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.
Hi Laura. Thanks for the reply. I very much agree effects on soil invertebrates resulting from land use changes have very uncertain magnitude and direction. However, they should not be neglected under the types of risk aversion you studied?
“Avoiding the worst” risk aversion: All else equal, we are averse to the worst states of the world arising and want to take actions that prevent them or lessen their badness.
Difference-making risk aversion: All else equal, we are averse to our actions doing no tangible good in the world or, worse, causing harm.
Ambiguity aversion: All else equal, we should be particularly cautious when taking actions for which the probabilities of the possible outcomes are unknown and quite uncertain.
“world” in 1 and 2, and “possible outcomes” in 3 should include effects soil invertebrates?
Do you agree that cage-free campaigns for laying hens may decrease the welfare of soil invertebrates much more than they increase the welfare of chickens, thus decreasing animal welfare a lot? I think this is very much on the table (although I can also see the effects on soil invertebrates being negligible). So I believe inaction is better than such campaigns for a sufficient level of “avoiding the worst”, or difference-making risk aversion. In addition, I infer inaction is better for a sufficient level of ambiguity aversion because the effects on soil invertebrates are very uncertain.
If we give extra weight to net harm over net benefits compared to inaction, as in typical difference-making [and “Avoiding the worst”] views, I think most animal interventions targeting vertebrates will look worse than doing nothing, considering only the effects on Earth or in the next 20 years, say. This is because:
there are possibly far larger effects on wild invertebrates (even just wild insects and shrimp, but also of course also mites, springtails, nematodes and copepods) through land use change and effects on fishing, and huge net harm is possible through harming them, and
there’s usually at least around as much reason to expect large net harm to wild animals as there is to expect large net benefit to them, and difference-making gives more weight to the former, so it will dominate.
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.
Because we have limited capacity, this is going to be my last comment about soil animals in particular.
I am replying in case anyone is interested, or you want to comeback to it later. Feel free not to reply, and thanks for the thoughts you have shared.
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.
Results are more certain in the situations you describe, but very uncertain results could still be informative. They may help identify the most important uncertainties. The results would not be useful even for this if they were sufficiently arbitrary. However, I do understand why you would believe this.
Based on the “More options” drop-down in question 1 of the Donor Compass, you are comparing the welfare of humans with that of chickens, shrimps, and “non-shrimp invertebrates”, which I assume includes BSFs (as these were covered in Bob’s book). So I would have expected you to be open to comparing the welfare of chickens with that of soil ants and termites, which are macroarthropods like shrimps and BSFs.
In addition, based on the “More options” drop-down in question 2 of the Donor Comass, you are modelling effects after 500 years, which I think are very uncertain. So I would have expected you to be open to modelling how changes in feed consumption resulting from improving the conditions of farmed animals impact the population of soil invertebrates.
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.
Makes sense. I just meant to illustrate accounting for effects on soil invertebrates could change recommendations even if there is large uncertainty about their magnitude and direction.
“In addition, based on the “More options” drop-down in question 2 of the Donor Comass, you are modelling effects after 500 years, which I think are very uncertain.”
That’s not a bad point. I would be kind of on the opposite end of the scale though. We are so clueless about both 500 years in the future and soil arthropods I wouldn’t be modeling either of those....
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.
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 is strong evidence that chickens in different conditions need a different amount of feed, and that this affects land use, and therefore soil animals? There is uncertainty about the number of soil-animal-years affected, and their welfare per animal-year, but I do not understand why it cannot simply be modelled with wide distributions, at least in the case of soil ants and termites (as there are other macroarthropods covered in Bob’s book about comparing welfare across species).
I don’t think welfare interventions targeting vertebrates will necessarily look worse than doing nothing on “Avoiding the worst” views, because they don’t specifically, AFAIK, increase the risks of worse cases than inaction. I think things that reduce land use for agriculture, including a lot of alt protein and veg advocacy work, plausibly do look bad in the near-term on “Avoiding the worst” views, because the near-term worst cases are where invertebrates have bad lives, modest/high moral weights and larger populations.
