I am a generalist quantitative researcher. I am open to volunteering and paid work. I welcome suggestions for posts. You can give me feedback here (anonymously or not).
Vasco Grilošø
Thanks, Ryan. Buck and Ryan, do you see any bet I could make with any of you against short transformative AI (TAI) timelines, or what they supposedly imply, that is beneficial for both of us?
Have you considered accounting for effects on soil invertebrates? One of the ākey takeawaysā from your work on risk aversion was that āSpending on corporate cage-free campaigns for egg-laying hens is robustly[8] cost-effective under nearly all reasonable types and levels of risk aversion considered hereā. However, I think such campaigns can easily increase or decrease animal welfare accounting for effects on soil ants and termites. I estimate their effects on ants and termites are much larger than those on chickens for the sentience-adjusted welfare ranges presented in Bob Fischerās book about comparing welfare across species. I suspect the vast majority of interventions perform worse than inaction accounting for effects on soil invertebrates under moderate levels of any type of risk aversion you considered (āavoiding the worstā risk aversion, difference-making risk aversion, and ambiguity aversion).
we aim to include more topics (which could include āmetaā work and research) in future iterations of the model
Good to know.
PreĀmaĀture decĀlaĀraĀtions on anĀiĀmal conĀsciousĀness hinĀder progress
Hi Nick. I agree it is important that the questions are easy to answer. However, I would say there are simple ways of letting users express their uncertainty. For the question about how much animal welfare should factor into funding decisions, there could be an option saying something like āI am very uncertain about which of the 4 views above I should pickā, or users could be allowed to give weights to each of the 4 views (instead of giving a weight of 1 to a single view). Then these answers could be used to define distributions for the probability of sentience, and welfare range conditional on sentience. Wider distributions would tend to result in a higher expected value of perfect information.
Thanks for the links, Laura. To clarify, I meant to ask whether you have considered making recommendations that specifically target decreasing the key uncertainties that matter for cross-cause prioritisation. For example, decreasing uncertainty about welfare comparisons across species by supporting work like RPās moral weight project.
Thanks for your work on this.
We modeled key uncertainties that matter for cross-cause prioritization: moral weights, time discounting, risk attitudes, aggregation across ethical views, AI-related uncertainty, and empirical uncertainty within each giving opportunity.
Have you considered making recommendations for people who do not have definitive views about the topics above? For example, question 1 of the Donor Compass asks about how much animal welfare whould factor into funding decisions, as illustrated below. I understand each option is represented by a set of point estimates describing welfare comparisons across species, but I do not undorse any particular set. I can see reasonable best guesses ranging from āOnly humans matter [in practice]ā to āAnimals matter, but somewhat less than humansā. So I think the priority should be decreasing uncertainty instead of acting based on a given set of best guesses.
Hi TFD. Great points.
Maybe? I personally would be much more discouraged from posting by a public comment than a private message. However itād be nice if we could have a live log somewhere where forum users could see that we dmād an anonymous user.
Yes, I can see many people preferring a private message. One could ask people whether they would be fine with the private message being made public.
I also think public messages are more valuable when the moderation is more contentious, and this makes them less embarrassing because it will be less clear that people did something wrong.
I can see both general guidance and specific examples about how to improve comments being useful. I believe they are often more effective together.
I appreciate being able to send someone a more casual message for butting up against guidelines rather than having to call them out publicly (which is a far more embarassing result of what can be a fairly minor infraction).
I feel like it should often be possible to give guidance to someone in public without embarassing them. One could explicitly say the infraction is minor, and use a casual tone. Public moderation may help reinforce good discussion norms in addition to increasing transparency.
One section from that post raises the concept of āAsymmetric effort ratiosā. This is definitely part of our moderation decision. At one point, if I remember correctly, you wrote almost a fifth of the words on the Forum in a week. You are very productive of long comments, which are often packed with difficult to dispel misunderstandings. This is part of why a rate-limit was the solution we arrived at. In small doses, you can be a valuable contributor, but without limit, it becomes unfairly taxing on your interlocutors.
I do not think the number of comments people can write should be greatly limited due to them having written long comments (supposedly) packed with difficult to dispel misunderstandings (I am not saying this applies to Yarrow or not; I do not know which comments you have in mind). Readers can lightly or strongly downvote comments they do not find useful, disagree with them, and point out clear mistake. What is a difficult to dispel misunderstanding is quite subjective. So effective moderation here seems difficult to me.
When writing a lot is the problem (I personally do not think this can by itself be a problem), it would be better for the rate limit to be defined in terms of words per week instead of number of comments per week? Alternatively, there could be a limit for the number of words, and a less strict limit for the number of comments.
