I always found the solution relatively obvious. People have diminishing marginal utility in their moral views, just as they do in their economic views.
We obviously value two things. Quality, and quantity. We can place a unit of value on each, and then compare the two to see if we’d be willing to sacrifice a huge amount of quality, for a large amount of quantity, but absent some level of discounting, you’ll get repugnant conclusions.
In microeconomics, a repugnant conclusion would be something like:
Option A: One week at a luxurious resort. You’ll have 10/10 enjoyment
Option B: 10 weeks of mildly pleasant stays in budget motels. Each week is a 2/10 experience.
Total enjoyment with option B is double option A, so it seems like you’d choose it. Fortunately we have complex values, and the more we have of something, the less we value more of it. The same can be said of moral terms. When we have conflicting values, it’s never a case of a linear total comparison. It’s a case of solving an optimization problem with hard constraints, as is the case in economics (and even then, these are just approximations of value, not the actual determinant itself).
This is also an argument against utility monsters. Would you value a millisecond of the most intense pleasure possible, a billion times better than anything anyone has ever experienced, or a highly-positive lifetime in the normal human sense?
Yes, although notice an important difference in the context of ethics and value that makes discounting much less plausible/obvious than in the economics case: presumably, there is no difference in *moral value added to the world* between one person enjoying a good life and another person enjoying a good life. While there plausibly is when it comes to me enjoying my 3rd day at a hotel. My faculties are limited in this way such that further days of joy are not created equal. But not at all clear that more *people* living the same lives should have less contributive moral value. And, as I say in the post, the biggest problem for these views is that they imply this odd fact that a world with people living only *slightly* worse lives could never have more value than a world with *50 trillion times less people*.
I think the disagreement is that economics is a theory for assigning value under constraints, while moral hypotheticals (those divorced from reality) are constructed without any constraints. In a moral hypothetical, the tradeoff between quantity and quality can be set at any arbitrary ratio, so no matter what level of discounting you imagine, there's always some ratio, with some large number of barely-worth-living lives, where the optimization breaks down.
But in applied economics, as in applied morality, we aren't considering quality and quantity in a vacuum. More quantity does mean less quality in some ratio, since there's only a finite number of atoms in the universe, and more practically, only a finite amount of resources and time available to do good things with.
The difference between the limitless scaling that moral hypotheticals like yours posit and the real world is that there's a fundamental part of reality that forces the first additional person and the 10^25th to have different moral value. Existence implies the consumption of resources (and therefore some tradeoff with quality of life). Once you start approaching the carrying capacity of the environment - whether that's a small island or the entire universe - each additional person necessarily reduces the average quality of life, however slightly. At some point, adding one more person, while valuable in isolation, subtracts value from everyone else in terms of marginally reduced quality, due to scarcer resources.
Essentially, just as in economics, ignoring cost gives you infinite demand curves. Ignoring physical constraints, where each person must consume some amount of resources to live a certain quality of life, gives you repugnant conclusions.
> the biggest problem for these views is that they imply this odd fact that a world with people living only *slightly* worse lives could never have more value than a world with *50 trillion times less people*.
I think the assumption here is that value scales linearly and indefinitely. If you stack up enough barely-worth-living lives, it must eventually outweigh a smaller group living slightly better lives. But this assumes constant moral returns on each additional life, which isn’t realistic. In any constrained system (planetary, ecological, social) scaling quantity degrades the conditions that support quality. So yes, it’s entirely possible that after a certain point, more lives just don’t add net value, even if they’re technically above the “barely-worth-living” line.
Apply this to the actual planet. Would it be better to have 50 trillion people living slightly worse lives, or 10 billion living slightly better ones? In a vacuum I would say the 50 Trillion, but in reality I would say the 10 Billion.
50 trillion people couldn’t live “only slightly worse” lives under real-world conditions. That would mean roughly 10 square meters per person—closer to 3 if you exclude oceans. The result wouldn't be 50 trillion lives at welfare level 99.9—it would be 50 trillion net-negative lives, starving and collapsing the biosphere. It seems it would be intuitive for the 10 billion better lives over the 50 trillion imploding ones.
The real question is: how much are we willing to trade off quality for quantity? And the answer isn’t fixed. It depends entirely on what level of quantity we can sustain before we start meaningfully sacrificing quality. Once you hit that threshold, more is worse. If you tell me "actually there is no threshold because we invented space travel and nuclear fusion so the only sacrifice of 50 Trillion lives is having slightly more light pollution from all the O'Neill cylinders in orbit" this moral framework would favor the 50 Trillion lives, despite applying a level of discounting, since the rate of discounting is dependent on the additional cost of each additional life lived.
I'm sympathetic to your argument here, but it can be avoided by doubling the number of planets along with the population. It's not 50 trillion people on/around one planet, but 5000 planets each with 10 billion people on it. Surely a universe with exactly 5000 planets, each with 10 billion moderately happy people on it, is slightly better than a universe with exactly one planet, even if the 10 billion people on it are very slightly happier than the people in the other universe. And since no one world is any more stressed than any other, your objection does not apply.
Their point is that the level of happiness/utility/whatever is stipulated, and the other facts about the world must be adjusted in whatever way necessary to support those numbers.
I appreciate articles like this one but I do think your case may be overstated. You say:
"And, contra what many people seem to think, avoiding the Repugnant Conclusion is everyone’s problem. It is not something that only arises if you are a utilitarian."
It's not everyone's problem. It's not a problem for me.
"The answer seems blindingly obvious: you should hope A is the case! It is preferable to Z!"
Who does it seem blindingly obvious to? It does not seem obvious to me. Neither answer seems obvious to me. I think the thought experiment involves underdescribed and highly complex scenarios that nobody could accurately model. Whatever people are doing, it's a bit like running a very bad simulation, those simulations likely vary between people, and I see no reason to think people are reliably converging on conclusions based on imagining the same sorts of things. I don't think anyone can accurately model either scenario to render any meaningful sort of judgment, and even if they did, I think this would only tell us about that person's psychology; it wouldn't tell us anything about what's true in any deeper sense.
There's a far simpler response to the repugnant conclusion than to try to address it: I simply deny there's anything repugnant in the first place. I don't find it repugnant. And I take whether something is repugnant or not to be a matter of individual psychology. Since I don't find it repugnant, I simply don't think there's a problem to solve.
"[...]I distrust people's intuitions (even my own!) about "lives barely worth living". I think we use that expression and similar ones in natural language to describe lives that actually aren't worth living. (But we don't say that, because it's taboo.)
I'm far from an antinatalist, but I think Benatar is partially correct, to the extent that there are many more net negative lives than we ordinarily admit, or, to put it another way, if we were to observe a bunch of lives that are *actually* barely worth living, this wouldn't strike our intuitions as repugnant at all."
Yes J agreed, our intuitions shouldn’t always give us valid conceptions of how things are. Just because we consider a situation to be “repugnant”, that shouldn’t mean it’s actually bad for that experiencer itself. It’s actually even defined to be good for it. But then I base this on the deeper concept of value itself rather than our various notions of “rightness” versus “wrongness” (which apparently concern social norms and whatnot, though the jury’s still out). Philosophers seem to like the concept so I guess it’s fine for them. Still it does surprise me that so many moral antirealists (which is what I am) would remain so committed to the institution of morality anyway. Why not consider things from the more fundamental concept of value existing as a defined subject? Whatever it is that prevents them from going deeper, I suspect this also helps keep the field of psychology so primitive.
That’s certainly one reasonable way of dealing with the repugnant conclusion Lance — deny that such universal repugnance exists. But what if the psychology of people in general, and even yours, were to foster strong general repugnance here? Would we then need to concede a point to Derek Parfit or other moral realists? I don’t think so.
My approach is to dismiss the usefulness of our various moral notions in general. I instead base my assessments on the value of existing in itself. Notice that there are lots of things that we consider “good” in some defined sense, but also repugnant. Why? Because the realities of what’s good can be horrible regardless of our various “rightness versus wrongness” notions.
