Consider peer review as a repugnant conclusion. A paper gets a bunch of comments from reviewers. Responding to each particular comment could make the paper better. But if you respond to every comment you’ll likely make the paper worse.
I think peer review is a a much better example than the others. It gets much closer to showing a contradiction between marginal improvements and overall optimization, which I doubt exists whatsoever in reality when it comes to burgers, walls, or lakes.
There's still might theoretically be an optimal version of each paper with optimal choices about which advice to take, but since there's a whole library of babel of potential things we could write and no perfect measurable standard by which to judge them all, or to determine which subjects are even worth writing about, I think it's a far trickier and more interesting problem.
With that said, I personally think it was an optimal choice to pay attention to comments asking for better examples, though obviously I'm biased since those included my own comments.
"Written on the subway walls, you could be forgiven for overlooking that thought experiment’s profundity." I am not written on the subway walls, thankfully. (This sentence begins with a misplaced modifier.)
I thought “were it written on the subway walls, you could be forgiven…” would be too stuffy. And I really wanted the allusion to the Sound of Silence :)
Very interesting article, came by way of the indominablte Soldo.
I am no philosopher, but the repugnant's dependence on trade-offs seems flawed, no? Why not have a bigger and deeper lake? A more tasty and less shitty burger?
It would seem day to day that there is no special law that dictates more at the cost of quality? In fact categorically, there are plenty of occasions where a person will either choose more and better quality (e.g. upgrading a house) or simply to opt out of more for less quality (McDonalds everyday).
I do think the analogy is apropos to what has happened largely in consumer technology (more convenience, but tradeoffs in things like privacy), electronics (more accessible to the masses, but at great cost to longevity, etc..) but i don't know if these were consciously made or rather just driven by techno-capitalism as opposed to clambering for goodness?
You're asking: isn't it possible to make things better along some dimension without making them worse along another dimension? The short answer is: I don't know. I suspect that it isn't possible, and that it only seems possible in some cases, like the ones you mention, because of ignorance (because we don't know or can't detect how we're making things worse). But it also doesn't matter, because the repugnant conclusion is like flypaper: maybe you can avoid it for a little while, but eventually you'll land on it and get stuck. It's just a matter of time.
I thoroughly enjoyed your essay on the logic of the repugnant conclusion [RC], especially your witty and irreverent style of writing! I take the following to be your most general conclusion:
"But the repugnant conclusion can be deployed against any natural relation – is-bigger-than, is-mightier-than, is-clearer-than, is-more-useful-than – and not to goodness (or is-better-than) in particular. The open-question argument and the repugnant conclusion are manifestations of the same general problem – a failure of analysis – a gap that forms whenever one seeks clarity by breaking wholes into parts."
This resonates with another 'failure of analysis' known as (quantum) contextuality*: https://en.wikipedia.org/wiki/Quantum_contextuality . Like the RC, though contextuality originated in one area of study (quantum physics), it has much broader applicability. The core insight of contextuality can be summarized as "local coherence but global incoherence".
A visual analogy is any of Escher's paradoxical structures, such as Ascending and Descending: https://en.wikipedia.org/wiki/Ascending_and_Descending . Any pair or triple of adjacent stairways is locally coherent, but all four stairways are globally incoherent. A logical example of contextuality is the the liar cycle: S1: S2 is true; S2: S3 is true; ... S(n-1): Sn is true; Sn: S1 is false. Any local subset of adjacent assertions is coherent, but the global set of assertions is incoherent.
It seems to me that the RC is another example of such contextuality: any local set of transitive comparisons is coherent, but the global set of comparisons is incoherent. As long as we remember to be pragmatically local in our observations and comparisons, we need not fear that the RC will "drag down us all". We only need remember that we can't always aggregate coherent local contexts into a holistic global context. (By the way, in the field of ethics, Pragmatists took this to heart with their localized concept of 'meliorism': focusing only on improving present conditions without aiming towards eventual perfection. I discuss such a 'directionless' melioristic approach here: https://erraticus.co/2022/10/27/what-can-pragmatists-hope-for-in-a-boundless-world/ .)
* Note that there are several bodies of work that go by the name "contextualism". I am speaking specifically about the body of work that originated in quantum physics, and which has been generalized using category theory by folks such as Samson Abramsky. For a good summary of Abramsky's work on contextuality, see 'At the Borders of Paradox' https://arxiv.org/pdf/2011.04899 .
Thanks for those links! But I think the solution to the RC, if there is one, is to resist the lure of pragmatic locality and think big instead. Thinking locally reveals opportunities for gradual improvement, the steady accumulation of which tend to make things worse.
"resist the lure of pragmatic locality and think big instead"
Depending on how "big" "thinking big" is, it may be compatible with pragmatically local meliorism. Centuries-long projects (such as cathedral building or reducing carbon emissions) can nonetheless be local and incremental.
"the steady accumulation of [gradual improvements] tend to make things worse"
I agree. But major (non-gradual) improvements tend to make things worse as well. Just as there's always an upside, there's always a downside: improvements in some dimensions are balanced by worsenings in other dimensions.
