Thanks for following up, Anthony.

My best guess about which of 2 identical objects has a larger mass in expectation will be arbitrary is their mass only differs by 10^-6 kg, and I have no way of assessing this small difference. However, this does not mean the expected mass of the 2 objects is fundamentally incomparable. Likewise, my best guess about which of 2 actions increases welfare more in expectation may be arbitrary without this implying that their expected change in welfare is incomparable.

I am not sure it matters whether one endorses precise expected values (EVs) or not. In practice, I still like to test different EVs when the underlying probability density function (PDF) is very arbitrary and uncertain, as it is the case for PDFs of welfare ranges. In such cases, I suspect decreasing uncertainty to find the best options has higher EV than the EVs of going with the current best option which feel reasonable.

Hi Michael.

Do you think it's reasonable for two people with all of the same evidence to disagree on precise probabilities and expected values?

It depends on what is included in "all of the same evidence". If 2 people had exactly the same evidence about everything, including internal states about the plausibility of the probabilities, they would be the same people, and therefore would agree on everything. In practice, different people share some evidence, but start with different priors, and therefore do not have to agree on precise probabilities and expected values. The stronger the evidence they share relative to their priors, the more they will agree.

Or would you just average yours and theirs in some way to get a new distribution? How?

If 2 probability density functions (PDFs) feel exactly as plausible, I would simply use the mean between them.

Side note. I often link to concepts in my comments that I am sure the person I am replying to is familiar with, but I do it anyway in case others find it relevant.

How would you choose the distributions for the model weights in a way that's not itself arbitrary? E.g. how do you choose their forms and parameters in a way that's not arbitrary?

I agree the distributions for the model weights would be arbitrary to some extent. However, I think probability density functions (PDFs) should be precise at a fundamental level, which implies precised expected values (EVs). If 2 PDFs feel exactly as plausible, I would simply use the mean between them.

I am not sure it matters whether one endorses precise EVs or not. In practice, I still like to test different EVs when the underlying PDF is very arbitrary and uncertain, as it is the case for PDFs of welfare ranges. In such cases, I suspect decreasing uncertainty to find the best options has higher EV than the EVs of going with the current best option which feel reasonable.

On there potentially being no fact of the matter, this may be helpful. It goes further than the issue of imprecise credences/EVs.

Here is a seemingly great summary from Gemini.

Thanks for the comment, Mal.

Unfortunately I think this post is otherwise not very relevant, mainly because no one uses Contrapest [...]

[...]

• Since I'm not sure Evolve works either, it may be that any fertility control product used in the future has an entirely different formulation, making speculation here quite difficult.

Do you agree replacing the rodenticide bait with ContraPest may impact soil animals much more than rodents for my estimate that it decreases cropland by 0.413 m²-year per initial rodent? If so, how much smaller do you think the change in cropland would have to be for replacing the rodenticide bait with Evolve to robustly increase animal welfare (in expectation, and accounting for all animals)?

Here is an illustration of how to quickly estimate the effects on soil animals for other fertility control baits. Imagine replacing the rodenticide bait with Evolve increases or decreases cropland by 10 %. I estimate full depopulation with the rodenticide bait increases cropland by 0.481 m²-year per initial rodent. So replacing the rodenticide bait with Evolve would increase or decrease cropland by 0.0481 m²-year per initial rodent (= 0.10*0.481). This means the effects on soil animals would be 11.6 % (= 0.0481/0.413) as large as for replacing the rodenticide bait with ContraPest. For effects on soil animals 10 % as large (a round fraction to simplify the calculations), depending on the biome replacing the cropland, my estimates for the change in the living time of soil animals per initial rodent for full depopulation with fertility control instead of rodenticide bait would range from:

• 1.52 to 43.7 soil-ant-years.
• 9.46 to 68.5 soil-termite-years.
• 65.1 to 3.00 k soil-springtail-years.
• 134 to 5.07 k soil-mite-years.
• 466 to 8.01 k soil-arthropod-years.

To clarify, the reason I viewed (at time of writing of the previous post) that rodenticide replacement on islands might be approximately ecologically inert is that conservationists use products until rats are entirely eradicated and then stop. [...]

[...]

• I strongly doubt that comparing "cost [cropland] to prevent one birth/cause one death" is the right comparison. The population dynamics of suppressing fertility are different than those from killing adults, so the timeline to eradication (in the island context) and the application volumes would likely be different, among other things.

Lorenzo Buonanno's comment made me update the post in agreement with the 1st paragraph above, although I only read this paragraph after my update. I am now using the amount of rodenticide and fertility control bait for full depopulation per initial rodent instead of the amount needed to kill one rodent, and prevent one rodent birth.

[...] You have this short duration of time during the eradication, and then a long period after where conditions on the island are basically the same. [...]

[...]

• If we're talking about use in settings other than islands, I want to reiterate that those are definitely not ecologically inert even to some approximation, and there are tons of things you'd have to look at besides production method if you wanted to estimate total animal effects. Age structure changes in rat populations, food web dynamics, most rodent fertility control agents appear to have reasonably strong insecticidal effects while brodifacoum doesn't affect most invertebrates we've studied, etc.