Ultimately I think the problem is that philosophers reside over the entire domain of value, and yet have been too timid to explore what lies deeper than our various moral notions. This is one of the reasons I believe a new breed of philosopher will need to emerge.
"You have three options: tell some crazy story as to why Trade Small-Quality for Big-Quantity is false, deny Transitivity (insane!), or accept the Repugnant Conclusion (gross!). There is literally no other option."
Not quite true. You can also reject the idea of a complete preference ordering over worlds. Why do we need to prefer one world over the other anyway? The idea that all possible worlds can necessarily be ranked implicitly assumes some form of consequentialism.
Well, to be clear, you don’t need a complete preference ordering over all worlds, just over the ones relevant to the spectrum. And I tried to state Trade Small-Quality for Big-Quantity as a proposition that *entailed* a preference ordering over the worlds (because you do need that, as you point out). I think I succeeded but if I did not (or if it’s just too unclear to tell since this was all an informal exposition) just add in that as a fourth proposition.
But, to get to your substantive point, the idea that we need to prefer one world over another does not at all assume some kind of consequentialism. Any plausible moral theory should have something to say about what we should want to be the case. It strikes me as absurd to say one has no need (at all!!) to say which states of affairs are more choiceworthy than others. Clearly, we should want it to be that 100 individuals live great lives than that 1000 live lives of torture. Those sorts of verdicts just are what a theory of the good tells you — as Rawls put it, any theory that cannot tell you what affairs are valuable is a theory not worth taking seriously.
I still think this is a fundamentally consequentialist way of reasoning. Non-consequentialist moralities simply do not rank worlds; ranking worlds by their properties, and assigning higher moral value to some than to others, is (as far as I can tell) basically the definition of consequentialism.
That doesn't mean non-consequentialist moralities have nothing to say about which worlds are more *choiceworthy*. If an individual in real life were to actually face the choice between the creation of torture-world and happy-world, many deontologies would prefer that he create happy-world, because creating unhappy beings is considered *in se* a bad act in the same way as causing direct harm. Other deontologies may not consider the creation of new beings to be within the realm of moral consideration at all, regardless of their happiness; you cannot harm or help a creature that does not exist yet. Either way, the worlds are not being ranked; rather, the acts of the individual are being ranked.
It isn’t for multiple reasons. First, non-consequentialist moralities do have something to say about what states of affairs are preferable: simply because states of affairs include actions! Deontologists prefer, eg, the state of affairs where I refrain from killing 1 to save 5 to the state of affairs where I don’t. Second, most, if not all, non-consequentialist moral theories have *theories of the good* or *theories of value*. These are theories of what is intrinsically valuable. Indeed, the way deontology is *defined* is as the view that, whatever our theory of the good is, the good cannot be the *only* thing relevant to determining right action. You should sometimes act in a way that minimizes total goodness because individuals have rights. In fact this is how absolute vs moderate deontology are distinguished: moderate deontologists, who say you can violate rights to save, eg, all of humanity, cast their view in terms of a background theory of goodness/value: you cannot kill someone, unless there is unbelievable amounts of good at stake.
With all that said, the Repugnant Conclusion and the Spectrum Argument are puzzles about total goodness: which world, A or Z, has more value? As an intuition pump for moderate deontologists: if I had to kill someone to maintain world Z’s existence, is it permissible to kill them?
The idea that all possible worlds can be ranked doesn't assume consequentialism. It would only assume consequentialism if you added onto this the assumption that you should always aim to bring about the highest-ranked possible world. Non-consequentialists agree with consequentialists that certain actions would have the best consequences - they just disagree on whether that means we should perform those actions.
You are right, though, about the general point that the repugnant conclusion assumes that you can compare the value of worlds with different population sizes. Although that doesn't need to be added as an additional premise - the first premise assumes it. So, if you deny that, it is one method for avoiding the repugnant conclusion. But denying it has a lot of counterintuitive consequences. For example, a world in which people are brought into existence solely to be tortured for their entire lives wouldn't be worse than a world in which this didn't happen. Also, it's going to imply that for virtually any action that you ever take, there is no option that produces the best consequences - the consequences will always be incomparable since they lead to slightly different population sizes in the distant future. This makes it impossible to do pro and con reasoning for any moral decision (and this doesn't assume consequentialism either. Other moral theories still take consequences into account - you'd have to be crazy not to).
I liked this post and I see what you’re gesturing at but I think the practical dynamics are very strange. It would be strange if pain is relational and not additive in the context of the global utility function. Though I share that the asymptote of total pain in the universe of a specific action could in principle be there, and even intuitively seems like it should be there, it seems like a big issue to have the global pain utility function not depend on subjective (additive) pain and instead interrelational pain.
That being said this post did make me think it’s plausible there is some sort of “special relativity analogue” where there is a sort of maximum pain bound ( like the speed of light). In this context it’s plausible that like in Newtonian relativity we say small velocities appear additive then small amounts of pain could be additive too. Perhaps the story is similar when you get two really painful objects where just like two photons moving away from each other at c is still only moving away from the other photon at c, rather than 2c, then relational pain could asymptote the same way.
In my view, if a version of consequentialism have a lexical threshold theory of aggregation, then it is not utilitarianism. Yet this is just a verbal dispute that does not matter.
There is a different way of arriving at the Repugnant Conclusion, which is much stronger, IMO.
Let's say we have a starting population A, which consists of 1000 people at welfare level 50. Let population A* be a population consisting of the very same 1000 people at welfare level 50, but with 1 extra person at welfare level 1 (which still indicates a life worth living). It seems like A < A*, because adding an extra person with a life worth living makes the world better.
Then, consider a world B, where we have the same 1001 people from A*, but where the welfare is distributed in an egalitarian way. So instead of having 1000 people at welfare level 50 and 1 person at welfare level of 1, B contains 1001 people, all at welfare level of ~49.9. It seems like A* < B, because B has shares all the same people with A*, and it contains just as much welfare as B, except for the fact that the welfare is distributed in a more fair/egalitarian way.
If we keep doing the procedure of adding one person with a life worth living and then redistributing the welfare in an egalitarian way, we will eventually arrive at the world Z, and by transitivity we will be forced to conclude that A<Z, which is the Repugnant Conclusion.
And this problem is hard to escape. Even if you bite the bullet on the fact that adding people with lives worth living doesn't make the world better (so A=A*) and another bullet on the fact that redistributing welfare in an egalitarian way doesn't make the world better (so A*=B), you would still only arrive at the conclusion that A is merely as good as Z (A=Z). But from my experience, most people who find Repugnant Conclusion to be a serious problem don't just think that A=Z, their pre-theoretical intuitions are strongly that A>Z. But in order to achieve that, one would either need to argue that adding people with lives worth living makes the world worse (which is crazy) or that redistributing welfare in an egalitarian way makes the world worse (which is crazy). No easy way out of this.
Yeah this is mere addition: bounded aggregation rejects the egalitarian step. Things can get worse by redistributing equally because you lower many people from a certain welfare level (50) that might be *much* better than a slightly lower one (49.9). Of course, that sounds unintuitive, but notice it is the exact unintuitive result I discuss in this article and try to explain away.
Ah, I see. I do think bounded aggregation would run into a different problem, then. This one is akin to the Egyptology objection against average utilitarianism.
Suppose that the cap on well-being from lives slightly worth living is 10,000 utils. You are presented with a big red button. This button, if pressed, creates Alex, with a life slightly worth living. You are informed in advance about every detail of Alex's life. Suppose that if you don't press the button, you get a free lollypop. Being a consequentialist, after reviewing Alex's life, you decide that pressing the button would be better than getting a lollypop. But as you are about to press it, an astrophysicist storms into your room. He informs you that the newest space station allowed him to discover a distant planet that is populated by aliens whose lives are slightly worth living, whose sum of lives add up to 10,000 utils.