By “think big” I mean follow the categorical imperative. Don’t ask what would make things better on some particular occasion but on what would happen if your maxim became universal law. Repugnancy is a contradiction in the will. It’s what you get when you think in terms of pebbles rather than boulders. I’ll explain in a future post.
"what would happen if your maxim became universal law"
From a contextualist perspective, maxims becoming universal laws would increase tragic suffering in some contexts and would decrease it in others. Universal laws, like parochial laws, are upayas (skillful means) that can be beneficial in some contexts and detrimental in others.
As much as I enjoyed this article which, so far as I can tell, is about what Cory Doctorow called "the enshittification of everything," but I can't help but suspect the conclusion. Doctorow describes a process endemic to digital capitalism; you see it as fundemental to reason. It seems distinctly anti-ameliorist, but very familiar in our political economy, to demand that in order to get better, you must be prepared to swallow some shit, and I worry that your conclusion engrains such a viewpoint as fundamental, not social. Is it really too utopian to think that a shitless burger is somehow possible?
It might be fundamental to anything that's trying to outrun its contradictions, which would include reason, capitalism, people, institutions, and maybe the universe itself. As for a shitless burger, I think it's possible.. It's even possible to run the repugnant conclusion in reverse, from Z to Y to X and all the way to A. But that's not the natural flow. It's swimming upstream. You can do it for a while but eventually the current gets you.
Applying the repugnant conclusion to ‘x-ness’ makes me that we are just incapable of imagining any concept where there aren’t diminishing marginal returns. We’re finite creatures reasoning about infinite increases. Doubling x is possible in mathematics, but in the physical world Zeno’s arrow hits its target. It’s impossible to imagine a burger doubling in deliciousness infinitely because at a certain point we don’t think we could tell the difference — we asymptotically approach the ‘most delicious’ burger. We can’t think imagine infinite doubling of happiness or clarity for the same reason. To a certain extent we can imagine someone is happier than another, but at a certain point we’d just say they’re both pretty happy.
Zeno’s arrow hits its target empirically (i.e. in reality) but not a priori (i.e. in reason). The repugnant conclusion is a problem in reason. But reason, more and more, is impinging on reality. And reason tells you that B is better than A, C is better than B, etc.
"Zeno’s arrow hits its target empirically (i.e. in reality) but not a priori (i.e. in reason)." Is that right, though?
That the sum of an infinite number of infinitely diminishing addenda can be a finite number is not (only) an empirical conclusion, but a feature of mathematics since Leibniz and Newton - I think there might be some a priori reasoning behind calculus...
Thank you for the link (though it doesn't seem to me that you have really answered to my enquiry), where, on swift perusal, I find this other possible objection to your line of reasoning:
"mathematicians value money in proportion to its quantity, and men of good sense in proportion to the usage that they may make of it. That which renders the mathematical expectation infinite, is the prodigious sum that I am able to receive, if the side of Heads falls only very late, the 100th or 1000th toss. Now this sum, if I reason as a sensible man, is not more for me, does not make more pleasure for me, does not engage me more to accept the game, than if it would be only 10 or 20 million coins. (Cramér to N. Bernouilli, 21 May 1728)"
Maybe there is something different in moral decisions...
Doesn’t that just confirm my view? He’s basically saying that calculations of expected value are one thing and common sense is another, which is just another way of saying that the empirical and the a priori can diverge.
Well, sounds like you are right. I guess i gots some readin' to do. Thanks for responding. I'm still thinking about how this reason generalizes. Great stuff. All the best.
"Humanity suffers from the crippling problem that it’s always possible to make things x-er (bigger, better, stronger, clearer)."
Rubbish. Real world problems usually depend on trade-offs which lead to an 'optimal' solution(s) which cannot be improved further. Intelligent people will wait until they reach a region of diminishing returns and stop bother with making things X-er.
As with gravity and natural selection, there are countervailing forces. But you can’t outrun the repugnant conclusion forever. In the end it turns everything to dust — civilizations, theories, people — A to B to C and all the way to Z.
Sorry Misha, but Tek Bunny is right and your burger example doesn’t work—for reasons that Economics sorted out in the 1860s, when it introduced the concept of the margin.
You can of course make a plain burger (burger A) more delicious by adding beetroot in burger B, and even more delicious by adding sauce in burger C. But you can’t in fact go on making burgers twice as delicious as their predecessors. Adding pepper may make burger D a little better, but not much. The marginal value of each addition at some point becomes less, and may even be negative.
On the other hand, the marginal values of the unpleasantness of the additional shit may rise from zero (when the shit is imperceptible) to very large when it becomes perceptible.
At that point, the marginal value for the burger as a whole becomes negative. Burger H is of less value than burger G, but burger G is definitely better than burger A. The burger example fails, and I suspect that the castle wall and lake examples do as well.
More generally, while I’m not familiar with the debate over Parfit, I suspect that one’s conclusions would depend on one’s views on Kantian claims that all reason is characterised by contradiction, or on post-modern claims that language is similarly afflicted, and/or on the paradoxes of Zeno and Russell. Parfit might be right, but the examples you give don’t adequately capture the problem.
