I think it basically depends on whether you’re a prospectist or deferentialist about decision theory. Prospectists (like me!) will say that you often ought to do things even if you know the result will be incomparable in value to the alternative results.
You address some similar ideas, but not quite the same. Part of the issue may just be that you stated things informally, but also this struck me as different from prospectism:
"when you normally act, because you don’t know which incomparable goods you are affecting, you get to ignore them."
In the usual "take the sugar cases," you *do* know which incomparable goods you're affecting. There's a fixed population, or a fixed tradeoff. In fact, you can state prospectism without having any reference to particular incomparable goods at all. The view just says "It's ok to do X iff X maximizes utility for at least *some* utility function that represents your preferences when they're made transitive."
A couple more comments.
i. About your button-lollipop case, which I liked. James Lenman has a similar case in his cluelessness paper -- "Spot the Dog on D-Day." In these "small tiebreaker" cases, I think the intuition that small goods can't break ties tends to go away once you stipulate that there aren't any deontological confounds (like "it would be ill-fitting to base such a momentous decision on such a trifling reason"). Small tiebreakers are basically unavoidable in ethics as long as you have real-valued utility functions playing a role somewhere or other.
ii. I wonder what you think about Richard Yetter Chappell's view that we can have incomparability (/parity) *and* transitivity. In particular, he thinks utility for A can be on par with utiltity B merely because they realize different final goods -- even though these goods can be precisely weighed up against each other.
//In the usual "take the sugar cases," you *do* know which incomparable goods you're affecting.//
But I gave arguments for why whether you know shouldn't affect it. I feel like I'm missing something.
//The view just says "It's ok to do X iff X maximizes utility for at least *some* utility function that represents your preferences when they're made transitive."//
Every act will be permissible if my argument goes through, but none will be better than others.
//Small tiebreakers are basically unavoidable in ethics as long as you have real-valued utility functions playing a role somewhere or other.//
But then how is this any different from just equality? If two things are such that neither is better than the other and tie-breaks outweigh, then this doesn't sound like incomparability.
// wonder what you think about Richard Yetter Chappell's view that we can have incomparability (/parity) *and* transitivity. //
I don't like that view because I don't believe in incomparability. But I'm not really sure how that has a unique advantage for the challenge I give here.
There are two distinct values at play, and you might feel rational regret, in some respect, for bringing about one rather than the other. By contrast, you wouldn't feel any rational regret for picking up a $20 bill rather than two $10 bills. They are completely fungible in terms of monetary value. It's not as if you lose monetary value in some respect, though gain in another, when you pick up the different denomination.
"I don't like that view because I don't believe in incomparability."
But when you say "incomparability," you mean something stronger than what Richard and I mean. You mean the conjunction of
(RYC-style) Incomparability: for some pair of options, neither is better, nor are they equally good.
Insensitivity to Sweetening: for any pair of incomparable options x, y, there is some x+ such that x+ > x and yet x+ is incomparable with y.
The motivation for RYC-style Incomparability is what I was saying earlier: different people's utilities (e.g.) don't seem like they're just means to a single good -- total utility -- but rather seem like distinct goods in themselves. So we can feel rationally torn between x and y even if they promote the same total utility, just because x is good for some people while y is good for other people.
Do you feel the pull at all towards that kind of pluralism about the goods?
I see. Yeah, I have the Richard pluralisticy view that things can be valuable in different ways so they're not just equivalent contributors to some overall value.
I think of the view you're describing as not really incomparability. It's not that the goods can't be precisely compared. They can, and one penny would perturb the comparison. It's that they're not dealing in the same units, and so though they're comparable in terms of amounts of value, the fitting attitude is to feel a kind of regret.
To be honest, I feel the force of the objections to insensitivity to sweetening. Your post is making me want to write a reply to explain why I can't bring myself to give it up...
This is what i came here to say. The bit at the end just read to me as urging us to be deferentialists rather than prospectists, rather than arguing that the result is independent of that debate. Caspar Hare's work in particular seems worth engaging, as he accepts some pretty wild theses in this neighborhood (eg, from our current perspective, the holocaust was not regrettable, because of how later people's identities depend on it), but he doesn't think it's paralyzing for forward looking decision making, precisely because of the distinction Daniel is pointing to.
I'm not sure about this. Suppose that I think happiness and suffering are incomparable, in the sense that there is no precise exchange rate between them. 10 minutes of happiness followed by 10 minutes of suffering is neither better nor worse than no experience, and 11 minutes of happiness followed by 10 minutes of suffering is neither better nor worse than no experience, but 11 minutes of happiness followed by 10 minutes of suffering is better than 10 minutes of happiness followed by 10 minutes of suffering.
I'm considering driving to the store. Driving to the store might have all kinds of hard-to-foresee effects on the world, but it seems equally likely to me that *not* driving to the store will instead have those same hard-to-foresee effects. So in expectation, the hard-to-foresee effects of driving to the store cause X minutes of happiness and Y minutes of suffering, and the hard-to-foresee effects of *not* driving to the store cause X minutes of happiness and Y minutes of suffering. But driving to the store has the easy-to-foresee effect of giving me 1 minute of happiness, and that's the tiebreaker. X+1 minutes of happiness and Y minutes of suffering is better than X minutes of happiness and Y minutes of suffering.
If you haven't already, you might check out "The Infectiousness of Nihilism" by MacAskill, who argues that nihilism is a special case of incomparability, and that views like nihilism (and all forms of radical incomparability) basically break our ability to make moral decisions (all actions have undefined choiceworthiness) in light of moral uncertainty. Later, he (with co-authors) develop an account of dealing with incomparability based on normalizing their variance, so I guess he thinks that fixes the infectiousness problem (see second half of Chapter 6 of Moral Uncertainty, which builds on chapter 5)
I personally think we can modify Ross' Modified Dominance over Theories principle to be in terms of contrastive reasons so that we can ignore theories that posit radical incomparability, including nihilism, no matter what our credence is in them (except if it's 1, which is plausibly irrational), which is great news. Nihilism then becomes irrelevant, even if I believe in it.
I read that a while ago but I think MacAskill's point is somewhat different from mine. His is about the absence of a general formula for comparing incomparable goods if I remember correctly. Mine follows even if you have no normative uncertainty and treat incomparability the standard way.
Yep, I would just flag though: the key thing that incomparability proponents should think is implausible is just that the “big” actions we take (saving a life) don’t matter. But it seems to me the key move you make to show this is that, if we look far enough into the future, any act we take will have huge incomparable consequences. But then maybe whatever (implausible) things people want to say about how we should discount long-term future facts will apply here too, such that the big actions we take still matter.
Also, I’m not so confident our actions don’t have infinite effects. So we might all have to be nihilists assuming certain views about infinite ethics. But I think you’ve made that point before.
I mean, you get enough incomparable stuff quickly enough that I don't think this will work. And most ways to discount the future get more incomparability rather than less.
I think this is charming but much less compelling than it should be because you’re acting like a debater, not a philosopher (to be sure, I think a lot of so-called philosophers do this.) Do you really actually think that the existence of incomparable goods implies nihilism? That would be really wild! I think you ought to sound legitimately confused and anxious about such a situation! But it feels more like you’re just stringing together a possibly-valid argument that you’re pleased with.
On the object level, it seems clear that I can’t be culpable for consequences of my actions I can’t possibly predict, so while I have no commitment on the existence of incomparability, if I did I wouldn’t feel at all worried about the driving to the store argument. I think it’s much harder to surmount this sense that this argument can’t possibly matter than to make the argument itself, which might be a more productive way of making my first point.
Yes I do think the existence of incomparable goods implies nihilism. One way to infer that I think that is that the blog post that you are commenting on is titled "Incomparability Implies Nihilism."
You say "On the object level, it seems clear that I can’t be culpable for consequences of my actions I can’t possibly predict, so while I have no commitment on the existence of incomparability, if I did I wouldn’t feel at all worried about the driving to the store argument." But I addressed that objection in the post.
Yeah, sorry for posting the comment without actually making clear my response to all of your responses. The style of argumentation, as I say, just feels fundamentally incorrect. It’s just obviously false that I can be culpable for things I don’t know about; I can pretty well predict that “Bentham has four different reasons why this self-evident fact isn’t true” isn’t going to lead to me being moved on it by one iota, and indeed having now re-read these four complaints, they do not move me one iota. As I say, this does not come across, tonally or structurally, like somebody who’s discovered a real fact about goodness. I think you have better things to do.
I mean, I think the objections are pretty decisive. But I pretty fundamentally disagree with your comment.