But I do think on many difference-making views that compare to some default option like inaction and weigh downsides at least linearly and more than upsides, most interventions will look worse than the default.
I don’t think welfare interventions targeting vertebrates will necessarily look worse than doing nothing on “Avoiding the worst” views, because they don’t specifically, AFAIK, increase the risks of worse cases than inaction.
Interventions which cost-effectively increase the welfare of vertebrates will change land use much more than inaction, and a greater change in land use increases the probability of causing lots of suffering? Are you assuming that i) such interventions would increase agricultural land, and that ii) this decreases suffering?
On i), such interventions may decrease agricultural land due to increasing the price of animal products, and therefore decreasing their consumption? I estimate that replacing Ross 308 (fast growth broiler breed) with Rebro (slower growth) decreases cropland by 0.102 m²-year/Ross-308-chicken-kg, and that replacing replacing eggs from battery cages with those from barns or aviaries decreases cropland by 0.529 m²-year/battery-cages-egg-kg. These estimates neglect increases in the consumption of other foods, but I believe accounting for this would increase uncertainty, and therefore further increase the risk of the worst outcomes.
On ii), increasing agricultural land may increase suffering by increasing the number of soil macroarthropods/nematodes? I think effects on these may dominate (given the large uncertainty about welfare comparisons across species), and they may have negative lives (although I can easily see them having positive lives too).
Interventions which cost-effectively increase the welfare of vertebrates will change land use much more than inaction, and a greater change in land use increases the probability of causing lots of suffering?
It might cause lots of suffering, but it could also prevent lots of suffering, too. I think you’re thinking in terms of difference-making, but “Avoiding the worst” risk aversion is not difference-making. Rather than thinking about what you cause, you should just look at both (distributions of) outcomes and ask which has more suffering in it, without privileging the results of inaction.
Unless you believe the expected amount of wild animal suffering is higher all-things-considered than with inaction, you shouldn’t really expect it to do worse according to “Avoiding the worst” risk aversion (as a heuristic; there could be exceptions).
It might cause lots of suffering, but it could also prevent lots of suffering, too.
I agree. So the worst case is that the campaigns cause lots of suffering (relative to inaction)?
[...] Rather than thinking about what you cause, you should just look at both (distributions of) outcomes and ask which has more suffering in it, without privileging the results of inaction.
Unless you believe the expected amount of wild animal suffering is higher all-things-considered than with inaction, you shouldn’t really expect it to do worse according to “Avoiding the worst” risk aversion (as a heuristic; there could be exceptions).
I understand I should look into the distributions of global welfare with and without the campaigns, and then assess their negative tails. I have little idea about which distribution has the highest expected value. However, I believe the distribution with the campaigns has longer positive and negative tails, and therefore the risk of the worst outcomes is higher with the campaigns (although the probability of the best outcomes is also higher).
Stepping back, regardless of how accounting for risk aversion changes recommendations, I would like greater reasoning transparency about why effects on soil invertebrates have been neglected. Roughly for the reasons @Marcus_A_Davismentioned replying to Nick’s concerns about conflicts of interest.
I [Marcus] think there is not a single EA organization I would consider unbiased on this question [cross-cause recommendations], including ourselves (despite our ongoing efforts not to be). That is exactly why we publish so much of our methodology and our assumptions openly. One of the main motivations for this work is concern about the effect of bias when assumptions and models are implicit or hidden. We would welcome more experts with broader backgrounds being involved in drafting and improving these estimates, which is part of what we hope this kind of public methodology enables.
In particular, I would like greater transparency about why effects on soil macroarthropods were neglected. They are covered in Bob’s book, unlike soil nematodes and microarthropods. Moreover, I estimate cage-free campaigns for laying hens change the welfare of soil ants and termites much more than they increase the welfare of chickens for the sentience-adjusted welfare ranges presented in Bob’s book.
I agree. So the worst case is that the campaigns cause lots of suffering (relative to inaction)?
I don’t think this is right. I think you’re still treating “Avoiding the worst” like a difference-making view. You shouldn’t be thinking in terms of “relative to inaction”, which itself has a highly uncertain distribution of outcomes. Just evaluate the distribution of outcomes for each option, without fixing any as a comparison option.