Whether you could in fact produce such a burger sequence is irrelevant. In principle, for any food, B could be tastier than A, C could be tastier than B, and so on. But A could be tastier than Z.
> Whether you could in fact produce such a burger sequence is irrelevant.
If you want to leave the comfortable confines of abstract theory and its frictionless surfaces, and apply them in the less forgiving world of object level reality (as seems to be what is implied in the essay) it is certainly not irrelevant.
Again, this demonstrates the power stories have over the human mind.
I'm not sure I follow completely, but I'd say that the "story" you're referring to just is the idea that the a priori world (of abstract theory and its frictionless surfaces, as you put it) and the empirical world are one and the same. And the repugnant conclusion, I believe, shows that they're different.
The story (stories) I am referring to is the story that the mind tells one about any given proposition or point of contention, and the related story it tells about the accuracy of its prior stories.
Kinda like "How is babby formed?", except instead of babby, reality.
Whether something like a soul continues to exist is an interesting and fun question to consider though, it can provide lots of opportunities to ask scientific ideologues if a perceived/asserted absence of evidence is proof of absence....and then watch them go into story telling mode, as is easily predicted.
LLM's may not be an exact repetition of what human are, but they sure do rhyme with them.
If it was not actually the way you describe it, would you necessarily be able to detect it?
Or another way of putting it: how do you know(!) that it is the way it seems to you? Have you a methodology for identifying objective optimality? How could such a thing be fact checked, by using the same claimed-to-be-optimal-yielding methodology?
No, intuition is a (typically subconscious) *prediction* of optimality. Typically, actual optimality cannot be measured, and in our culture when facts are not available, we hallucinate ~"cultural/personal facts" into existence. There are all sorts of weird hallucinations like this within our culture.
That is not correct! Optimality is always defined w/ reference to some higher order good. For the wall, the higher order good is mightiness (effectiveness at keeping people out). The supremely thin and tall wall is obviously not mighty, but there is a wall which is just the right height and thickness to be the mightiest wall, given constraints. Note that superiority of this wall to ones which are slightly taller and thinner or thicker and shorter does NOT violate transitivity - this bundle of width and height is transitively better than those alternatives.
As Misha points out, pretty much everything in the world is like this.
But Parfit is not myopic for thinking that goodness itself should be different. There can be no higher order good than goodness itself, so a simple “more is more” philosophy should suffice for finding the transitively optimal bundle. The repugnance of his conclusion, and the preference intransitivity it implies, shows that utility cannot be such an ultimate good.
Utilitarianism is an important moral philosophy, so this problematisation is a worthy subject of discussion. But it doesn’t follow that our actual moral preferences are intransitive and that moral philosophical inquiry is therefore doomed to fail, as Misha seems to suggest.
In particular, the possibility remains that there is some other higher order good that you can maximise in a transitive way by optimally trading off average utility against total utility.
What this higher order good might be and how to maximise it is a tough question, but I think it’s clearly a normal and sensible question that we can come up with various plausible answers to. Indeed, Parfit seemed satisfied that he had come up with an answer in the final years of his life!
So the RC is a useful problem for moral philosophers to discuss, but it is not so problematic that it portends doom for all efforts to solve problems that look even vaguely like it. In other words, Parfit was fully justified in spending a lot of time working on it!
Likewise, Parfit seems to have been perfectly capable of arriving at good conclusions to the important and difficult question of how much philosophical clarity it was worth his while to provide. He was indeed a perfectionist and probably should have worried less about loose ends, but he’s clearly been pretty well vindicated by history! RAP and OWM are excellent, well-regarded books, and I’m glad that Parfit spent a lot of time working on them before ultimately calling it a day and sending them off to the presses, warts and all.
Parfit was a good, sensible, and successful person who was correctly willing to eat a certain amount of shit in order to make progress. How many of us can say the same for ourselves?
Well like obviously it depends on the circumstance. The point is that for utilitarians, the best amount of utility should not depend on anything - more should always be better. But in the RC, that seems not to be the case. That is a problem for utilitarians, but it is a problem that can be solved, for example by conceding that goodness depends on more than just total utility. If this is the case, then there are conditions under which some outcome between A and Z is optimal.
For example, you could think that if adding an extra person would bring average utility below a certain critical level, then you should not add an extra person, even if doing so would increase total utility. This is like saying that there's no use in adding bricks to a wall if the wall is still too thin to serve its purpose.
In philosophy and in life, Parfit seemed to accept a version of this. He talked about how some profound experience were lexically better than less profound types, so that no amount of lizards basking in the sun could make up for the loss of one human moment of true love, for example. In his writing, he recognised that his two books were better off published than perfect.
It's true that he went further along the path from A to Z than most in evaluating such trade-offs, but given his monumental impact, I don't think we can say that this approach ultimately left him eating shit.
It's the other way round. Where is any evidence for the nonsense of the statement I quoted. The norm of every system we study, whether physical or biological or social, is to have complex non linear dynamics. Making things incrementally and continuously X-er is something we have no example of.
> It's the other way round. Where is any evidence for the nonsense of the statement I quoted.
You have it backwards. What you are trying to do is "shift the burden of proof".
