Seems Wenar was talking about credit assignment (where you want to share a fixed pie of credit fairly, which something like Aumann Shapley handles and says each person gets credit for one life) not a decision rule (which Parfit’s counterfactual view handles). These are two different things.
Sure, but it does seem likely that the origin of the disagreement was a misunderstanding of what MacAskill was weighing (what decision to take) and that Wenar might have thought he was talking about post facto credit assignment. Which is still a dumb mistake for Wener to make and I can almost smell the motivated reasoning (wanting to dunk on those dang smug EAs). But maybe that common conflation is the more important thing to address first. (Still, I enjoyed your post and the examples in it.)
I've been thinking about this for a while. I'd get confused whenever I assessed counterfactual credit and found it bigger (in sum) than counterfactual impact. Now I know I should read more "moral mathematics"
I was also confused by the article and wondered, "but the kid would counterfactually die without the donation?" It's weird how Wenar thinks a donor saying "my donation saved a life" takes away from the parent who allowed it. Neither GiveWell nor any other org has ever claimed that ALL the credit is theirs, only that they counterfactually save lives.
Seems like the share of total view confuses personal responsibility with ideal decision making - but it's easy to see why. The Talmud (Bava Kama 10b) considers a case where 5 people sat on a bench and then a 6th person joined them and the bench broke. The Talmud is concerned with who pays for the bench. Do we divide the cost among the 6 or does the last guy pay it all? But now ask a different question: What is less bad, for the 6th guy to join them on the big bench and break it, or for this person to sit down on a cheaper bench and somehow single handedly break that? Obviously the second is less bad; there's less total damage. But here's the thing: Say the big bench costs $300 and the small one is $100. Now imagine yourself telling this person "Sit on the cheaper bench and you'll pay $100 in damages rather than $50, but hey you're a better person!" Silly example, but you know what I mean.
The 6th person to sit on the big bench should pay to repair it though, or at least we anchor on that and negotiate contributions from the others (the non-marginal breakers)
This is in large part why I think the fixation on moral responsibility is really weird. Ethics has for a long time had a strong fixation on moral desert and who gets to take credit for things. But it just doesn’t seem to matter at least from a utilitarian perspective all that matters is the counterfactuals. So in the sense of making rational assessments of what to do this is very important. But in another sense, that of actually dividing the credit it’s totally trivial and meaningless, which is another reason why we really should not be bothered if it multiplies the credit well beyond a 1 to 1 accounting of the impact caused.
In practice, doesn't this just trivially give bad answers if you try to do charity evaluation?
Imagine donating $3000 to Charity A that saves 1 life per marginal donation, vs donating to Charity B that saves 2 lives at $300k (and there's an assurance contract where 99 other people donate to it iff you also donate $3000).
It seems to me like donating to Charity A is much better! Do you disagree?
Does it in this scenario? It seems like it is highly unlikely to be the case that donating to charity B is worth it unless the other 99 people are using their 297,000 to contribute to net negative utility things. As long as they engage in reasonably positive or neutral utility behavior it seems likely that this much money saves at least 1 life in its normal operation through the economy.
Obviously to his point it’s really hard to do this kind of math but still.
You can estimate this through several different means, one of which is US willingness to pay for a QALY.
Generally you shouldn't expect normal economic operation to be more effective than optimized health spending, so that creates some bounds.
Another way to think about it is that US GDP per capita is something like 94k/year. It'd be surprising if normal economic operation for 3 years here saves a life! Would indicate substantially increasing marginal returns for economic operation, which flies in the face of everything we know about the relationship between growth and health.
From my brief survey of some paper abstracts it seems like the U.S. willingness to pay for QALY is 100-150000 so it doesn’t immediately rule out that you would get 1+ QALY out of 297k.
Secondly the 297k is specifically NOT optimized health spending. In this example the 300k for 2 lives is a very inefficient non optimal use of funds.
I think if normal economic operation falls below the level of 297k for one life it’s gonna because there are substantial negative externalities to normal economic operation. Not because that normal operation isn’t contributing to wellbeing.
You are correct, but also kind of fighting the hypothetical. Your theory does have the problem that under perfectly imaginable conditions, it gives the wrong answer in these kinds of thought experiments. After all this perfectly possible to imagine a situation where those are the 99 people behave in ways where the first charity is still better as a target for donation.
Well in the full analysis of the counterfactual what matters I think is the full amount of lives saved not the amount of credit every individual gets. I’m not 100% sure how to explain that in such a way as it’s perfectly clear though.
My thinking is that in the optimal incentives set up urine incentives should not move away from what you want to maximise since they would then cause other people not to maximise what you want them to maximise. Not sure whether this just means that a truly optimal incentive structure is just impossible or instead that the total impact ignoring all shares theory of responsibility is recommending incorrect behaviour that doesn’t maximise what an ideal society would want to maximise. I am certainly not suggesting that social credit is what actually matters, more that optimal social credit at least given some spherical cow assumptions should match what morality wants you to do, and I have a very strong intuition to that effect.
I have a rather strong intuition that this isn’t the case actually. It seems like proper moral mathematics in utilitarianism is rather difficult and complex and it may be the case that Sidgwick was right.
It’s also a serious question in ideal theory how much contingency should sanction, and when I get more into my studies I plan to write that paper.
The problem with the credit sharing view exists only with the assumption that one should maximize their own personal credit instead of overall good results. It's completely coherent to believe that I deserve credit only for a fair share of lives saved and that I still should do action that saves collectively the most lives - conditionally on other people also doing their part - of course.
No sure how much credit for this confusion between "credit" and "moral imperative" should be attributed to BB and how much to Parfit. But the total isn't supposed to exceed 100%!
Also, I keep noticing that quite often when BB tries to show us an example of philosophical progress it turns out to be yet another case of being entangled in a semantic confusion. Not a good look.
Interesting arguments, but I think it would be better if you actually went into depth on all the counter examples you think are decisive. My worry is that if you think of credit as an incentive provided to people to get them to do the right thing, then not dividing by your share of the contribution will recommend completely unreasonable levels of incentivising good deeds by multiple people when they are all necessary for the good deed, at least when compared to the incentive for doing a comparable good deed alone. Imagine two situations, one were a single person can save a life and another way to people need to work together to do it. In both cases. The actual value gained is the same. Yet in the second case, we should apparently expend twice the resources to incentivise the action, which doesn’t seem intuitive and in each cases can make the incentive costlyer than the benefit gained by society. Although obviously that’s not true of the example of saving a life.
To get an idea of why I intuitively, think this is a bad approach to incentivising. Imagine you are a person trying to incentivise people doing you a favour. Imagine one person does you a favour on their own afterwards a group of people do a comparable favour, but one that requires all their assistance as a matter of fact, the group is very large say 100 people. It seems absolutely crazy to say that you should give each member of the group. The full incentive that you would give someone who had done the deed on their own. So obviously you have to reduce the incentive because of the number of people necessary to do it. Praise and credit are after all at the end of the day incentives at least in part and most people do think that it would be strange if they totally came apart from incentives. If you are an economist trying to design the optimal incentives to make people do the right thing, you would not reward purely based on counterfactual impact that would lead to over incentivising things where multiple people are involved compare to the optimum amount of resource expenditure on incentives.
