please ask for more structure than an abstract set of people when you do infinite ethics
In desperate times like these, I think it is a good idea to review what is already known11 According to me and a large pile of remarkably dubious philosophical assumptions I and those I discourse with have slowly accumulated over the years which I have chosen to provisionally accept without question for the purposes of this blog post and put into practice regarding infinite ethics all the time.
One might think they have never, in practice, had to deal with situations involving more than a finite number of people. This is incorrect. Supposing that the Universe is continuous rather than discrete, you are constantly, in each moment, dealing with an uncountably infinite number of people. You see, according to the Many-Worlds Interpretation, the Universe is continuously splitting off into different quantum branches, with macroscopic decoherence occurring over the course of zeptoseconds. It’s not remotely like some sort of “when you go flip a two-sided quantum coin you split the world in twain22 If it were like that, would you have a moral obligation to flip as many quantum coins as you possibly can? Answer in the comments below.” or whatever thing. Every action you take impacts uncountably infinitely many people! So the questions of infinite ethics are actually of utmost practical importance.
How does anyone ever cope? Well, mostly you just need a measure determining how much of any given sort of person there will be. I suppose one could object that choosing any particular measure is unprincipled and you might as well use something totally incomparable to the standard Born rule’s opinion on how common different sorts of people are, but like, eh. The entire concept that one might need to “count” branches seems to me about as misguided as the idea of “counting” states of a thermodynamic system: one should concern themselves as much as possible with properties that do not depend on exactly what course graining you use or precisely how similar two worlds need to be before you consider them to be the same world and sum their measures together.
Fundamentally you just really gotta ask for more structure here than just a set. And I don’t think that that’s all that difficult a thing to ask for. Some person-states are more similar to each other than others are, there’s clearly enough here to start working with. So just say something handwavey about how much measure different worlds and people have, it’s fine. You don’t have to believe in uncountably many equally real people. There’s no way that’s possible. I don’t believe in it. You can just believe in uncountably many people, some of whom are less real than others, or whatever.
Also, like, if we are insisting on just considering the cardinalities of sets, it’s not like it stops at “uncountable infinity”. If this is how we’re saying the SIA works, then surely it is infinitely more likely that there are $2^{|\mathbb{R}|}$ many people than that there are $|\mathbb{R}|$ many people. And you can just keep stacking those power sets. So if we’re admitting just some totally unstructured set of people, it’s going to end up being, like, class-sized. Obviously ludicrous. Please demand more structure than raw sets here, I beg of you. Just Stipulate The Existence Of The Measure, it will make a much better axiom than SIA or whatever. Different people exist different amounts, or different parts of person-space have different existiness-densities, or something like that.
According to me and a large pile of remarkably dubious philosophical assumptions I and those I discourse with have slowly accumulated over the years which I have chosen to provisionally accept without question for the purposes of this blog post
↩If it were like that, would you have a moral obligation to flip as many quantum coins as you possibly can? Answer in the comments below.
↩
and no i don't think an ordered set of people really does the job either
"If it were like that, would you have a moral obligation to flip as many quantum coins as you possibly can? Answer in the comments below."
This depends on whether you have person-affecting views, whether you believe in an asymmetry between creating people with positive experiences versus creating people with negative experiences, and whether you're an average utilitarian or a total utilitarian.
Decompressing that:
Having person-affecting views means that you believe that in order for something to be wrong, it has to wrong someone who exists, and in order for something to be good, it has to benefit someone who exists. (This seems pretty obvious to me, personally. You can’t harm a non-person, or someone who doesn’t exist. You can’t help a non-person, or someone who doesn’t exist.)
An implication of having person-affecting views is that there's no intrinsic merit in creating someone just to fulfill their preferences. No one asked for you to do it, and everyone would have been perfectly happy if you hadn't. This is in line with my moral intuitions. If a mad scientist creates a utility monster, I'm not thrilled with the situation, even if (after warring with the mad scientist to prevent this ever happening again) I might feed a lot of utility to the innocent monster.
But, creating someone who will have unmet preferences is obviously horrible and bad. That's the asymmetry.