Here is a question, I am interested in whether you have the ability to coerce your mind into attempting an answer to it: "If it was not actually the way you describe it, would you necessarily be able to detect it?"
Now, having reposed the very same question, I can now observe if you behave any differently than the last time I asked....and for extra fun: I am explicitly revealing the experiment I am running!
> The norm of every system we study, whether physical or biological or social, is to have complex non linear dynamics. Making things incrementally and continuously X-er is something we have no example of.
Do you genuinely believe that you possess genuine omniscient knowledge (on this claim, and your prior ones)?
Interesting post. I do take the repugnant conclusion seriously, and I also think it has relevance beyond the specific case of population growth, but I think most of the examples here are distractingly poor. Optimal design of, for example, a wall, obviously depends on much more than height, meaning there's decreased marginal utility for each foot of wall added, especially if it comes directly at the expense of other important qualities like thickness. Scott Alexander's "Meditations on Moloch" expands on similar ideas more effectively than this, with compelling examples. (He's a psychiatrist, but he studied philosophy in undergrad, so it's not like he has no clue what he's talking about.)
On the subject of the repugnant conclusion itself, I always thought antinatalism had some relevance. Basically, if there's some threshold beneath which life is not worth living, then there seems to be, in most humans, an even lower threshold above which we are hesitant to end a life, whether our own or someone else's, or even to prevent a new life from forming, as in the case of abortion or population control. It seems to me that it's possible to be in the unpleasant area between those two thresholds. In other words, we could already be living in the repugnant conclusion, or a stage beyond it, where some, most, or possibly all human lives have negative utility. So when Parfit asks us to imagine a world where life is barely worth living, our instinctual biases cause us to imagine a world worse than our own, when for all we know, maybe we should be imagining a better one. Now, that's just an idea I came up with years ago as a poorly educated, drug-addled, and extremely depressed teenager, so I'm not sure if it stands up to scrutiny, but it's the best way I've found so far to explain the discrepancy between the compelling logic of the thought experiment with the, well, repugnancy of its conclusion.
Oh and I had one more thought on this post, which I've saved for last since it's harshest: The Simon and Garfunkel reference reads as clumsy and unnecessary to me, and I'm one of the biggest Simon and Garfunkel fans on earth.
(3) The S&G reference is a subtle hint that the repugnant conclusion is better understood as prophecy than paradox. (1) Optimality depends on the available resources. Start with "optimal" walls, but now imagine unlimited resources so that I could double their height at just a very slight decrease in their width. That's all you need to get the repugnant conclusion going. (2) As for whether we're living in the Z-world, well, I don't know, but I definitely think that, on multiple fronts, we're hurtling towards it.
I still find the wall example pretty poor, sorry. If I were defending against a medieval siege, and I had infinite wall supplies, the first thing I'd do is estimate the minimum thickness able to resist my enemies strongest battering ram, and set that as my minimum. If I went above it, I'd make the wall shorter.
You could think of different defeat conditions in terms of probabilities, like 40% chance we starve to death, 5% chance of mutiny, 5% chance of major plague outbreak, 15% chance the enemies scale the walls, 10% chance the enemies break through the walls, and a remaining 25% or so chance that we win. Increasing the wall height is the optimal strategy for as long as enemies scaling the walls is the highest risk, or at least for as long as it's greater than the chance of enemies breaking through the walls. As long as there's a tradeoff between height and thickness, the usual marginal principles still apply, so I don't really see how it's an interesting hypothetical.
It seems to me that 'the repugnant conclusion' boils down to an optimization problem for each decision which imply a tradeoff where (a) a desired good can increase (for example) geometrically, and (b) an inevitable thing we don't like increases with it, but (i.e.) arithmetically. Thus, we can model it beforehand and discuss which of the A-...-Z options is optimal.
But even so, I think not many real-life (including philosophical) problems match those assumptions, let alone every problem. Your examples are about abstract rationality (sb offers you options, you chose the one you prefer), but in real life I can't think of any such dilemma, maybe I just lack imagination, yet is an empirical matter in the end.
Optimality would be a situation in which no amount of Y could compensate for the loss of any amount of X. In that situation you’d effectively be stuck and thus safe from the RC. But the RC would beckon if you came unstuck. That’s why I describe it as the form of change.
And real life examples abound. The RC is why colleges are so expensive (you keep adding amenities and each time slightly raising tuition), why bureaucracies become stagnant and inefficient, why the Soviet Union collapsed, why you hate your favorite band’s new album. More speculatively, it may also explain why we age. And why we die.
Consider peer review as a repugnant conclusion. A paper gets a bunch of comments from reviewers. Responding to each particular comment could make the paper better. But if you respond to every comment you’ll likely make the paper worse.
I think peer review is a a much better example than the others. It gets much closer to showing a contradiction between marginal improvements and overall optimization, which I doubt exists whatsoever in reality when it comes to burgers, walls, or lakes.
There's still might theoretically be an optimal version of each paper with optimal choices about which advice to take, but since there's a whole library of babel of potential things we could write and no perfect measurable standard by which to judge them all, or to determine which subjects are even worth writing about, I think it's a far trickier and more interesting problem.
With that said, I personally think it was an optimal choice to pay attention to comments asking for better examples, though obviously I'm biased since those included my own comments.