The question is only whether worst-case outcomes are more or less likely with action or inaction.
FWIW, the actual worst cases are s-risks, and I’d expect “Avoiding the worst” views to prioritize their mitigation, as long as we’re not clueless about that.
You shouldn’t be thinking in terms of “relative to inaction”, which itself has a highly uncertain distribution of outcomes. Just evaluate the distribution of outcomes for each option, without fixing any as a comparison option.
I had understood this. As I said, “I understand I should look into the distributions of global welfare with and without the campaigns, and then assess their negative tails”. My phrasing “cause lots of suffering (relative to inaction)” was confusing. However, I meant “increase the probability of outcomes with lots of suffering (the worst) relative to the probability under inaction”.
Ah, sorry, I was too quick and should have read more carefully.
However, I believe the distribution with the campaigns has longer positive and negative tails
Why do you believe this?
As chicken and egg prices increase from these welfare reforms, I would expect:
some shifts between crops and nature (including through substitution), but I’m clueless about which involves more suffering, so this doesn’t clearly favour one or the other.
substitution towards beef and other pasture products, which reduces invertebrate populations substantially, and probably without making lives much worse. This would mean less suffering and so do better in the worst case.
Here is how I am thinking. Imagine the world has welfare 0.
With inaction, the final welfare will be 0 with probability 100 %.
With campaigns, there is lots of uncertainty, but here is a simplified set of outcomes:
Agricultural land will increase with probability 75 %, and decrease with probability 25 %.
If agricultural land increases, the final welfare will be −1 with probability 25 %, and 1 with probability 75 %. So the final welfare will be −1 with probability of 18.75 % (= 0.75*0.25), and 1 with probability 56.25 % (= 0.75*0.75) considering outcomes where agricultural land increases.
If agricultural land decreases, the final welfare will be −1 with probability 75 %, and 1 with probability 25 %. So the final welfare will be −1 with probability 18.75 % (= 0.25*0.75), and 1 with probability 6.25 % (= 0.25*0.25) considering outcomes where agricultural land decreases.
As a result, final welfare will be −1 with probability 37.5 % (= 0.1875*2), 1 with probability 62.5 % (= 0.5625 + 0.0625), and 0.25 (= 0.375*(-1) + 0.625*1) in expectation.
The worst possible outcome across the 2 interventions is a final welfare of −1. With inaction, it has a probability of 0. With campaigns, it has a probability of 37.5 %. So the campaigns make the worst possible outcome more likely.
Unless you believe the expected amount of wild animal suffering is higher all-things-considered than with inaction, you shouldn’t really expect it to do worse according to “Avoiding the worst” risk aversion (as a heuristic; there could be exceptions).
The intervention which can decrease welfare the most is the one leading to the lowest possible final welfare.
With inaction, the final welfare will be 0 with probability 100 %.
This is exactly a procedure you could follow for difference-making risk aversion; it’s equivalent to taking the statewise difference with inaction. The welfare of the world with inaction isn’t 0 with probability 100%.
RP has a model/procedure for avoiding the worst risk aversion here.
The welfare of the world with inaction isn’t 0 with probability 100%.
Why? I was assuming a world with initial welfare 0 with probability 100 %. However, I think my point stands for any distribution describing the initial welfare.
RP has a model/procedure for avoiding the worst risk aversion here.
If you instead set the campaign option to 0 welfare and defined the welfare of the world with inaction relative to the campaign option, you’d end up with the opposite conclusion, that only inaction reaches −1.
Avoiding the worst is meant to treat each option symmetrically. It doesn’t depend (in theory) on which option you single out to define things relative to.
(RP’s practical procedure does start with inaction, but if you end up with the same probability distributions for each option in the end, the results will be the same as if you started with a different option to define all distributions relative to. I think their procedure helps ensure consistent probability assignments and is less work, compared to directly estimating each distribution independently.)
What exactly do you mean by this? The campaign has many potential effects. So it cannot result in a final welfare of X with probability 100 %, where X can be 0 or any other number.