First you say it seems "obviously false that I can be culpable for things I don’t know about." But you *do* know about them. You know your actions affect lots of incomparable goods. You don't know exactly which ones, but one can obviously be culpable for knowing their action has a class of effects even if they don't know which specific effect it has. For example, if you knew that by walking to the store you would cause something very bad to happen, then you shouldn't drive to the store, even though you don't know exactly what will happen.
I think this actually does look structurally like it might be a right insight. One thing about false theories is they often have deep structural problems that apply weird incoherence. So if we think that incomparability is non-existent--as, of course, I do--then we should expect something like this. Of course, maybe you reject that, but it's weird to reject an argument against A on grounds that you simply know at the outset that A is true.
When I talk to consequentialists who believe in incomparable goods, I get the sense that they think:
(1) Actions A has many different consequences a1, a2, a3...
(2) Action A is better than action B iff:
(a) some consequence of A is better than some consequence of B
(b) no consequence of B is better than a consequence of A
Of course, (1) assumes more non-normative structure than in classical consequentialism, where actions have a single consequence W -- either the state of the world (in objective utilitarianism) or a distribution over worlds (in expectational utilitarianism).
So incomparable consequentialists assume some way to slice W into discrete consequences.
Formally, you can think of this as a family of functions from W to partially ordered sets.
- You could slice consequences up by people, which implies that action A beats B iff a is better than B for someone and not worse for anyone (i.e. pareto).
- You could slice consequences up by *sets* of people, which implies that action A beats B iff the populations are the same and the aggregate welfare is higher in A than in B.
- You could slice consequences up spatio-temporal regions.
- You could slice consequences up by 1st-order, 2nd-order, 3rd-order, etc (maybe).
- You could slice things up by an finitary consequence and an infinitary consequence (maybe), and similarly for each infinite cardinal.
- You could slice up by 'goods', e.g. how much justice there is vs how much beauty.
I'm not sure if the claim is that this gets around the issue of incomparability leading to nihilism. The problem still recurs even if you slice the world, and say that we'll measure the goodness of our action according to its consequences, when considered as a function into the partial orderings listed here.
However, I like the slices that you've proposed, I think I can explain how "giving a lollipop" to a girl" becomes incomparable in at least some of these slices. The stories become pretty similar pretty quickly though, so I'll only do the first 3 proposals.
- We slice up by people, and A beats B iff A is better than B for someone and not worse for anyone.
You give the girl a lollipop, she grows up to become dictator of the world, and loves the red colour of the lollipop so much that everyone with (natural) red hair gets treated as though in a utopia. Everyone else is treated far worse. If you don't give her the lollipop, the converse will happen, everyone with red hair will be punished due to her hatred of the colour red ever since she didn't get that red lollipop. Since in each case, different categories of people are worse off, it's neither good nor bad to give her the lollipop.
- We slice by sets of people, that A beats B iff the populations are the same and aggregate welfare is higher in A than B.
You give the girl a lollipop. Later on, she decides to have a child with Mr Rogers, the candy-store salesman, due to her fond recollections of the lollipop-gifter. This child is happy. If you don't give her a lollipop, she will have a child with Mr Gorers, the anti-candy-store activist due to her recollections of not being given a lollipop by that terrible person. That child is also happy overall, though less happy, since they don't get any candy. Two different people exist in these cases, and they're both worse off by not existing if you chose the other option, it's neither good nor bad to give her the lollipop
- We slice consequences by spatio-temporal regions.
You give the girl a lollipop. Later on, she becomes a billionaire and decides to invest massively in the area where you gave her that lollipop, causing everyone's welfare in that region to go massively up. However, this prices out people living the neighbourhood down the road who are worse off. You don't give her the lollipop, and she keeps walking and gets a lollipop in that next neighbourhood. She invests in that area instead, causing the precise opposite to occur. In either case, one spatial region will be worse off, and one better off depending on whether you give her the lollipop or not. So it's neither good nor bad to give her the lollipop.
I don't think that helps. I mean, first of all, that just doesn't work with how incomparability is explicitly supposed to work. This would imply that replacing Jon with Fred and giving a child a lollipop is good. But then how is incomparability different to two things being equally good? The thing that's supposed to make it distinct is that it can't be outweighed by small improvements.
Also, your criterion would rule out any actions being good. After all, all actions have many good and bad effects for butterfly reasons, so it's never true that "no consequence of B is better than a consequence of A." In fact, this doesn't even work with incomparability because incomparable goods always have unique advantages. Fred might go on to make great music--so some consequences of swapping out Fred for Jon are good and some are bad.
I don't really get the latter half of your comment.
Couldn't you say that *some* but not *all* small improvements can make a difference between incomparable things? Like, if a change is made to some feature of the world that is in some way separable from the two incomparables, it counts, but not otherwise?
Obviously this is tricky because in this case the things aren't totally separable: which set of people exists depends on whether you give the lollipop. But again, it seems to me there's something in the idea of looking at this casually that helps: improvements that casually precede the two incomparables or are causally disconnected do count, but improvements that casually follow (i.e., "ok we'll create Alice but not Bob, but conditional on creating Alice, Alice gets a lollipop") don't?
Is it built into incomparables that *all* sweeteners make no difference?
If you treat sweeteners differently by whether they're separable or not then you'll get weird results. Suppose we have the following options of which to create:
1 Bob + a child gets a lollipop
2 Bob who lives an extra day
3 Fred
Assume Bob and Fred are incomparable. If the lollipop is extrinsic but the extra day is not, then 1>3, 2 is incomparable with 3, but presumably 2>1. This is very strange.
It also seems highly arbitrary to have to group acts into whether they're extrinsic or not. And doesn't seem morally significant.
Lastly, the incomparable intuitions are supposed to hold for both extrinsic and intrinsic sweetenings. You're not supposed to swap out your child for another incomparable child and a nickel!
If those are your only options, the the child getting the lollipop isn't separable--they only get the lollipop in Bob world, not in Fred world. So 1 and 3 could remain incomparable. Then you'd just have 2 > 1 and each of them incomparable to 3, which seems... fine?
I think extrinsic vs intrinsic isn't the right distinction, which is why it feels morally arbitrary; I think you want to rather think of causality: it's much more intuitive to me that benefits that only accrue conditional on some scenario coming to pass might be "discounted" somehow by virtue of having to pass through that conditional bottleneck: if we create A, then A + B is better than A, but if we don't create A to begin with, then A+B might have an unclear relationship to ~A.
I'm not sure what you mean by "swap out", but I think one of the advantages of building the incompatibility relation to know about causal relationships is it can screen out cases where "swapping out" is possible: once you've had your child, you can't "swap" to a world where you didn't have that child and have a different one instead--you're stuck in worlds where you had that child, and so the only relevant considerations are those that condition on that.
In all these cases you're causing all the events in question. So causality can't be the relevant metric.
I've argued that normal actions are analogous to swapping out Fred for Bob and a lollipop, so if that action is neither good nor bad then neither is ordinary action.
But your creation of the lollipop is casually dependent on Bob being created, otherwise your option 1 couldn't be "Bob plus lollipop"--whether or not the kid gets the lollipop is casually downstream of whether Bob is created or not. You can't create Fred, and then give the kid a lollipop without leaving the set of options you've laid out.
What I'm arguing is that if you had option 4. Lollipop, and then afterwards decide what to do about Bob and Fred
then the lollipop is casually disconnected, and so it's comparable to anything you do downstream of it.
I think you're going to argue that this violates irrelevance of independent alternatives, since the seeming casual dependence on Bob's existence is an artifact of the fact that you've excluded the possibility 5. Fred and lollipop, and then I'm going to insist on making a distinction between deciding on the lollipop separately, and then deciding on Bob and Fred vs vice versa, or treating all the decisions as simultaneous.
I agree I don't love those features, especially the latter, but given that we actually do make choices constrained by causality in real life, I'm not sure how bad this can get in practice: if my choices are aligned with a causal graph, I can leave money on the table, but I don't think I can be money-pumped by someone who doesn't have a time machine, so my guess is you'd only get the mildest possible problems that can arise from rejecting irrelevance of alternatives. Anyone who believes in incomparables is probably already ok with that.
You need some Zen.Eventually everything is null and void because we live in a universe that will eventually become null and void. Who cares about the eventualities. What matters is the here and now bro. All the philosophical ramblings aside, if I can travel back in time in 1938 and assassinate Hitler I don't give a s***bought some hypothetical theoretical notion of what could or couldn't happen because of that action. That action right then and there would save 6 million people at least from the horrors of the Holocaust. Now whether that would cause some other horrible thing to happen in the future who gives hoot. We can never control the future so, stop all this Western thinking.