I also think you are approach cannot be Universalised. Which is problem since most people think that everybody reasoning and acting in accordance with morality should not lead to silly results that could be improved upon. To get an idea of why I think this look at the people who most often think like you do in Effective altruist circles, it is not at all uncommon to see people. Try to promote some cause by offering to match your donations if you donate, and not donating, if you do not, but this kind of donation matching appear sub optimal from your side if your own contribution through donations is less than the contribution from donating to something else minus the matching in your opinion, and in their opinion, their donation to this cause is the highest value thing then in an ideal world, they should donate to the cause, regardless of what you do, and you should donate to the cause you think is higher priority. But if everybody thought about responsibility, the way you do it leads to a bad incentive because they might be better off if they hold their own contribution to the cause they think is highest priority hostage unless you donate with them. I’m not putting this well into words, but I just have an intuition that the correct theory of moral responsibility would not recommend falling for this kind of behaviour, which is obviously bad for your own values.
To be clear, when I'm talking about credit, I'm just talking about the strength of your reasons to perform an act, not how much social credit people should give.
Re getting multiple people to do one good deed, I talk about why that's not a counterexample at the end.
I kind of have an intuition that your reasons for doing a thing and the optimum level of incentivising at least given the same level of selfishness throughout society and other spherical cow assumptions should match, and also I kind of think that moral responsibility isn’t really useful for anything much. If it’s not useful for allocating the socially optimum level of praise and blame. Also, as I mentioned in my edit to my original comment, I am also doubtful whether your approach can be Universalised. My objection with getting large numbers of people to do something that appears lower value is this. You should not end up further ahead in terms of credit by doing blatantly sub optimal things. However, your approach appears to recommend that in some cases at least on first gland, suppose you think that wild animals suffering is the most important cause and BOB thinks AI risk is the most important cause from your perspective, assuming that you think AI risk is still important, you would prefer you donating to animal welfare and BOB donating to AI risk, however, unless I am misunderstanding, your approach, actually say that if Bob decides to hold his own donation to AI risk hostage by refusing to donate anything to any cause, if you don’t donate or more realistically offers to match your donation to AI risk, and if you don’t donate to spend the money on something useless or burn it then you are approved to recommend going along with him, but this seems stupid because you prefer the scenario where Bob doesn’t pull this kind of shit on you and the only reason he’s pulling this kind of shit on you is because he knows you’ll give in, he won’t be trying this shit if you knew he would only get you telling him to fuck off, resulting in him, having to burn all that money given this, it appears like you will get better moral results by rejecting your theory of moral responsibility and I and many others have the intuition that this is not a property. You expect from the correct theory of moral responsibility. It should not predictably under, not that unusual circumstances start recommending that you adopt a different theory of moral responsibility. In any case, what is the correct theory of moral responsibility sounds like a practical rather than a descriptive question about states of affairs and I think for many people including me. The intuitive answer is the correct theory is the theory that you should adopt, and if you think the correctness of action is determined by their consequences. That means the correct theory is the one that gives you best consequences.
Ali, in general, if you want me to respond to comments, try to make them shorter, because I have a lot of commenters and a limited amount of willpower to read through long comments.
Maybe you should precommit to not giving into Bob's bullshit. Alternatively, maybe you have some decision theory where you think of your action as being sort of like a precommitment. But that is about how to give into threats, not about decision theory.
Wait what’s the incentive structure specifically here? We haven’t defined an ideal incentive structure? It’s also not clear why the posts analysis would differ from a hypothetical ideal.
Yes that’s also what I am saying. But what’s going on in the post has nothing to do with incentives? It’s purely about how to reason with regard to joint action scenario.
It’s getting hard to parse what you actually are getting at with these long posts. Can you be clearer and give
1. what this ideal incentive- ethical behavior relationship looks like
2. How the ideal incentive structure differs from what’s in the post
3. Why actual conditions don’t you think justify a different relationship.
Because the example of the donations doesn’t seem relevant since here we just have one person trying to behave ethically and another who is just being obstinate and behavior sub optimally by refusing to donate if you don’t. That’s as far as I have understood what you’re saying just him behaving wrongly.
Also you probably cannot get an incentive structure to perfectly match moral behavior under utilitarianism because what is correct is highly
Contextual and very complicated, where as policy (incentives) has to be set down in advance and so can only imperfectly act as guide.
The donation example is irrelevant that was for the case where people are trying to threaten you, and as I mentioned, I was incorrect there.
Part the reason I’m so confusing is because I’m kind of confused myself as I mentioned. I just have a conflicting intuition here and I’m having a hard time putting it into words. An ideal incentive structure is the incentive structure that utilitarianism would recommend that people ought to adopt. The intuition is this suppose you have a selfish member of society. It appears very strange that a Perfect utilitarian is willing to spend less resources to reward him for an action compare to another if they both are equally valuable. If a perfectly agent would be willing to spend less resources to make something happen, maybe that means that thing happening is less valuable, otherwise, the utilitarian will be willing to trade more to make it happen, and also if a utilitarian is not willing to expend as much resources to make it happen in you if you’re put in the position that the hypothetical selfish agent is in you should also perhaps on account of sharing exactly the same moral philosophy also treat the action as less valuable the way the utilitarian does. I agree that an actual reality a perfect incentive structure is not realistically achievable, but that’s not really the issue that I think between conflict with the post since the same issue appears even when you imagine the simplified situation where a perfect insect incentive structure is in fact quite feasible. Honestly, I think at this point, I am mostly confused rather than definitely disagree with the post and should read more about this topic before I put my foot in my mouth worse.
“An ideal incentive structure is the incentive structure that utilitarianism would recommend that people ought to adopt.”
Ok got that.
“It appears very strange that a Perfect utilitarian is willing to spend less resources to reward him for an action compare to another if they both are equally valuable.”
So what is causing this outcome to occur here? Basically how does this happen in the framework in the post?
Also I have to point if the agent is acting selfishly (which I take to mean more than mild self interest) they aren’t also going to be acting as a utilitarian. Those are mutually exclusive (in most cases).
You may be confused but I suspect there is a point in here even if you’re struggling to articulate it.
The selfish agent is merely hypothetical, what I mean is, suppose you have an action that 10 people can take together to save a life. A person is considering whether he should drop his plan to do something else, which should also sabe a life. In this case, I’m imagining that hypothetically if the person had been selfish, what would an ideal utilitarian be willing to do to incentivise them to pick one or the other option, so the selfish agent is never meant to be utilitarian. It’s more that I am using a hypothetical selfish agent to get an idea of how much utilitarianism would recommend you be willing to do to get someone to perform a particular action. In fact, selfishness is not necessary. The point is how much would utilitarianism recommend you be willing to trade to make something happen?