"Written on the subway walls, you could be forgiven for overlooking that thought experiment’s profundity." I am not written on the subway walls, thankfully. (This sentence begins with a misplaced modifier.)
I thought “were it written on the subway walls, you could be forgiven…” would be too stuffy. And I really wanted the allusion to the Sound of Silence :)
Lessons we might learn from Parfit and Valdman:
Relations are not always transitive.
Quantities are not always commensurable.
Scaling is not always linear.
Results may be path-dependent.
If logic and intuition conflict, they may be trying to tell you something.
This is beautifully written. Thank you!
Very interesting article, came by way of the indominablte Soldo.
I am no philosopher, but the repugnant's dependence on trade-offs seems flawed, no? Why not have a bigger and deeper lake? A more tasty and less shitty burger?
It would seem day to day that there is no special law that dictates more at the cost of quality? In fact categorically, there are plenty of occasions where a person will either choose more and better quality (e.g. upgrading a house) or simply to opt out of more for less quality (McDonalds everyday).
I do think the analogy is apropos to what has happened largely in consumer technology (more convenience, but tradeoffs in things like privacy), electronics (more accessible to the masses, but at great cost to longevity, etc..) but i don't know if these were consciously made or rather just driven by techno-capitalism as opposed to clambering for goodness?
Happily now a subscriber!
You're asking: isn't it possible to make things better along some dimension without making them worse along another dimension? The short answer is: I don't know. I suspect that it isn't possible, and that it only seems possible in some cases, like the ones you mention, because of ignorance (because we don't know or can't detect how we're making things worse). But it also doesn't matter, because the repugnant conclusion is like flypaper: maybe you can avoid it for a little while, but eventually you'll land on it and get stuck. It's just a matter of time.
I thoroughly enjoyed your essay on the logic of the repugnant conclusion [RC], especially your witty and irreverent style of writing! I take the following to be your most general conclusion:
"But the repugnant conclusion can be deployed against any natural relation – is-bigger-than, is-mightier-than, is-clearer-than, is-more-useful-than – and not to goodness (or is-better-than) in particular. The open-question argument and the repugnant conclusion are manifestations of the same general problem – a failure of analysis – a gap that forms whenever one seeks clarity by breaking wholes into parts."
This resonates with another 'failure of analysis' known as (quantum) contextuality*: https://en.wikipedia.org/wiki/Quantum_contextuality . Like the RC, though contextuality originated in one area of study (quantum physics), it has much broader applicability. The core insight of contextuality can be summarized as "local coherence but global incoherence".
A visual analogy is any of Escher's paradoxical structures, such as Ascending and Descending: https://en.wikipedia.org/wiki/Ascending_and_Descending . Any pair or triple of adjacent stairways is locally coherent, but all four stairways are globally incoherent. A logical example of contextuality is the the liar cycle: S1: S2 is true; S2: S3 is true; ... S(n-1): Sn is true; Sn: S1 is false. Any local subset of adjacent assertions is coherent, but the global set of assertions is incoherent.
It seems to me that the RC is another example of such contextuality: any local set of transitive comparisons is coherent, but the global set of comparisons is incoherent. As long as we remember to be pragmatically local in our observations and comparisons, we need not fear that the RC will "drag down us all". We only need remember that we can't always aggregate coherent local contexts into a holistic global context. (By the way, in the field of ethics, Pragmatists took this to heart with their localized concept of 'meliorism': focusing only on improving present conditions without aiming towards eventual perfection. I discuss such a 'directionless' melioristic approach here: https://erraticus.co/2022/10/27/what-can-pragmatists-hope-for-in-a-boundless-world/ .)
* Note that there are several bodies of work that go by the name "contextualism". I am speaking specifically about the body of work that originated in quantum physics, and which has been generalized using category theory by folks such as Samson Abramsky. For a good summary of Abramsky's work on contextuality, see 'At the Borders of Paradox' https://arxiv.org/pdf/2011.04899 .
Thanks for those links! But I think the solution to the RC, if there is one, is to resist the lure of pragmatic locality and think big instead. Thinking locally reveals opportunities for gradual improvement, the steady accumulation of which tend to make things worse.
You're welcome!
"resist the lure of pragmatic locality and think big instead"
Depending on how "big" "thinking big" is, it may be compatible with pragmatically local meliorism. Centuries-long projects (such as cathedral building or reducing carbon emissions) can nonetheless be local and incremental.
"the steady accumulation of [gradual improvements] tend to make things worse"
I agree. But major (non-gradual) improvements tend to make things worse as well. Just as there's always an upside, there's always a downside: improvements in some dimensions are balanced by worsenings in other dimensions.
By “think big” I mean follow the categorical imperative. Don’t ask what would make things better on some particular occasion but on what would happen if your maxim became universal law. Repugnancy is a contradiction in the will. It’s what you get when you think in terms of pebbles rather than boulders. I’ll explain in a future post.
"what would happen if your maxim became universal law"
From a contextualist perspective, maxims becoming universal laws would increase tragic suffering in some contexts and would decrease it in others. Universal laws, like parochial laws, are upayas (skillful means) that can be beneficial in some contexts and detrimental in others.
