman manentioned an electron travels back in time as a positron and then starts again as another electron. Similarly a single soul is experiencing all living things. If you are that soul you may compute your utility accordingly to your taste.

[-]TAG10

Utilitarianism is not based on the sole axiom that suffering exists. It also requires it to be measurable, to be commensurable between subjects and so on.

For example, take the the rogue surgeon thought experiment. If you only care about maximising the number of living people, it could make sense for surgeons to go around kidnapping healthy people and butchering them for their organs, which can then be transplanted into terminal patients, ultimately saving more people than are killed. However, this doesn’t take into account all the collateral effects caused by the fear and insecurity that this kind of practice would unleash on the general population, not to mention the violent deaths of the victims

A utilitarian society wouldn't have rogue surgeons, but would have organ harvesting. The maximum utility is gained by harvesting organs in some organised , predictable way, removing the fear and uncertainty.