Now, the obvious question you raise is, why would there be a difference in how the utilitarian treats the situations and the reason I think there might be a difference is what I mentioned in my original comment, because if you were as winning to bribe each of the 10 people to do the right thing as you were willing to bribe one, you’d end up spending a Ludacris amount socially you must scale back considering their multiple people. The issue of course, the issue anybody would obviously point out is that a actual reason why we have the actual issue is because We are imagining one of these people becomes perfectly selfish, so you have to bribe him maybe in actual reality. He would have done it anyways, and you don’t need to do any such thing. Actually, I’m beginning to think the whole perfect selfishness thing was a stupid addition. I had originally used it as a simplifying assumption, but in real life, perhaps I should go with what I’d be willing to pay the actual person if I were a perfect utilitarian but that is no longer a measure of what I would be willing to trade off to get the result since in actual bargains with real people involve concerns like how much would be the minimum price they would be willing to settle for which can often be quite a bit lower than what you would be willing to pay required or reputational concerns of not wanting to take too many unfavourable deals. Honestly, I think it’s obvious from my rambling that I’m kind of confused, but my best guess right now is that the reason the utilitarian is willing to spend less on each is because the guarantee of an action to save the life is only obtained when he bribe 10 people in the situation where are required to save a life, but in the alternate option where that guy goes to single handedly save a life, the guarantee is only purchased when one person agrees, which means that the utilitarian is simply not buying the same thing, but if the other nine people are already set on helping, then the decision becomes equivalent, and it’s just that in a joint action situation, complexities, like whether other people will be willing to do their part and how hard it will be to persuade them. Make the situations not comparable, but now I’m confused whether this actually tied to anything in the post. Sorry for being completely in coherent. But I guess my takeaway is that if you’re one of a group performing an action, maybe given that nature doesn’t actually define an agent. You can sort of conceptualise the group of 10 people as a single agent and individual 10 people as sub components when they are saving the life and so getting one of them to act is a bit like getting part of what you want, but not be entirety and so obviously you don’t actually consider them comparable. Honestly, I think my primary take away now is that I should stop this discussion and go read EA writing on this topic before I embarrass myself even more because I keep getting confused over how different views on the joint action scenario and moral responsibility would impact actual concrete decision-making. I wonder if this has anything to do with the conflict between functional decision theory and causal decision theory when it comes to trading with other agents as the latter treats agents who are not you as essentially the environment outside your control while the Former treats them as thinking agents who can be changed by how you think, but that’s just shooting in the dark and there might be no connection between this question of moral responsibility, and the decision theory question.
This I think may be reason to adopt something like a share of the total view as some kind of shallow scheme of compensation.
But it’s true in a deep sense.
For the donation example I think I understand what you’re getting at but here is the situation that I think deals with it
1. Will agent B donate at all if he doesn’t donate to charity 2?
2. How many lives will Agent A save if he donates his money to charity 1 (his preferred highest impact charity)?
3. How many lives will the combined donation of A &B save at charity 2?
4. Can B be convinced to donate to charity 1?
If the answer to 1. And 4. Is no (or yes for 1. but to somewhere really suboptimal) and the number of lives in 3 > 2 then this is a sensible thing for the EA to do.
That said I don’t know why you think this view of moral responsibility would lead to situations where you hold donations hostage like this? Moral desert or whatever you are getting in greater quantity by forcing a huge joint donation to get everyone the credit isn’t something any (traditional) utilitarian is attempting to maximize.
I agree with what you mention in your points. I am specifically imagining a situation where neither of you can persuade the other, and your view is that him donating to his charity, and you donating to your charity is better than you donating to his charity and him matching you, but if you don’t donate as a consequence of him matching you, he just waste his money. There is a difference between helping to coordinate people donating by offering to match which I think is completely pro social and the scenario I am imagining which is deliberately meant to be hostile behaviour. When you will something as a universal law, you don’t have to imagine. Everyone else is a angel. For example when thinking of whether you should lie to the ax murderer, you don’t imagine that actually a moral person would never be an ax murder unless the target genuinely had done something to deserve it. You just note that everybody adopting your policy would lead to particular consequences and depending on your view of those consequences, you either think it can be willed as a universal law to lie in that situation or not. Unless I am misunderstanding the suggested theory, everybody adopting it as their theory of moral responsibility would incentivise extortionary behaviour of the type eye outline. Let me put it like this this theory of moral responsibility, unless I may understand would recommend that if you think of the scenario that I outline, you should reject this theory of all responsibility in favour of a modified version, which you tell BOB to fuck off, but if you don’t think of it in time, and BOB does actually come and make his threat, you should give, except Bob never would pull that stunt if he knew that you tell him to fuck off. So another words, I think this theory of moral responsibility recommends that you replace it with a different theory of moral responsibility which gives different advice when it comes to how to behave which is not a property. I expect from the correct theory of moral responsibility. It also means that it is not the theory that would have the best consequences if you adopted it which if you think that the correct theory is the one with the best consequences or even simply that the correct theory is the one you should adopt and the right of behaviour, including mental behaviour is determined by its consequences. Just rule it out as the correct theory.
Also part of the reason why I don’t like the correct theory of moral responsibility becoming separate from the correct theory of incentives is that incentives are just a way of getting people to do what you want. So the correct responsibility theory should not come away from the correct theory of incentives, because that means what you want people to do and what they ought to do are coming apart, which is ridiculous. For example to take the example from the original post, if four people together can save 100 people and one of them can save 50 people, then you would not want to incentivise that guy to save the 50 people instead, and if you did end up incentivising that then your incentive is suboptimal. Optimal incentive from the point of view of a perfectly moral agent or society should always be in agreement with morality and moral responsibility.
I am unclear why we are consider a scenario where someone is operating deliberately hostile. That person is just acting wrongly and it’s not supported by the theory it seems like. Because you have multiple decision points
1. Where he makes his threat- he should make it
2. After you tell him no- when he should donate despite his threat.
Ah so here is the thing, you don’t need a theory of moral responsibility under this scenario, that’s not what it’s about. Utilitarianism doesn’t require it. You need to not understand any of this as a theory of moral responsibility but merely as a way of reasoning through joint action scenarios. You can have moral obligation without moral responsibility.
BB correctly pointed out that actually the problem in my scenario doesn’t have anything to do with moral responsibility. It’s a pure question of how you react to threats. So actually I was mistaken there in thinking his theory requires the incorrect answer in that situation.
