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 Grilođ¸
Karma: 10,994
Alfredo commented on the post, and I replied.
Hi Alfredo. Thanks for sharing your thoughts.
A human body can be described by the state of all its fundamental particles.
I am not sure the state of a human body is fully defined by the state of its fundamental particles, but I used this to mean all of its physical properties.
The key issue is that continuous variation in underlying physical parameters does not entail continuous variation in the properties that emerge from them.
For my argument to work, it is enough for pain intensity to be quantitatively comparable for infinitesimally different physical parameters. The pain intensity does not have to be a continuous function of the physical parameters. I actually suspect physical parameters can only vary discretely, in agreement with quantum mechanics. So I believe both physical parameters and pain intensity vary in infinitesimally small jumps.
When you heat a ferromagnet continuously, nothing dramatic happens for a while. Then, at a specific critical temperature (the Curie temperature), ferromagnetism doesnât just decrease, but it vanishes entirely. The property âbeing able to attract ironâ is not something that fades smoothly to zero. It disappears at a sharp threshold, even though the underlying control parameter (temperature) was varied continuously.
This is not true. I wonder whether Claude hallucinated it. Spontaneous magnetisation (M) just below the Curie temperature is proportional to (âCurie temperatureâââtemperatureâ)^âexponent (0.34 for iron)â. So spontaneous magnetisation smoothly goes to 0 as the temperature approaches the Curie temperature. In any case, spontaneous magnetisation would be comparable across any 2 temperatures even if it abrupty dropped to 0 near the Curie temperature. The spontaneous magnetisation for a temperature at least as high as the Curie temperature is 0 times that for a temperature below it.
For example, steam can do mechanical work via expansion (it drives turbines); liquid water cannot.
The mechanism via which steam does mechanical work is fundamentally the same as that through which liquid water does mechanical work in hydropower. In both cases, molecules of water collide with turbines, and make them spin. For the case of steam, water vapour molecules are accelerated by temperature. For the case of hydropower, liquid water molecules are accelerate by gravity. In any case, the mechanical work done by steam and liquid water is quantitatively comparable.
The density of water also varies smoothly as it boils. It increases linearly with the vapour quality, which is the mass of water vapour as a fraction of all the mass. For a vapour quality of 0, there is saturated liquid water, and the density matches that of liquid water at the boiling point. For a vapour quality of 1, there is saturated water vapour, and the density matches that of water vapour at the boiling point.
In the example above, an infinitesimal change in a physical property (temperature) leads to an abrupt change in another physical property (density), but the underlying physical state does not change infinitesimally (at the boiling point, an infinitesimal change in vapour quality results in an infinitesimal change in density). In any case, a physical property varying abrupty for an infinitesimal change in the underlying physical state would not undermine my argument. The physical property in one state would have to be quantitatively incomparable with that on another state. The density of water at different states is quantitatively comparable.
Liquid water has surface tension (it forms droplets, menisci, capillary action); steam doesnât.
Surface tension is technically not a property of liquid water or water vapour. It is a property of liquidâair interfaces, and it varies smoothly with temperature. The surface tension for the interface between liquid water and water vapour is different than that between liquid water and other gases.
Topological transformations
Are you practically arguing that some pain intensities are not quantitatively comparable, even if their underlying physical states only differ infinitesimally, because there are mathematical functions which are not continuous? I do not understand why the space of possible mathematical functions would provide any meaningful empirical evidence about pain intensities.
I think the burden of proof is on showing that pain intensity actually behaves like temperature
My core point is that infinitesimal changes in physical reality do not make pain intensities quantitatively incomrable. Physical reality is infinitesimally changing all the time, and personally experienced pain intensities seem very much quantitatively comparable. So I would say the burden of proof is on showing this is not the case.
I hope this helps clarify where Iâm at!
Likewise. Thanks for the opportunity to clarify my position.
Thanks for the clarifying comment, Jim. I agree with all your points. For individual (expected hedonistic) welfare per fully-healthy-animal-year proportional to âindividual number of neuronsâ^âexponent 1âł, and âexponent 1â from 0.5 to 1.5, which I believe covers reasonable best guesses, I estimate that the absolute value of the total welfare of:
Farmed shrimps ranges from 2.82*10^-7 to 0.282 times that of humans.
Soil nematodes ranges from 0.00252 to 902 k times that of humans.
Moreover, the above ranges underestimate uncertainty due to considering a single type of model for the individual welfare per fully-healthy-animal-year. At the same time, the results for individual welfare per fully-healthy-animal-year proportional to âindividual number of neuronsâ^âexponent 1â can be used to get results for individual welfare per fully-healthy-animal-year proportional to âproxyâ^âexponent 2â if âproxyâ is proportional to âindividual number of neuronsâ^âexponent 3â. All of âexponent 1â, âexponent 2â, and âexponent 3â can vary. I am using different numbers because they are not supposed to be the same.