This reminds me of trait-based embryo selection, which Scott Alexander wrote about a few months ago. There is something intuitively frightening about it, as if for the first time, you can actively choose whether to create Jon or Fred. The reality is less dramatic; after all, parents already exercise great influence over what sort of people their children become through the combination of partner selection and upbringing. Changing someone’s eye color (or even IQ by a few points) is pretty trivial by comparison. Yet the technology is still controversial. Some of the controversy has to do with practical downsides, but one objection is more fundamental: the idea that choosing these traits is playing God, doing something that humans are not meant to do. Said objection is usually raised in a literal religious context, but it has intuitive resonance even for me as an atheist. Not enough to make me think trait-based embryo selection is wrong (certainly not with the milquetoast effects of current technology), but enough to make me wonder whether a moral boundary is being approached and what that boundary might be.
But of course, trait-based embryo selection is simply replacing a random choice with an informed choice. In regular IVF, the parent is already selecting an embryo, knowing that there is a cause-and-effect relationship between which embryo they choose and what sort of person is created. The new technology just gives them somewhat better information about what that relationship is.
If I follow the intuition that there is some moral boundary here (that would be breached by a hypothetical more-precise future version of the technology), then I’m led directly to “a distinction between incomparable goods you know about and ones you don’t know about”. As well as a need to regard an act as good (namely, selecting an embryo, let’s say before seeing the screening results) despite knowing, with certainty, that if one knew more (i.e. saw the screening results) then one would consider it bad.
I’m not sure I should follow the intuition. Perhaps this is more unsalvageable mess. But I’m not sure it’s not an unavoidable mess.
When computing total utility of an action, just ignore all the incomparable goods whether we know about them or not. Simply discard everything incomparable and reason about comparable. Then everything adds up to normality.
This is, in fact, a standard practice in data processing.
This doesn't work because incomparable goods contain lots of other sub goods. For example, suppose that Bob and Fred are incomparable. Well, Bob is also incomparable with the last 80 years of Fred's life, so you can ignore those and just take into account the first year. There's no natural way of carving goods up.
Posting on the Internet that it doesn't work is like speaking out loud that speech is impossible. With all likelihood, data pipelines for this very site use such calculations.
> incomparable goods contain lots of other sub goods.
That doesn't matter at all. Just discard all that is incomparable. In the example from the post: The action of going to the shop leads to three consequences:
1. Satisfying my desire to get to the shop - lets say it's 2 utilons
2. Non-existence of one person - incomparable - NaN utilons
3. Existence of the other person - incomparable - NaN utilons
The total utility is calculated by discarding all the NaN's and summing all the numeric values. Which is 2 utilons. So the action is slightly positive.
You are saying something doesn't work. I show you that it does, pointing to the actual thing in the reality that is using this principle. Then, I show you how we can execute this algorithm, with the specific example from the post, to make it even clearer.
On the other hand, your counterargument is completely irrelevant to the point I'm making. Nothing in my reasoning was based on the fact that contents of incomparable stuff has to be comparable. At no point I said that we should ignore only the first year of Bobs life so I don't think what it has to do with anything.
We can't in a general case divide incomparable goods into comparable parts. This doesn't mean that we can't divide sets of goods into packages at all. When we have a set we can talk about it individual elements and construct subsets from them. Some of the subsets can be *non-measurable*. It doesn't mean that the notion of measure suddenly looses all meaning.
NaN + n = NaN for any number n. So, if we simply have a NaN value we can't derive anything from it.
But if we have a list of values, some of which may be NaNs, nothing prevents us from adding only the non-NaNs together and using the result in further calculations. It's a simple function:
def sum(l):
s = 0
for v in l:
if v is not np.nan:
s += v
return s
I understand why it can be counterintuitive for people without mathematical or programming background, but surely you are eventually able to notice a counterexample when it's right in front of you? The problem you claim to be unsolvable, in principle, is solved so much that we've been building technology on this solution for decades.
What is the point of worrying about consequences past my horizon of knowledge, beyond which all causality is incomprehensible anyway?
Why should I feel culpable, as a decision-making creature, for an infinite array of hypothetical ripple effects propagating from even the tiniest action?
It seems like in the face of this kind of illegibility of any final accounting, all I can do is *try* to do the right thing to the best of my understanding, knowing full well that my moral reasoning is but a dim flashlight in a dark ocean.
I do not think stocking the world with "completely different people" leads to a meaningfully different world on aggregate if structural forces are the same. To use your example of driving to the store, if you have political, economic, geographic and environmental factors unchanged between group A and B, then those factors are severely constraining the outcomes, and group A and B will be in fact comparable and very similar regardless of whether you drive to the store or not. That some people have different life trajectories does not matter at all if statistically speaking you get the same number of doctors, engineers, drug addicts, and so on.
To use the well-worn Hitler example, to the extent that his rise to power was dependent on contingent factors, those factors extended far beyond his person and were themselves dependent on structural forces. Weimar Germany was unstable and radical far-right groups were sprouting like mushrooms. Hitler's rise depended on the Versailles treaty, the results of the 1932 Reichstag election, the arrogance of the ruling elite thinking they could control him, the corruption of the Hindenburg family by Junker agrarian property owners, the inability of Kurt Von Scheicher to do his own counter-coup first (which may or may not have led to a quite similar state for Germany anyway). While those factors were certainly contingent, they were independent of Hitler being born or not, and we have good reason that another agitator in a similar position would lead to a very comparable outcome.
Even at an individual level, the idea that different people being born lead to incomparability is weak regardless of whether it leads to nihilism. Let's assume John is born instead of Fred. if I do not know Fred or John, and their respective lives are not meaningfully better or worse (one is a doctor, the other an engineer, let's say, instead of a doctor and a criminal), then their respective moral value to me is not incomparable: it is equal. I have no reason to prefer John over Fred. If I do know John or Fred, and let's say John is my friend, then they are very comparable: I have a clear reason to prefer John being born. They are comparable in both cases.
I address the thing you say at the end at the end.
As for the rest, even if you have a pretty strong view that the broad arc of history is inevitable, clearly changing the identity of everyone in the world will have a much bigger effect than your decision to go to the store.
I always approach the chaos theory stuff with a lot of skepticism. Isn't it possible states of the world are relatively resilient (eg, if Hitler had gotten into art school, someone else would've harnessed racial chauvinism in Germany)? Is there any proof of chaos theory that isn't purely theoretical?
It's possible but in any case the differences from rescrambling the identities of everyone in the world will obviously swamp the first-order effects of going to the store.
I guess my object-level question (I have not thought about this as much as you have to be sure) is how much actual rescrambling is going on. You just hand-wave it as "the identities of everyone in the world" because traffic patterns impact the timing/happenings of sex on the margins. I'm not sure you're right though. All the times I've had sex, a marginal (we are talking about seconds/minutes at most) change to traffic patterns would not have altered the event.
Bit of a side-question here, but what in your view determines someone's identity? What exactly is it that ensures that if "you" had been fertilized by a different sperm and egg, you would be a different person?
I'm not a philosopher and am probably missing something, but the argument seems nuts to me, and an example of what can go wrong with analytic philosophy.
To try to be slightly less unhelpful, here are two thoughts.
1. "A first problem is: this doesn’t really seem to be how incomparability works. If you knew that by pressing a button, there’d be a huge switch in the distribution of incomparable goods, and a child would get a lollipop, seems like the action is neither good nor bad. After all, the lollipop shouldn’t be the tiebreaker. "
This seems obviously wrong to me. Of course it's fine to give the child a lollipop.
2. And if it's not, you're in deep trouble, even if you ditch incomparability. Because now you have two *comparable* states -- with vast differences -- and one may really be much better than the other. And every action you take is going to affect which of them happens. Now you can't do anything without really vastly changing the world, and possibly making it much *worse* (rather than merely incomparable).
Of course you get to ignore that, because you don't know how it will affect it one way or the other. And that seems obviously true, whether you say that one state is actually worse, or that the states are incomparable.
The actual issue, to me, seems to be the assumption of ethical theories that require one to play god, not whether different states are comparable or not.
But quite probably I'm missing something through my admitted great ignorance.
//Of course it's fine to give the child a lollipop.//
It's fine to do it but not good IF INCOMPARABILITY IS TRUE. After all, by definition an incomparable good plus a small boost is neither good nor bad. So to be clear: I'm not saying that we should have the intuition that it's neither good nor bad to give the kid a lollipop. I'm saying that follows from incomparability.