I think part of the issue here is that since as you pointed out traditional moral responsibility does not really add anything given utilitarianism, I mentally use the term moral responsibility for instead talking about the moraly recommended amount of praise and blame. It is possible that this use of the same word for different things is causing me to incorrectly think that there should be a strong connection between optimal incentives and correct reasoning in joint action scenarios, but I don’t think so. After all your moral obligations and the optimal incentive structure should point in the direction of reasoning correctly through joint action scenarios instead of pushing you to act in ways in such situations which are suboptimal, but perhaps I am confused there or mistaken in thinking that you can in theory actually set up an incentive structure that always incentivises doing what you want.
The question is whether people can be relied on to reason effectively, given people as they are it seems possible even likely that the most reliable optimum is constituted when the incentive structure and correct moral reasoning differs. I think this is likely the case since we all suffer from substantial cognitive biases that prevent optimal reasoning and most people (potentially all people) lack the capacity to do reasonably correct moral mathematics in most scenarios.
Where I think the issue is, is that you think that moral obligation + incentive structure = correct moral reasoning, but there is potentially good reason to think the incentive structure shouldn’t be calibrated to correct moral reasoning in order to actually fulfill our moral obligations.
Well, the thing is, I think, given human cognitive limitations and biases The incentive structure should be calibrated so as 2 try to get the best possible response that you can get from the humans, although obviously, this may still fall shot of what an ideal agent would actually do in that situation because the human is not some perfect angel, nor perfectly rational. However, doesn’t matter in any case because it’s not like your theory of moral responsibility is only applicable conditional on cognitive limitations. As I presume you would agree that it would equally be applicable if humans had no such limitations. That’s why for is analysis I keep talking about spherical cow assumptions. If you assume, for example that all humans are perfectly rational, selfish agents and God set out the perfect logically possible incentive structure shouldn’t that in structure match what God with his perfect ethical reasoning, thinks should be done as it’s after all, trying to just get the humans to do what he wants. Yes, this is a very far-fetched hypothetical, but my point is that in the simplified situation where it’s much easier to get a handle on the perfect incentive structure it becomes immediately obvious that your theory moral responsibility doesn’t match the correct incentive structure. I’m not sure whether the conclusion is that even in this hypothetical which I outlined a perfect incentive structure that gets the humans to do exactly what God wants is simply impossible or whether it’s your theory of moral responsibility which is wrong. Obviously, this is a hypothetical for ease of getting a handle on what the correct incentive would be in actual reality. You have to change the incentives because humans are not perfectly rational or perfectly selfish. So the optimal incentive structure would be different in reality, the thought experiment is just for ease of analysis.
Let's assume that if Trump won in 2020 he would've killed USAID earlier and millions of people would've died that survived due to aid given in 2021-2025 (I'm not sure this is exactly right but the EV is probably not far off and scenarios like this definitely exist regardless). Wouldn't your theory say that I deserve full credit for saving all of these lives by voting for Biden in 2020, and that all of my charitable donations are a rounding error in light of that?
Alternatively, if you say that this isn't true because Biden would've won without my vote, does that mean that I shouldn't feel like I did a good thing in voting for him or that people who voted for Trump shouldn't feel bad?
I feel like both of these are obviously wrong, and a useful approximation is that people should feel responsible for their share of the vote (though to be clear this is very much an approximation in multiple ways, I just think it's better than either of the extremes).
Seems like this is obviously wrong in the case of impact certificates. If I buy impact certificates at the rate of $1M per life saved, I am happy to buy from one person who saved a life on their own, but unhappy to give $1M to each of 100 doctors who did one step in a 100-step surgery that saved a life (since this would result in me having to spend $100M in impact certificates to retroactively save one life). It seems like I should be willing to give one impact certificate to the hospital, to be divided up among those 100 doctors, which reduces to the view you reject. But impact certificates are the clearest case of when you would actually care how much credit somebody should get. If we admit it's wrong in that case, isn't that ceding most of the question?
But staying within the philosophical frame, consider the following: there's a pool and a tarpit. It takes one person to drag a drowning child out of the pool, and ten people to drag a child out of the tarpit. Dragging a child from the pool takes 10 minutes, dragging a child from the tarpit takes 9 minutes. There are an arbitrarily large but finite number of rescuers, and 10x that many drowning children, distributed evenly across both locations. You have ninety minutes to devote to rescuing children. It seems like under your view, you should rescue children from the tarpit (since you'll save 10 lives, opposed to the counterfactual 9). But if everyone makes that choice, ~9x fewer children will be saved. Here you can't argue that your choosing the pool would make everyone else choose the pool, because there are an arbitrarily large number of rescuers, and their chance of finding a group of ten at the tarpit is unaffected. And I don't think you can argue that you only get credit at the tarpit if you are the tenth person and therefore cause a new group of ten to be formed. After all, you have an arbitrarily high chance of actually joining a group of ten (there are arbitrarily many rescuers, and at most nine people can be left out). And since you are actually literally pulling children out of the tarpit, what does it mean not to credit you for those children? I think you would have to say "you only deserve one-tenth share for each child you pulled out, because there was only one chance in ten that you counterfactually caused this group of ten people to form". But now we're back to only giving you 1/10th credit! Tell me what I'm getting wrong.
The share of the total view is counterintuitive on its face when considering the converse; if I pay an assassin to murder someone, it seems I am guilty of murder and the assassin is also guilty of murder.
Well yes, but you have to take into account the opportunity costs. If you get someone to tell his friend to save the drowning people, you are losing the value of whatever else those people would have done with their time. Now maybe this is not so important in the drowning children scenario, since maybe nobody involved was doing anything very important with their time. But in the context of effective altruism I think many people have something like a fixed budget of time and effort that they are willing or able to spend on altruism. So the counterfactual may be costly, changing the calculus in roughly the same direction to how the "sharing the credit" view changes the calculus.
Pretty interesting puzzle I’ve not heard before. Seems like Parfit is correct in the positive case in which lives are saved. I wonder if this idea has implications for the negative case too. Imagine 10 guys beat the crap out of someone and would do so if you weren’t there. I would want to say if you participate you are at least 1/10th or more responsible for beating him up.
I think that's wrong. Maybe you're responsible if it makes you more vicious, but doing it wouldn't harm the person extra. One way to see this: imagine that joining in would allow you to heal him after. Seems you should do that. But if you got 1/10 of the responsibility for the beating that would be false.
As applied to EA, the case is even stronger because of scale. Two people saving two lives only if both act is intuitively murkier, but with EA, it's more like 1000 people saving 1000 lives or 999 people saving 999 lives.
I liken reading Bentham's Bulldog to going to church. It's not enjoyable, but enjoyment isn't the point. Just take 10 minutes to at least think about being a better person. Okay, back to my veal parmesan
Seems Wenar was talking about credit assignment (where you want to share a fixed pie of credit fairly, which something like Aumann Shapley handles and says each person gets credit for one life) not a decision rule (which Parfit’s counterfactual view handles). These are two different things.
It was in response to MacAskill's claim that your action saved a life, not about how to give people social credit.
