Great response!

Unfortunately, building on Parfit’s work philosophers have produced formal proofs, known as impossibility theorems, showing that every ethical position has some repugnant-seeming implication or other.

Maybe this is a nitpick, but I think this is overestated and that the consequences could actually be quite important for cause prioritization, largely through EAs taking person-affecting views and other alternatives to the total view much more seriously: incomparability/incompleteness and violations of "the independence of irrelevant alternatives" (and maybe intransitivity) are plausible ways around these impossibility theorems, and I don't think people who object to them typically describe them as repugnant, only (perhaps very) counterintuitive. I think they are counterintuitive if you start from mathematical assumptions on the structure of comparisons first, e.g. a single complete and transitive preference relation. However, I think we should put empathy first and let the structure of comparisons follow from that, and some important cases of incomparability/incompleteness and violations of the independence of irrelevant alternatives can be explained from this frame.

Even the name "independence of irrelevant alternatives" and examples people often use to object to its violation (e.g. picking what to eat) seem misleading, because if a theory violating the independence of irrelevant alternatives is reasonably well-motivated and justified, it would give us plausible explanations for violations and the relevance of the alternatives. So the alternatives wouldn't in fact be irrelevant!

I have heard some describe person-affecting views and similar views in quite derisive terms, though, so presumably find them and some of their implications repugnant.

I definitely appreciate how seriously you and many others in the community take moral uncertainty, moderation and cooperation.

Do you think you've ever taken an action (or sequence of actions) that, with your current understanding of cluelessness and unwarenesss, should have been ex ante c-preferred to another (or doing nothing, specifically)? Like thinking a bit more about specific backfire risks or cluelessness, or breathing?

EDIT: Also some more weirder things, like not kicking a puppy when given the chance and no one else would know. Or, say, if you've already stepped on a snail and it's clearly going to die, should you put it out of its misery?

Here's the draft!

I think it's helpful to separate positive impact and negative impact causal paths specifically. By still grouping them together in 3, it could stack the deck in favour of ignoring them together, rather than say grouping different items together and ignoring them together. So, instead, we could have something like:

1. I get a tasty snack right away.
2. Some broken glass is left somewhere. I am confident it will remain buried for 100 years, but after that I think it is highly likely to resurface and cause someone a nasty injury, due to where it is located.
3. increase the probability of a dystopian lock in via a set of causal paths X.
4. decrease the probability of a dystopian lock in via a set of causal paths Y.
5. decrease the probability of a dystopian lock in via a set of causal paths Z.
6. other good things.
7. other bad things.

And, of course, we could divide up each of 3-7 into further items. Or we could combine different subsets of them into one item.

So, my question would be, how should we divide and group these items together to decide which to ignore, and why?

I agree with the meta point, but my reaction is that if this particular theory is practically never action-guiding, I'll just move onto other theories that could be action-guiding, say via some kind of metanormative bracketing or wager.

I think treating tiny probabilities of severe suffering (or any large difference for an individual) as lower would probably work to often give determinate comparisons, as you suggest. We could get a range of views that set the thresholds between higher and lower in different places and have moral uncertainty over them.

It could go too far or turn on empirical facts that in my view it shouldn't really, although this is pulling in nonidentity/counterpart intuitions, so basically question begging.

Generally any moral patients who do not yet exist / are not yet conceived could be dramatically discounted because each such pre-specified individual is extremely unlikely to exist, based on the huge number of sperm cells in the race to fertilize an egg. The view could practically be a strict person-affecting view like presentism or necessitarianism.

All of our apparently cost-effective animal welfare interventions could only help animals to the lower level, because none of them help already conceived individuals, given the delay between the decision to support an intervention and its actual implementation.

Causing millions more chickens to be factory farmed per year would only reach the lower level for each resulting chicken ex ante.

Plausibly each specific far future moral patient is far far far less likely to exist than any specific chicken in the next 10 years. Each individual's probability of existing could roughly decrease exponentially with their generation number from the present, at least for biological moral patients. So, there could be thresholds where chickens over the next few years reach higher, but far future moral patients only reach lower. But should the view turn on these particulars?