One thing that you are missing is the other few objections I give to this proposal right below the one you discussed, which are more decisive.
Sorry for the slow reply. It's been an extremely hectic period for me.
I disagree with your retort. First of all, you say that incomparability is the belief that *some* states of affairs are incomparable. In some of your post, however, you act as though all states were.
In the instant case, the point is that the button-presser has no idea whether the states that result will be incomparable or not. As you say, a deontologist should care about good or bad consequences, when they're known. But a deontologist is *not* required to care about unknowable incomparable consequences. You make your little "deontologists care about consequences too" move -- which is true -- and then thereafter act as if you've reduced deontology to utilitarianism.
In fact, humans aren't God. We don't know a lot about the consequences of our actions. And when we don't, it's OK to just go ahead and do the good thing with the good consequences that we do know. In fact, it's good to do so.
Now, you direct me to your further objections. The next objection you raise ("a second, bigger problem") is basically that the objector's (your interlocutor)'s position would imply that knowing a little more could radically change the moral calculus. You then give the example of walking to the store, and learning that your actions will bring about the birth of Steve and/or Mary, and then changing your moral calculus.
Well, this brings us to another problem with the article -- *all of this applies without incomparability, too!* You think it's really weird? I'm sorry, but any actual human who *does* learn that his actions will bring about the birth of this or that person instead of another *would* *and does* become paralyzed and unable to pursue the decision. Far from being weird, it's what just about any actual human would really experience in that scenario. Faced with a view of the life each of Steve or Mary would have, the decision about the store would become intolerable and incalculable (even if theoretically comparable).
Fortunately, we are blessed with ignorance, a fundamental aspect of our existence. We don't know the chaotic consequences of our actions, and are thus meant to simply do the good things in front of us.
None of this changes if all states of affairs *are* strictly comparable. As you point out, our actions *do* have such dramatic effects, and you believe that all states of affairs *are* comparable, and so it's very possible that *your* giving a child a lollipop will bring about the existence of the next Hitler, and yet I don't see you advocating a moral calculus based on this fact. It's not because the state of affairs where you do give the lollipop is better -- in fact, it's far worse! But all the reasons why you don't (or anyway shouldn't) advise refraining from the gift on your actual ethical views apply just as well if one does assume incomparability.
Next, you go on to say that "Third, the ocean of incomparability is large enough that all our acts end up being neither good nor bad if incomparable goods exist." But, sorry, that's nonsense. It could be that large states of affairs are incomparable, and yet acts or events within them are very good. Certainly a deontologist or a virtue ethicist will insist on this. In the same way, even in a very good state of affairs, I might do evil (or do good in an evil state of affairs).
If I take two vast states of the world, each filled with misery and suffering and glory and beauty, and assume they are incommensurable, there is still nothing to prevent me from pointing to specific acts in each (absent in the other) and saying, "That is noble and good and beautiful."
Not everything can be turned into math. We are finite and it is of the essence of who we are.
I don't think this deals with every objection you raise, but my first thought is: most of the changes a small action of yours will make in who exists will be small, at least in the short term--there's a sort of continuity in change of identities as you change actions. Perhaps there's some measure of identity of potential future people where people only become incomparable past some threshold (or maybe better, how much extra good you need to add to break ties grows as a function of that distance), so that if the difference your action makes is that the same sperm and egg fertilize one second later than they otherwise would have, that means the resulting child is 99.9% the same, so the two potential children are only .1% incomparable, so the good you get of going to the store is sufficient to break the tie.
But if the difference is a different couple completely wins the lottery of who gets to be the once-a-year lucky couple to reproduce in your weird dystopian society, the two potential different kids are 99% incomparable.
Not sure if that helps at all? The idea that there ought to be some distance measure on potential future people sounds pretty reasonable to me regardless of anything else
First of all, the factual claim is wrong. You get totally different people in very short order. Like ten years from now, because of your decision to go to the store, I'd guess there are millions of different people.
Second, even if the only affects were in the far future, far future effects exist. So the puzzle still arises for the future.
Third, I think you get a different sperm and egg meeting by slightly delaying sex, so a totally different person.
What I'm imagining is that you integrate the effects of your actions over time, multiplied by some probability distribution that reflects your uncertainty over effects into the future; this should downweight future effects so your calculation is dominated by short-term effects (except where you have reasonable certainty, or the far future effect is big enough to dominate the uncertainty).
And I don't think different sperm and egg necessarily means "completely different person" in my framework: presumably the sperm and egg, as long as they come from the same parents, will share a lot of generic material, for example... So you could still have the person be "60% the same" as the alternate-world child. Obviously the details of this would matter a lot, and I have no deep thoughts on details.
I mean, I just don't see how that helps at all. You don't have uncertainty concerning whether your actions will affect distribution of incomaprable goods. See also the bit that begins with "There’s one objection you might be thinking of. Perhaps, when you normally act, because you don’t know which incomparable goods you are affecting, you get to ignore them."
Yeah you're right, I think to really make this work you need a sort of path-dependency: if you're deciding whether to have a lollipop, you're allowed to ignore the effects of incomparable goods and treat the lollipop as decisive if you think in expectation the decision to have a lollipop will be neutral in it's effects on incomparable goods; but if you're deciding between two incomparable goods and then add the lollipop, now you can't treat the lollipop as decisive.
I don't think an idea like this is totally unworkable: it really does seem like an important difference to say "I want a lollipop, and I can't quantify the weird effects it'll have on all these other things so I can ignore them" vs "I want to decide between two of those effects, can I use the decision to have a lollipop as a tiebreaker"--in particular there's a causal direction here that I think treats one differently: if the decision to have the lollipop causes the difference in incompletely incomparable goods, then it's impossible to use the lollipop as a tiebreaker if you're just trying to decide between the two incomparable goods, since by hypothesis the decision to have the lollipop is only compatible with one of the options.
But I agree this adds a lot more complication at best, and probably runs into worse problems at worst.
> It seems obvious that some actions we take are good. It’s a good thing to take a walk if doing so makes you happier.
It seems obvious that some actions we take are good *with respect to local goals*. It’s not at all obvious to me that some actions we take are better for total welfare, impartially considered. Why would my intuitions be well-calibrated to a metric as extremely complex and crucial consideration-prone as that?
(Indeed, I don't think you need incomparability to get very counterintuitive verdicts about local-scale decisions, from an impartial perspective. See Greaves's examples of complex cluelessness about global health interventions: https://philpapers.org/archive/GREC-38.pdf.)
ETA: Quoting you here (https://benthams.substack.com/p/the-world-cant-be-weird-fallacy), "You shouldn’t start out certain that some action is right, and then conclude that its effects must not be so bad. Instead, you should first consider the effects it likely has and whether they’re good or bad. You should then use that to determine if the action is right."
Hmm. I don't think nihilism follows from incomparability. Exchange of goods appears to me to be a counterexample.
Here it goes: heterogeneous goods are incommensurable. In fact, an equivalence relation cannot even be established over (direct) exchanges. For example: suppose you have three different goods A, B, & C. And suppose the following real-valued exchange ratios between them, a = (x units of A)/(y units of B), b = (u units of B)/(v units of C), c = (w units of A)/(z units of C). Then, since exchanges are mutually independent and subjective, it is not guaranteed that a*b=c. Good B occupies two different roles in the relation, but its quantities do not necessarily cancel.
So direct exchange is inconsistent. Since one and the same good can assume two different roles in the relation, there is no reason to suppose that transitivity will be a general feature of the relation. And since the relation isn't transitive, it isn't an equivalence relation.
However, it is still the case that heterogeneous goods are compared in reality (even in barter - so this generalizes outside of money-economies!), even if the comparison can't be "meaningfully" described or operationalized. So, summing this all up: there are exchanges between incommensurable goods, and people clearly aren't nihilistic about such exchanges between states of affairs! In fact, they seem to care a lot about them, even if this "caring" is only trivially meaningful.
I think it basically depends on whether you’re a prospectist or deferentialist about decision theory. Prospectists (like me!) will say that you often ought to do things even if you know the result will be incomparable in value to the alternative results.
See the bit at the end where I address this.
You address some similar ideas, but not quite the same. Part of the issue may just be that you stated things informally, but also this struck me as different from prospectism:
"when you normally act, because you don’t know which incomparable goods you are affecting, you get to ignore them."
In the usual "take the sugar cases," you *do* know which incomparable goods you're affecting. There's a fixed population, or a fixed tradeoff. In fact, you can state prospectism without having any reference to particular incomparable goods at all. The view just says "It's ok to do X iff X maximizes utility for at least *some* utility function that represents your preferences when they're made transitive."