I neglect differences in the population of rodents between the 2 depopulation methods for simplicity, and underestimating the effects on soil animals.

although I do wonder if the limited duration of these eradications and amount of use for island conservation relative to overall production is high enough to make a meaningful impact on land use? I don't think these things are continuous so a small enough demand shift signal might not have any impact? Just speculating...

I agree cropland changes in steps, but I do not think this changes the overall picture. Relatedly, it may naively seem that decreasing the consumption of chicken by 0.1 kg does not change the production of chicken if this can only be adjusted by multiples of e.g. 1 k kg. However, in this case, a better model would be that decreasing the consumption of chicken by 0.1 kg would increase by roughly 0.01 pp (= 0.1/(1*10^3)) the probability of the production of chicken decreasing by 1 k kg. So the expected reduction in the production of chicken would still be roughly 0.1 kg (= 1*10^-4*1*10^3).

• The dosages are difficult to determine. [...]

[...]

• Other complications: The dosing requirements also seem to be quite different between species of rats. [...]

Accounting for considerations like these would increase uncertainty. However, the final conclusion would still be that controlling the fertility of rodents instead of killing them can easily increase or decrease welfare, even on islands? 

• I think I'm just generally more comfortable than you balancing my uncertainties with a combination of direct work on arthropods, and work on things that have (1) high non-arthropod upsides and (2) where I'm clueless about the effects on arthropods. [...]

I agree.

@Dane Valerie, would it make sense for moderators to mention the reasons for comments being deleted? I guess you deleted the comment I replied to because it was not published by a human.

Thanks for starting this thread, Lorenzo. I strongly upvoted your initial comment because it prompted me to significantly update the post. Thanks for the relevant context, Mal. I am now using the amount of rodenticide and fertility control bait for full depopulation per initial rodent instead of the amount needed to kill one rodent, and prevent one rodent birth. I think my current approach is better because full depopulation (in a given area) is the practical goal of using rodenticide or fertility control bait. In my past approach, the goal was preventing one rodent birth via fertility control for each rodent that would be killed with rodenticide, which I worry has little connection with the practical implementation of the intervention.

The conclusions of the post have not changed. I had calculated an increase in cropland of 0.621 m²-year per rodent birth prevented. Now I estimate a decrease in cropland of 0.413 m²-year per initial rodent. So the change in the living time of soil animals per initial rodent is -66.5 % (= -0.413/0.621) that I had obtained per rodent birth prevented. However, my conclusions only depended (and depend) on the ratio between the (size of the) effects on soil animals and rodents. As I said (and say) in the post, "I have very little idea about whether it [controlling the fertility of rodents instead of killing them] increases or decreases the welfare of soil animals".

Hi Ben and Richard.

Next year, as well as continued engagement on urban infrastructure, we’ll work on new policy areas such as fertility control and pesticide policy. [...]

[...]

We selected these focus areas – after extensive research and consultation with wild animal welfare experts – because we believe some policy options look realistic, robust and helpful. That is:

• Tractable in the near term (e.g. wild animal welfare scientists have a recommendation, and there’s an upcoming consultation or policy window), and;
• Robustly positive across a range of worldviews (i.e. we seek to minimize backfire risks where possible, including individual and population-level risks), and;
• High in expected value, and/or helpful for spreading our values.

I believe controlling the fertility of rodents instead of killing them may impact soil ants and termites much more than rodents even if the population of rodents remains constant. So I do not think it robustly increases welfare accounting for all animals.

Hi Mal.

As a result, these individuals hope to identify "ecologically inert" interventions that don't affect population dynamics or have cascading effects. Corporate welfare campaigns might be one sort of intervention that clears this bar.

I think chicken welfare reforms may impact soil ants and termites much more than chickens.

I (and several others) think we could reasonably view a handful of ["ecologically inert"] interventions as worth pursuing under this mindset. Mostly, these sorts of interventions change how humans kill animals or control populations, such that suffering is decreased without changing the net population outcome. Examples might include stunning wild-caught fish before slaughter or replacing rodenticides with fertility control on islands.

I believe controlling the fertility of rodents instead of killing them may impact soil ants and termites much more than rodents even if the population of rodents remains constant.

I would be curious to know your thoughts on this discussion between me and Anthony DiGiovanni about imprecise expected values.

Thanks, Michael. Do not worry about not having replied earlier.

I agree that the weights/coefficients in the model could end up quite arbitrary, and I would expect them to if someone tried to set them precisely.

I am still thinking that expected values should be precise, or at least practically precise. However, I think the weights of models should be modelled as distributions instead of constants as in Bob's book about comparing welfare across species, and Rethink Priorities' (RP's) digital consciousness model (DCM).

We may be able to give some arguments for some bounds on the weights, and some structural constraints on how the weights relate to each other

I agree.

Within these constraints, the choices are very subjective and highly arbitrary.

I agree.

there may be no fact of the matter at all

I disagree.