Upon learning this information, it seems like you would completely change your mind and decide to not press the button for the lollypop. But this seems very wrong, because, we can stipulate, Alex would never causally interact with people on that planet. Their well-being is completely irrelevant to his well-being. Intuitively, the choice of bringing Alex into the existence shouldn't depend upon the well-being of aliens located potentially billions of light-years away from him, but rather on how good or bad his own life is.
No, because bounded aggregation doesn’t entail that there is a cap on well-being. It just entails that the total value for any particular kind of burden asymptotically approaches a bound. Broadly speaking, lexical threshold views face a choice: they can either violate separability (and thus face the Egyptology objection) or they can accept that some goods that are only slightly better than others are lexically superior (Strong Lexicality). See this paper if you’d like to see the proof: https://philpapers.org/archive/THOADF-4.pdf
Hmmmm, perhaps I am misunderstanding you (sorry if that's the case). In the original post you say that you think there is an upper limit to how good a population of lives barely worth living can be.
Let's say that this limit is 1000 utils. Let's also say that if the population consists of just one person leading a life barely worth living, then it would count for 1 util. To avoid the conclusion that a population made up of, say, 2000 lives would count for more than 1000 utils, you either have to 1) make utility of every person after the 1000th one count for 0 utils; 2) make utility of every subsequent person after the first one count for less than 1 util (perhaps asymptotically approaching 0 utils); 3) make *everyone* count for less utils, each time a new person is added to the population.
It seems that in all 3 cases, how much a person's life counts for, crucially depends on the existence of other people (and so the Egyptology problem reappears).
That’s not right. I explicitly propose (2). But that alone though doesn’t lead to violations of separability (Egyptology problems) because it might be (and, indeed, this is what most bounded aggregationists propose) that *all* value is context sensitive in this way. The universe is like one large geometric series. What you have to show isn’t merely that the value of particular lives is dependent on others; it’s that that would lead to *ordinal* differences, ie, differences in how you would rank various states of affairs. It can be that: *whether* I should A or B does not change depending on what else is out in the universe, even if the *particular amounts of value* brought about by A and B are both different depending on what else is out in the universe.
If I understand it correctly what you mean by ordinal differences, then I think it can be demonstrated using something like my original thought experiment with pressing a button.
Suppose that pressing the button creates Alex, whose util contribution would range from 1 to ~0 depending on whether he is the first person or one of the subsequent ones. And suppose that by not pressing the button you would get a lollipop, which would give you 0.6 utils.
It seems like the choice of whether to press the button (and consequently, the decision about which world is better between the Lollipop-world and the Alex-world) depends on the existence of other people (like distant aliens). If no other people with lives barely worth living exist, then you should press the button (because you gain 1 util as opposed to 0.6). If 2000 people with lives barely worth living exist (which would mean that Alex's life contributes, say, 0.01 util), then you shouldn't press the button (because you would only get 0.01 util as opposed to 0.6).
I am not exactly sure whether problems with Egyptology entail violations of Separability, but it seems like at least my example above has problems with what I take to be the most standard formulation of Egyptology ("the choice you make depends upon the existence of aliens in distant galaxies").
All the better! One less problem for you. Although: do you think a billion instances of torture could be outweighed by some number of dust specks? Similar problem/considerations apply there
But, in essence, aren't you just accepting the repugnant conclusion? Parfit, after all, grants that you eventually hit a Z-world: one in which harm along one axis can no longer be offset by benefits along another. And your only disagreement with him, as far as I can tell, is that you think that the Z-world is reached sooner than he thinks.
The repugnant conclusion is the proposition that a world with lots of people living lives barely worth living is better than a world with less people living perfect lives. I do not accept that. I maintain almost the opposite view: there’s a population with less people but more average well being that is better than a population with more people but less average well being
Sure, but Parfit would accept your conclusion provided that the world with more people but less average well being had an average well being at or below the “life worth living” threshold. So, basically, your position is that there comes a point at which adding more people, no matter how many, isn’t worth *any* decrease in average well being. For Parfit that’s true once we reach the “life is just barely worth living” threshold. And, for you, it’s also true but just a bit sooner, no?
Just a few notes (for clarity, perhaps just for myself). *Parfit* did not accept the Repugnant Conclusion. He considered it a deep paradox, but did not accept it. He held out hope for some solution. Additionally, the Repugnant Conclusion is the view that there *is* some amount of people you can add to improve things, as long as the lives are worth living. So, my solution: rejects the RC (it is the opposite conclusion) and might be something Parfit would like to accept. We will never know :(
Also: Parfit agreed that at the life-barely-worth-living stage, A is always better no matter how many people are in Z. But that’s not a *solution*. That is just to assert the negation of the RC. A *solution* is some independent theory that rejects one of the premises in the Spectrum Argument above.
Ah, okay, that's helpful. I think we may have been talking past each other. I was using "repugnant conclusion" to refer to the idea that it's possible for a situation to get progressively better and better but, in the end, be worse than it was originally. And your position, I think, is compatible with that possibility.
I'd add, though, that it's premature to seek a solution because the problem hasn't yet been fully understood. Parfit thought that he had discovered a weird paradox about goodness: that you can make things worse by progressively making them better. But the repugnant conclusion, I'd argue, has nothing to do with goodness per se and everything to do with analysis (i.e. the method of reducing wholes to parts). You can run similar spectrum arguments for all kinds of relations, moral and non-moral alike (e.g. is stronger than, is healthier than, is clearer than, etc.).
“The pain is extremely slight: just barely noticeable. But it exists. The world is slightly worse because of it.”
I just don’t see how you take this as a given. I don’t think you can automatically take any and all instances of pain and say they make the world worse. They’re literally part of the fabric of reality. When I experience pain (and I have experienced a lot) I have never just taken it as a given that it’s existentially or universally bad. It just seems like an unrealistic position that never gets properly defended in these thought experiments.
What do you mean “existentially or universally” bad? All I mean to say is that pain is bad. That does not mean that it is *really* bad, or that it doesn’t contribute to various good things such that in many contexts it’s worth having. All I mean is that if we could remove some instance of pain, and maintain all of the consequences of it with 0 downsides whatsoever, we should do that, because pain in and of itself is not a good thing.
The trick isn't to solve the Spectrum Argument by accepting both it and the Repugnant Conclusion. The trick is to solve the RC by rejecting the Spectrum Argument!
The stronger versions of the argument for the RC that I've seen don't actually assume Trade Small Quality for Large Quantity as an axiom. Instead, they assume these two axioms, from which TSQLQ follows, given transitivity:
Addition: Adding extra people with net positive lives to a population, with no other effects, makes the world better.
Equality: If two worlds, W1 and W2, both have the same population with the same average wellbeing, but wellbeing is more equally distributed in W2 than W1, then W2 is at least as good as W1.
To see how these two entail TSQLQ, consider a world A with 100 people living net-positive lives - let's call their utility 100. Then we can imagine B* which is exactly like A except that we add 100 more people with utility 99. And finally, imagine B, containing 200 people each with utility 99.5. Addition says that B*>A, and Equality says that B≥B*. Transitivity says that therefore B>A. We can plug in different numbers to get all the comparisons from A through Z and end up with the Repugnant Conclusion.
Equality is a very weak premise - it's agreed to both by people who put no fundamental value on equality or diminishing returns to wellbeing (e.g. utilitarians) but also by people who fundamentally value equality and people who think improving the wellbeing of the worse-off matters more than that of the well-off. The only way to deny it would be to say that equality can be inherently bad, or that the marginal wellbeing of the worse-off matters less than that of the well-off, or that worlds with exact same population and average wellbeing can be incomparable. Also, even if it is denied, it can be significantly weakened and the repugnant conclusion will still follow - you just have to believe that there's some possible set of populations and some possible distributions of wellbeing where every step of the argument works. So people who deny Equality because they think some people inherently deserve higher wellbeing are still not going to be able to deny the RC this way.