A couple more comments.
i. About your button-lollipop case, which I liked. James Lenman has a similar case in his cluelessness paper -- "Spot the Dog on D-Day." In these "small tiebreaker" cases, I think the intuition that small goods can't break ties tends to go away once you stipulate that there aren't any deontological confounds (like "it would be ill-fitting to base such a momentous decision on such a trifling reason"). Small tiebreakers are basically unavoidable in ethics as long as you have real-valued utility functions playing a role somewhere or other.
ii. I wonder what you think about Richard Yetter Chappell's view that we can have incomparability (/parity) *and* transitivity. In particular, he thinks utility for A can be on par with utiltity B merely because they realize different final goods -- even though these goods can be precisely weighed up against each other.
//In the usual "take the sugar cases," you *do* know which incomparable goods you're affecting.//
But I gave arguments for why whether you know shouldn't affect it. I feel like I'm missing something.
//The view just says "It's ok to do X iff X maximizes utility for at least *some* utility function that represents your preferences when they're made transitive."//
Every act will be permissible if my argument goes through, but none will be better than others.
//Small tiebreakers are basically unavoidable in ethics as long as you have real-valued utility functions playing a role somewhere or other.//
But then how is this any different from just equality? If two things are such that neither is better than the other and tie-breaks outweigh, then this doesn't sound like incomparability.
// wonder what you think about Richard Yetter Chappell's view that we can have incomparability (/parity) *and* transitivity. //
I don't like that view because I don't believe in incomparability. But I'm not really sure how that has a unique advantage for the challenge I give here.
"how is this any different from just equality?"
There are two distinct values at play, and you might feel rational regret, in some respect, for bringing about one rather than the other. By contrast, you wouldn't feel any rational regret for picking up a $20 bill rather than two $10 bills. They are completely fungible in terms of monetary value. It's not as if you lose monetary value in some respect, though gain in another, when you pick up the different denomination.
"I don't like that view because I don't believe in incomparability."
But when you say "incomparability," you mean something stronger than what Richard and I mean. You mean the conjunction of
(RYC-style) Incomparability: for some pair of options, neither is better, nor are they equally good.
Insensitivity to Sweetening: for any pair of incomparable options x, y, there is some x+ such that x+ > x and yet x+ is incomparable with y.
The motivation for RYC-style Incomparability is what I was saying earlier: different people's utilities (e.g.) don't seem like they're just means to a single good -- total utility -- but rather seem like distinct goods in themselves. So we can feel rationally torn between x and y even if they promote the same total utility, just because x is good for some people while y is good for other people.
Do you feel the pull at all towards that kind of pluralism about the goods?
I see. Yeah, I have the Richard pluralisticy view that things can be valuable in different ways so they're not just equivalent contributors to some overall value.
I think of the view you're describing as not really incomparability. It's not that the goods can't be precisely compared. They can, and one penny would perturb the comparison. It's that they're not dealing in the same units, and so though they're comparable in terms of amounts of value, the fitting attitude is to feel a kind of regret.
I <3 the puralisticy view.
To be honest, I feel the force of the objections to insensitivity to sweetening. Your post is making me want to write a reply to explain why I can't bring myself to give it up...
Maybe neither here nor there but: I would rather have a $20 bill than twenty $1 coins.
It’s a good point - but one way to put it is that you don’t just care about *monetary* value in that case. You also care about convenience.
This is what i came here to say. The bit at the end just read to me as urging us to be deferentialists rather than prospectists, rather than arguing that the result is independent of that debate. Caspar Hare's work in particular seems worth engaging, as he accepts some pretty wild theses in this neighborhood (eg, from our current perspective, the holocaust was not regrettable, because of how later people's identities depend on it), but he doesn't think it's paralyzing for forward looking decision making, precisely because of the distinction Daniel is pointing to.
I'm not sure about this. Suppose that I think happiness and suffering are incomparable, in the sense that there is no precise exchange rate between them. 10 minutes of happiness followed by 10 minutes of suffering is neither better nor worse than no experience, and 11 minutes of happiness followed by 10 minutes of suffering is neither better nor worse than no experience, but 11 minutes of happiness followed by 10 minutes of suffering is better than 10 minutes of happiness followed by 10 minutes of suffering.
I'm considering driving to the store. Driving to the store might have all kinds of hard-to-foresee effects on the world, but it seems equally likely to me that *not* driving to the store will instead have those same hard-to-foresee effects. So in expectation, the hard-to-foresee effects of driving to the store cause X minutes of happiness and Y minutes of suffering, and the hard-to-foresee effects of *not* driving to the store cause X minutes of happiness and Y minutes of suffering. But driving to the store has the easy-to-foresee effect of giving me 1 minute of happiness, and that's the tiebreaker. X+1 minutes of happiness and Y minutes of suffering is better than X minutes of happiness and Y minutes of suffering.
Responded on substack notes.
If you haven't already, you might check out "The Infectiousness of Nihilism" by MacAskill, who argues that nihilism is a special case of incomparability, and that views like nihilism (and all forms of radical incomparability) basically break our ability to make moral decisions (all actions have undefined choiceworthiness) in light of moral uncertainty. Later, he (with co-authors) develop an account of dealing with incomparability based on normalizing their variance, so I guess he thinks that fixes the infectiousness problem (see second half of Chapter 6 of Moral Uncertainty, which builds on chapter 5)
I personally think we can modify Ross' Modified Dominance over Theories principle to be in terms of contrastive reasons so that we can ignore theories that posit radical incomparability, including nihilism, no matter what our credence is in them (except if it's 1, which is plausibly irrational), which is great news. Nihilism then becomes irrelevant, even if I believe in it.
I read that a while ago but I think MacAskill's point is somewhat different from mine. His is about the absence of a general formula for comparing incomparable goods if I remember correctly. Mine follows even if you have no normative uncertainty and treat incomparability the standard way.
Yep, I would just flag though: the key thing that incomparability proponents should think is implausible is just that the “big” actions we take (saving a life) don’t matter. But it seems to me the key move you make to show this is that, if we look far enough into the future, any act we take will have huge incomparable consequences. But then maybe whatever (implausible) things people want to say about how we should discount long-term future facts will apply here too, such that the big actions we take still matter.
Also, I’m not so confident our actions don’t have infinite effects. So we might all have to be nihilists assuming certain views about infinite ethics. But I think you’ve made that point before.
Agree with second half.
I mean, you get enough incomparable stuff quickly enough that I don't think this will work. And most ways to discount the future get more incomparability rather than less.
I think this is charming but much less compelling than it should be because you’re acting like a debater, not a philosopher (to be sure, I think a lot of so-called philosophers do this.) Do you really actually think that the existence of incomparable goods implies nihilism? That would be really wild! I think you ought to sound legitimately confused and anxious about such a situation! But it feels more like you’re just stringing together a possibly-valid argument that you’re pleased with.
On the object level, it seems clear that I can’t be culpable for consequences of my actions I can’t possibly predict, so while I have no commitment on the existence of incomparability, if I did I wouldn’t feel at all worried about the driving to the store argument. I think it’s much harder to surmount this sense that this argument can’t possibly matter than to make the argument itself, which might be a more productive way of making my first point.
Yes I do think the existence of incomparable goods implies nihilism. One way to infer that I think that is that the blog post that you are commenting on is titled "Incomparability Implies Nihilism."
You say "On the object level, it seems clear that I can’t be culpable for consequences of my actions I can’t possibly predict, so while I have no commitment on the existence of incomparability, if I did I wouldn’t feel at all worried about the driving to the store argument." But I addressed that objection in the post.
Yeah, sorry for posting the comment without actually making clear my response to all of your responses. The style of argumentation, as I say, just feels fundamentally incorrect. It’s just obviously false that I can be culpable for things I don’t know about; I can pretty well predict that “Bentham has four different reasons why this self-evident fact isn’t true” isn’t going to lead to me being moved on it by one iota, and indeed having now re-read these four complaints, they do not move me one iota. As I say, this does not come across, tonally or structurally, like somebody who’s discovered a real fact about goodness. I think you have better things to do.
I mean, I think the objections are pretty decisive. But I pretty fundamentally disagree with your comment.
First you say it seems "obviously false that I can be culpable for things I don’t know about." But you *do* know about them. You know your actions affect lots of incomparable goods. You don't know exactly which ones, but one can obviously be culpable for knowing their action has a class of effects even if they don't know which specific effect it has. For example, if you knew that by walking to the store you would cause something very bad to happen, then you shouldn't drive to the store, even though you don't know exactly what will happen.