And you said things that suggests you agree in this recent interview. You seemed to have deviated from your previous ânematodes obviously dominateâ view. Or did I miss something?
I think my previous view was more âit is very difficult for effects on soil animals or ones with a similar number of neurons not to be the major driver of the overall effect in expectationâ. I would say this was my view in this post. The bullet of the summary starting with the following summarises it well. âI believe effects on soil animals are much larger than those on target beneficiariesâ. In any case, I was certainly overconfident about the dominance of effects on soil animals or ones with a similar number of neurons.
1. In all the cases youâre describing there was a shift from battery cages to enriched cages. So that is two eras of cages potentially in one overall housing unit. 2. I think there is also additional years in advance of a ban to be considered, markets, particularly now on cage-free in Europe have shifted significantly in advance of the bans announced and I think the same thing is happening in countries we will soon see bans from.
Makes sense.
Hi. You tested prior distributions for the probability of consciousness with means ranging from 10 % (= 1/â(1 + 9)) for âlowâ to 90 % (= 9/â(1 + 9)) for âhighâ. I feel like this underestimates the uncertainty considering âThe choice of a prior is often somewhat arbitraryâ. Have you considered testing prior distributions with means ranging, for example, from 10^-6 to 99.999 %? Relatedly, how much would the uncertainty increase if you had considered a wider range of possible updates via allowing for stronger levels of support and demandingness, whose corresponding likelohood ratios range from 2 % (= 1â50) to 50 (from Table 2 of the report)? I worry the range of results one gets is overwhelmingly determined by the range of priors and updates. I suspect the final probabilities of consciousness with span many orders of magnitude for priors with means ranging from 10^-6 to 99.999 %, and possible updates with likelihood ratios randing from 10^-6 to 10^6. You focus on updates in the probability of consciousness in light of the evidence, not on the values of the probability of consciousness. However, it might be better to neglect these even more.
Hi Alfredo. I would be curious to know your thoughts on my post All pains are comparable?. You are welcome to comment there.
Summary
I agree there are pains which feel qualitatively different in the sense of having distinct properties. For example, annoying and excruciating pain as defined by the Welfare Footprint Institute (WFI).
Some think there are pains whose intensity is incomparably/âqualitatively worse than others. For instance, some believe averting an arbitrarily short time of excruciating pain in humans is better than averting an arbitrarily long time of annoying pain in humans. In contrast, I would prefer warming up slightly cold patches of soil for sufficiently many nematodes over averting 1 trillion human-years of extreme torture.
Consider a human body as described by the state of all of its fundamental particles. Are there any 2 states which are only infinitesimally different whose pain intensities are not quantitatively comparable? I do not see how this could be possible. So I conclude the pain intensities for any 2 states of a human body are quantitatively comparable.
Hi Wladimir.
Point 2: The Risks of Aggregating Intensities and Durations
I would be curious to know your thoughts on my post All pains are comparable?. You are welcome to comment there.
Summary
I agree there are pains which feel qualitatively different in the sense of having distinct properties. For example, annoying and excruciating pain as defined by the Welfare Footprint Institute (WFI).
Some think there are pains whose intensity is incomparably/âqualitatively worse than others. For instance, some believe averting an arbitrarily short time of excruciating pain in humans is better than averting an arbitrarily long time of annoying pain in humans. In contrast, I would prefer warming up slightly cold patches of soil for sufficiently many nematodes over averting 1 trillion human-years of extreme torture.
Consider a human body as described by the state of all of its fundamental particles. Are there any 2 states which are only infinitesimally different whose pain intensities are not quantitatively comparable? I do not see how this could be possible. So I conclude the pain intensities for any 2 states of a human body are quantitatively comparable.
Hi Alex. Thanks for the summary and additional points.
Furnished cages typical lifespans are considered to be 15-25 years.
I agree with this range. This assessment of the economic impact of phasing out furnished cages in Flanders analyses depreciation periods of 20 and 25 years in Figure 7, whose translation to English is below.
This report exploring the consequences of banning enriched cages in the Netherlands (here is an English translation) discusses âa depreciation period of 15 yearsâ.
Beyond just the systems, the houses they are built in have lifespans of around 50 years.
I assume this is not a determinant factor. Otherwise, I would have expected economic assessments to consider a longer depreciation period than 15 to 25 years, and transition periods longer than this too. Across 10 bans in Europe, I got a time between the annoucement of the ban until it starts applying to all systems ranging from 4 to 28 years, with the mean being 12.2 years.