Addition is the more contentious one, but if you already think TSQLQ is intuitive, Addition should be much more intuitive. Here, you don't have to trade off anything at all to get the extra quantity - you just add more happy lives. And you can't get around it by saying that the value of a happy life asymptotically approaches zero - it has to actually be zero (or less, or incomparable) at some step of the argument. In fact, if you want to conclude that A > Z, rather than A = Z, and still accept Transitivity and Equality, you have to argue that at some point adding extra lives that are net-positive actually makes the world worse.
I would be interested to hear which of these premises you deny (assuming you're still not convinced by this stronger version of the argument) and if you have any reasons that make the denial more intuitive.
I would also say that I think the argument you gave about there being two different dimensions of value - quality and quantity - is not quite right. Quantity has no value in and if itself - it's only good if the extra people you're adding have good lives, and bad if they have bad lives. So it can't possibly be an independent dimension from quality. But more importantly, it seems to me that there are actually way more than two dimensions - the correct way to think of the value-space for a population is to treat each individual person's wellbeing as its own dimension, since everyone's wellbeing is valuable for its own sake. But if you do this, it doesn't really seem like you can say that quantity matters less if there's already a lot of it - that means that you are saying that individual people are less valuable just because a lot of other people already exist. This would mean nobody has inherent value - your value is dependent on the number of other people who exist. But this conclusion itself seems just as repugnant as the one we are trying to avoid.
"You have three options: tell some crazy story as to why Trade Small-Quality for Big-Quantity is false, deny Transitivity (insane!), or accept the Repugnant Conclusion (gross!). There is literally no other option."
Accept that there is an inflection point past which trading small-quality for big-quantity is no longer worth it, which does not violate transitivity. The fact that this would be difficult to define rigorously seems like a bad reason to assume it does not exist.
I don’t think the difficulty is that we can’t define it rigorously. It is that each pair, even the ones in the middle, really do seem better and better. It does seem better to have 50 trillion times as many people exist who are experiencing almost the same level of well-being as the prior population. But that is what we’ll have to say!
What happens when you zoom out, comparing not adjacent quantities but the log 25th and 75th percentile of the distribution you've created? 100 quadrillion people living below poverty level lives versus to 1 billion people living modestly to moderately good lives - does either obviously stand out as better? If so, why?
Well, whether one stands out as better or not, the puzzle here is that we seem to have really strong views about each individual pair in the sequence. And the trouble is that that alone, alongside transitivity, gets us into trouble.
I notice that you choose not to answer whether one is better or not, why? Wouldn't transitivity require that you be able to compare any two intermediate qualities - not just adjacent ones - and determine which is better, whether you're looking at 10 trillion vs 100 trillion or quintillions compared to billions? What does transitivity mean if it requires adjacent pairings to be relevant?
You asked whether one “obviously” stood out as better than the other. My point is that that is going to be orthogonal to the main issue at play concerning the Repugnant Conclusion. Because it does not require that one clearly stand out as better than the other.
I didn’t claim that there *isnt* one that is better. There surely is. Assuming you specify the values correctly, I would think the one later down in the sequence is better, but it’s not as though this is clear or obvious or anything.
" Because it does not require that one *clearly* stand out as better than the other." is hard to square for me with your assertion from the prior comment - "Well, whether one stands out as better or not, the puzzle here is that *we seem to have really strong views* about each individual pair in the sequence. And the trouble is that *that alone, alongside transitivity*, gets us into trouble."
Is it important that we have strong feelings - draw obvious conclusions - about what goodness is in this discussion, or is it orthogonal to the trilemma you lay out in this post?
My understanding based on your comments is that transitivity alone actually is insufficient to be represented in its current position in your tripartite dilemma, and must instead be caveated with statements about adjacency, intuition, etc. That does make for a much less difficult logical question to engage with than a mathematically and rationally strict idea like transitivity, however. Have I misunderstood your post or replies?
Do you think that given a world with one person slightly above the threshold, adding one slightly below is bad? If yes, I think Valdman is right and you just have a different threshold for "life worth living". If no, then theres a choice where doubling the population in both worlds flips your decision.
By “threshold” do you mean threshold for “life worth living”? If so, I take it that adding a person slightly above and slightly below keeps the world’s total value exactly the same.
By "threshold" I mean the highest level of utility that you think can not outweigh a slightly higher level just by having more people. And I dont mean adding one person above and one below, I mean there already is one above, and is it good to add one thats slightly below?
Ok, so if Im understanding it correctly, you think there are some thresholds where people above are lexically prefered to people below, but those below can still be positive.
In that case, you will contradict individuals choices about things that effect only them. Eg if Alice is ε utils above a threshold, and 3ε below the next one, and she can take a gamble 10% -2ε 90% +2ε, then you would want to stop her from doing that.
Actually, we’d probably be best to not introduce risk here — I think risk and bounded aggregation is a really tricky issue. I’m a fanatic and generally a huge fan of expected value theory (in finite contexts at least), and BA basically entails that some populations are infinitely more valuable than others.
Additionally, though, note my solution quantified over *worlds*. So it can’t merely be an individuals decision to improve their own well-being — it has to be along both dimensions (their well being, and the number of people involved).
I dont think the worlds avoid anything here? We could always say that Alice is the only person in that world. Alternatively, if there is a large number of people all offered the bet with independent randomness, then the aggregate result is quasi-deterministic, and you again think their choice makes things worse.
We could even think of a version where the people unanimously agree to redistribute their resources in such a way that 90% gain 2ε and 10% lose 2ε, with the ratio fixed and randomness only determining who the winners and loser are.
Generally, I think expected value theory forces your treatment of risk based on your treatment of riskless comparisons, and all these problems can be created in a finite context.
I'm not a philosopher or an economist so my comment may be misinformed, irrelevant, and if so, mildly embarrassing, but here it is anyway.
Your solution to the repugnant conclusion reminded me of the economic law of diminishing returns—as one input in a production process is increased, while other inputs are held constant, the marginal increase in output will eventually decrease.
I imagine it works out like this:
Population A is starting at the end of Stage 1, where welfare per individual is at it's peak. As you work toward population Z, increasing the number of people leads to a decrease in welfare per individual. This is still producing a positive output (in populations prior to Z), because, even after the law of diminishing marginal returns kicks in, and welfare per individual declines, average welfare continues to increase because the individual welfare still exceeds the average. Once the welfare per individual is equal to the average total welfare, then our population would be in the middle of Stage 2.
The repugnant conclusion seems to take place at the end of Stage 2, the point where the total number of people is at it's peak, welfare per individual is at it's lowest, and average welfare is declining. This would be the moment before moving into Stage 3 where individual welfare would become negative.
If we think of individuals and their welfare as inputs to the productivity of life then it would make sense that, holding welfare as a fixed variable and constantly dividing it among more and more individuals would reduce life productivity for everyone once welfare is decreased beyond a particular point.
IDK, maybe this isn’t a proper application of the law.
There is a way easier answer to this puzzle; human intuitions have evolved to be useful cues in "environment of evolutionary adaptedness" (EEA). They are not going to be useful in hypothetical possible worlds immensely different than EEA and deciding fates of two planets (one with a hundred thousand habitants and the other with one trillion or whatever huge number you choose) is definitely not something any human in EEA had ever to do. This thought experiment is misusing a tool evolved to be useful in environment A in the incomparable environment B, like for example using a kitchen scale to weight the sun. The whole enterprise is fried
Many women accept level 10 pain (or at least the risk of it) for the pleasure of having a baby.
Nobody (OK, surrogates excepted!) would accept level 10 pain for somebody else to have a baby. So there's something fundamentally wrong in the whole trade-off thing.
I don't personally consider a world with a million people living grade X lives any better or worse than one with a billion living similar lives.
I always found the solution relatively obvious. People have diminishing marginal utility in their moral views, just as they do in their economic views.
We obviously value two things. Quality, and quantity. We can place a unit of value on each, and then compare the two to see if we’d be willing to sacrifice a huge amount of quality, for a large amount of quantity, but absent some level of discounting, you’ll get repugnant conclusions.
In microeconomics, a repugnant conclusion would be something like:
Option A: One week at a luxurious resort. You’ll have 10/10 enjoyment
Option B: 10 weeks of mildly pleasant stays in budget motels. Each week is a 2/10 experience.