"I’ll explain in a future post."
I look forward to it!
The issue is moreso that improvements in one dimension tend to cause worsenings in the very same dimension.
Rather than thinking big or thinking small, it would probably be best to avoid thinking at all.
As much as I enjoyed this article which, so far as I can tell, is about what Cory Doctorow called "the enshittification of everything," but I can't help but suspect the conclusion. Doctorow describes a process endemic to digital capitalism; you see it as fundemental to reason. It seems distinctly anti-ameliorist, but very familiar in our political economy, to demand that in order to get better, you must be prepared to swallow some shit, and I worry that your conclusion engrains such a viewpoint as fundamental, not social. Is it really too utopian to think that a shitless burger is somehow possible?
It might be fundamental to anything that's trying to outrun its contradictions, which would include reason, capitalism, people, institutions, and maybe the universe itself. As for a shitless burger, I think it's possible.. It's even possible to run the repugnant conclusion in reverse, from Z to Y to X and all the way to A. But that's not the natural flow. It's swimming upstream. You can do it for a while but eventually the current gets you.
This was very enjoyable to read, thanks for writing it, Misha.
One of the most interesting things I've learned in the realm of philosophy. Thank you for writing this.
Also fun to read.
Applying the repugnant conclusion to ‘x-ness’ makes me that we are just incapable of imagining any concept where there aren’t diminishing marginal returns. We’re finite creatures reasoning about infinite increases. Doubling x is possible in mathematics, but in the physical world Zeno’s arrow hits its target. It’s impossible to imagine a burger doubling in deliciousness infinitely because at a certain point we don’t think we could tell the difference — we asymptotically approach the ‘most delicious’ burger. We can’t think imagine infinite doubling of happiness or clarity for the same reason. To a certain extent we can imagine someone is happier than another, but at a certain point we’d just say they’re both pretty happy.
Zeno’s arrow hits its target empirically (i.e. in reality) but not a priori (i.e. in reason). The repugnant conclusion is a problem in reason. But reason, more and more, is impinging on reality. And reason tells you that B is better than A, C is better than B, etc.
"Zeno’s arrow hits its target empirically (i.e. in reality) but not a priori (i.e. in reason)." Is that right, though?
That the sum of an infinite number of infinitely diminishing addenda can be a finite number is not (only) an empirical conclusion, but a feature of mathematics since Leibniz and Newton - I think there might be some a priori reasoning behind calculus...
Check out the St. Petersburg paradox: https://plato.stanford.edu/entries/paradox-stpetersburg/
The empirical world and the a priori aren’t always in sync.
Thank you for the link (though it doesn't seem to me that you have really answered to my enquiry), where, on swift perusal, I find this other possible objection to your line of reasoning:
"mathematicians value money in proportion to its quantity, and men of good sense in proportion to the usage that they may make of it. That which renders the mathematical expectation infinite, is the prodigious sum that I am able to receive, if the side of Heads falls only very late, the 100th or 1000th toss. Now this sum, if I reason as a sensible man, is not more for me, does not make more pleasure for me, does not engage me more to accept the game, than if it would be only 10 or 20 million coins. (Cramér to N. Bernouilli, 21 May 1728)"
Maybe there is something different in moral decisions...
Doesn’t that just confirm my view? He’s basically saying that calculations of expected value are one thing and common sense is another, which is just another way of saying that the empirical and the a priori can diverge.
Well, sounds like you are right. I guess i gots some readin' to do. Thanks for responding. I'm still thinking about how this reason generalizes. Great stuff. All the best.
I think I read the z version of this essay
Law bomb lobbed!
The East Indies, genius
"Humanity suffers from the crippling problem that it’s always possible to make things x-er (bigger, better, stronger, clearer)."
Rubbish. Real world problems usually depend on trade-offs which lead to an 'optimal' solution(s) which cannot be improved further. Intelligent people will wait until they reach a region of diminishing returns and stop bother with making things X-er.
As with gravity and natural selection, there are countervailing forces. But you can’t outrun the repugnant conclusion forever. In the end it turns everything to dust — civilizations, theories, people — A to B to C and all the way to Z.
Sorry Misha, but Tek Bunny is right and your burger example doesn’t work—for reasons that Economics sorted out in the 1860s, when it introduced the concept of the margin.
You can of course make a plain burger (burger A) more delicious by adding beetroot in burger B, and even more delicious by adding sauce in burger C. But you can’t in fact go on making burgers twice as delicious as their predecessors. Adding pepper may make burger D a little better, but not much. The marginal value of each addition at some point becomes less, and may even be negative.
On the other hand, the marginal values of the unpleasantness of the additional shit may rise from zero (when the shit is imperceptible) to very large when it becomes perceptible.
At that point, the marginal value for the burger as a whole becomes negative. Burger H is of less value than burger G, but burger G is definitely better than burger A. The burger example fails, and I suspect that the castle wall and lake examples do as well.
More generally, while I’m not familiar with the debate over Parfit, I suspect that one’s conclusions would depend on one’s views on Kantian claims that all reason is characterised by contradiction, or on post-modern claims that language is similarly afflicted, and/or on the paradoxes of Zeno and Russell. Parfit might be right, but the examples you give don’t adequately capture the problem.
