I am a Researcher at Rethink Priorities, working mostly on cross-cause prioritization (though I occasionally do global health and development research as well). I am passionate about farmed animal welfare, global development, and economic growth/progress studies. Previously, I worked in U.S. budget and tax policy as a policy analyst for the Progressive Policy Institute. I earned a B.S. in Statistics from the University of Chicago, where I volunteered as a co-facilitator for UChicago EA’s Introductory Fellowship.
Laura Duffy
Karma: 1,220
Hi Vasco and Nick,
On the future weights specifically, I think you raise a really interesting point. I tentatively think there are some differences between the GCR modeling and soil animal effects modeling, at least at this stage, but I understand if one wants to zero out their effects (and I suggest below ways to do so).As a way of background, in the model for GCRs, we largely adopt the Time of Perils perspective, wherein we’re living in an unusually risky period of time, but if we survive past it, existential risk drops to some lower level. The expected value of the future is partially governed by the probability that we survive into the far future. Then, interventions that reduce/increase risk now–even if they only have temporary effects–raise/lower the probability of civilization surviving until a specific year and realizing the value of civilization in that year.
Based on this structure, I do think that the case of modeling the effects of GCR interventions in the far future is meaningfully different enough from modeling the effects of interventions on soil animals to justify including the former in the model but not the latter, at least at this point. For one, we do have some existing frameworks from within the GCR and cause-prioritization academic communities for modeling them (which we drew upon in doing the modeling, see here and here for example). Consistent with working on the terms of the fields themselves, it’s my understanding that at least people think it would be bad if human civilization ended, so that reducing x-risk is good (if we can figure out interventions that actually succeed at this). By contrast, it’s my understanding that how to incorporate effects on soil animals into our models is pretty uncertain, as we’re highly uncertain about all of the mechanisms that affect animal welfare that are involved. And, as you’ve pointed out, we’re also highly uncertain about the direction of the impact on soil animals.
There are many possible improvements that we could make to the modeling of GCR interventions’ long-term effects (see below). Yet, being able to draw upon existing frameworks within the time that we built the first version and having some consensus about the directional value of reducing x-risks made us more confident about including their effects on the longer-term than we feel about soil animals.
Nevertheless, I entirely understand being very skeptical of effects 500+ years out (or even 100-500 years out), as Nick is. I’d recommend using the Advanced User mode and setting the weights for effects 100-500 and 500+ years out to zero. In our baseline inputs, we do include five (out of 14) “clueless” worldviews that assign zero weight to impacts over 100 years out, taking up 30% of the overall credence. Of course, I think reasonable people can prefer to give more weight to clueless worldviews, which they can do with the tool linked above.
I hope that this answer is clear enough – again, it’s an area of significant uncertainty and there can be many different legitimate approaches.Addendum: Improvements we could make to our modeling of GCR impacts in the far future
Going forward, I think there are many different improvements that we might want to make to our models of the effects of GCR interventions in the far future.
Though we adopt wide uncertainty levels for all of the parameters involved, which result in very different expected values of the future, one weakness of the model is that it’s entirely possible that the future does not follow a Time of Perils trajectory at all (i.e. that risk is flat, or that we go through many periods of high-then-low-risk). Part of the reason we did so is because it makes integrating into the far future tractable, and also because we wanted to represent GCRs as a field in a way that those working inside of it would agree is fair, and that we’re in a Time of Perils is a quite common point of view. But I think a real area of improvement for the model in the future would be to figure out how to incorporate different risk trajectories.
Additionally, I think we could consider implementing some kind of Bayesian discount (like that which David Bernard discussed here in RP’s CURVE sequence) to the expected value of the impact of the GCR interventions for each year in the future. I don’t think we have a very good intuition for what order of magnitude discount to apply to the 500+ year period overall (since our uncertainty could grow very rapidly). I also think that the discount we apply to the impact in the 501st year should be different from the discount in the 10^6th year. Finally, an upside of this approach would be that it separates out cluelessness from person-affecting worldviews, which the time discounts currently don’t do very well.
Hi Vasco,
Generally, I think that modeling is most useful in situations where we know enough about an issue to construct a solid framework for what effects to include, how to analyze it, and how to provide some evidence-based justification for key parameter values.Given the enormous, many-layered uncertainties that surround second-order effects (like those on soil animals), and given that we don’t have yet a framework for analyzing such second-order effects comprehensively and equally across all interventions, I think it wouldn’t be responsible for me to speculate on either how various kinds of risk aversion can or should apply to them, or what the impact on our recommendations would be.
