Beautifully written, and thanks for your work!

Hi Vasco,

When Bob was selecting the species, he was thinking of adult insects as the edge cases for the model (bees, BSF). He included juveniles to see what the model implies, not because he really thought the model should be extended to them. You'll notice that, in the book, the species list narrows considerably partly for this reason.

On the points related to sentience-conditioned welfare ranges, e.g. "So an organism having 0 neurons only decreases its welfare range conditional on sentience, and the rate of subjective experience of humans by 1/9. I understand having no neurons at all would also lead to a lower probability of sentience, but I think it should directly imply a much larger decrease in the welfare range conditional on sentience."
I think it's a mistake to point to a hypothetical sentience-conditioned welfare range, which is an intermediate step in the calculations, for an animal that has zero neurons as indicative of an issue with the methodology overall for animals with complex brains. 

Put straightforwardly, if an animal has zero neurons, it would have a welfare range of 0 overall, because I would give it a zero percent chance of being sentient, which affects all the models.

I also am not going to put a precise probability of sentience on nematodes, but I do think it's much much closer to zero and crosses the threshold of being Pascal's mugged.

I'm finding these discussions very draining and not productive at this point, so will not be engaging further in this debate. 

Hi Vasco, 

I just want to make a few points: 

1. We didn’t do the welfare range calculations for plants, protists, nematodes, etc, because we don’t think the methodology is appropriate for organisms that lack a complex brain and/or nervous system. There are a lot of methodological complexities with even applying them to complex farmed animals like chickens, and if we were to try to do something similar for very simple organisms, we might take a quite different approach.
2. We don’t really put much stock in the probability of sentience estimates, which weren't the focus of the project and are subject to much more uncertainty than the welfare range estimates themselves conditional on sentience (which themselves are highly uncertain). If you read the welfare range report’s footnotes, you’ll find that the 6.8% probability of sentience estimate for c elegans is driven substantially by my interpretation there that “probably not sentient” meant 10-35%, which was really just an off-the-cuff judgment. The other people whose views were included in that assessment gave under 1% or under 2% probabilities of sentience, and updating based on the proxies didn’t budge the priors much. On reflection, I think lower numbers are more appropriate than 6.8%, and I really would not anchor on that as “RP’s own lights”. 
3. I think part of this stems from a misunderstanding about the spreadsheets that I mistakenly linked to in the welfare range report. The vast majority of the calculations in the spreadsheet you were working off of were from a very early draft of the project, before we had ironed out a methodology and which animals we thought the methodology could apply to. Since they were a first draft and lack the full context of decisions we made along the way, I really would not consider them as our official position. I am sorry for any confusion that may have caused with respect to our methodology, opinions, or the scope of the project. Here are updated tables containing the proxies: Public Welfare Range Data and Public sentience table (Though, please note that the sentience proxies do very little work at all in the sentience probability assessments, which, again, we don’t put that much stock in, particularly for simple animals)
4. On the 0.00027 welfare range being high: 1) this was just an example to illustrate that Nick isn't correct about the structure of the model guaranteeing high numbers, not to show that it's a suitable welfare range estimate for nematodes per se. We’re not claiming that it’s actually what we would arrive at if we did assess nematodes under a more appropriate framework. And 2) it’s only high if you think you can multiply very small numbers by very big numbers and then act on that, which is a separate point. 
5. I think it’s fine if you or others have a different approach to weighing lean/likely no proxies, that was a judgment call. All of the code is public if you’d like the run it, and I created the ability for you to weigh likely/lean nos differently. That being said, they’re not creating very high estimates for many animals because there were relatively few “lean/likely no” judgments, we have many more models than just the pain/pleasure model that give lower welfare ranges, and we were quite conservative by assuming that all “Unknown” proxies were in fact absent. We’d love to have the chance to come up with new models using a more Bayesian framework, and in doing so, we might make different choices. But the point still holds that the models currently do not guarantee high welfare ranges. 
6. Speaking personally (though I know some others on the team agree), I also reject approaches to meta-normative uncertainty that can easily lead you to be dominated by one fanatical theory. If you resolve meta-normative uncertainty by maximizing choiceworthiness, you're equally susceptible to Pascal's mugging. So, if (like me) you don't want to go all-in on expected value maximization because of the Pascal’s mugging worry, you aren't going to accept strategies for resolving meta-normative uncertainty that recreate that exact problem. In this case, then, the argument that we should still think that nematode welfare dominates our calculations even if we put a small credence on total hedonic utilitarianism doesn’t move me that much. 

Overall, I encourage you and others on the EA Forum to not view our first version of the welfare range estimates as our final word on this. The book version, Weighing Animal Welfare, is more systematic, and we hope to improve on the methods in the future. But even still, I don’t think that the original version commits one to the view that very simple animals should dominate our calculations absent other highly controversial normative and meta-normative assumptions. 

Hi Vasco, 

Thanks for the question (as a researcher, I greatly appreciate the depth of your interest in our work!). It appears as though you're right about the mix-up in the formula in the spreadsheet you referenced, so I have corrected that. 

Importantly, however, I would note that the quantitative model is not one that we included in our welfare range estimates (it came from an earlier draft version of the project), and we wouldn't endorse using its results over our all-things-considered welfare range estimates that we've published here. 

Thanks again for the comment!