Suppose the initial welfare is 0 with probability 100 %. Inaction would lead to a final welfare of 0 with probability 100 %. Imagine an intervention which decreases welfare by 1 with probability 50 %, and increases welfare by 1 with probability 50 %. The intervention leads to a final welfare of 0 in expectation. However, it leads to a final welfare of −1 with probability 50 %, and 1 with probability 50 %. The the lowest possible welfare of −1 is more likely with the intervention?
Inaction also does not in fact lead to welfare of 0 with probability 100%. There will be lots of animals suffering and many possible outcomes if we do nothing. So it’s not correct to assume total welfare of 0.
I think my point stands for any distribution describing the initial welfare. Imagine the minimum initial welfare W_min has probability p. With inaction, the minimum final welfare would still be W_min, and have probability p. With an intervention which decreases welfare by 1 with probability 50 %, and increases welfare by 1 with probability 50 %, the minimum final welfare would be W_min − 1 with probability 0.5*p. So the lowest possible welfare of W_min − 1 would be lower and more likely with the intervention?
No, I don’t think this is the right way to model this. This looks a lot like the typical error people make for the original two envelopes problem.
Initial welfare (what does that mean?) and final welfare after inaction can differ, because the world, e.g. land use, will change even if you do nothing, and campaigns take time for their effects to materialize.
If you swapped the roles of campaign and inaction, you would flip the conclusion, too.
This looks a lot like the typical error people make for the original two envelopes problem.
The moral two envelopes problem is not problematic if there is a common scale to compare the welfare per unit time (as there is to compare temperature)?
Initial welfare (what does that mean?) and final welfare after inaction can differ, because the world, e.g. land use, will change even if you do nothing, and campaigns take time for their effects to materialize.
Suppose that inaction leads to a distribution for the future welfare (integral of the welfare per unit time across all future time) whose minimum value W_min has probability p. With an intervention that decreases future welfare by 1 with probability 50 %, and increases it by 1 with probability 50 %, the minimum future welfare would be W_min − 1 with probability 0.5*p. So I think the lowest possible future welfare of W_min − 1 would be lower and more likely with the intervention (although the intervention would not change future welfare in expectation).
If you swapped the roles of campaign and inaction
What do you mean by this? By definition, inaction does not change the distribution of the future welfare?
I see. Thanks for the patience. I could equally say that an intervention leads to a distribution for the future welfare whose minimum value W_min has probability p, and that inaction decreases it by 1 with probability 50 %, and increases it by 1 with probability 50 %, thus implying a minimum future welfare of W_min − 1 with probability 0.5*p. This is the exact opposite of what I concluded above, and suggests the lowest possible future welfare of W_min − 1 would be lower and more likely with inaction.
I agree both models are wrong. I cannot assume that the change in future welfare caused by the intervention is independent from the future welfare under inaction (as I did in my past comments), or that the change in future welfare caused by inaction is independent from the future welfare caused by the intervention (as I did just above).
Unless you believe the expected amount of wild animal suffering is higher all-things-considered than with inaction, you shouldn’t really expect it to do worse according to “Avoiding the worst” risk aversion (as a heuristic; there could be exceptions).
I agree that increasing welfare in expectation is a good heuristic for better performance under “avoiding the worst” risk aversion. I have very little idea about whether cage-free campaigns for laying hens increase or decrease welfare in expectation. So I do not know whether they are favoured or not under “avoiding the worst” risk aversion. They are still disfavoured under difference-making and ambiguity risk aversion, and this could make them worse than inaction. In addition, they may be worse than inaction under no risk aversion of any type.
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.
Hi Laura. Thanks for the reply. I very much agree effects on soil invertebrates resulting from land use changes have very uncertain magnitude and direction. However, they should not be neglected under the types of risk aversion you studied?
“world” in 1 and 2, and “possible outcomes” in 3 should include effects soil invertebrates?
Do you agree that cage-free campaigns for laying hens may decrease the welfare of soil invertebrates much more than they increase the welfare of chickens, thus decreasing animal welfare a lot? I think this is very much on the table (although I can also see the effects on soil invertebrates being negligible). So I believe inaction is better than such campaigns for a sufficient level of “avoiding the worst”, or difference-making risk aversion. In addition, I infer inaction is better for a sufficient level of ambiguity aversion because the effects on soil invertebrates are very uncertain.