I second Impartial Partially's answer, but figured I'd address another issue here re presentation which stacks the deck in one direction. I think it will seem less intuitively objectionable if you name particular ways someone at risk of the broken glass could benefit from 3, and separate those into separate points, rather than (fairly arbitrarily, imo) grouping all harms and benefits together in 3 besides 1 and 2.

I'd be curious about how you think about transworld identity and personal identity and to get your take on the following proposal for counterpart relations. I expect that bottom-up bracketing would rarely be action-guiding in practice without counterpart relations, because of huge numbers of contingent moral patients pointing in each direction, and counterpart relations could make some version action-guiding in practice by allowing us to bracket out far more of them.

This is a pretty rough sketch.

One way I'm thinking of grounding counterpart relations is by looking for measures that naturally generalize our concepts of transworld and personal identity. For example, Alice in A is the same person as Alice in B and her past self because of common causes of the features important to Alice's identity, like her memories, psychological dispositions, etc.. Maybe we'd say specifically due to the preservation or inheritance of the physical structures (her brain) responsible for her psychological features from her past selves.

All of this is a matter of degree, because Alice changes over time and could diverge between worlds A and B. Almost all of the particles in her neurons and elsewhere in her body are replaced over time, and which particles make her up will diverge across worlds. In cases of amnesia, brain damage and normal human aging and experience, people can change fairly dramatically in their psychological features. We might say that transworld and personal identity matching are themselves graded. Alice is kind of both the same person and kind of a different person, and more a different person the larger the divergence and the less is inherited.

Then, Alice and Bob, if both human, share more psychological dispositions than Alice and a chicken, and this can be traced back to more recent common causes of their psychological dispositions, i.e. more recent common ancestry between any two humans than between a human and a chicken, passed through genes. This could let us preferentially match counterparts that are more closely related/trace more of their genes that account for psychological features back to the same more recent ancestors.

Transworld identity is just a special case of this kind of causal counterpart relation where the common causes of psychological features are especially potent or of a particular kind, e.g. the "same" brain, which we can trace through spacetime. The gap between distinct persons is admittedly generally much larger than between the same person across worlds or at different points in time, but if it's more a difference in degree than kind since both are graded, then this would give us reason to accept or reject both transworld/personal identity and causal counterpart relations together, rather than accept transworld/personal identiy and reject causal counterpart relations.

One potential supporting argument for the generalization from direct inheritance of psychological features to common causes of psychological features for counterpart relations:

1. While adult Alice inherits some psychological features from her infant self and infant brain, many of her cognitive capacities as an adult — which are plausibly important to her identity — were not directly inherited from her infant self or infant brain (or fetal brain), but were caused in part by general and often region-specific DNA-driven processes to restructure her brain. Furthermore, the vast majority of the particles in her neurons and brain get replaced even if most neurons are preserved, and (I'd guess) the DNA in those neurons are necessary for ensuring their function even as the particles in them turn over. So, Alice's psychological features and identity over time and across worlds come in large part from her DNA. The same has been typically true for conscious biological beings on Earth so far.
2. The persistent DNA particles over time cause Alice's psychological features over time and across worlds. Copies of them from her ancestors cause many of her relative's similar psychological features over time and across worlds. What grounds her identity across time and worlds is common causes of her psychological features, but going back up the causal chain, we find common causes of psychological features between moral patients, too, not just within, through copies of the same DNA.^[1]
3. If common causes, like someone's DNA, of shared psychological features across worlds are what ground transworld identity, then we should be able to use common causes, like a common ancestor's DNA from which copies were inherited, of shared psychological features to ground counterpart relations across worlds to generalize transworld identity, too.

We can use other common causes of shared psychological features to ground counterparts more generally. Or even common causes of individuals' existences across worlds, e.g. if someone is deciding between dog breeders to breed a puppy for them (or buying a puppy from them and creating demand for further puppies), they are causally downstream of each of the possible resulting puppies, and this further common cause supports matching the possible resulting puppies across worlds as counterparts.