I think this actually does look structurally like it might be a right insight. One thing about false theories is they often have deep structural problems that apply weird incoherence. So if we think that incomparability is non-existent--as, of course, I do--then we should expect something like this. Of course, maybe you reject that, but it's weird to reject an argument against A on grounds that you simply know at the outset that A is true.
That's helpful for suggesting a way to engage more deeply, will try to come back to this.
When I talk to consequentialists who believe in incomparable goods, I get the sense that they think:
(1) Actions A has many different consequences a1, a2, a3...
(2) Action A is better than action B iff:
(a) some consequence of A is better than some consequence of B
(b) no consequence of B is better than a consequence of A
Of course, (1) assumes more non-normative structure than in classical consequentialism, where actions have a single consequence W -- either the state of the world (in objective utilitarianism) or a distribution over worlds (in expectational utilitarianism).
So incomparable consequentialists assume some way to slice W into discrete consequences.
Formally, you can think of this as a family of functions from W to partially ordered sets.
- You could slice consequences up by people, which implies that action A beats B iff a is better than B for someone and not worse for anyone (i.e. pareto).
- You could slice consequences up by *sets* of people, which implies that action A beats B iff the populations are the same and the aggregate welfare is higher in A than in B.
- You could slice consequences up spatio-temporal regions.
- You could slice consequences up by 1st-order, 2nd-order, 3rd-order, etc (maybe).
- You could slice things up by an finitary consequence and an infinitary consequence (maybe), and similarly for each infinite cardinal.
- You could slice up by 'goods', e.g. how much justice there is vs how much beauty.
I'm not sure if the claim is that this gets around the issue of incomparability leading to nihilism. The problem still recurs even if you slice the world, and say that we'll measure the goodness of our action according to its consequences, when considered as a function into the partial orderings listed here.
However, I like the slices that you've proposed, I think I can explain how "giving a lollipop" to a girl" becomes incomparable in at least some of these slices. The stories become pretty similar pretty quickly though, so I'll only do the first 3 proposals.
- We slice up by people, and A beats B iff A is better than B for someone and not worse for anyone.
You give the girl a lollipop, she grows up to become dictator of the world, and loves the red colour of the lollipop so much that everyone with (natural) red hair gets treated as though in a utopia. Everyone else is treated far worse. If you don't give her the lollipop, the converse will happen, everyone with red hair will be punished due to her hatred of the colour red ever since she didn't get that red lollipop. Since in each case, different categories of people are worse off, it's neither good nor bad to give her the lollipop.
- We slice by sets of people, that A beats B iff the populations are the same and aggregate welfare is higher in A than B.
You give the girl a lollipop. Later on, she decides to have a child with Mr Rogers, the candy-store salesman, due to her fond recollections of the lollipop-gifter. This child is happy. If you don't give her a lollipop, she will have a child with Mr Gorers, the anti-candy-store activist due to her recollections of not being given a lollipop by that terrible person. That child is also happy overall, though less happy, since they don't get any candy. Two different people exist in these cases, and they're both worse off by not existing if you chose the other option, it's neither good nor bad to give her the lollipop
- We slice consequences by spatio-temporal regions.
You give the girl a lollipop. Later on, she becomes a billionaire and decides to invest massively in the area where you gave her that lollipop, causing everyone's welfare in that region to go massively up. However, this prices out people living the neighbourhood down the road who are worse off. You don't give her the lollipop, and she keeps walking and gets a lollipop in that next neighbourhood. She invests in that area instead, causing the precise opposite to occur. In either case, one spatial region will be worse off, and one better off depending on whether you give her the lollipop or not. So it's neither good nor bad to give her the lollipop.
I don't think that helps. I mean, first of all, that just doesn't work with how incomparability is explicitly supposed to work. This would imply that replacing Jon with Fred and giving a child a lollipop is good. But then how is incomparability different to two things being equally good? The thing that's supposed to make it distinct is that it can't be outweighed by small improvements.
Also, your criterion would rule out any actions being good. After all, all actions have many good and bad effects for butterfly reasons, so it's never true that "no consequence of B is better than a consequence of A." In fact, this doesn't even work with incomparability because incomparable goods always have unique advantages. Fred might go on to make great music--so some consequences of swapping out Fred for Jon are good and some are bad.
I don't really get the latter half of your comment.
Couldn't you say that *some* but not *all* small improvements can make a difference between incomparable things? Like, if a change is made to some feature of the world that is in some way separable from the two incomparables, it counts, but not otherwise?
Obviously this is tricky because in this case the things aren't totally separable: which set of people exists depends on whether you give the lollipop. But again, it seems to me there's something in the idea of looking at this casually that helps: improvements that casually precede the two incomparables or are causally disconnected do count, but improvements that casually follow (i.e., "ok we'll create Alice but not Bob, but conditional on creating Alice, Alice gets a lollipop") don't?
Is it built into incomparables that *all* sweeteners make no difference?
If you treat sweeteners differently by whether they're separable or not then you'll get weird results. Suppose we have the following options of which to create:
1 Bob + a child gets a lollipop
2 Bob who lives an extra day
3 Fred
Assume Bob and Fred are incomparable. If the lollipop is extrinsic but the extra day is not, then 1>3, 2 is incomparable with 3, but presumably 2>1. This is very strange.
It also seems highly arbitrary to have to group acts into whether they're extrinsic or not. And doesn't seem morally significant.
Lastly, the incomparable intuitions are supposed to hold for both extrinsic and intrinsic sweetenings. You're not supposed to swap out your child for another incomparable child and a nickel!
If those are your only options, the the child getting the lollipop isn't separable--they only get the lollipop in Bob world, not in Fred world. So 1 and 3 could remain incomparable. Then you'd just have 2 > 1 and each of them incomparable to 3, which seems... fine?
I think extrinsic vs intrinsic isn't the right distinction, which is why it feels morally arbitrary; I think you want to rather think of causality: it's much more intuitive to me that benefits that only accrue conditional on some scenario coming to pass might be "discounted" somehow by virtue of having to pass through that conditional bottleneck: if we create A, then A + B is better than A, but if we don't create A to begin with, then A+B might have an unclear relationship to ~A.
I'm not sure what you mean by "swap out", but I think one of the advantages of building the incompatibility relation to know about causal relationships is it can screen out cases where "swapping out" is possible: once you've had your child, you can't "swap" to a world where you didn't have that child and have a different one instead--you're stuck in worlds where you had that child, and so the only relevant considerations are those that condition on that.
In all these cases you're causing all the events in question. So causality can't be the relevant metric.
I've argued that normal actions are analogous to swapping out Fred for Bob and a lollipop, so if that action is neither good nor bad then neither is ordinary action.
But your creation of the lollipop is casually dependent on Bob being created, otherwise your option 1 couldn't be "Bob plus lollipop"--whether or not the kid gets the lollipop is casually downstream of whether Bob is created or not. You can't create Fred, and then give the kid a lollipop without leaving the set of options you've laid out.
What I'm arguing is that if you had option 4. Lollipop, and then afterwards decide what to do about Bob and Fred
then the lollipop is casually disconnected, and so it's comparable to anything you do downstream of it.
I think you're going to argue that this violates irrelevance of independent alternatives, since the seeming casual dependence on Bob's existence is an artifact of the fact that you've excluded the possibility 5. Fred and lollipop, and then I'm going to insist on making a distinction between deciding on the lollipop separately, and then deciding on Bob and Fred vs vice versa, or treating all the decisions as simultaneous.
I agree I don't love those features, especially the latter, but given that we actually do make choices constrained by causality in real life, I'm not sure how bad this can get in practice: if my choices are aligned with a causal graph, I can leave money on the table, but I don't think I can be money-pumped by someone who doesn't have a time machine, so my guess is you'd only get the mildest possible problems that can arise from rejecting irrelevance of alternatives. Anyone who believes in incomparables is probably already ok with that.
You need some Zen.Eventually everything is null and void because we live in a universe that will eventually become null and void. Who cares about the eventualities. What matters is the here and now bro. All the philosophical ramblings aside, if I can travel back in time in 1938 and assassinate Hitler I don't give a s***bought some hypothetical theoretical notion of what could or couldn't happen because of that action. That action right then and there would save 6 million people at least from the horrors of the Holocaust. Now whether that would cause some other horrible thing to happen in the future who gives hoot. We can never control the future so, stop all this Western thinking.