Here is a related comment from @Michael St Jules 🔸.
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.
I am replying in case anyone is interested, or you want to comeback to it later. Feel free not to reply, and thanks for the thoughts you have shared.
Results are more certain in the situations you describe, but very uncertain results could still be informative. They may help identify the most important uncertainties. The results would not be useful even for this if they were sufficiently arbitrary. However, I do understand why you would believe this.
Based on the “More options” drop-down in question 1 of the Donor Compass, you are comparing the welfare of humans with that of chickens, shrimps, and “non-shrimp invertebrates”, which I assume includes BSFs (as these were covered in Bob’s book). So I would have expected you to be open to comparing the welfare of chickens with that of soil ants and termites, which are macroarthropods like shrimps and BSFs.
In addition, based on the “More options” drop-down in question 2 of the Donor Comass, you are modelling effects after 500 years, which I think are very uncertain. So I would have expected you to be open to modelling how changes in feed consumption resulting from improving the conditions of farmed animals impact the population of soil invertebrates.
Makes sense. I just meant to illustrate accounting for effects on soil invertebrates could change recommendations even if there is large uncertainty about their magnitude and direction.
“In addition, based on the “More options” drop-down in question 2 of the Donor Comass, you are modelling effects after 500 years, which I think are very uncertain.”
That’s not a bad point. I would be kind of on the opposite end of the scale though. We are so clueless about both 500 years in the future and soil arthropods I wouldn’t be modeling either of those....
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.
Thanks for the context, Laura.
There is strong evidence that chickens in different conditions need a different amount of feed, and that this affects land use, and therefore soil animals? There is uncertainty about the number of soil-animal-years affected, and their welfare per animal-year, but I do not understand why it cannot simply be modelled with wide distributions, at least in the case of soil ants and termites (as there are other macroarthropods covered in Bob’s book about comparing welfare across species).
I don’t think welfare interventions targeting vertebrates will necessarily look worse than doing nothing on “Avoiding the worst” views, because they don’t specifically, AFAIK, increase the risks of worse cases than inaction. I think things that reduce land use for agriculture, including a lot of alt protein and veg advocacy work, plausibly do look bad in the near-term on “Avoiding the worst” views, because the near-term worst cases are where invertebrates have bad lives, modest/high moral weights and larger populations.
But I do think on many difference-making views that compare to some default option like inaction and weigh downsides at least linearly and more than upsides, most interventions will look worse than the default.
I also have a similar comment here and a piece critiquing and exploring different versions of difference-making more generally here. A version of difference-making that wouldn’t let invertebrates dominate would be one that discounts both more extreme upsides and more extreme downsides (especially symmetrically) relative to a comparison option (more).
Hi Michael. Thanks for sharing your thoughts.
Interventions which cost-effectively increase the welfare of vertebrates will change land use much more than inaction, and a greater change in land use increases the probability of causing lots of suffering? Are you assuming that i) such interventions would increase agricultural land, and that ii) this decreases suffering?
On i), such interventions may decrease agricultural land due to increasing the price of animal products, and therefore decreasing their consumption? I estimate that replacing Ross 308 (fast growth broiler breed) with Rebro (slower growth) decreases cropland by 0.102 m²-year/Ross-308-chicken-kg, and that replacing replacing eggs from battery cages with those from barns or aviaries decreases cropland by 0.529 m²-year/battery-cages-egg-kg. These estimates neglect increases in the consumption of other foods, but I believe accounting for this would increase uncertainty, and therefore further increase the risk of the worst outcomes.
On ii), increasing agricultural land may increase suffering by increasing the number of soil macroarthropods/nematodes? I think effects on these may dominate (given the large uncertainty about welfare comparisons across species), and they may have negative lives (although I can easily see them having positive lives too).
It might cause lots of suffering, but it could also prevent lots of suffering, too. I think you’re thinking in terms of difference-making, but “Avoiding the worst” risk aversion is not difference-making. Rather than thinking about what you cause, you should just look at both (distributions of) outcomes and ask which has more suffering in it, without privileging the results of inaction.