Whether you could in fact produce such a burger sequence is irrelevant. In principle, for any food, B could be tastier than A, C could be tastier than B, and so on. But A could be tastier than Z.
> Whether you could in fact produce such a burger sequence is irrelevant.
If you want to leave the comfortable confines of abstract theory and its frictionless surfaces, and apply them in the less forgiving world of object level reality (as seems to be what is implied in the essay) it is certainly not irrelevant.
Again, this demonstrates the power stories have over the human mind.
I'm not sure I follow completely, but I'd say that the "story" you're referring to just is the idea that the a priori world (of abstract theory and its frictionless surfaces, as you put it) and the empirical world are one and the same. And the repugnant conclusion, I believe, shows that they're different.
The story (stories) I am referring to is the story that the mind tells one about any given proposition or point of contention, and the related story it tells about the accuracy of its prior stories.
Kinda like "How is babby formed?", except instead of babby, reality.
Object level reality is abstract theory.
This is an interesting belief, can you cite any academic material that supports it?
Whether something like a soul continues to exist is an interesting and fun question to consider though, it can provide lots of opportunities to ask scientific ideologues if a perceived/asserted absence of evidence is proof of absence....and then watch them go into story telling mode, as is easily predicted.
LLM's may not be an exact repetition of what human are, but they sure do rhyme with them.
If it was not actually the way you describe it, would you necessarily be able to detect it?
Or another way of putting it: how do you know(!) that it is the way it seems to you? Have you a methodology for identifying objective optimality? How could such a thing be fact checked, by using the same claimed-to-be-optimal-yielding methodology?
The measure of optimality is intuition; the repugnant conclusion shows that our intuitions do not obey transitivity. Do with that what you will.
> The measure of optimality is intuition
No, intuition is a (typically subconscious) *prediction* of optimality. Typically, actual optimality cannot be measured, and in our culture when facts are not available, we hallucinate ~"cultural/personal facts" into existence. There are all sorts of weird hallucinations like this within our culture.
'Our intuitions are not transitive'?? Thanks for the simplistic profundity Yoda. Maybe look up Arrow's Theorem sometime.
That is not correct! Optimality is always defined w/ reference to some higher order good. For the wall, the higher order good is mightiness (effectiveness at keeping people out). The supremely thin and tall wall is obviously not mighty, but there is a wall which is just the right height and thickness to be the mightiest wall, given constraints. Note that superiority of this wall to ones which are slightly taller and thinner or thicker and shorter does NOT violate transitivity - this bundle of width and height is transitively better than those alternatives.
As Misha points out, pretty much everything in the world is like this.
But Parfit is not myopic for thinking that goodness itself should be different. There can be no higher order good than goodness itself, so a simple “more is more” philosophy should suffice for finding the transitively optimal bundle. The repugnance of his conclusion, and the preference intransitivity it implies, shows that utility cannot be such an ultimate good.
Utilitarianism is an important moral philosophy, so this problematisation is a worthy subject of discussion. But it doesn’t follow that our actual moral preferences are intransitive and that moral philosophical inquiry is therefore doomed to fail, as Misha seems to suggest.
In particular, the possibility remains that there is some other higher order good that you can maximise in a transitive way by optimally trading off average utility against total utility.
What this higher order good might be and how to maximise it is a tough question, but I think it’s clearly a normal and sensible question that we can come up with various plausible answers to. Indeed, Parfit seemed satisfied that he had come up with an answer in the final years of his life!
So the RC is a useful problem for moral philosophers to discuss, but it is not so problematic that it portends doom for all efforts to solve problems that look even vaguely like it. In other words, Parfit was fully justified in spending a lot of time working on it!
Likewise, Parfit seems to have been perfectly capable of arriving at good conclusions to the important and difficult question of how much philosophical clarity it was worth his while to provide. He was indeed a perfectionist and probably should have worried less about loose ends, but he’s clearly been pretty well vindicated by history! RAP and OWM are excellent, well-regarded books, and I’m glad that Parfit spent a lot of time working on them before ultimately calling it a day and sending them off to the presses, warts and all.
Parfit was a good, sensible, and successful person who was correctly willing to eat a certain amount of shit in order to make progress. How many of us can say the same for ourselves?
>there is a wall which is just the right height and thickness to be the mightiest wall
Ok, what height and thickness is that?
Well like obviously it depends on the circumstance. The point is that for utilitarians, the best amount of utility should not depend on anything - more should always be better. But in the RC, that seems not to be the case. That is a problem for utilitarians, but it is a problem that can be solved, for example by conceding that goodness depends on more than just total utility. If this is the case, then there are conditions under which some outcome between A and Z is optimal.
For example, you could think that if adding an extra person would bring average utility below a certain critical level, then you should not add an extra person, even if doing so would increase total utility. This is like saying that there's no use in adding bricks to a wall if the wall is still too thin to serve its purpose.
In philosophy and in life, Parfit seemed to accept a version of this. He talked about how some profound experience were lexically better than less profound types, so that no amount of lizards basking in the sun could make up for the loss of one human moment of true love, for example. In his writing, he recognised that his two books were better off published than perfect.