This reminds me of trait-based embryo selection, which Scott Alexander wrote about a few months ago. There is something intuitively frightening about it, as if for the first time, you can actively choose whether to create Jon or Fred. The reality is less dramatic; after all, parents already exercise great influence over what sort of people their children become through the combination of partner selection and upbringing. Changing someone’s eye color (or even IQ by a few points) is pretty trivial by comparison. Yet the technology is still controversial. Some of the controversy has to do with practical downsides, but one objection is more fundamental: the idea that choosing these traits is playing God, doing something that humans are not meant to do. Said objection is usually raised in a literal religious context, but it has intuitive resonance even for me as an atheist. Not enough to make me think trait-based embryo selection is wrong (certainly not with the milquetoast effects of current technology), but enough to make me wonder whether a moral boundary is being approached and what that boundary might be.
But of course, trait-based embryo selection is simply replacing a random choice with an informed choice. In regular IVF, the parent is already selecting an embryo, knowing that there is a cause-and-effect relationship between which embryo they choose and what sort of person is created. The new technology just gives them somewhat better information about what that relationship is.
If I follow the intuition that there is some moral boundary here (that would be breached by a hypothetical more-precise future version of the technology), then I’m led directly to “a distinction between incomparable goods you know about and ones you don’t know about”. As well as a need to regard an act as good (namely, selecting an embryo, let’s say before seeing the screening results) despite knowing, with certainty, that if one knew more (i.e. saw the screening results) then one would consider it bad.
I’m not sure I should follow the intuition. Perhaps this is more unsalvageable mess. But I’m not sure it’s not an unavoidable mess.
You are missing the most obvious option:
When computing total utility of an action, just ignore all the incomparable goods whether we know about them or not. Simply discard everything incomparable and reason about comparable. Then everything adds up to normality.
This is, in fact, a standard practice in data processing.
This doesn't work because incomparable goods contain lots of other sub goods. For example, suppose that Bob and Fred are incomparable. Well, Bob is also incomparable with the last 80 years of Fred's life, so you can ignore those and just take into account the first year. There's no natural way of carving goods up.
> This doesn't work
Posting on the Internet that it doesn't work is like speaking out loud that speech is impossible. With all likelihood, data pipelines for this very site use such calculations.
See this: https://en.wikipedia.org/wiki/NaN
> incomparable goods contain lots of other sub goods.
That doesn't matter at all. Just discard all that is incomparable. In the example from the post: The action of going to the shop leads to three consequences:
1. Satisfying my desire to get to the shop - lets say it's 2 utilons
2. Non-existence of one person - incomparable - NaN utilons
3. Existence of the other person - incomparable - NaN utilons
The total utility is calculated by discarding all the NaN's and summing all the numeric values. Which is 2 utilons. So the action is slightly positive.
You just ignored the argument and repeated yours
You are saying something doesn't work. I show you that it does, pointing to the actual thing in the reality that is using this principle. Then, I show you how we can execute this algorithm, with the specific example from the post, to make it even clearer.
On the other hand, your counterargument is completely irrelevant to the point I'm making. Nothing in my reasoning was based on the fact that contents of incomparable stuff has to be comparable. At no point I said that we should ignore only the first year of Bobs life so I don't think what it has to do with anything.
If there’s no way of divvying up sets of goods into packages that are then compared then you can’t subtract out the incomparable sets.
We can't in a general case divide incomparable goods into comparable parts. This doesn't mean that we can't divide sets of goods into packages at all. When we have a set we can talk about it individual elements and construct subsets from them. Some of the subsets can be *non-measurable*. It doesn't mean that the notion of measure suddenly looses all meaning.
https://www.google.com/url?sa=t&source=web&rct=j&opi=89978449&url=https://en.wikipedia.org/wiki/Non-measurable_set&ved=2ahUKEwj50aaIy-SQAxXj1gIHHZrVJJgQFnoECBsQAQ&usg=AOvVaw3QJMNHfCs_RhFI8-Y_Do8b
Consider this:
NaN + n = NaN for any number n. So, if we simply have a NaN value we can't derive anything from it.
But if we have a list of values, some of which may be NaNs, nothing prevents us from adding only the non-NaNs together and using the result in further calculations. It's a simple function:
def sum(l):
s = 0
for v in l:
if v is not np.nan:
s += v
return s
I understand why it can be counterintuitive for people without mathematical or programming background, but surely you are eventually able to notice a counterexample when it's right in front of you? The problem you claim to be unsolvable, in principle, is solved so much that we've been building technology on this solution for decades.
What is the point of worrying about consequences past my horizon of knowledge, beyond which all causality is incomprehensible anyway?
Why should I feel culpable, as a decision-making creature, for an infinite array of hypothetical ripple effects propagating from even the tiniest action?
It seems like in the face of this kind of illegibility of any final accounting, all I can do is *try* to do the right thing to the best of my understanding, knowing full well that my moral reasoning is but a dim flashlight in a dark ocean.
I address this at the end.
I do not think stocking the world with "completely different people" leads to a meaningfully different world on aggregate if structural forces are the same. To use your example of driving to the store, if you have political, economic, geographic and environmental factors unchanged between group A and B, then those factors are severely constraining the outcomes, and group A and B will be in fact comparable and very similar regardless of whether you drive to the store or not. That some people have different life trajectories does not matter at all if statistically speaking you get the same number of doctors, engineers, drug addicts, and so on.
To use the well-worn Hitler example, to the extent that his rise to power was dependent on contingent factors, those factors extended far beyond his person and were themselves dependent on structural forces. Weimar Germany was unstable and radical far-right groups were sprouting like mushrooms. Hitler's rise depended on the Versailles treaty, the results of the 1932 Reichstag election, the arrogance of the ruling elite thinking they could control him, the corruption of the Hindenburg family by Junker agrarian property owners, the inability of Kurt Von Scheicher to do his own counter-coup first (which may or may not have led to a quite similar state for Germany anyway). While those factors were certainly contingent, they were independent of Hitler being born or not, and we have good reason that another agitator in a similar position would lead to a very comparable outcome.
Even at an individual level, the idea that different people being born lead to incomparability is weak regardless of whether it leads to nihilism. Let's assume John is born instead of Fred. if I do not know Fred or John, and their respective lives are not meaningfully better or worse (one is a doctor, the other an engineer, let's say, instead of a doctor and a criminal), then their respective moral value to me is not incomparable: it is equal. I have no reason to prefer John over Fred. If I do know John or Fred, and let's say John is my friend, then they are very comparable: I have a clear reason to prefer John being born. They are comparable in both cases.
I address the thing you say at the end at the end.
As for the rest, even if you have a pretty strong view that the broad arc of history is inevitable, clearly changing the identity of everyone in the world will have a much bigger effect than your decision to go to the store.
I always approach the chaos theory stuff with a lot of skepticism. Isn't it possible states of the world are relatively resilient (eg, if Hitler had gotten into art school, someone else would've harnessed racial chauvinism in Germany)? Is there any proof of chaos theory that isn't purely theoretical?
It's possible but in any case the differences from rescrambling the identities of everyone in the world will obviously swamp the first-order effects of going to the store.
I guess my object-level question (I have not thought about this as much as you have to be sure) is how much actual rescrambling is going on. You just hand-wave it as "the identities of everyone in the world" because traffic patterns impact the timing/happenings of sex on the margins. I'm not sure you're right though. All the times I've had sex, a marginal (we are talking about seconds/minutes at most) change to traffic patterns would not have altered the event.
It would have altered the event if you had a child! It would have totally changed their identity. https://www.globalprioritiesinstitute.org/wp-content/uploads/2019/MacAskill_Mogensen_Paralysis_Argument.pdf
Bit of a side-question here, but what in your view determines someone's identity? What exactly is it that ensures that if "you" had been fertilized by a different sperm and egg, you would be a different person?
Ok, but then it's not obvious that I (a deontologist) should care--I don't believe in anyone's "right to be unscrambled."
I address that in the piece.
I'm not a philosopher and am probably missing something, but the argument seems nuts to me, and an example of what can go wrong with analytic philosophy.
To try to be slightly less unhelpful, here are two thoughts.
1. "A first problem is: this doesn’t really seem to be how incomparability works. If you knew that by pressing a button, there’d be a huge switch in the distribution of incomparable goods, and a child would get a lollipop, seems like the action is neither good nor bad. After all, the lollipop shouldn’t be the tiebreaker. "
This seems obviously wrong to me. Of course it's fine to give the child a lollipop.
2. And if it's not, you're in deep trouble, even if you ditch incomparability. Because now you have two *comparable* states -- with vast differences -- and one may really be much better than the other. And every action you take is going to affect which of them happens. Now you can't do anything without really vastly changing the world, and possibly making it much *worse* (rather than merely incomparable).