Unless you believe the expected amount of wild animal suffering is higher all-things-considered than with inaction, you shouldn’t really expect it to do worse according to “Avoiding the worst” risk aversion (as a heuristic; there could be exceptions).
I agree. So the worst case is that the campaigns cause lots of suffering (relative to inaction)?
I understand I should look into the distributions of global welfare with and without the campaigns, and then assess their negative tails. I have little idea about which distribution has the highest expected value. However, I believe the distribution with the campaigns has longer positive and negative tails, and therefore the risk of the worst outcomes is higher with the campaigns (although the probability of the best outcomes is also higher).
Stepping back, regardless of how accounting for risk aversion changes recommendations, I would like greater reasoning transparency about why effects on soil invertebrates have been neglected. Roughly for the reasons @Marcus_A_Davis mentioned replying to Nick’s concerns about conflicts of interest.
In particular, I would like greater transparency about why effects on soil macroarthropods were neglected. They are covered in Bob’s book, unlike soil nematodes and microarthropods. Moreover, I estimate cage-free campaigns for laying hens change the welfare of soil ants and termites much more than they increase the welfare of chickens for the sentience-adjusted welfare ranges presented in Bob’s book.
I don’t think this is right. I think you’re still treating “Avoiding the worst” like a difference-making view. You shouldn’t be thinking in terms of “relative to inaction”, which itself has a highly uncertain distribution of outcomes. Just evaluate the distribution of outcomes for each option, without fixing any as a comparison option.
The question is only whether worst-case outcomes are more or less likely with action or inaction.
FWIW, the actual worst cases are s-risks, and I’d expect “Avoiding the worst” views to prioritize their mitigation, as long as we’re not clueless about that.
I had understood this. As I said, “I understand I should look into the distributions of global welfare with and without the campaigns, and then assess their negative tails”. My phrasing “cause lots of suffering (relative to inaction)” was confusing. However, I meant “increase the probability of outcomes with lots of suffering (the worst) relative to the probability under inaction”.
Ah, sorry, I was too quick and should have read more carefully.
Why do you believe this?
As chicken and egg prices increase from these welfare reforms, I would expect:
some shifts between crops and nature (including through substitution), but I’m clueless about which involves more suffering, so this doesn’t clearly favour one or the other.
substitution towards beef and other pasture products, which reduces invertebrate populations substantially, and probably without making lives much worse. This would mean less suffering and so do better in the worst case.
Here is how I am thinking. Imagine the world has welfare 0.
With inaction, the final welfare will be 0 with probability 100 %.
With campaigns, there is lots of uncertainty, but here is a simplified set of outcomes:
Agricultural land will increase with probability 75 %, and decrease with probability 25 %.
If agricultural land increases, the final welfare will be −1 with probability 25 %, and 1 with probability 75 %. So the final welfare will be −1 with probability of 18.75 % (= 0.75*0.25), and 1 with probability 56.25 % (= 0.75*0.75) considering outcomes where agricultural land increases.
If agricultural land decreases, the final welfare will be −1 with probability 75 %, and 1 with probability 25 %. So the final welfare will be −1 with probability 18.75 % (= 0.25*0.75), and 1 with probability 6.25 % (= 0.25*0.25) considering outcomes where agricultural land decreases.
As a result, final welfare will be −1 with probability 37.5 % (= 0.1875*2), 1 with probability 62.5 % (= 0.5625 + 0.0625), and 0.25 (= 0.375*(-1) + 0.625*1) in expectation.
The worst possible outcome across the 2 interventions is a final welfare of −1. With inaction, it has a probability of 0. With campaigns, it has a probability of 37.5 %. So the campaigns make the worst possible outcome more likely.
The intervention which can decrease welfare the most is the one leading to the lowest possible final welfare.
This is exactly a procedure you could follow for difference-making risk aversion; it’s equivalent to taking the statewise difference with inaction. The welfare of the world with inaction isn’t 0 with probability 100%.
RP has a model/procedure for avoiding the worst risk aversion here.