Total enjoyment with option B is double option A, so it seems like you’d choose it. Fortunately we have complex values, and the more we have of something, the less we value more of it. The same can be said of moral terms. When we have conflicting values, it’s never a case of a linear total comparison. It’s a case of solving an optimization problem with hard constraints, as is the case in economics (and even then, these are just approximations of value, not the actual determinant itself).
This is also an argument against utility monsters. Would you value a millisecond of the most intense pleasure possible, a billion times better than anything anyone has ever experienced, or a highly-positive lifetime in the normal human sense?
Yes, although notice an important difference in the context of ethics and value that makes discounting much less plausible/obvious than in the economics case: presumably, there is no difference in *moral value added to the world* between one person enjoying a good life and another person enjoying a good life. While there plausibly is when it comes to me enjoying my 3rd day at a hotel. My faculties are limited in this way such that further days of joy are not created equal. But not at all clear that more *people* living the same lives should have less contributive moral value. And, as I say in the post, the biggest problem for these views is that they imply this odd fact that a world with people living only *slightly* worse lives could never have more value than a world with *50 trillion times less people*.
I think the disagreement is that economics is a theory for assigning value under constraints, while moral hypotheticals (those divorced from reality) are constructed without any constraints. In a moral hypothetical, the tradeoff between quantity and quality can be set at any arbitrary ratio, so no matter what level of discounting you imagine, there's always some ratio, with some large number of barely-worth-living lives, where the optimization breaks down.
But in applied economics, as in applied morality, we aren't considering quality and quantity in a vacuum. More quantity does mean less quality in some ratio, since there's only a finite number of atoms in the universe, and more practically, only a finite amount of resources and time available to do good things with.
The difference between the limitless scaling that moral hypotheticals like yours posit and the real world is that there's a fundamental part of reality that forces the first additional person and the 10^25th to have different moral value. Existence implies the consumption of resources (and therefore some tradeoff with quality of life). Once you start approaching the carrying capacity of the environment - whether that's a small island or the entire universe - each additional person necessarily reduces the average quality of life, however slightly. At some point, adding one more person, while valuable in isolation, subtracts value from everyone else in terms of marginally reduced quality, due to scarcer resources.
Essentially, just as in economics, ignoring cost gives you infinite demand curves. Ignoring physical constraints, where each person must consume some amount of resources to live a certain quality of life, gives you repugnant conclusions.
> the biggest problem for these views is that they imply this odd fact that a world with people living only *slightly* worse lives could never have more value than a world with *50 trillion times less people*.
I think the assumption here is that value scales linearly and indefinitely. If you stack up enough barely-worth-living lives, it must eventually outweigh a smaller group living slightly better lives. But this assumes constant moral returns on each additional life, which isn’t realistic. In any constrained system (planetary, ecological, social) scaling quantity degrades the conditions that support quality. So yes, it’s entirely possible that after a certain point, more lives just don’t add net value, even if they’re technically above the “barely-worth-living” line.
Apply this to the actual planet. Would it be better to have 50 trillion people living slightly worse lives, or 10 billion living slightly better ones? In a vacuum I would say the 50 Trillion, but in reality I would say the 10 Billion.
50 trillion people couldn’t live “only slightly worse” lives under real-world conditions. That would mean roughly 10 square meters per person—closer to 3 if you exclude oceans. The result wouldn't be 50 trillion lives at welfare level 99.9—it would be 50 trillion net-negative lives, starving and collapsing the biosphere. It seems it would be intuitive for the 10 billion better lives over the 50 trillion imploding ones.
The real question is: how much are we willing to trade off quality for quantity? And the answer isn’t fixed. It depends entirely on what level of quantity we can sustain before we start meaningfully sacrificing quality. Once you hit that threshold, more is worse. If you tell me "actually there is no threshold because we invented space travel and nuclear fusion so the only sacrifice of 50 Trillion lives is having slightly more light pollution from all the O'Neill cylinders in orbit" this moral framework would favor the 50 Trillion lives, despite applying a level of discounting, since the rate of discounting is dependent on the additional cost of each additional life lived.
I'm sympathetic to your argument here, but it can be avoided by doubling the number of planets along with the population. It's not 50 trillion people on/around one planet, but 5000 planets each with 10 billion people on it. Surely a universe with exactly 5000 planets, each with 10 billion moderately happy people on it, is slightly better than a universe with exactly one planet, even if the 10 billion people on it are very slightly happier than the people in the other universe. And since no one world is any more stressed than any other, your objection does not apply.
Their point is that the level of happiness/utility/whatever is stipulated, and the other facts about the world must be adjusted in whatever way necessary to support those numbers.
I appreciate articles like this one but I do think your case may be overstated. You say:
"And, contra what many people seem to think, avoiding the Repugnant Conclusion is everyone’s problem. It is not something that only arises if you are a utilitarian."
It's not everyone's problem. It's not a problem for me.
"The answer seems blindingly obvious: you should hope A is the case! It is preferable to Z!"
Who does it seem blindingly obvious to? It does not seem obvious to me. Neither answer seems obvious to me. I think the thought experiment involves underdescribed and highly complex scenarios that nobody could accurately model. Whatever people are doing, it's a bit like running a very bad simulation, those simulations likely vary between people, and I see no reason to think people are reliably converging on conclusions based on imagining the same sorts of things. I don't think anyone can accurately model either scenario to render any meaningful sort of judgment, and even if they did, I think this would only tell us about that person's psychology; it wouldn't tell us anything about what's true in any deeper sense.
There's a far simpler response to the repugnant conclusion than to try to address it: I simply deny there's anything repugnant in the first place. I don't find it repugnant. And I take whether something is repugnant or not to be a matter of individual psychology. Since I don't find it repugnant, I simply don't think there's a problem to solve.
I still like the phrasing I wrote here a few years ago:
https://open.substack.com/pub/fakenous/p/in-defense-of-repugnance?utm_campaign=comment-list-share-cta&utm_medium=web&comments=true&commentId=12337907
"[...]I distrust people's intuitions (even my own!) about "lives barely worth living". I think we use that expression and similar ones in natural language to describe lives that actually aren't worth living. (But we don't say that, because it's taboo.)
I'm far from an antinatalist, but I think Benatar is partially correct, to the extent that there are many more net negative lives than we ordinarily admit, or, to put it another way, if we were to observe a bunch of lives that are *actually* barely worth living, this wouldn't strike our intuitions as repugnant at all."
Yes J agreed, our intuitions shouldn’t always give us valid conceptions of how things are. Just because we consider a situation to be “repugnant”, that shouldn’t mean it’s actually bad for that experiencer itself. It’s actually even defined to be good for it. But then I base this on the deeper concept of value itself rather than our various notions of “rightness” versus “wrongness” (which apparently concern social norms and whatnot, though the jury’s still out). Philosophers seem to like the concept so I guess it’s fine for them. Still it does surprise me that so many moral antirealists (which is what I am) would remain so committed to the institution of morality anyway. Why not consider things from the more fundamental concept of value existing as a defined subject? Whatever it is that prevents them from going deeper, I suspect this also helps keep the field of psychology so primitive.
Here’s a post on value from me: https://eborg760.substack.com/p/i-value-therefore-i-am
That’s certainly one reasonable way of dealing with the repugnant conclusion Lance — deny that such universal repugnance exists. But what if the psychology of people in general, and even yours, were to foster strong general repugnance here? Would we then need to concede a point to Derek Parfit or other moral realists? I don’t think so.
My approach is to dismiss the usefulness of our various moral notions in general. I instead base my assessments on the value of existing in itself. Notice that there are lots of things that we consider “good” in some defined sense, but also repugnant. Why? Because the realities of what’s good can be horrible regardless of our various “rightness versus wrongness” notions.
Ultimately I think the problem is that philosophers reside over the entire domain of value, and yet have been too timid to explore what lies deeper than our various moral notions. This is one of the reasons I believe a new breed of philosopher will need to emerge.