Sure, but it does seem likely that the origin of the disagreement was a misunderstanding of what MacAskill was weighing (what decision to take) and that Wenar might have thought he was talking about post facto credit assignment. Which is still a dumb mistake for Wener to make and I can almost smell the motivated reasoning (wanting to dunk on those dang smug EAs). But maybe that common conflation is the more important thing to address first. (Still, I enjoyed your post and the examples in it.)
I've been thinking about this for a while. I'd get confused whenever I assessed counterfactual credit and found it bigger (in sum) than counterfactual impact. Now I know I should read more "moral mathematics"
I was also confused by the article and wondered, "but the kid would counterfactually die without the donation?" It's weird how Wenar thinks a donor saying "my donation saved a life" takes away from the parent who allowed it. Neither GiveWell nor any other org has ever claimed that ALL the credit is theirs, only that they counterfactually save lives.
Seems like the share of total view confuses personal responsibility with ideal decision making - but it's easy to see why. The Talmud (Bava Kama 10b) considers a case where 5 people sat on a bench and then a 6th person joined them and the bench broke. The Talmud is concerned with who pays for the bench. Do we divide the cost among the 6 or does the last guy pay it all? But now ask a different question: What is less bad, for the 6th guy to join them on the big bench and break it, or for this person to sit down on a cheaper bench and somehow single handedly break that? Obviously the second is less bad; there's less total damage. But here's the thing: Say the big bench costs $300 and the small one is $100. Now imagine yourself telling this person "Sit on the cheaper bench and you'll pay $100 in damages rather than $50, but hey you're a better person!" Silly example, but you know what I mean.
The 6th person to sit on the big bench should pay to repair it though, or at least we anchor on that and negotiate contributions from the others (the non-marginal breakers)
This is in large part why I think the fixation on moral responsibility is really weird. Ethics has for a long time had a strong fixation on moral desert and who gets to take credit for things. But it just doesn’t seem to matter at least from a utilitarian perspective all that matters is the counterfactuals. So in the sense of making rational assessments of what to do this is very important. But in another sense, that of actually dividing the credit it’s totally trivial and meaningless, which is another reason why we really should not be bothered if it multiplies the credit well beyond a 1 to 1 accounting of the impact caused.
In practice, doesn't this just trivially give bad answers if you try to do charity evaluation?
Imagine donating $3000 to Charity A that saves 1 life per marginal donation, vs donating to Charity B that saves 2 lives at $300k (and there's an assurance contract where 99 other people donate to it iff you also donate $3000).
It seems to me like donating to Charity A is much better! Do you disagree?
Now in theory this problem can be solved by chaining counterfactuals all the way through (https://forum.effectivealtruism.org/posts/myp9Y9qJnpEEWhJF9/linch-s-shortform?commentId=tXautB29NTGKnAqTe), but in practice this is often computationally intractable, which is why people often turn to Shapley values or impact equity instead.
It depends on what the other people would have done with their money if you hadn't donated.
Does it in this scenario? It seems like it is highly unlikely to be the case that donating to charity B is worth it unless the other 99 people are using their 297,000 to contribute to net negative utility things. As long as they engage in reasonably positive or neutral utility behavior it seems likely that this much money saves at least 1 life in its normal operation through the economy.
Obviously to his point it’s really hard to do this kind of math but still.
tbc, normal operation in the US economy does less good than $297k/life, on average.
Do you have some receipts on that? It seems quite possible but I don’t want to take that on faith.
You can estimate this through several different means, one of which is US willingness to pay for a QALY.
Generally you shouldn't expect normal economic operation to be more effective than optimized health spending, so that creates some bounds.
Another way to think about it is that US GDP per capita is something like 94k/year. It'd be surprising if normal economic operation for 3 years here saves a life! Would indicate substantially increasing marginal returns for economic operation, which flies in the face of everything we know about the relationship between growth and health.
From my brief survey of some paper abstracts it seems like the U.S. willingness to pay for QALY is 100-150000 so it doesn’t immediately rule out that you would get 1+ QALY out of 297k.
Secondly the 297k is specifically NOT optimized health spending. In this example the 300k for 2 lives is a very inefficient non optimal use of funds.
I think if normal economic operation falls below the level of 297k for one life it’s gonna because there are substantial negative externalities to normal economic operation. Not because that normal operation isn’t contributing to wellbeing.
You are correct, but also kind of fighting the hypothetical. Your theory does have the problem that under perfectly imaginable conditions, it gives the wrong answer in these kinds of thought experiments. After all this perfectly possible to imagine a situation where those are the 99 people behave in ways where the first charity is still better as a target for donation.
Well in the full analysis of the counterfactual what matters I think is the full amount of lives saved not the amount of credit every individual gets. I’m not 100% sure how to explain that in such a way as it’s perfectly clear though.
My thinking is that in the optimal incentives set up urine incentives should not move away from what you want to maximise since they would then cause other people not to maximise what you want them to maximise. Not sure whether this just means that a truly optimal incentive structure is just impossible or instead that the total impact ignoring all shares theory of responsibility is recommending incorrect behaviour that doesn’t maximise what an ideal society would want to maximise. I am certainly not suggesting that social credit is what actually matters, more that optimal social credit at least given some spherical cow assumptions should match what morality wants you to do, and I have a very strong intuition to that effect.
I have a rather strong intuition that this isn’t the case actually. It seems like proper moral mathematics in utilitarianism is rather difficult and complex and it may be the case that Sidgwick was right.
It’s also a serious question in ideal theory how much contingency should sanction, and when I get more into my studies I plan to write that paper.
The problem with the credit sharing view exists only with the assumption that one should maximize their own personal credit instead of overall good results. It's completely coherent to believe that I deserve credit only for a fair share of lives saved and that I still should do action that saves collectively the most lives - conditionally on other people also doing their part - of course.
No sure how much credit for this confusion between "credit" and "moral imperative" should be attributed to BB and how much to Parfit. But the total isn't supposed to exceed 100%!
Also, I keep noticing that quite often when BB tries to show us an example of philosophical progress it turns out to be yet another case of being entangled in a semantic confusion. Not a good look.
Interesting arguments, but I think it would be better if you actually went into depth on all the counter examples you think are decisive. My worry is that if you think of credit as an incentive provided to people to get them to do the right thing, then not dividing by your share of the contribution will recommend completely unreasonable levels of incentivising good deeds by multiple people when they are all necessary for the good deed, at least when compared to the incentive for doing a comparable good deed alone. Imagine two situations, one were a single person can save a life and another way to people need to work together to do it. In both cases. The actual value gained is the same. Yet in the second case, we should apparently expend twice the resources to incentivise the action, which doesn’t seem intuitive and in each cases can make the incentive costlyer than the benefit gained by society. Although obviously that’s not true of the example of saving a life.