Why? I was assuming a world with initial welfare 0 with probability 100 %. However, I think my point stands for any distribution describing the initial welfare.
I have read the section Avoiding the Worst Risk Aversion: A Model, and I do not understand why you think it undermines my point.
If you instead set the campaign option to 0 welfare and defined the welfare of the world with inaction relative to the campaign option, you’d end up with the opposite conclusion, that only inaction reaches −1.
Avoiding the worst is meant to treat each option symmetrically. It doesn’t depend (in theory) on which option you single out to define things relative to.
(RP’s practical procedure does start with inaction, but if you end up with the same probability distributions for each option in the end, the results will be the same as if you started with a different option to define all distributions relative to. I think their procedure helps ensure consistent probability assignments and is less work, compared to directly estimating each distribution independently.)
What exactly do you mean by this? The campaign has many potential effects. So it cannot result in a final welfare of X with probability 100 %, where X can be 0 or any other number.
Suppose the initial welfare is 0 with probability 100 %. Inaction would lead to a final welfare of 0 with probability 100 %. Imagine an intervention which decreases welfare by 1 with probability 50 %, and increases welfare by 1 with probability 50 %. The intervention leads to a final welfare of 0 in expectation. However, it leads to a final welfare of −1 with probability 50 %, and 1 with probability 50 %. The the lowest possible welfare of −1 is more likely with the intervention?
It was illustrative.
Inaction also does not in fact lead to welfare of 0 with probability 100%. There will be lots of animals suffering and many possible outcomes if we do nothing. So it’s not correct to assume total welfare of 0.
I think my point stands for any distribution describing the initial welfare. Imagine the minimum initial welfare W_min has probability p. With inaction, the minimum final welfare would still be W_min, and have probability p. With an intervention which decreases welfare by 1 with probability 50 %, and increases welfare by 1 with probability 50 %, the minimum final welfare would be W_min − 1 with probability 0.5*p. So the lowest possible welfare of W_min − 1 would be lower and more likely with the intervention?
No, I don’t think this is the right way to model this. This looks a lot like the typical error people make for the original two envelopes problem.
Initial welfare (what does that mean?) and final welfare after inaction can differ, because the world, e.g. land use, will change even if you do nothing, and campaigns take time for their effects to materialize.
If you swapped the roles of campaign and inaction, you would flip the conclusion, too.
The moral two envelopes problem is not problematic if there is a common scale to compare the welfare per unit time (as there is to compare temperature)?
Suppose that inaction leads to a distribution for the future welfare (integral of the welfare per unit time across all future time) whose minimum value W_min has probability p. With an intervention that decreases future welfare by 1 with probability 50 %, and increases it by 1 with probability 50 %, the minimum future welfare would be W_min − 1 with probability 0.5*p. So I think the lowest possible future welfare of W_min − 1 would be lower and more likely with the intervention (although the intervention would not change future welfare in expectation).
What do you mean by this? By definition, inaction does not change the distribution of the future welfare?
In your model and your answers here, just replace inaction with campaign and campaign with inaction.
I see. Thanks for the patience. I could equally say that an intervention leads to a distribution for the future welfare whose minimum value W_min has probability p, and that inaction decreases it by 1 with probability 50 %, and increases it by 1 with probability 50 %, thus implying a minimum future welfare of W_min − 1 with probability 0.5*p. This is the exact opposite of what I concluded above, and suggests the lowest possible future welfare of W_min − 1 would be lower and more likely with inaction.
I agree both models are wrong. I cannot assume that the change in future welfare caused by the intervention is independent from the future welfare under inaction (as I did in my past comments), or that the change in future welfare caused by inaction is independent from the future welfare caused by the intervention (as I did just above).
I agree that increasing welfare in expectation is a good heuristic for better performance under “avoiding the worst” risk aversion. I have very little idea about whether cage-free campaigns for laying hens increase or decrease welfare in expectation. So I do not know whether they are favoured or not under “avoiding the worst” risk aversion. They are still disfavoured under difference-making and ambiguity risk aversion, and this could make them worse than inaction. In addition, they may be worse than inaction under no risk aversion of any type.