(There's also a funkiness to having person-affecting views where if you believe that creating new people is Fine as long as they have positive net utility, you can't be said to be wronging anyone by knowingly and willingly creating someone with half as much net utility as someone else you could have created. (For instance, a mad scientist is thinking about creating a servitor species of house elf that are all fantastically content regardless of circumstances, but is discouraged by how they'd lack drive to make sure his house gets cleaned in a timely manner, so instead he makes house elves who are distressed by untidiness. If the house elves are positive net utility, can he be said to have wronged anyone? (I think philosophers usually give an example about a mother choosing between two embryos, one of whom will live 40 years in expectation and the other 80 years. It seems like you should pick the one who'll get twice as many years of happy life, but can you be said to have wronged the one who gets 40 years, since the child wouldn't have existed if the parent had picked the other embryo?) I resolve this funkiness by saying a creator doesn't get any merit for the met preferences of someone they create, but does get demerits for their creation's unmet preferences. Conjuring a new human is still net-positive utility in expectation, in most cases historically or today, because the conjurer gets merit for improving the world by summoning someone into it who improves the lives of the people around them. But in cases where resource constraints mean that creating a person would make the lives of existing people worse on average, there's no merit to it. This accords with my moral intuitions and avoids the repugnant conclusion.)
Someone could argue that, by flipping a quantum coin, you're forking people, not creating new people, so the asymmetry doesn't hold. People who already exist might like to do twice as many things. However, if you fork people and send them into alternate universes, they'll have different kids in different universes, which means that you're creating new people, which means that the ethics of the action depends on whether you have person-affecting views, asymmetry views, and whether you're an average utilitarian or a total utilitarian.
Plus, it's not clear that forking people without their consent and sending them into alternate universes where they can't communicate with each other is actually something that people want, and also not clear that it's actually improving their lives in any way.
I feel like person-affecting views as you describe them partially capture some of my ethical intuitions but clearly aren't actually fully the case. Like, if the mad scientist takes full responsibility for supplying their utility monster with utility to the point where no one else is particularly obliged to help, I suppose I think the mad scientist did a good but supererogatory thing?
Some people feel this way, but I very much don't. Part of what makes a utility monster a utility monster is its bottomless appetite, like a human's. Suppose a mad scientist grew a human child--since pro and proper nouns make conversation easier, let's say that the scientist is Lex Luthor and the child Conner Kent. And because this explanation is easier to follow if there are two created Kents, let's say Conner has a twin named Courtney.
Prior to growing Conner and Courtney, Lex was entitled to a fair share of the cosmic endowment, and a fair say over what happens in the rest of it, in whatever way maximizes the utility of people who exist, this according to the principle of fairness.
Now that Conner and Courtney exist, they're also owed their own equal claims to the cosmic endowment. If Lex had grown only one child, sapient-kind could demand that the Kents' share of the universe come out of their father's, but because there are two, Lex's share isn't large enough to pay them both.
(Another part of what makes a utility monster a utility monster is that the utility monster is happier to have something than other people are sad to lack it; that's why they're called "monsters." Utilitarianism says that utility monsters should be given utility at the expense of other people, since that's what maximizes average utility (for average utilitarians) or total utility (for total utilitarians). The Kent twins make a good example of a utility monster; they're human, sure, but there are two of them.)
There's no amount of utility Lex can feed his utility monster which will make it the case that no one else is "particularly obliged to help"; the monster is a pair of innocent children, and the obligations good people have towards innocent children are very large. There's no amount of Lex's split of the cosmic endowment he can feed Conner or Courtney that won't make it the case that they, by the principle of fairness, are entitled to their own equal claims to the cosmic endowment.
So much for the mad scientist taking full responsibility for a utility monster. What about for a sapient creation that's not a utility monster?
If a mad scientist creates a new sapient, that sapient either has a utility function that saturates or a utility function that doesn't. (As in, either you could spend a star to make the universe a better place from the utility monster's point of view, or you couldn't.)