To get an idea of why I intuitively, think this is a bad approach to incentivising. Imagine you are a person trying to incentivise people doing you a favour. Imagine one person does you a favour on their own afterwards a group of people do a comparable favour, but one that requires all their assistance as a matter of fact, the group is very large say 100 people. It seems absolutely crazy to say that you should give each member of the group. The full incentive that you would give someone who had done the deed on their own. So obviously you have to reduce the incentive because of the number of people necessary to do it. Praise and credit are after all at the end of the day incentives at least in part and most people do think that it would be strange if they totally came apart from incentives. If you are an economist trying to design the optimal incentives to make people do the right thing, you would not reward purely based on counterfactual impact that would lead to over incentivising things where multiple people are involved compare to the optimum amount of resource expenditure on incentives.
I also think you are approach cannot be Universalised. Which is problem since most people think that everybody reasoning and acting in accordance with morality should not lead to silly results that could be improved upon. To get an idea of why I think this look at the people who most often think like you do in Effective altruist circles, it is not at all uncommon to see people. Try to promote some cause by offering to match your donations if you donate, and not donating, if you do not, but this kind of donation matching appear sub optimal from your side if your own contribution through donations is less than the contribution from donating to something else minus the matching in your opinion, and in their opinion, their donation to this cause is the highest value thing then in an ideal world, they should donate to the cause, regardless of what you do, and you should donate to the cause you think is higher priority. But if everybody thought about responsibility, the way you do it leads to a bad incentive because they might be better off if they hold their own contribution to the cause they think is highest priority hostage unless you donate with them. I’m not putting this well into words, but I just have an intuition that the correct theory of moral responsibility would not recommend falling for this kind of behaviour, which is obviously bad for your own values.
To be clear, when I'm talking about credit, I'm just talking about the strength of your reasons to perform an act, not how much social credit people should give.
Re getting multiple people to do one good deed, I talk about why that's not a counterexample at the end.
I kind of have an intuition that your reasons for doing a thing and the optimum level of incentivising at least given the same level of selfishness throughout society and other spherical cow assumptions should match, and also I kind of think that moral responsibility isn’t really useful for anything much. If it’s not useful for allocating the socially optimum level of praise and blame. Also, as I mentioned in my edit to my original comment, I am also doubtful whether your approach can be Universalised. My objection with getting large numbers of people to do something that appears lower value is this. You should not end up further ahead in terms of credit by doing blatantly sub optimal things. However, your approach appears to recommend that in some cases at least on first gland, suppose you think that wild animals suffering is the most important cause and BOB thinks AI risk is the most important cause from your perspective, assuming that you think AI risk is still important, you would prefer you donating to animal welfare and BOB donating to AI risk, however, unless I am misunderstanding, your approach, actually say that if Bob decides to hold his own donation to AI risk hostage by refusing to donate anything to any cause, if you don’t donate or more realistically offers to match your donation to AI risk, and if you don’t donate to spend the money on something useless or burn it then you are approved to recommend going along with him, but this seems stupid because you prefer the scenario where Bob doesn’t pull this kind of shit on you and the only reason he’s pulling this kind of shit on you is because he knows you’ll give in, he won’t be trying this shit if you knew he would only get you telling him to fuck off, resulting in him, having to burn all that money given this, it appears like you will get better moral results by rejecting your theory of moral responsibility and I and many others have the intuition that this is not a property. You expect from the correct theory of moral responsibility. It should not predictably under, not that unusual circumstances start recommending that you adopt a different theory of moral responsibility. In any case, what is the correct theory of moral responsibility sounds like a practical rather than a descriptive question about states of affairs and I think for many people including me. The intuitive answer is the correct theory is the theory that you should adopt, and if you think the correctness of action is determined by their consequences. That means the correct theory is the one that gives you best consequences.
Ali, in general, if you want me to respond to comments, try to make them shorter, because I have a lot of commenters and a limited amount of willpower to read through long comments.
Maybe you should precommit to not giving into Bob's bullshit. Alternatively, maybe you have some decision theory where you think of your action as being sort of like a precommitment. But that is about how to give into threats, not about decision theory.
Wait what’s the incentive structure specifically here? We haven’t defined an ideal incentive structure? It’s also not clear why the posts analysis would differ from a hypothetical ideal.
Yes that’s also what I am saying. But what’s going on in the post has nothing to do with incentives? It’s purely about how to reason with regard to joint action scenario.
It’s getting hard to parse what you actually are getting at with these long posts. Can you be clearer and give
1. what this ideal incentive- ethical behavior relationship looks like
2. How the ideal incentive structure differs from what’s in the post
3. Why actual conditions don’t you think justify a different relationship.
Because the example of the donations doesn’t seem relevant since here we just have one person trying to behave ethically and another who is just being obstinate and behavior sub optimally by refusing to donate if you don’t. That’s as far as I have understood what you’re saying just him behaving wrongly.
Also you probably cannot get an incentive structure to perfectly match moral behavior under utilitarianism because what is correct is highly
Contextual and very complicated, where as policy (incentives) has to be set down in advance and so can only imperfectly act as guide.
The donation example is irrelevant that was for the case where people are trying to threaten you, and as I mentioned, I was incorrect there.
Part the reason I’m so confusing is because I’m kind of confused myself as I mentioned. I just have a conflicting intuition here and I’m having a hard time putting it into words. An ideal incentive structure is the incentive structure that utilitarianism would recommend that people ought to adopt. The intuition is this suppose you have a selfish member of society. It appears very strange that a Perfect utilitarian is willing to spend less resources to reward him for an action compare to another if they both are equally valuable. If a perfectly agent would be willing to spend less resources to make something happen, maybe that means that thing happening is less valuable, otherwise, the utilitarian will be willing to trade more to make it happen, and also if a utilitarian is not willing to expend as much resources to make it happen in you if you’re put in the position that the hypothetical selfish agent is in you should also perhaps on account of sharing exactly the same moral philosophy also treat the action as less valuable the way the utilitarian does. I agree that an actual reality a perfect incentive structure is not realistically achievable, but that’s not really the issue that I think between conflict with the post since the same issue appears even when you imagine the simplified situation where a perfect insect incentive structure is in fact quite feasible. Honestly, I think at this point, I am mostly confused rather than definitely disagree with the post and should read more about this topic before I put my foot in my mouth worse.
“An ideal incentive structure is the incentive structure that utilitarianism would recommend that people ought to adopt.”
Ok got that.
“It appears very strange that a Perfect utilitarian is willing to spend less resources to reward him for an action compare to another if they both are equally valuable.”
So what is causing this outcome to occur here? Basically how does this happen in the framework in the post?
Also I have to point if the agent is acting selfishly (which I take to mean more than mild self interest) they aren’t also going to be acting as a utilitarian. Those are mutually exclusive (in most cases).
You may be confused but I suspect there is a point in here even if you’re struggling to articulate it.