It's true that he went further along the path from A to Z than most in evaluating such trade-offs, but given his monumental impact, I don't think we can say that this approach ultimately left him eating shit.
It's the other way round. Where is any evidence for the nonsense of the statement I quoted. The norm of every system we study, whether physical or biological or social, is to have complex non linear dynamics. Making things incrementally and continuously X-er is something we have no example of.
> It's the other way round. Where is any evidence for the nonsense of the statement I quoted.
You have it backwards. What you are trying to do is "shift the burden of proof".
Here is a question, I am interested in whether you have the ability to coerce your mind into attempting an answer to it: "If it was not actually the way you describe it, would you necessarily be able to detect it?"
Now, having reposed the very same question, I can now observe if you behave any differently than the last time I asked....and for extra fun: I am explicitly revealing the experiment I am running!
> The norm of every system we study, whether physical or biological or social, is to have complex non linear dynamics. Making things incrementally and continuously X-er is something we have no example of.
Do you genuinely believe that you possess genuine omniscient knowledge (on this claim, and your prior ones)?
Interesting post. I do take the repugnant conclusion seriously, and I also think it has relevance beyond the specific case of population growth, but I think most of the examples here are distractingly poor. Optimal design of, for example, a wall, obviously depends on much more than height, meaning there's decreased marginal utility for each foot of wall added, especially if it comes directly at the expense of other important qualities like thickness. Scott Alexander's "Meditations on Moloch" expands on similar ideas more effectively than this, with compelling examples. (He's a psychiatrist, but he studied philosophy in undergrad, so it's not like he has no clue what he's talking about.)
On the subject of the repugnant conclusion itself, I always thought antinatalism had some relevance. Basically, if there's some threshold beneath which life is not worth living, then there seems to be, in most humans, an even lower threshold above which we are hesitant to end a life, whether our own or someone else's, or even to prevent a new life from forming, as in the case of abortion or population control. It seems to me that it's possible to be in the unpleasant area between those two thresholds. In other words, we could already be living in the repugnant conclusion, or a stage beyond it, where some, most, or possibly all human lives have negative utility. So when Parfit asks us to imagine a world where life is barely worth living, our instinctual biases cause us to imagine a world worse than our own, when for all we know, maybe we should be imagining a better one. Now, that's just an idea I came up with years ago as a poorly educated, drug-addled, and extremely depressed teenager, so I'm not sure if it stands up to scrutiny, but it's the best way I've found so far to explain the discrepancy between the compelling logic of the thought experiment with the, well, repugnancy of its conclusion.
Oh and I had one more thought on this post, which I've saved for last since it's harshest: The Simon and Garfunkel reference reads as clumsy and unnecessary to me, and I'm one of the biggest Simon and Garfunkel fans on earth.
(3) The S&G reference is a subtle hint that the repugnant conclusion is better understood as prophecy than paradox. (1) Optimality depends on the available resources. Start with "optimal" walls, but now imagine unlimited resources so that I could double their height at just a very slight decrease in their width. That's all you need to get the repugnant conclusion going. (2) As for whether we're living in the Z-world, well, I don't know, but I definitely think that, on multiple fronts, we're hurtling towards it.
I still find the wall example pretty poor, sorry. If I were defending against a medieval siege, and I had infinite wall supplies, the first thing I'd do is estimate the minimum thickness able to resist my enemies strongest battering ram, and set that as my minimum. If I went above it, I'd make the wall shorter.
You could think of different defeat conditions in terms of probabilities, like 40% chance we starve to death, 5% chance of mutiny, 5% chance of major plague outbreak, 15% chance the enemies scale the walls, 10% chance the enemies break through the walls, and a remaining 25% or so chance that we win. Increasing the wall height is the optimal strategy for as long as enemies scaling the walls is the highest risk, or at least for as long as it's greater than the chance of enemies breaking through the walls. As long as there's a tradeoff between height and thickness, the usual marginal principles still apply, so I don't really see how it's an interesting hypothetical.
It seems to me that 'the repugnant conclusion' boils down to an optimization problem for each decision which imply a tradeoff where (a) a desired good can increase (for example) geometrically, and (b) an inevitable thing we don't like increases with it, but (i.e.) arithmetically. Thus, we can model it beforehand and discuss which of the A-...-Z options is optimal.
But even so, I think not many real-life (including philosophical) problems match those assumptions, let alone every problem. Your examples are about abstract rationality (sb offers you options, you chose the one you prefer), but in real life I can't think of any such dilemma, maybe I just lack imagination, yet is an empirical matter in the end.
Optimality would be a situation in which no amount of Y could compensate for the loss of any amount of X. In that situation you’d effectively be stuck and thus safe from the RC. But the RC would beckon if you came unstuck. That’s why I describe it as the form of change.
And real life examples abound. The RC is why colleges are so expensive (you keep adding amenities and each time slightly raising tuition), why bureaucracies become stagnant and inefficient, why the Soviet Union collapsed, why you hate your favorite band’s new album. More speculatively, it may also explain why we age. And why we die.