Of course you get to ignore that, because you don't know how it will affect it one way or the other. And that seems obviously true, whether you say that one state is actually worse, or that the states are incomparable.
The actual issue, to me, seems to be the assumption of ethical theories that require one to play god, not whether different states are comparable or not.
But quite probably I'm missing something through my admitted great ignorance.
//Of course it's fine to give the child a lollipop.//
It's fine to do it but not good IF INCOMPARABILITY IS TRUE. After all, by definition an incomparable good plus a small boost is neither good nor bad. So to be clear: I'm not saying that we should have the intuition that it's neither good nor bad to give the kid a lollipop. I'm saying that follows from incomparability.
One thing that you are missing is the other few objections I give to this proposal right below the one you discussed, which are more decisive.
Sorry for the slow reply. It's been an extremely hectic period for me.
I disagree with your retort. First of all, you say that incomparability is the belief that *some* states of affairs are incomparable. In some of your post, however, you act as though all states were.
In the instant case, the point is that the button-presser has no idea whether the states that result will be incomparable or not. As you say, a deontologist should care about good or bad consequences, when they're known. But a deontologist is *not* required to care about unknowable incomparable consequences. You make your little "deontologists care about consequences too" move -- which is true -- and then thereafter act as if you've reduced deontology to utilitarianism.
In fact, humans aren't God. We don't know a lot about the consequences of our actions. And when we don't, it's OK to just go ahead and do the good thing with the good consequences that we do know. In fact, it's good to do so.
Now, you direct me to your further objections. The next objection you raise ("a second, bigger problem") is basically that the objector's (your interlocutor)'s position would imply that knowing a little more could radically change the moral calculus. You then give the example of walking to the store, and learning that your actions will bring about the birth of Steve and/or Mary, and then changing your moral calculus.
Well, this brings us to another problem with the article -- *all of this applies without incomparability, too!* You think it's really weird? I'm sorry, but any actual human who *does* learn that his actions will bring about the birth of this or that person instead of another *would* *and does* become paralyzed and unable to pursue the decision. Far from being weird, it's what just about any actual human would really experience in that scenario. Faced with a view of the life each of Steve or Mary would have, the decision about the store would become intolerable and incalculable (even if theoretically comparable).
Fortunately, we are blessed with ignorance, a fundamental aspect of our existence. We don't know the chaotic consequences of our actions, and are thus meant to simply do the good things in front of us.
None of this changes if all states of affairs *are* strictly comparable. As you point out, our actions *do* have such dramatic effects, and you believe that all states of affairs *are* comparable, and so it's very possible that *your* giving a child a lollipop will bring about the existence of the next Hitler, and yet I don't see you advocating a moral calculus based on this fact. It's not because the state of affairs where you do give the lollipop is better -- in fact, it's far worse! But all the reasons why you don't (or anyway shouldn't) advise refraining from the gift on your actual ethical views apply just as well if one does assume incomparability.
Next, you go on to say that "Third, the ocean of incomparability is large enough that all our acts end up being neither good nor bad if incomparable goods exist." But, sorry, that's nonsense. It could be that large states of affairs are incomparable, and yet acts or events within them are very good. Certainly a deontologist or a virtue ethicist will insist on this. In the same way, even in a very good state of affairs, I might do evil (or do good in an evil state of affairs).
If I take two vast states of the world, each filled with misery and suffering and glory and beauty, and assume they are incommensurable, there is still nothing to prevent me from pointing to specific acts in each (absent in the other) and saying, "That is noble and good and beautiful."
Not everything can be turned into math. We are finite and it is of the essence of who we are.
I don't think this deals with every objection you raise, but my first thought is: most of the changes a small action of yours will make in who exists will be small, at least in the short term--there's a sort of continuity in change of identities as you change actions. Perhaps there's some measure of identity of potential future people where people only become incomparable past some threshold (or maybe better, how much extra good you need to add to break ties grows as a function of that distance), so that if the difference your action makes is that the same sperm and egg fertilize one second later than they otherwise would have, that means the resulting child is 99.9% the same, so the two potential children are only .1% incomparable, so the good you get of going to the store is sufficient to break the tie.
But if the difference is a different couple completely wins the lottery of who gets to be the once-a-year lucky couple to reproduce in your weird dystopian society, the two potential different kids are 99% incomparable.
Not sure if that helps at all? The idea that there ought to be some distance measure on potential future people sounds pretty reasonable to me regardless of anything else
I don't think this helps.
First of all, the factual claim is wrong. You get totally different people in very short order. Like ten years from now, because of your decision to go to the store, I'd guess there are millions of different people.
Second, even if the only affects were in the far future, far future effects exist. So the puzzle still arises for the future.
Third, I think you get a different sperm and egg meeting by slightly delaying sex, so a totally different person.
What I'm imagining is that you integrate the effects of your actions over time, multiplied by some probability distribution that reflects your uncertainty over effects into the future; this should downweight future effects so your calculation is dominated by short-term effects (except where you have reasonable certainty, or the far future effect is big enough to dominate the uncertainty).
And I don't think different sperm and egg necessarily means "completely different person" in my framework: presumably the sperm and egg, as long as they come from the same parents, will share a lot of generic material, for example... So you could still have the person be "60% the same" as the alternate-world child. Obviously the details of this would matter a lot, and I have no deep thoughts on details.
I mean, I just don't see how that helps at all. You don't have uncertainty concerning whether your actions will affect distribution of incomaprable goods. See also the bit that begins with "There’s one objection you might be thinking of. Perhaps, when you normally act, because you don’t know which incomparable goods you are affecting, you get to ignore them."
Yeah you're right, I think to really make this work you need a sort of path-dependency: if you're deciding whether to have a lollipop, you're allowed to ignore the effects of incomparable goods and treat the lollipop as decisive if you think in expectation the decision to have a lollipop will be neutral in it's effects on incomparable goods; but if you're deciding between two incomparable goods and then add the lollipop, now you can't treat the lollipop as decisive.
I don't think an idea like this is totally unworkable: it really does seem like an important difference to say "I want a lollipop, and I can't quantify the weird effects it'll have on all these other things so I can ignore them" vs "I want to decide between two of those effects, can I use the decision to have a lollipop as a tiebreaker"--in particular there's a causal direction here that I think treats one differently: if the decision to have the lollipop causes the difference in incompletely incomparable goods, then it's impossible to use the lollipop as a tiebreaker if you're just trying to decide between the two incomparable goods, since by hypothesis the decision to have the lollipop is only compatible with one of the options.
But I agree this adds a lot more complication at best, and probably runs into worse problems at worst.
See the bit in the article where I give objections to this proposal.
> It seems obvious that some actions we take are good. It’s a good thing to take a walk if doing so makes you happier.
It seems obvious that some actions we take are good *with respect to local goals*. It’s not at all obvious to me that some actions we take are better for total welfare, impartially considered. Why would my intuitions be well-calibrated to a metric as extremely complex and crucial consideration-prone as that?
(Indeed, I don't think you need incomparability to get very counterintuitive verdicts about local-scale decisions, from an impartial perspective. See Greaves's examples of complex cluelessness about global health interventions: https://philpapers.org/archive/GREC-38.pdf.)
ETA: Quoting you here (https://benthams.substack.com/p/the-world-cant-be-weird-fallacy), "You shouldn’t start out certain that some action is right, and then conclude that its effects must not be so bad. Instead, you should first consider the effects it likely has and whether they’re good or bad. You should then use that to determine if the action is right."
Hmm. I don't think nihilism follows from incomparability. Exchange of goods appears to me to be a counterexample.
Here it goes: heterogeneous goods are incommensurable. In fact, an equivalence relation cannot even be established over (direct) exchanges. For example: suppose you have three different goods A, B, & C. And suppose the following real-valued exchange ratios between them, a = (x units of A)/(y units of B), b = (u units of B)/(v units of C), c = (w units of A)/(z units of C). Then, since exchanges are mutually independent and subjective, it is not guaranteed that a*b=c. Good B occupies two different roles in the relation, but its quantities do not necessarily cancel.
So direct exchange is inconsistent. Since one and the same good can assume two different roles in the relation, there is no reason to suppose that transitivity will be a general feature of the relation. And since the relation isn't transitive, it isn't an equivalence relation.
However, it is still the case that heterogeneous goods are compared in reality (even in barter - so this generalizes outside of money-economies!), even if the comparison can't be "meaningfully" described or operationalized. So, summing this all up: there are exchanges between incommensurable goods, and people clearly aren't nihilistic about such exchanges between states of affairs! In fact, they seem to care a lot about them, even if this "caring" is only trivially meaningful.