The selfish agent is merely hypothetical, what I mean is, suppose you have an action that 10 people can take together to save a life. A person is considering whether he should drop his plan to do something else, which should also sabe a life. In this case, I’m imagining that hypothetically if the person had been selfish, what would an ideal utilitarian be willing to do to incentivise them to pick one or the other option, so the selfish agent is never meant to be utilitarian. It’s more that I am using a hypothetical selfish agent to get an idea of how much utilitarianism would recommend you be willing to do to get someone to perform a particular action. In fact, selfishness is not necessary. The point is how much would utilitarianism recommend you be willing to trade to make something happen?
Now, the obvious question you raise is, why would there be a difference in how the utilitarian treats the situations and the reason I think there might be a difference is what I mentioned in my original comment, because if you were as winning to bribe each of the 10 people to do the right thing as you were willing to bribe one, you’d end up spending a Ludacris amount socially you must scale back considering their multiple people. The issue of course, the issue anybody would obviously point out is that a actual reason why we have the actual issue is because We are imagining one of these people becomes perfectly selfish, so you have to bribe him maybe in actual reality. He would have done it anyways, and you don’t need to do any such thing. Actually, I’m beginning to think the whole perfect selfishness thing was a stupid addition. I had originally used it as a simplifying assumption, but in real life, perhaps I should go with what I’d be willing to pay the actual person if I were a perfect utilitarian but that is no longer a measure of what I would be willing to trade off to get the result since in actual bargains with real people involve concerns like how much would be the minimum price they would be willing to settle for which can often be quite a bit lower than what you would be willing to pay required or reputational concerns of not wanting to take too many unfavourable deals. Honestly, I think it’s obvious from my rambling that I’m kind of confused, but my best guess right now is that the reason the utilitarian is willing to spend less on each is because the guarantee of an action to save the life is only obtained when he bribe 10 people in the situation where are required to save a life, but in the alternate option where that guy goes to single handedly save a life, the guarantee is only purchased when one person agrees, which means that the utilitarian is simply not buying the same thing, but if the other nine people are already set on helping, then the decision becomes equivalent, and it’s just that in a joint action situation, complexities, like whether other people will be willing to do their part and how hard it will be to persuade them. Make the situations not comparable, but now I’m confused whether this actually tied to anything in the post. Sorry for being completely in coherent. But I guess my takeaway is that if you’re one of a group performing an action, maybe given that nature doesn’t actually define an agent. You can sort of conceptualise the group of 10 people as a single agent and individual 10 people as sub components when they are saving the life and so getting one of them to act is a bit like getting part of what you want, but not be entirety and so obviously you don’t actually consider them comparable. Honestly, I think my primary take away now is that I should stop this discussion and go read EA writing on this topic before I embarrass myself even more because I keep getting confused over how different views on the joint action scenario and moral responsibility would impact actual concrete decision-making. I wonder if this has anything to do with the conflict between functional decision theory and causal decision theory when it comes to trading with other agents as the latter treats agents who are not you as essentially the environment outside your control while the Former treats them as thinking agents who can be changed by how you think, but that’s just shooting in the dark and there might be no connection between this question of moral responsibility, and the decision theory question.
This I think may be reason to adopt something like a share of the total view as some kind of shallow scheme of compensation.
But it’s true in a deep sense.
For the donation example I think I understand what you’re getting at but here is the situation that I think deals with it
1. Will agent B donate at all if he doesn’t donate to charity 2?
2. How many lives will Agent A save if he donates his money to charity 1 (his preferred highest impact charity)?
3. How many lives will the combined donation of A &B save at charity 2?
4. Can B be convinced to donate to charity 1?
If the answer to 1. And 4. Is no (or yes for 1. but to somewhere really suboptimal) and the number of lives in 3 > 2 then this is a sensible thing for the EA to do.
That said I don’t know why you think this view of moral responsibility would lead to situations where you hold donations hostage like this? Moral desert or whatever you are getting in greater quantity by forcing a huge joint donation to get everyone the credit isn’t something any (traditional) utilitarian is attempting to maximize.
I agree with what you mention in your points. I am specifically imagining a situation where neither of you can persuade the other, and your view is that him donating to his charity, and you donating to your charity is better than you donating to his charity and him matching you, but if you don’t donate as a consequence of him matching you, he just waste his money. There is a difference between helping to coordinate people donating by offering to match which I think is completely pro social and the scenario I am imagining which is deliberately meant to be hostile behaviour. When you will something as a universal law, you don’t have to imagine. Everyone else is a angel. For example when thinking of whether you should lie to the ax murderer, you don’t imagine that actually a moral person would never be an ax murder unless the target genuinely had done something to deserve it. You just note that everybody adopting your policy would lead to particular consequences and depending on your view of those consequences, you either think it can be willed as a universal law to lie in that situation or not. Unless I am misunderstanding the suggested theory, everybody adopting it as their theory of moral responsibility would incentivise extortionary behaviour of the type eye outline. Let me put it like this this theory of moral responsibility, unless I may understand would recommend that if you think of the scenario that I outline, you should reject this theory of all responsibility in favour of a modified version, which you tell BOB to fuck off, but if you don’t think of it in time, and BOB does actually come and make his threat, you should give, except Bob never would pull that stunt if he knew that you tell him to fuck off. So another words, I think this theory of moral responsibility recommends that you replace it with a different theory of moral responsibility which gives different advice when it comes to how to behave which is not a property. I expect from the correct theory of moral responsibility. It also means that it is not the theory that would have the best consequences if you adopted it which if you think that the correct theory is the one with the best consequences or even simply that the correct theory is the one you should adopt and the right of behaviour, including mental behaviour is determined by its consequences. Just rule it out as the correct theory.
Also part of the reason why I don’t like the correct theory of moral responsibility becoming separate from the correct theory of incentives is that incentives are just a way of getting people to do what you want. So the correct responsibility theory should not come away from the correct theory of incentives, because that means what you want people to do and what they ought to do are coming apart, which is ridiculous. For example to take the example from the original post, if four people together can save 100 people and one of them can save 50 people, then you would not want to incentivise that guy to save the 50 people instead, and if you did end up incentivising that then your incentive is suboptimal. Optimal incentive from the point of view of a perfectly moral agent or society should always be in agreement with morality and moral responsibility.
I am unclear why we are consider a scenario where someone is operating deliberately hostile. That person is just acting wrongly and it’s not supported by the theory it seems like. Because you have multiple decision points
1. Where he makes his threat- he should make it
2. After you tell him no- when he should donate despite his threat.
Ah so here is the thing, you don’t need a theory of moral responsibility under this scenario, that’s not what it’s about. Utilitarianism doesn’t require it. You need to not understand any of this as a theory of moral responsibility but merely as a way of reasoning through joint action scenarios. You can have moral obligation without moral responsibility.
BB correctly pointed out that actually the problem in my scenario doesn’t have anything to do with moral responsibility. It’s a pure question of how you react to threats. So actually I was mistaken there in thinking his theory requires the incorrect answer in that situation.