Because we have limited capacity, this is going to be my last comment about soil animals in particular. However, if you have questions about other aspects of the Cross Cause Fund, we would be happy to engage with them.
Hi Vasco,
Right now, we’re not including second-order effects of interventions, and in this case, we’re deeply uncertain about the magnitude or direction of effects like those you mention. I think you’re right to point out that more research is needed here to address questions concerning the welfare capacity, baseline welfare of, and effects of any interventions on invertebrates.As mentioned above, we’ve for the time being chosen to focus on funds that already exist, have a strong track record, and can absorb a considerable amount of funding this year. In general, we’d like to expand the set to include individual interventions. Whether that will include more research on soil animals, I really can’t say right now, because we don’t yet know how granular we’re going to get with the included projects. Please look out for more updates in the coming months, as this is an evolving project.
I think I see now, thanks for the clarification. We don’t currently include funds/interventions that specifically work on research (to reduce key uncertainties or otherwise). We do think this kind of work is important, and we aim to include more topics (which could include “meta” work and research) in future iterations of the model.
Hi Vasco,
Thanks for the question! We’ve designed the Donor Compass to be more streamlined, but we certainly appreciate and share your moral uncertainty. Our recommendations for the Cross-Cause Fund take into consideration 14 distinct worldviews, which take into account a wide range of animal moral weights (among other variables). You can investigate the assumptions for each here, along with the credences placed on each. If the range of views you place credence on significantly differ, however, you can use the Advanced Mode of the Cross-Cause fund to tailor it to your specific needs. (The definitions of each term can be found in this spreadsheet)
Hi Benton,
Thanks for the comment! To clarify, our model and recommendations do draw upon the same approaches to moral uncertainty as our moral parliament. Our Cross-Cause Fund draws upon 14 different worldviews (see here for the exact details) and uses a weighted average of recommendations across nine methods of aggregating across them.Additionally, though our main Donor Compass tool is simplified, you can use a more Advanced Mode here to incorporate moral uncertainty across several combinations of worldviews. (I’m not familiar with Ross’ prima facie duties theory, but hopefully you could represent it adequately in the model.) Then, just like the Moral Parliament, you can specify which aggregation method you want to apply to resolve disagreements between worldviews.
Of course! We’re also happy to answer any additional methodological questions you may have in the future
Hi Clara,
Thanks for the good question!
Short answer: yes, you’re understanding this correctly. And, yes, the model’s allocations can be sensitive to putting any non-zero weight in for the value of impacts 500+ years out (but worldview diversification can mitigate this factor’s impact on the outcome).
The model computes a “score” for each fund based on the sum of: (impact in time period t)*(weight in time period t) over all time periods.
In our estimates of the impacts of GCR funds, we do take the approach of estimating the impact of avoiding an existential catastrophe over many generations in the future. As such, if you were to give full weight to further-out time periods, the vast majority of the expected impact of avoiding an existential catastrophe is in the long-run future. So any weight that is meaningfully above zero will make these projects have a large-in-magnitude score, all else equal, compared to putting a weight of zero. (We’re preparing a more detailed sensitivity analysis that we’ll release soon, and which will address your question in more detail.)
Nevertheless, it’s important to note that our recommended allocations are influenced by 14 worldviews, some of which assign zero weight to the far future. As such, we’re not forced to choose between putting zero weight on the 500+ year period vs. some non-trivial amount that swamps the calculations. This creates a meaningful amount of diversification within the model’s results, allowing other factors like moral weights to be influential.
Additionally, there’s an interaction between risk attitudes and the amount of weight that one puts on the far future, such that putting a non-zero amount on the 500+ year period doesn’t automatically recommend you spend 100% of your budget on GCR funds. For instance, if you’re highly risk-averse and put a weight of 1% on the 500+ year period, then you might avoid funding certain GCR funds that have a high enough chance of raising existential risk, because permanently harming the long-run future would have such a considerable impact.
If you’re interested in reading more about the worldviews we’ve used in the model, please feel free to reference this link.
Thanks again for the question!
Digital Consciousness Model Results and Key Takeaways
Digital Consciousness Model launch events
[Question] Cause prio cruxes in 2026?
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: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.
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”.
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)
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.
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.
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.
Doing Prioritization Better
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!
Beyond Extinction: Revisiting the Question and Broadening Our View
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, thanks for the question! You can find the Navigation Fund and all the others on the main page of our website, under the section header “The Rethink Priorities Cross-Cause Fund currently covers the following funds”.