"You have three options: tell some crazy story as to why Trade Small-Quality for Big-Quantity is false, deny Transitivity (insane!), or accept the Repugnant Conclusion (gross!). There is literally no other option."
Not quite true. You can also reject the idea of a complete preference ordering over worlds. Why do we need to prefer one world over the other anyway? The idea that all possible worlds can necessarily be ranked implicitly assumes some form of consequentialism.
Well, to be clear, you don’t need a complete preference ordering over all worlds, just over the ones relevant to the spectrum. And I tried to state Trade Small-Quality for Big-Quantity as a proposition that *entailed* a preference ordering over the worlds (because you do need that, as you point out). I think I succeeded but if I did not (or if it’s just too unclear to tell since this was all an informal exposition) just add in that as a fourth proposition.
But, to get to your substantive point, the idea that we need to prefer one world over another does not at all assume some kind of consequentialism. Any plausible moral theory should have something to say about what we should want to be the case. It strikes me as absurd to say one has no need (at all!!) to say which states of affairs are more choiceworthy than others. Clearly, we should want it to be that 100 individuals live great lives than that 1000 live lives of torture. Those sorts of verdicts just are what a theory of the good tells you — as Rawls put it, any theory that cannot tell you what affairs are valuable is a theory not worth taking seriously.
I still think this is a fundamentally consequentialist way of reasoning. Non-consequentialist moralities simply do not rank worlds; ranking worlds by their properties, and assigning higher moral value to some than to others, is (as far as I can tell) basically the definition of consequentialism.
That doesn't mean non-consequentialist moralities have nothing to say about which worlds are more *choiceworthy*. If an individual in real life were to actually face the choice between the creation of torture-world and happy-world, many deontologies would prefer that he create happy-world, because creating unhappy beings is considered *in se* a bad act in the same way as causing direct harm. Other deontologies may not consider the creation of new beings to be within the realm of moral consideration at all, regardless of their happiness; you cannot harm or help a creature that does not exist yet. Either way, the worlds are not being ranked; rather, the acts of the individual are being ranked.
It isn’t for multiple reasons. First, non-consequentialist moralities do have something to say about what states of affairs are preferable: simply because states of affairs include actions! Deontologists prefer, eg, the state of affairs where I refrain from killing 1 to save 5 to the state of affairs where I don’t. Second, most, if not all, non-consequentialist moral theories have *theories of the good* or *theories of value*. These are theories of what is intrinsically valuable. Indeed, the way deontology is *defined* is as the view that, whatever our theory of the good is, the good cannot be the *only* thing relevant to determining right action. You should sometimes act in a way that minimizes total goodness because individuals have rights. In fact this is how absolute vs moderate deontology are distinguished: moderate deontologists, who say you can violate rights to save, eg, all of humanity, cast their view in terms of a background theory of goodness/value: you cannot kill someone, unless there is unbelievable amounts of good at stake.
With all that said, the Repugnant Conclusion and the Spectrum Argument are puzzles about total goodness: which world, A or Z, has more value? As an intuition pump for moderate deontologists: if I had to kill someone to maintain world Z’s existence, is it permissible to kill them?
The idea that all possible worlds can be ranked doesn't assume consequentialism. It would only assume consequentialism if you added onto this the assumption that you should always aim to bring about the highest-ranked possible world. Non-consequentialists agree with consequentialists that certain actions would have the best consequences - they just disagree on whether that means we should perform those actions.
You are right, though, about the general point that the repugnant conclusion assumes that you can compare the value of worlds with different population sizes. Although that doesn't need to be added as an additional premise - the first premise assumes it. So, if you deny that, it is one method for avoiding the repugnant conclusion. But denying it has a lot of counterintuitive consequences. For example, a world in which people are brought into existence solely to be tortured for their entire lives wouldn't be worse than a world in which this didn't happen. Also, it's going to imply that for virtually any action that you ever take, there is no option that produces the best consequences - the consequences will always be incomparable since they lead to slightly different population sizes in the distant future. This makes it impossible to do pro and con reasoning for any moral decision (and this doesn't assume consequentialism either. Other moral theories still take consequences into account - you'd have to be crazy not to).
I liked this post and I see what you’re gesturing at but I think the practical dynamics are very strange. It would be strange if pain is relational and not additive in the context of the global utility function. Though I share that the asymptote of total pain in the universe of a specific action could in principle be there, and even intuitively seems like it should be there, it seems like a big issue to have the global pain utility function not depend on subjective (additive) pain and instead interrelational pain.
That being said this post did make me think it’s plausible there is some sort of “special relativity analogue” where there is a sort of maximum pain bound ( like the speed of light). In this context it’s plausible that like in Newtonian relativity we say small velocities appear additive then small amounts of pain could be additive too. Perhaps the story is similar when you get two really painful objects where just like two photons moving away from each other at c is still only moving away from the other photon at c, rather than 2c, then relational pain could asymptote the same way.
Hmm your analogy with special relativity and Newtonian mechanics is very clever. I want to think about that more
I was also hoping to explore this and make a response to your post. Perhaps we could cowrite something together :)
Sounds fun!
Good work. Would you classify this view as version of a lexical threshold, though not necessarily negative, utilitarianism?
It entails a lexical threshold, yep.
In my view, if a version of consequentialism have a lexical threshold theory of aggregation, then it is not utilitarianism. Yet this is just a verbal dispute that does not matter.
There is a different way of arriving at the Repugnant Conclusion, which is much stronger, IMO.
Let's say we have a starting population A, which consists of 1000 people at welfare level 50. Let population A* be a population consisting of the very same 1000 people at welfare level 50, but with 1 extra person at welfare level 1 (which still indicates a life worth living). It seems like A < A*, because adding an extra person with a life worth living makes the world better.
Then, consider a world B, where we have the same 1001 people from A*, but where the welfare is distributed in an egalitarian way. So instead of having 1000 people at welfare level 50 and 1 person at welfare level of 1, B contains 1001 people, all at welfare level of ~49.9. It seems like A* < B, because B has shares all the same people with A*, and it contains just as much welfare as B, except for the fact that the welfare is distributed in a more fair/egalitarian way.
If we keep doing the procedure of adding one person with a life worth living and then redistributing the welfare in an egalitarian way, we will eventually arrive at the world Z, and by transitivity we will be forced to conclude that A<Z, which is the Repugnant Conclusion.
And this problem is hard to escape. Even if you bite the bullet on the fact that adding people with lives worth living doesn't make the world better (so A=A*) and another bullet on the fact that redistributing welfare in an egalitarian way doesn't make the world better (so A*=B), you would still only arrive at the conclusion that A is merely as good as Z (A=Z). But from my experience, most people who find Repugnant Conclusion to be a serious problem don't just think that A=Z, their pre-theoretical intuitions are strongly that A>Z. But in order to achieve that, one would either need to argue that adding people with lives worth living makes the world worse (which is crazy) or that redistributing welfare in an egalitarian way makes the world worse (which is crazy). No easy way out of this.
Yeah this is mere addition: bounded aggregation rejects the egalitarian step. Things can get worse by redistributing equally because you lower many people from a certain welfare level (50) that might be *much* better than a slightly lower one (49.9). Of course, that sounds unintuitive, but notice it is the exact unintuitive result I discuss in this article and try to explain away.
Ah, I see. I do think bounded aggregation would run into a different problem, then. This one is akin to the Egyptology objection against average utilitarianism.
Suppose that the cap on well-being from lives slightly worth living is 10,000 utils. You are presented with a big red button. This button, if pressed, creates Alex, with a life slightly worth living. You are informed in advance about every detail of Alex's life. Suppose that if you don't press the button, you get a free lollypop. Being a consequentialist, after reviewing Alex's life, you decide that pressing the button would be better than getting a lollypop. But as you are about to press it, an astrophysicist storms into your room. He informs you that the newest space station allowed him to discover a distant planet that is populated by aliens whose lives are slightly worth living, whose sum of lives add up to 10,000 utils.