@JamesĂz đ¸ points about advocacy cost and public support resonate with me. My understanding is where we have seen furnished/âlarger cages pop up, like in Europe, Canada, Australia, South Korea briefly in the US. This hasnât been an advocate push. The push has been for cage-free and the industry has lobbied to have that lowered to furnished/âlarger cages.
I wonder whether some sympathy from animal advocates towards furnished cages relative to conventional cages was needed to get furnished cages. Without that sympathy, in cases where cage-free was not really on the table, the outcome could have been conventional instead furnished cages? If so, animal advocates could still advocate for cage-free, but make it clear that furnished cages are better than conventional cages.
@Mia Fernyhough point about any cage putting a cap on welfare improvement resonates. I see
Did you mean to add something else? The sentence above does not end with a dot.
@Joren Vuylsteke shared a chart of the change and I think by starting from 2009, it misses some of the shift to cage-free which was already happening in advance of that. I can share a chart in DM if anyone would like. ~50M was cage-free in 2003, ~75M in 2006. Countries like Germany leading this.
Eyeballing Figure 3 of this report from Compassion in World Farming (CIWF), the one below which Joren was referring to, there were around 80 M cage-free hens at the start of 2009 in the countries in the EU today, 60 M in barns, and 20 M in free-range. In addition, it looks like there were 220 M in cages, which implies a total of 300 M (= (220 + 80)*10^6), of which 26.7 % (= 80*10^6/â(300*10^6)) were cage-free. The EU banned conventional cages 3 years later. It looks like around 10 % of laying hens in China are cage-free. So I wonder whether there are tractable ways of pushing for a ban on conventional cages there, or more states in India.
Overall, one of the strongest things points from my perspective is being aware that there are definitely waves of change in the systems that we can benefit on focussing on.
Agreed.
All pains are comÂpaÂrable?
Hi Michael. That is fair. On the other hand, what readers ignore or not depends on how the results are communicated. It would be harder to ignore uncertainty about AI timelines if AI2027 was e.g. AI2027-2047, and this would still undercommunicate uncertainty. The difference between the 90th and 10th percentile dates of artificial superintelligence (ASI), as defined below, is more than 100 years for Daniel Kokotajlo and Eli Lifland, the 2 main forecasters of the AI Futures Model (which superseded AI2027).
Thanks for the post, Zoe.
Thanks for the very relevant sources you have been sharing too. I strongly upvoted your initial comment because I have found this thread valuable.
The report you linked exploring the consequences of banning enriched cages in the Netherlands (here is an English translation) says conventional cages had fully depreciated in 2012.
Generally speaking, the majority of cage rearing systems were built between 1995 and 1998. 1999. With a depreciation period of 15 years, these systems had an average book value of zero in 2012.
2012 is when the ban on conventional cages in the EU started. So the above supports your take that cages will only be banned when they are near the end of their lifetime. However, I do not think this means a ban on cages in the EU will start, for example, in either 2032 or 2047 (= 2032 + 15). I think it just means the ban will have to be announced 15 years before it enters into force such that the economic loss is minimised. This is in agreement with the report above.
The total financial loss from the inventory of enriched cages, cages to be enriched, and rearing cages is âŹ11.8 million. The loss calculation is based on a ban effective in 2012. If the period of use is shorter or longer, the financial loss will also be proportionally higher or lower. If the ban were to take effect in 2017, the financial loss would be âŹ2.1 million. If the end date were to be postponed to 2020, the financial loss would be âŹ0.7 million. In 2022 [15 years after 2007, when the report was published], the financial loss will be zero because the inventory, after a 15-year depreciation period, will have a residual value of zero.
As a result, if the EU announces a ban on furnished cages in 2026, I guess it will only start applying to all cages (instead of just new cages) in 2041 (= 2026 + 15) or so. Here is an estimate of the economic loss from shortening the transition period. From Table 1.1 of van Horne and Bondt (2023), the housing cost for furnished cages is 3.39 2021-âŹ/âhen, 4.84 $/âhen (= 3.39*1.22*1.17). For hens with a lifespan of 70 weeks (WFI assumes â60 to 80 weeks for all systemsâ), 1.34 hen-years (= 70*7/â365.25), the housing cost of furnished cages is 3.61 $/âhen-year (= 4.84/â1.34). I estimate there were 149 M hens in furnished cages in the EU in 2024. So I think renewing all furnished cages in the EU would cost 538 M$ (= 3.61*149*10^6), 1.20 $/âcitizen (= 538*10^6/â(450*10^6)). I speculate 50 % of the value can be recovered via exporting the cages to countries outside the EU. Consequently, for cages fully depreciating in 15 years, the cost of shortening the transition period by 1 year would be 17.9 M$ (= 538*10^6*(1 â 0.50)/â15), 0.0398 $/âcitizen (= 17.9*10^6/â(450*10^6)).