If the utility function doesn't saturate, I think Lex Luthor has wronged everyone else in the universe, as expressed above, when that wasn't necessary and didn't improve anyone's life.
(A (total) utilitarian without person-affecting views will protest that it did improve someone's life: it improved Conner's life, and Courtney's! I think the total utilitarian without person-affecting views has an incomprehensible viewpoint. This might bottom out in my moral intuitions, which aren't shared by everyone? But in general I think that self-justifying moral actions (that is, actions which no one wants, or that are unwanted on net, but which will bootstrap into a situation that shouldn't be reversed) are bad. If you're allowed to create or destroy moral patients and justify your moral action based on the end result after the moral agents patients have been created or destroyed, it's the acme of ethics to kill everyone in the universe and replace them with different, happier people, and only slightly less ethical to kill everyone in the universe except yourself and replace them with different, happier people, who serve their creator in all things. If this is morality, why should anyone want to be moral? It seems like we should all agree to be moral* instead, with the asterisk, since that's better for literally everyone.)
If the utility function *does* saturate, and the mad scientist is paying to saturate the utility function out of their own share of the cosmic endowment, I am suspicious about this entire situation. As I said in the parenthetical aside above, I don't find it intuitive that creating a sapient and satisfying their preferences has made anyone better off. Lex made a problem and solved it--I guess that's better than making a problem and not solving it, but it doesn't have me clapping him on the back. And I have some revulsion towards the idea of creating people for the gratification of others, and even more revulsion towards creating people whose utility functions saturate at a lower level than their parents. It doesn't feel to me like a parent should get to specify a child's destiny, and especially doesn't feel to me like a parent should get to specify that their child has a smaller destiny than the parent. But I don't have objections on utilitarian grounds, to mad scientists creating sapients with utility functions that saturate at an affordable level, unless you count the utilitarian argument that I personally dislike it.
I think I'm not sure Conner and Courtney are owed equal claims to the cosmic endowment? Like, if Conner prefers (by some substantial amount) existence over nonexistence even supposing that Conner is only given however much Lex feels like giving him (and isn't given anything by anyone else), and Lex prefers to create Conner, I think probably it's good for Lex to do so? Like, that looks like a clear pareto improvement to me, unless we count the utilitarian argument that you personally dislike it.
But if it *isn't* a pareto improvement, because Lex doesn't (selfishly) prefer to create Conner, I'm no longer sure it's a good thing. Because if we start saying that Lex ought to sacrifice his preferences for the sake of altruistically creating Conner, maybe this roads leads us to instead caring about whether he ought* to.
I don't think it makes sense to describe Conner, prior to being born, as preferring existence to nonexistence. Prior to being born, Conner has no preferences, because there's no such person. After being born, Conner probably prefers existence to death, but that's an importantly different question. And so you can call creating Conner a "Pareto improvement", if it makes Lex better off, since if you make one person better off without making anyone worse off, that's a Pareto improvement, but I think it'd be more natural to say that creating Conner makes Lex Luthor better off, nevermind how Conner feels about it.
And I personally feel that once Conner is born, he has just as strong a claim on the cosmic endowment as I have. I'm not more deserving than Superboy. So that means that if Conner is born, I have to split the universe with him. (Although this is describing a scenario where the universe is near carrying capacity. On Earth in the present day, or historically, adding a new person to your continent makes you *better* off, such that of course you'd be willing to pay them into the LVT-funded UBI or social services.) If Lex doesn't want to give Conner very much universe to live on, but Conner still prefers impoverished existence to death, I don't smile and shrug at the Pareto improvement. The principle of fairness obliges me to help. (Especially because there are diminishing returns to resources; I'd prefer to be run at the fastest possible speed, but if I'm run much slower than that, I'm still glad to be running at superhuman speeds, and if I'm run slower still, I'm still happy to be alive.)