1. ^{^}

   I also don't think it would make a difference if all of Alice's DNA got replaced by exact copies and none of her particles were preserved over time. Either causal relationships or continuity are doing the work here.

Is this a version of (weak) actualism as a person-affecting theory? In worlds where Alice doesn't exist, she has no properties like welfare to provide us any reasons or welfare-based objections.

Weak actualism, as in the literature (Hare, 2007, Spencer, 2021 section 6), would say to pick among the options that maximize "self-conditional value". An option is permissible only if it's no worse than any other for the people who (ever) exist in the resulting outcome. In other words, assuming we choose the option A, we compare all the options pairwise on the basis of the reasons provided by the people who would exist given our choice of A,^[1] and if A is no worse overall than any other option on these comparisons, then it maximizes self-conditional value.^[2]

It sounds similar to what you wrote here, too:

Perhaps, on a person-centered approach to ethics, that is simply true (cf. Boonin 2014 and Pummer 2024): If we choose B, we will have made it the case that (i) no actual person has negative welfare, which they could object to on personal welfare grounds, and (ii) all actual persons have the maximum amount of positive welfare they could have had. (Person p2, once she exists, will have no welfare-based objection to our choosing B: We could not have made her better off. Nor will p1, who will not exist.) On a person-centered approach, we may therefore have maximized people’s welfare in the sense that matters.

1. ^{^}

   Weighing the reasons by the probability that those reasons exist at all, say.

2. ^{^}

   Weak actualism doesn't need to say that all options that maximize self-conditional value are permissible. There are some cases where we can plausibly recommend choices between them.

   I also think maximizing self-conditional value as formalized in the literature is the wrong approach for an actualist when deciding between at least 3 options, because it's myopic, and we should use backward induction on behalf of the people who would exist given an option. This can avoid the repugnant conclusion. I have an unpublished draft based on the idea of backward induction for weak actualism I'd be happy to share, and I discuss the problem a bit here. 

Hmm, I think I made some unjustified leaps in the above comment.

Some contingent moral patients in A could be determinately better off in expectation (conditional on existing) than the statistical average contingent moral patient in B, and some contingent moral patients in A could be determinately worse off in expectation (conditional on existing) than the statistical average contingent moral patient in B. So, with counterparts:

1. on bottom-up bracketing, we wouldn't bracket the determinately better off than average contingent moral patients and determinately worse off than average contingent moral patients out. We do get to bracket out those where the comparison is indeterminate, by overlapping in range of welfare with the average. I'd guess it's possible to specify some that are far enough away from the average, but this could be wrong.
2. on top-down bracketing, whether they get bracketed out probably depends a lot more on the specifics.

There's also a question of whether we should handle counterparts iteratively, and then if order matters. If we bracket out Alice in A with her statistical counterparts in B, then it doesn't seem like Alice should be left as an available counterpart to those in B (after pulling out one statistically average contingent moral patient from B), so Alice shouldn't be part of the statistical average for counterparts of those left in B. So the statistical average of remaining moral patients in A needs to be recomputed. But if we bracket out the contingent moral patient Alex in A first instead of Alice, could this change the verdict of the comparison?

Very interesting!

How do you recommend choosing between somewhere-uniquely-optimal actions if there are multiple? I'm imagining the worst case where we still have several left that we're clueless between.^[1]

And are all convex combinations of (lotteries over) somewhere-uniquely-optimal actions also somewhere-uniquely-optimal? (Let’s suppose our set of actions is closed under convex combinations, so somewhere-unique-optimality is defined with respect to that set.)

EDIT: The answer is no. 60-For-Sure, 100-If-True and 100-If-False are all somewhere-uniquely-optimal (under convex combinations), but 60-For-Sure dominates the convex combination of picking one of 100-If-True or 100-If-False at random, with 50% probability each.

1. ^{^}

   I suppose the actual worst case is where inaction is ~always somewhere uniquely-optimal, and especially when building a portfolio to give across organizations, but that will depend on the particulars involved.