I think part of the issue here is that since as you pointed out traditional moral responsibility does not really add anything given utilitarianism, I mentally use the term moral responsibility for instead talking about the moraly recommended amount of praise and blame. It is possible that this use of the same word for different things is causing me to incorrectly think that there should be a strong connection between optimal incentives and correct reasoning in joint action scenarios, but I don’t think so. After all your moral obligations and the optimal incentive structure should point in the direction of reasoning correctly through joint action scenarios instead of pushing you to act in ways in such situations which are suboptimal, but perhaps I am confused there or mistaken in thinking that you can in theory actually set up an incentive structure that always incentivises doing what you want.
The question is whether people can be relied on to reason effectively, given people as they are it seems possible even likely that the most reliable optimum is constituted when the incentive structure and correct moral reasoning differs. I think this is likely the case since we all suffer from substantial cognitive biases that prevent optimal reasoning and most people (potentially all people) lack the capacity to do reasonably correct moral mathematics in most scenarios.
Where I think the issue is, is that you think that moral obligation + incentive structure = correct moral reasoning, but there is potentially good reason to think the incentive structure shouldn’t be calibrated to correct moral reasoning in order to actually fulfill our moral obligations.
Well, the thing is, I think, given human cognitive limitations and biases The incentive structure should be calibrated so as 2 try to get the best possible response that you can get from the humans, although obviously, this may still fall shot of what an ideal agent would actually do in that situation because the human is not some perfect angel, nor perfectly rational. However, doesn’t matter in any case because it’s not like your theory of moral responsibility is only applicable conditional on cognitive limitations. As I presume you would agree that it would equally be applicable if humans had no such limitations. That’s why for is analysis I keep talking about spherical cow assumptions. If you assume, for example that all humans are perfectly rational, selfish agents and God set out the perfect logically possible incentive structure shouldn’t that in structure match what God with his perfect ethical reasoning, thinks should be done as it’s after all, trying to just get the humans to do what he wants. Yes, this is a very far-fetched hypothetical, but my point is that in the simplified situation where it’s much easier to get a handle on the perfect incentive structure it becomes immediately obvious that your theory moral responsibility doesn’t match the correct incentive structure. I’m not sure whether the conclusion is that even in this hypothetical which I outlined a perfect incentive structure that gets the humans to do exactly what God wants is simply impossible or whether it’s your theory of moral responsibility which is wrong. Obviously, this is a hypothetical for ease of getting a handle on what the correct incentive would be in actual reality. You have to change the incentives because humans are not perfectly rational or perfectly selfish. So the optimal incentive structure would be different in reality, the thought experiment is just for ease of analysis.
Let's assume that if Trump won in 2020 he would've killed USAID earlier and millions of people would've died that survived due to aid given in 2021-2025 (I'm not sure this is exactly right but the EV is probably not far off and scenarios like this definitely exist regardless). Wouldn't your theory say that I deserve full credit for saving all of these lives by voting for Biden in 2020, and that all of my charitable donations are a rounding error in light of that?
Alternatively, if you say that this isn't true because Biden would've won without my vote, does that mean that I shouldn't feel like I did a good thing in voting for him or that people who voted for Trump shouldn't feel bad?
I feel like both of these are obviously wrong, and a useful approximation is that people should feel responsible for their share of the vote (though to be clear this is very much an approximation in multiple ways, I just think it's better than either of the extremes).
Seems like this is obviously wrong in the case of impact certificates. If I buy impact certificates at the rate of $1M per life saved, I am happy to buy from one person who saved a life on their own, but unhappy to give $1M to each of 100 doctors who did one step in a 100-step surgery that saved a life (since this would result in me having to spend $100M in impact certificates to retroactively save one life). It seems like I should be willing to give one impact certificate to the hospital, to be divided up among those 100 doctors, which reduces to the view you reject. But impact certificates are the clearest case of when you would actually care how much credit somebody should get. If we admit it's wrong in that case, isn't that ceding most of the question?
But staying within the philosophical frame, consider the following: there's a pool and a tarpit. It takes one person to drag a drowning child out of the pool, and ten people to drag a child out of the tarpit. Dragging a child from the pool takes 10 minutes, dragging a child from the tarpit takes 9 minutes. There are an arbitrarily large but finite number of rescuers, and 10x that many drowning children, distributed evenly across both locations. You have ninety minutes to devote to rescuing children. It seems like under your view, you should rescue children from the tarpit (since you'll save 10 lives, opposed to the counterfactual 9). But if everyone makes that choice, ~9x fewer children will be saved. Here you can't argue that your choosing the pool would make everyone else choose the pool, because there are an arbitrarily large number of rescuers, and their chance of finding a group of ten at the tarpit is unaffected. And I don't think you can argue that you only get credit at the tarpit if you are the tenth person and therefore cause a new group of ten to be formed. After all, you have an arbitrarily high chance of actually joining a group of ten (there are arbitrarily many rescuers, and at most nine people can be left out). And since you are actually literally pulling children out of the tarpit, what does it mean not to credit you for those children? I think you would have to say "you only deserve one-tenth share for each child you pulled out, because there was only one chance in ten that you counterfactually caused this group of ten people to form". But now we're back to only giving you 1/10th credit! Tell me what I'm getting wrong.
The share of the total view is counterintuitive on its face when considering the converse; if I pay an assassin to murder someone, it seems I am guilty of murder and the assassin is also guilty of murder.
Well yes, but you have to take into account the opportunity costs. If you get someone to tell his friend to save the drowning people, you are losing the value of whatever else those people would have done with their time. Now maybe this is not so important in the drowning children scenario, since maybe nobody involved was doing anything very important with their time. But in the context of effective altruism I think many people have something like a fixed budget of time and effort that they are willing or able to spend on altruism. So the counterfactual may be costly, changing the calculus in roughly the same direction to how the "sharing the credit" view changes the calculus.
Yes, I say that.
Pretty interesting puzzle I’ve not heard before. Seems like Parfit is correct in the positive case in which lives are saved. I wonder if this idea has implications for the negative case too. Imagine 10 guys beat the crap out of someone and would do so if you weren’t there. I would want to say if you participate you are at least 1/10th or more responsible for beating him up.
I think that's wrong. Maybe you're responsible if it makes you more vicious, but doing it wouldn't harm the person extra. One way to see this: imagine that joining in would allow you to heal him after. Seems you should do that. But if you got 1/10 of the responsibility for the beating that would be false.
Hmm, that seems right.
As applied to EA, the case is even stronger because of scale. Two people saving two lives only if both act is intuitively murkier, but with EA, it's more like 1000 people saving 1000 lives or 999 people saving 999 lives.
I liken reading Bentham's Bulldog to going to church. It's not enjoyable, but enjoyment isn't the point. Just take 10 minutes to at least think about being a better person. Okay, back to my veal parmesan