Hi Henry! The reason why the intervals are so wide is because they're mixing together several models. I've explained more about this modeling choice and result here: https://forum.effectivealtruism.org/posts/rLLRo9C4efeJMYWFM/welfare-ranges-per-calorie-consumption?commentId=Wc2xksAF3Ctmi4cXY

Hi Vasco, 

Thanks for this interesting post, and in general for the amount of time and consideration you’ve given to analyzing animal welfare issues here on the Forum. I want to reiterate the points others in this comment section, and urge you to consider much more explicitly the wide range of uncertainty involved in asking a question like this. In particular, the following model choices are in my opinion deserving of a more careful uncertainty treatment in your analysis:

• The probability of sentience and welfare capacities of mosquitoes. 
  • This may be substantively much different than that of black soldier flies, whose sentience and welfare capacities we are also deeply uncertain about.
• The duration of different types of pain experienced by mosquitoes as they die, conditioned on them having valenced states. 
  • I think we should be much more uncertain about the accuracy of the GPT pain track tool at capturing the true experiences of mosquitoes dying from ITNs. The estimates included in your spreadsheet vary quite a lot. 
• The number of mosquitoes killed per hour by a net 
• The weight you should give to excruciating pain relative to longer, less-intense suffering and, to a lesser extent, the weight you give to disabling pain relative to a DALY. 
  • I think it’s an open question, which is quite relevant to this analysis, how pain intensity scales with duration. 
• Whether the suffering experienced by those who have fatal cases of malaria is accurately accounted for in the analysis. 
  • In particular, pain-track estimates and GBD DALY estimates are different tools for comparing suffering. When I’ve combined/compared them in previous reports, I did so mainly by calibrating the weights for harmful and disabling pain that’s experienced by animals on a more routine basis to the descriptions of different maladies analysed by the GBD. The weight I set on excruciating pain was not high enough to overwhelm the weight given to harmful and disabling pain. 
  • But I think that comparing these two methodologies might break down when you’re giving extremely high weight to short-lived extreme pain since it’s less clear how such pain is incorporated into the DALY estimates. One would need to do a pain-track analysis for the suffering experienced if one has a fatal case of malaria for a true apples-to-apples comparison. I think this would be a fruitful area for more research. 
• The counterfactual life outcomes and welfare experienced by mosquitoes

Though you mention there is uncertainty in each of these variables, I think that it’s important to consider how they multiplicatively add up when combined and their aggregate effect on the range of plausible results. Otherwise, there’s a good risk of arriving at a directionally incorrect conclusion that can have big consequences if we act too quickly on it. This, in my view, is especially true if you’re bringing a set of controversial assumptions to bear on a sensitive and morally important topic. 

Hi Vasco, thanks for the question. 

Even though we ourselves are skeptical of the neuron count theory, many people in EA do put significant credence on it. As such, we chose to present the results that includes the neuron count model in this particular diagram.  Additionally, the differences between the results including and excluding the neuron count model are small. As we've mentioned in this post, our estimates are not meant to be precise -- rather, we think that order-of-magnitude comparisons are probably more appropriate given our significant uncertainty in theories of welfare and how best to represent them in a model.

Hi Vasco, 
Thanks for the good question! I think it's important to note that there are (at least) 3 types of model choices and uncertainty at work:
a) we have a good deal of uncertainty about each theory of welfare represented in the model,
b) we don't have a ton of confidence that the function we included to represent each theory of welfare is accurate (especially the undiluted experiences function, which partially drives the high mean results),
a) we could have uncertainty that our approach to estimating welfare ranges in general is correct, but we've not included this overall model uncertainty.  For instance, our model has no "prior" welfare ranges for each species, so the distribution output by the calculation entirely determines our judgement of the welfare range of the species involved. We also might be uncertain that simply taking a weighted mixture of each theory of welfare is a good way to arrive at an overall judgement of welfare ranges. Etc. 

Our preliminary method used in this project incorporates model uncertainty in the form of (a) by mixing together the separate distributions generated by each theory of welfare, but we don't incorporate model uncertainty in the ways specified by (b) or (c). I think these additional layers of uncertainty are epistemically important, and incorporating them would likely serve to "dampen" the effect that the mean result of the model affects our all-things-considered judgement about the welfare capacity of any species. Using the median is a quick (though not super rigorous or principled) of encoding that conservatism/additional uncertainty into how you apply the moral weight project's results in real life. But there are other ways to aggregate the estimates, which could (and likely would) be better than using the median. 
 

Seconding this question, and wanted to ask more broadly: 

A big component/assumption of the example given is that we can "re-run" simulations of the world in which different combinations of actors were present to contribute, but this seems hard in practice. Do you know of any examples where Shapley values have been used in the "real world" and how they've tackled this question of how to evaluate counterfactual worlds?

(Also, great post! I've been meaning to learn about Shapley values for a while, and this intuitive example has proven very helpful!)

Hi Michael, here are some additional answers to your questions: 

1. I roughly calibrated the reasonable risk aversion levels based on my own intuition and using a Twitter poll I did a few months ago: https://x.com/Laura_k_Duffy/status/1696180330997141710?s=20. A significant number (about a third of those who are risk averse) of people would only take the bet to save 1000 lives vs. 10 for certain if the chance of saving 1000 was over 5%. I judged this a reasonable cut-off for the moderate risk aversion level. 

4. The reason the hen welfare interventions are much better than the shrimp stunning intervention is that shrimp harvest and slaughter don't last very long. So, the chronic welfare threats that ammonia concentrations battery cages impose on shrimp and hens, respectively, outweigh the shorter-duration welfare threats of harvest and slaughter.

The number of animals for black soldier flies is low, I agree. We are currently using estimates of current populations, and this estimate is probably much lower than population sizes in the future. We're only somewhat confident in the shrimp and hens estimates, and pretty uncertain about the others. Thus, I think one should feel very much at liberty to plug in different numbers for population sizes for animals like black soldier flies.

More broadly, I think this result is likely a limitation of models based on total population size, versus models that are based more on the number of animals affected per campaign. Ideally, as we gather more information about these types of interventions, we could assess the cost-effectiveness using better estimates of the number of animals affected per campaign. 

Thanks for the thorough questions!