As a side note, the calculations for the Netherlands did not account for the possibility of exporting the cages.
Because the systems are permitted in other EU countries, it is possible to sell enriched cages, and to a lesser extent, enriched cages, on the international market. Any potential proceeds from such a sale have not been taken into account in these calculations.
Hi Angelina.
It is also really interesting and encouraging to hear that you think welfare in some cage-free systems is continuing to improve over time.
Relatedly, Schuck-Paim et al. (2021) âconducted a large meta-analysis of laying hen mortality in conventional cages, furnished cages and cage-free aviaries using data from 6040 commercial flocks and 176 million hens from 16 countriesâ. Here is how they describe their findings in the abstract.
We show that except for conventional cages, mortality gradually drops as experience with each system builds up: since 2000, each year of experience with cage-free aviaries was associated with a 0.35â0.65% average drop in cumulative mortality, with no differences in mortality between caged and cage-free systems in more recent years. As management knowledge evolves and genetics are optimized, new producers transitioning to cage-free housing may experience even faster rates of decline. Our results speak against the notion that mortality is inherently higher in cage-free production and illustrate the importance of considering the degree of maturity of production systems in any investigations of farm animal health, behaviour and welfare.
Thanks for the great post, Ajeya. I agree.
Thanks for sharing, Max. Have you reached out to GiveWell about this? You may want to collaborate on further work.
Thanks for looking into this, Mia.
This relies on the premise that welfare is a linear scale where âless sufferingâ equals âadequate welfare.â
I do not rely on the concept of âadequate welfareâ in my analysis. I estimate welfare from âtime with positive experiencesâ*âintensity of positive experiencesâ - (âtime in annoying painâ*âintensity of annoying painâ + âtime in hurtful painâ*âintensity of hurtful painâ + âtime in disabling painâ*âintensity of disabling painâ + âtime in excruciating painâ*âintensity of excruciating painâ. My assumptions for the pain intensities imply each of the following individually neutralise 1 fully-healthy-chicken-day:
10 days of annoying pain, which I assume is 10 % as intense as hurtful pain.
1 day of hurtful pain, which I assume is as intense as fully healthy life.
2.40 h of disabling pain, which I assume is 10 times as intense as hurtful pain.
0.864 s of excruciating pain, which I assume is 100 k times as intense as hurtful pain.
It [a furnished cage] provides a slightly less bad life, but it does not provide a life worth living.
âI estimate that hens in conventional (battery) and furnished (enriched) cages, and cage-free aviaries (barns) have a welfare of â1.79, â1.09, and â0.798 chicken-QALY/âchicken-yearâ. Values below 0 imply more suffering than happiness, and, in this sense, lives not worth living. At the same time, I estimate the welfare per chicken-year increases by 39.1 % (= (-1.09 - (-1.79))/â1.79) when chickens go from conventional to furnished cages.
It [a furnished cage] offers almost no opportunity for positive experiences or pleasure (let alone basic needs), which are critical components of any welfare assessment. Iâm interested to understand how you accounted for positive experiences?
I speculated chickens have positive experiences when they are awake, and not experiencing hurtful, disabling, or excruciating pain. In addition, I guessed the positive experiences to be as intense as hurtful pain. WFI will publish a book this year with estimates for the duration of positive experiences for 4 levels of intensity. I am looking forward to these, and may use them to produce updated estimates for the welfare of layers.
You mention that furnished cages in the EU require specific resources, such as âat least 250 cm² of littered area per henâ. This is incorrect. That specific requirement is for non-cage systems.
Great catch. I copy-pasted from the wrong place. I have corrected that sentence of the post to the following.
Each laying hen in a furnished cage in the EU must have âa nestâ, âlitter such that pecking and scratching are possibleâ, and âappropriate perches of at least 15 cmâ.
Furnished cages must have âlitter such that pecking and scratching are possibleâ, but no minimum area is specified.
Advocating for furnished cages would amount to welfare washing. It allows the industry to claim they have âreformedâ the system by adding token resources that do not meaningfully improve the birdâs subjective experience.