Lex could say, "Well, that's on you Fair Folk, then. I think a universe that has 100 happy people plus Conner living on muzak and potatoes, is better than a universe that only has 100 happy people, but if the Fair coalition feels otherwise and wants to subsidize me, I'll just keep cranking these Kryptonians out until there are billions of you and all of you are living on potatoes." This isn't even a repackaged version of the repugnant conclusion, it's the original version from Reasons and Persons. You take a universe with a hundred happy people, note that adding some number of less happy people would be an improvement from a total utilitarian point of view (if you don't have person-affecting views), note that dehappying the hundred people and redistributing their joy would be an improvement from a total utilitarian point of view, and then you rinse and repeat.
Every step in the process improves total utility, half the steps in the process improves average utility, and if everyone involved are Fair Folk, the whole thing would be voluntary. (If not everyone believes that Conner and Courtney are just as deserving as the original 100, only the Fair Folk get repugnified, but this doesn't seem like a redeeming quality of the setup.)
To this I say: I don't think it makes sense to describe Conner, prior to being born, as preferring existence to nonexistence. I think prior to being born, Conner has no preferences at all; there's no such person.
Personally I want to be ran as slow as possible so as to maximize the speed of light relative to my subjective rate of time and enable prompt-seeming communication with people in other star systems.
I'm not a stellar engineer, but it seems possible to me that we'll eventually want to drag most of our stars nearer together, to make co-op games less laggy and shorten the commute.
Dragging a solar system from the edge of the local group to the center wouldn't cost very much in terms of a star's lifetime output, although it'd cost a lot in absolute terms and it's possible we'd have better things to spend those joules on.
I know if I we're willing to adopt unprincipled rules it would solve all my problems! aiui you just have to bite one of the prongs of the impossibility theorem for infinite ethics and you're chilling, but I don't want to adopt an unprincipled (I guess convenience is a principle but whatever) rule.
I'd be modifying my preferences to do that, my preferences weren't already about "measure", they were about like some other shit. I don't want to change my preferences to suit unprincipled rules!
Your preferences are *totally* already about measure. You think it's half as bad if I torture a shrimp conditional on a quantum coin coming up heads compared to if I torture a shrimp unconditionally.
well, maybe I think many worlds generates large but finite numbers of worlds, or maybe I don't agree with MWI
I think you still care about measure in this case. You still think the shrimp torture is half as bad if it's a classical coin, and indeed you think it's p as bad if it's a classical coin that comes up heads with probability p, and that's enough that you're using a measure.
I think it's different. If we stipulate some fair coin object or some bell curve object or whatever, we're also basically stipulating the right measure theory assumptions to use. Like the thing about a classical coin is that half the time it lands heads and half the time it lands tails.
And with the real world we're not dealing with hypothetical objects and there's probably an actual fact of the matter about if I can take actions that will increase or decrease the total amount of suffering in the multiverse. I object to just making whatever assumptions make it the case that my actions matter by my values, because I want to learn the true fact that my actions don't matter by my values if they don't.
Either that and/or I don't understand measure theory.
If my actions don't matter for some concrete reason I want to know that, but if my actions don't matter due to some abstract mathematical argument about whether it's possible for my actions to matter given some particular choice of what mathematical structure to use to model the multiverse then I think I want a better mathematical structure? Unless there is a very compelling argument that the particular mathematical structure being proposed is correct. So I think "probably we want to stipulate i.e. some sort of measure" feels like an improvement to me on just dealing with abstract sets.
I don't want to hold a higher standard for mathematical structures that prove my actions don't matter than for mathematical structures that prove my actions do matter
I mean I'm not sure if I think it's a dumb choice because it proves my actions don't matter per se, rather than one of the many consequences of it being a dumb choice being that it can be used to argue my actions don't matter
Part of it might be that I've just thought about set theory enough to come to the conclusion that you really really want at least a topology or something whenever you start talking about infinities. Or like, ordinals are fine in some contexts, but they don't do the job here.
Where do multiverse ethics differ from EV ethics with normal probabilistic reasoning?
I think mostly nowhere?
Though it's possible that you like, care about a certain sort of thing happening in at least one world, but don't care if it's in the multiverse branch you end up in, so you're okay with it being quantum RNG-unlikely but not with it being classical PRNG-unlikely. I worry that this sort of preference might run into dilemmas of some sort but eh.