Upon learning this information, it seems like you would completely change your mind and decide to not press the button for the lollypop. But this seems very wrong, because, we can stipulate, Alex would never causally interact with people on that planet. Their well-being is completely irrelevant to his well-being. Intuitively, the choice of bringing Alex into the existence shouldn't depend upon the well-being of aliens located potentially billions of light-years away from him, but rather on how good or bad his own life is.
No, because bounded aggregation doesn’t entail that there is a cap on well-being. It just entails that the total value for any particular kind of burden asymptotically approaches a bound. Broadly speaking, lexical threshold views face a choice: they can either violate separability (and thus face the Egyptology objection) or they can accept that some goods that are only slightly better than others are lexically superior (Strong Lexicality). See this paper if you’d like to see the proof: https://philpapers.org/archive/THOADF-4.pdf
The latter option is what I take in this post
Hmmmm, perhaps I am misunderstanding you (sorry if that's the case). In the original post you say that you think there is an upper limit to how good a population of lives barely worth living can be.
Let's say that this limit is 1000 utils. Let's also say that if the population consists of just one person leading a life barely worth living, then it would count for 1 util. To avoid the conclusion that a population made up of, say, 2000 lives would count for more than 1000 utils, you either have to 1) make utility of every person after the 1000th one count for 0 utils; 2) make utility of every subsequent person after the first one count for less than 1 util (perhaps asymptotically approaching 0 utils); 3) make *everyone* count for less utils, each time a new person is added to the population.
It seems that in all 3 cases, how much a person's life counts for, crucially depends on the existence of other people (and so the Egyptology problem reappears).
That’s not right. I explicitly propose (2). But that alone though doesn’t lead to violations of separability (Egyptology problems) because it might be (and, indeed, this is what most bounded aggregationists propose) that *all* value is context sensitive in this way. The universe is like one large geometric series. What you have to show isn’t merely that the value of particular lives is dependent on others; it’s that that would lead to *ordinal* differences, ie, differences in how you would rank various states of affairs. It can be that: *whether* I should A or B does not change depending on what else is out in the universe, even if the *particular amounts of value* brought about by A and B are both different depending on what else is out in the universe.
If I understand it correctly what you mean by ordinal differences, then I think it can be demonstrated using something like my original thought experiment with pressing a button.
Suppose that pressing the button creates Alex, whose util contribution would range from 1 to ~0 depending on whether he is the first person or one of the subsequent ones. And suppose that by not pressing the button you would get a lollipop, which would give you 0.6 utils.
It seems like the choice of whether to press the button (and consequently, the decision about which world is better between the Lollipop-world and the Alex-world) depends on the existence of other people (like distant aliens). If no other people with lives barely worth living exist, then you should press the button (because you gain 1 util as opposed to 0.6). If 2000 people with lives barely worth living exist (which would mean that Alex's life contributes, say, 0.01 util), then you shouldn't press the button (because you would only get 0.01 util as opposed to 0.6).
I am not exactly sure whether problems with Egyptology entail violations of Separability, but it seems like at least my example above has problems with what I take to be the most standard formulation of Egyptology ("the choice you make depends upon the existence of aliens in distant galaxies").
I just don’t get why it should be obvious to me that A > Z.
All the better! One less problem for you. Although: do you think a billion instances of torture could be outweighed by some number of dust specks? Similar problem/considerations apply there
But, in essence, aren't you just accepting the repugnant conclusion? Parfit, after all, grants that you eventually hit a Z-world: one in which harm along one axis can no longer be offset by benefits along another. And your only disagreement with him, as far as I can tell, is that you think that the Z-world is reached sooner than he thinks.
The repugnant conclusion is the proposition that a world with lots of people living lives barely worth living is better than a world with less people living perfect lives. I do not accept that. I maintain almost the opposite view: there’s a population with less people but more average well being that is better than a population with more people but less average well being
Sure, but Parfit would accept your conclusion provided that the world with more people but less average well being had an average well being at or below the “life worth living” threshold. So, basically, your position is that there comes a point at which adding more people, no matter how many, isn’t worth *any* decrease in average well being. For Parfit that’s true once we reach the “life is just barely worth living” threshold. And, for you, it’s also true but just a bit sooner, no?
Just a few notes (for clarity, perhaps just for myself). *Parfit* did not accept the Repugnant Conclusion. He considered it a deep paradox, but did not accept it. He held out hope for some solution. Additionally, the Repugnant Conclusion is the view that there *is* some amount of people you can add to improve things, as long as the lives are worth living. So, my solution: rejects the RC (it is the opposite conclusion) and might be something Parfit would like to accept. We will never know :(
Also: Parfit agreed that at the life-barely-worth-living stage, A is always better no matter how many people are in Z. But that’s not a *solution*. That is just to assert the negation of the RC. A *solution* is some independent theory that rejects one of the premises in the Spectrum Argument above.
Ah, okay, that's helpful. I think we may have been talking past each other. I was using "repugnant conclusion" to refer to the idea that it's possible for a situation to get progressively better and better but, in the end, be worse than it was originally. And your position, I think, is compatible with that possibility.
I'd add, though, that it's premature to seek a solution because the problem hasn't yet been fully understood. Parfit thought that he had discovered a weird paradox about goodness: that you can make things worse by progressively making them better. But the repugnant conclusion, I'd argue, has nothing to do with goodness per se and everything to do with analysis (i.e. the method of reducing wholes to parts). You can run similar spectrum arguments for all kinds of relations, moral and non-moral alike (e.g. is stronger than, is healthier than, is clearer than, etc.).
Check this out if you're curious: https://zworld.substack.com/p/could-i-make-you-eat-shit
“The pain is extremely slight: just barely noticeable. But it exists. The world is slightly worse because of it.”
I just don’t see how you take this as a given. I don’t think you can automatically take any and all instances of pain and say they make the world worse. They’re literally part of the fabric of reality. When I experience pain (and I have experienced a lot) I have never just taken it as a given that it’s existentially or universally bad. It just seems like an unrealistic position that never gets properly defended in these thought experiments.
What do you mean “existentially or universally” bad? All I mean to say is that pain is bad. That does not mean that it is *really* bad, or that it doesn’t contribute to various good things such that in many contexts it’s worth having. All I mean is that if we could remove some instance of pain, and maintain all of the consequences of it with 0 downsides whatsoever, we should do that, because pain in and of itself is not a good thing.
The trick isn't to solve the Spectrum Argument by accepting both it and the Repugnant Conclusion. The trick is to solve the RC by rejecting the Spectrum Argument!
https://ramblingafter.substack.com/p/the-repugnant-conclusion-messed-with
The stronger versions of the argument for the RC that I've seen don't actually assume Trade Small Quality for Large Quantity as an axiom. Instead, they assume these two axioms, from which TSQLQ follows, given transitivity:
Addition: Adding extra people with net positive lives to a population, with no other effects, makes the world better.
Equality: If two worlds, W1 and W2, both have the same population with the same average wellbeing, but wellbeing is more equally distributed in W2 than W1, then W2 is at least as good as W1.
To see how these two entail TSQLQ, consider a world A with 100 people living net-positive lives - let's call their utility 100. Then we can imagine B* which is exactly like A except that we add 100 more people with utility 99. And finally, imagine B, containing 200 people each with utility 99.5. Addition says that B*>A, and Equality says that B≥B*. Transitivity says that therefore B>A. We can plug in different numbers to get all the comparisons from A through Z and end up with the Repugnant Conclusion.
Equality is a very weak premise - it's agreed to both by people who put no fundamental value on equality or diminishing returns to wellbeing (e.g. utilitarians) but also by people who fundamentally value equality and people who think improving the wellbeing of the worse-off matters more than that of the well-off. The only way to deny it would be to say that equality can be inherently bad, or that the marginal wellbeing of the worse-off matters less than that of the well-off, or that worlds with exact same population and average wellbeing can be incomparable. Also, even if it is denied, it can be significantly weakened and the repugnant conclusion will still follow - you just have to believe that there's some possible set of populations and some possible distributions of wellbeing where every step of the argument works. So people who deny Equality because they think some people inherently deserve higher wellbeing are still not going to be able to deny the RC this way.