Very interesting. Does that mean you very much disagree with WFIâs estimates implying that chickens experience significantly less pain in furnished than conventional cages (illustrated in the 2nd graph of my post)? They calculate there is 64.0 % (= (431 â 155)/â431) less disabling pain per hen in furnished cages than in conventional cages. Maybe you think WFIâs estimates only hold water under idealised conditions which are rarely present in practice? @cynthiaschuck, do you have any thoughts on how having more realistic generalisable studies would change the comparison between conventional and furnished cages?
Producers operate on long investment cycles. If we convince a producer in a developing market to invest millions in furnished cages today, we are not creating a stepping stone; we are cementing a ceiling for the next 20 years. Once that capital is sunk, the economic incentive to upgrade again to cage-free vanishes.
I agree. However, advocating for furnished cages could still make sense in regions which are only expected to become cage-free in more than 20 years, like some countries in Africa and Asia?
Figure 3 of the report you linked from Compassion in World Farming (CIWF) does show the transition from conventional to furnished cages happened just in 3 years in the EU, from 2009 to 2011. Very interesting. I did not know it happened so fast.
However, I do not think this implies keeping hens in cages in the EU will be banned from either 2032 or 2052 on. If this was the case, I would have expected bans in the EU to all start in 2032, whereas there are many timelines. 1992 for Switzerland (this one is not relevant for our discussion because it happened before 2012), 2020 for Austria, 2021 or earlier for Luxembourg, 2023 for Iceland, 2027 for the Czech Republic, 2028 for Wallonia (a region of Belgium), 2029 for Germany (2026 for non-exceptional cases) and Slovenia, 2030 for Slovakia, 2035 for Denmark, and 2036 for Flanders (a region of Belgium). Colony cages are still allowed in the Netherlands, and I am not aware of a ban on all cages having been announced there. Did I miss it?
In any case, do you think EUâs global influence is sufficiently strong to determine what to advocate for in Africa and Asia? I agree it would not make sense to advocate for furnished cages, for example, in China if this could undermine a ban on cages there from 2032 on (because the new furnished would still be very early in their lifetime then). Yet, I do not see situations like this coming to pass. I expect Africa and Asia to become cage-free at least 20 years after the EU does. So I believe there is time for new furnished cages there to operate for their full lifetime of 20 years.
Hi Max. I very much agree WFIâs estimates alone will not be persuasive to the public. However, I wonder whether it would be possible to communicate the importance of nests, perches, and litter. The EU banned conventional cages. So at least some decision-makers and citizens in the EU had to prefer furnished cages over conventional cages. Maybe these were people that engaged more with the topic, and such level of engagement cannot be reached as a result of public campaigns targeting companies. In this case, as I say in the post, âthere may still be room to advocate for political change in more authoritarian countries like China where companies are less subject to public pressureâ.
Thanks for looking into this, Joren.
You seem to assuming that furnished cages are overwhelmingly built every 20 years, and were last build just before 2012. If this was the case, the vast majority of furnished cages would be renewed just before 2032 without a ban to end furnished caged before then. So I would agree that a ban would as a result happen in 2032 or 2052. However, in reality, I expect furnished cages to be built or renewed gradually, not all at once across the EU in a few years. So I believe a ban on furnished cages does not have to start in very specific years further apart by 20 years like 2032 or 2052.
In any case, I do not think it would make sense to advocate for furnished cages in the EU. Only 38.2 % of layers were in cages in the EU in 2024. So there is already significant momentum for cage-free.
I believe advocating for furnished cases in Africa and Asia would be better (although I am not confident it would be a good idea). The timeline I suggested above of 15 years until full implementation of furnished cages would allow for 75 % (= 15â20) of conventional cages to operate for a lifetime of 20 years. The remaining 25 % could be compensated for the 5 years (= 20 â 15) of fixed costs that did not get amortised. Moreover, there would only be rigid timelines applying to a whole country for political work. For work targeting companies, there could be different timelines, and therefore less need for some producers to end the operations of farms before they operate for their whole lifetime. Companies transitioning to furnished cages earlier could source their eggs from producers which transitioned to furnished cages earlier due to having started with older conventional cages. In addition, they could source eggs produced in furnished cages in the EU.
Hi Anthony. Do you think the expected welfare of 2 states of the world which only differ infinitesimally (e.g. an electron is ) can be incomparable? I do not see how this could be possible. For example, it feels super counterintuitive to me that, given 2 identical states, moving an electron by 10^-100 m in one of the states would make their expected welfare incomparable. I guess one can get from any state of the universe to another in an astronomical number of infinitesimal steps, and I believe any 2 states which only differ infinitesimally are comparable. So I conclude any 2 states are comparable too, even if it is very hard to compare them, to the point that I do not know if electrically stunning shrimps increases welfare in expectation.