Hi Sasha. I have only made such bets with Greg Colbourn and David Manheim. I have proposed similar bets many times. People with short timelines usually say they had better take a loan from a bank, which is fair. They do not expect to pay it fully due to a global catastrophe, or expect to pay part of it super easily due to extreme abundance. On the other hand, I suspect many of them have not taken loans up to the point of marginal borrowed money being neutral (instead of beneficial) under their views about AI.
Thanks for sharing, Bob. I think this research is valuable.
What are your thoughts on the pen and paper argument against computational functionalism (CF)? I am open to consciousness being possible in non-biological systems, but I reject CF. I do not see how any set of AND, OR, and NOT operations could itself be conscious (even if implemented with pen and paper). Digital computers are implementations of these operations. So I do not think they can be conscious.
Hi Michelle.
Preventing humans from going extinct in the next decade continues to affect the future indefinitelyāhuman extinction seems like a clear ālock-inā event.
Why so? Something being locked in forever is not sufficient for longterm benefits?
Suppose the European Unionās (EUās) ban on housing chickens in battery cages is locked in forever, in the sense the vast majority of farmed chickens in the countries which currently belong to the EU will forever be outside cages. This does not mean the marginal advocacy for the ban had longterm benefits. I think the number of chickens in cages in 1 M years would have been the same in expectation if the spending advocating for the ban had been e.g. 10 k$ smaller. I would estimate the benefits from āincrease in the probability of the banā*āacceleration in the transition to cage-free conditional on the ban being passedā. I believe this last factor is like 3 to 10 years, thus preventing astronomical benefits.
Why departing from the logic above for ālongtermist interventionsā? One could define these interventions as ones which depart from the logic above, but the question then is whether such interventions exist.
Thanks for the clarifications, Toby. They made sense to me.
Thanks for sharing, Yarrow. @Toby Tremlettš¹, have you considered being transparent about the reasons for soft and hard bans? I think the EA Forum used to have a thread where they were announced. I believe transparency is important to keep the moderation team accountable.
In my understanding (which could be wrong), they want the EA Forum to be more of a collaborative space where contributors are uplifted, and not have too much emphasis on debate or criticism. And/āor they just fundamentally disagree with what I think and the way I think (e.g., they think that what I take to be the cruxes of disagreements in many cases is not whatās really the crux), to such an extent they donāt think itās productive to discuss on the EA Forum. So, an editorial decision.
I have limited context. However, based on the above, I get the impression a soft ban is not justified. I do not think people should be banned due to disagreements with their thinking.
Hi Yarrow. Great post.
The GoodĀhart Singularity
ACE influencing significantly more funding than we distribute directly, this influence varying by charity
You have estimates for the additional funding charities get as a result of your recommendation (accounting for additional grants from ACE, and donations from individual donors)? If so, you could analyse the cost-effectiveness of that additional funding.
it being especially difficult to estimate for charities we havenāt recommended before
Have you considered assessing the marginal cost-effectiveness of charities you have recommended in the past, and the overall cost-effectiveness of charities you have never recommended?
[...] For THLās plans to expand the Open Wing Alliance, for instance, we assessed the cost-effectiveness (and theory of change) of the OWA itself, not THLās overall cost-effectiveness.
[...] For these reasons and others, Iām not sure investigating the marginal cost-effectiveness of OWA Europe and Africa specifically would have been the best use of limited evaluation time, but we do that kind of analysis when we think itās decision-relevant.
The programs from THL you assessed had a cost of 7.96 M$ (= (7.65 + 0.311)*10^6), 19.9 (= 7.96*10^6/ā(399*10^3)) times the amount you granted to THL in 2025. I understand you made more funding go to THL than what you granted, but still significantly less than 19.9 times as much? If so, I think the cost-effectiveness analysis could have focussed on more marginal spending. I believe the marginal cost-effectiveness of large programs could be significantly lower than their overall cost-effectiveness. THLās programs are quite large. So I would have thought that assessing their marginal cost-effectiveness is especially important.
Hi Nick. Do you know about any public explanations of why indirect effects on macroarthropods like soil ants and termites are not included? I suspect you have in mind the comments below. However, they apply to plants, microorganisms, nematodes, and microarthropods, not to macroarthropods like soil ants and termites.
Laura Duffy said on 17 July 2025:
Bob Fischer said on 28 July 2025:
Bob said on 21 November 2025:
From the doc linked above:
As I commented above, I estimate the effects of chicken welfare campaigns on ants and termites are much larger than those on chickens for the sentience-adjusted welfare ranges presented in Bobās book. For this not to hold, I think the sentience-adjusted welfare ranges of ants and termites would have to be much lower than that of black soldier flies (BSFs). I would be surprised if this was the case under the methodology of Bobās book. Godfrey et al. (2021) estimated 90 k neurons for a desert ant, and 92.5 k for a fruit fly (āvinegar flyā), and āindividual number of neuronsā^0.188 explains pretty well the welfare ranges in Bobās book, as illustrated below.