Addition is the more contentious one, but if you already think TSQLQ is intuitive, Addition should be much more intuitive. Here, you don't have to trade off anything at all to get the extra quantity - you just add more happy lives. And you can't get around it by saying that the value of a happy life asymptotically approaches zero - it has to actually be zero (or less, or incomparable) at some step of the argument. In fact, if you want to conclude that A > Z, rather than A = Z, and still accept Transitivity and Equality, you have to argue that at some point adding extra lives that are net-positive actually makes the world worse.
I would be interested to hear which of these premises you deny (assuming you're still not convinced by this stronger version of the argument) and if you have any reasons that make the denial more intuitive.
I would also say that I think the argument you gave about there being two different dimensions of value - quality and quantity - is not quite right. Quantity has no value in and if itself - it's only good if the extra people you're adding have good lives, and bad if they have bad lives. So it can't possibly be an independent dimension from quality. But more importantly, it seems to me that there are actually way more than two dimensions - the correct way to think of the value-space for a population is to treat each individual person's wellbeing as its own dimension, since everyone's wellbeing is valuable for its own sake. But if you do this, it doesn't really seem like you can say that quantity matters less if there's already a lot of it - that means that you are saying that individual people are less valuable just because a lot of other people already exist. This would mean nobody has inherent value - your value is dependent on the number of other people who exist. But this conclusion itself seems just as repugnant as the one we are trying to avoid.
"You have three options: tell some crazy story as to why Trade Small-Quality for Big-Quantity is false, deny Transitivity (insane!), or accept the Repugnant Conclusion (gross!). There is literally no other option."
Accept that there is an inflection point past which trading small-quality for big-quantity is no longer worth it, which does not violate transitivity. The fact that this would be difficult to define rigorously seems like a bad reason to assume it does not exist.
I don’t think the difficulty is that we can’t define it rigorously. It is that each pair, even the ones in the middle, really do seem better and better. It does seem better to have 50 trillion times as many people exist who are experiencing almost the same level of well-being as the prior population. But that is what we’ll have to say!
What happens when you zoom out, comparing not adjacent quantities but the log 25th and 75th percentile of the distribution you've created? 100 quadrillion people living below poverty level lives versus to 1 billion people living modestly to moderately good lives - does either obviously stand out as better? If so, why?
Well, whether one stands out as better or not, the puzzle here is that we seem to have really strong views about each individual pair in the sequence. And the trouble is that that alone, alongside transitivity, gets us into trouble.
I notice that you choose not to answer whether one is better or not, why? Wouldn't transitivity require that you be able to compare any two intermediate qualities - not just adjacent ones - and determine which is better, whether you're looking at 10 trillion vs 100 trillion or quintillions compared to billions? What does transitivity mean if it requires adjacent pairings to be relevant?
You asked whether one “obviously” stood out as better than the other. My point is that that is going to be orthogonal to the main issue at play concerning the Repugnant Conclusion. Because it does not require that one clearly stand out as better than the other.
I didn’t claim that there *isnt* one that is better. There surely is. Assuming you specify the values correctly, I would think the one later down in the sequence is better, but it’s not as though this is clear or obvious or anything.
" Because it does not require that one *clearly* stand out as better than the other." is hard to square for me with your assertion from the prior comment - "Well, whether one stands out as better or not, the puzzle here is that *we seem to have really strong views* about each individual pair in the sequence. And the trouble is that *that alone, alongside transitivity*, gets us into trouble."
Is it important that we have strong feelings - draw obvious conclusions - about what goodness is in this discussion, or is it orthogonal to the trilemma you lay out in this post?
My understanding based on your comments is that transitivity alone actually is insufficient to be represented in its current position in your tripartite dilemma, and must instead be caveated with statements about adjacency, intuition, etc. That does make for a much less difficult logical question to engage with than a mathematically and rationally strict idea like transitivity, however. Have I misunderstood your post or replies?
Do you think that given a world with one person slightly above the threshold, adding one slightly below is bad? If yes, I think Valdman is right and you just have a different threshold for "life worth living". If no, then theres a choice where doubling the population in both worlds flips your decision.
By “threshold” do you mean threshold for “life worth living”? If so, I take it that adding a person slightly above and slightly below keeps the world’s total value exactly the same.
By "threshold" I mean the highest level of utility that you think can not outweigh a slightly higher level just by having more people. And I dont mean adding one person above and one below, I mean there already is one above, and is it good to add one thats slightly below?
Ah gotcha. Yes, it would be good to add one slightly below.
Ok, so if Im understanding it correctly, you think there are some thresholds where people above are lexically prefered to people below, but those below can still be positive.
In that case, you will contradict individuals choices about things that effect only them. Eg if Alice is ε utils above a threshold, and 3ε below the next one, and she can take a gamble 10% -2ε 90% +2ε, then you would want to stop her from doing that.
Actually, we’d probably be best to not introduce risk here — I think risk and bounded aggregation is a really tricky issue. I’m a fanatic and generally a huge fan of expected value theory (in finite contexts at least), and BA basically entails that some populations are infinitely more valuable than others.
Additionally, though, note my solution quantified over *worlds*. So it can’t merely be an individuals decision to improve their own well-being — it has to be along both dimensions (their well being, and the number of people involved).
I dont think the worlds avoid anything here? We could always say that Alice is the only person in that world. Alternatively, if there is a large number of people all offered the bet with independent randomness, then the aggregate result is quasi-deterministic, and you again think their choice makes things worse.
We could even think of a version where the people unanimously agree to redistribute their resources in such a way that 90% gain 2ε and 10% lose 2ε, with the ratio fixed and randomness only determining who the winners and loser are.
Generally, I think expected value theory forces your treatment of risk based on your treatment of riskless comparisons, and all these problems can be created in a finite context.
I'm not a philosopher or an economist so my comment may be misinformed, irrelevant, and if so, mildly embarrassing, but here it is anyway.
Your solution to the repugnant conclusion reminded me of the economic law of diminishing returns—as one input in a production process is increased, while other inputs are held constant, the marginal increase in output will eventually decrease.
I imagine it works out like this:
Population A is starting at the end of Stage 1, where welfare per individual is at it's peak. As you work toward population Z, increasing the number of people leads to a decrease in welfare per individual. This is still producing a positive output (in populations prior to Z), because, even after the law of diminishing marginal returns kicks in, and welfare per individual declines, average welfare continues to increase because the individual welfare still exceeds the average. Once the welfare per individual is equal to the average total welfare, then our population would be in the middle of Stage 2.
The repugnant conclusion seems to take place at the end of Stage 2, the point where the total number of people is at it's peak, welfare per individual is at it's lowest, and average welfare is declining. This would be the moment before moving into Stage 3 where individual welfare would become negative.
If we think of individuals and their welfare as inputs to the productivity of life then it would make sense that, holding welfare as a fixed variable and constantly dividing it among more and more individuals would reduce life productivity for everyone once welfare is decreased beyond a particular point.
IDK, maybe this isn’t a proper application of the law.
There is a way easier answer to this puzzle; human intuitions have evolved to be useful cues in "environment of evolutionary adaptedness" (EEA). They are not going to be useful in hypothetical possible worlds immensely different than EEA and deciding fates of two planets (one with a hundred thousand habitants and the other with one trillion or whatever huge number you choose) is definitely not something any human in EEA had ever to do. This thought experiment is misusing a tool evolved to be useful in environment A in the incomparable environment B, like for example using a kitchen scale to weight the sun. The whole enterprise is fried
Many women accept level 10 pain (or at least the risk of it) for the pleasure of having a baby.
Nobody (OK, surrogates excepted!) would accept level 10 pain for somebody else to have a baby. So there's something fundamentally wrong in the whole trade-off thing.
I don't personally consider a world with a million people living grade X lives any better or worse than one with a billion living similar lives.