Hi Evan. I agree forgotten pain is still bad.

Nitpick. Dead people do not have memory, and therefore cannot remember or forget anything. So I would not say people forget their past pains once they die.

Hello ABlank. You are right. ACE has been using Ambitious Impact's (AIM's) pain intensities since 2024, which was when AIM started estimating benefits to animals in suffering-adjusted days (SADs) averted.

Nice. Note the estimates I linked in my comment above refer to estimates for the whole spending of programs, not to their marginal cost-effectiveness.

Hello Nick. Good point. I think the most agnostic assumption is supposing that the ratios between the intensities of 2 pain categories are the same across species. For example, that "intensity of excruciating pain in humans"/"intensity of hurtful pain in humans" is equal to "intensity of excruciating pain in chickens"/"intensity of hurtful pain in chickens". I believe we should expect humans to have a wider range of pain intensities. For example, that "range of pain intensities in humans" = "intensity of excruciating pain in humans" - "intensity of annoying pain in humans" is much larger than "range of pain intensities in chickens" = "intensity of excruciating pain in chickens" - "intensity of annoying pain in chickens". However, the measure ACE uses to measure benefits to animals, scaled suffering-adjusted days (SADs), is proportional to the sentience-adjusted welfare range of the target animals. So the range of pain intensities in the target animals is also proportional to their sentience-adjusted welfare range. For example, "range of pain intensities in chickens" = "sentience-adjusted welfare range of chickens"*"range of pain intensities in humans". This means the sentience-adjusted welfare ranges already contribute towards animals having a narrower range of pain intensities. I feel like making them even narrower by supposing that the ratio "intensity of excruciating pain"/"intensity of annoying pain" is lower for animals would be some sort of double counting.

Hello Stien. Thanks for the comment.

From what I understand, small changes to modeling assumptions can swing estimates by an order of magnitude

I agree with this in general. However, the CEA seems to suggest there is not room for the best guess for GBD's cost-effectiveness to be much higher. It is mentioned there that, "Depending on our assumptions of this [most impactful] engagement, the result [best guess for the cost-effectiveness?] ranges from 0.12 to 2.48 SADs per dollar", 5.61 % to 116 % of the best guess you presented of 2.14 SADs per $. This range only accounts for uncertainty in the impact from the most impactful engagement, but I believe you would get a similar range accounting for uncertainty in the impact from all engagements. This is because the most impactful engagement accounts for the vast majority of GBD's impact. In particular, at least 94.4 % (= 1 - 0.0561) of GBD's impact because the cost-effectiveness could be as low as 5.61 % for the most pessimistic assumptions about the most impactful engagement, and I assume this engagement has a positive impact under such assumptions. I suggest you mention in the CEA which fraction of GBD's impact comes from the most impactful engagement (which I understand is at least 94.4 %).

The team found GBD has a very strong MEL process that aims to measure actual substitution of animal products for plant-based ones at the point of consumption. Although there is still significant uncertainty in their cost-effectiveness estimate, GBD bases some of their key assumptions and strategies on real-world consumption data rather than theoretical projections.

This does not seem to apply to the most impactful engagement, which I think is responsible for at least 94.4 % of GBD's impact. It is mentioned in the CEA that "The largest institution included in this program [which I assume is linked to the most impactful engagement] does not collect procurement data, which were instead assessed by proxy based on data from other institutions".

They also have a track record of clear achievements, with no expectation of diminishing returns in the near future. 

I understand the best guess for the cost-effectiveness of GBD would become 5.61 % as high under the most pessimistic assumptions for the most impactful engagement, which may be difficult to rule out considering the lack of procurement data for it. So GBD's cost-effectiveness relies a lot on a single very impactful engagement instead of a track record of many impactful engagements.

What I personally found quite compelling was that we expect that counterfactually GBD will reduce SARP because many of their clients were planning to have more climate-friendly menus, and GBD steers menu changes into a plant-based rather than small-animal-based direction. GBD also monitors for any inadvertent increases in procurement of animal products wherever possible, investigates such increases and then suggests plant-based alternative strategies. 

I think this makes GBD more likely to increase the welfare of farmed animals than diet change interventions which are less focussed on mitigating the small animal replacement problem (SARP). At the same time, GBD could robustly increase the welfare of farmed animals in a way that is not sufficiently cost-effective.

Thanks for the reply.

It makes sense the differences in cost-effectiveness become smaller if you use similar values for inputs besides the population of layers affected. I suggest you link to this sensitivity analysis. However, I understand the best guesses for the inputs in your CEAs are your actual best guesses. Do you have specific reasons to believe the inputs are actually much more similar?

I understand estimates for the number of layers helped per $ are uncertain, but the estimates for Sinergia Animal are much lower than for ÇHKD, AWO, and THL. In particular, there is very little overlap between your ranges for Sinergia Animal and THL, and your estimates for these charities both refer to accountability work targeting layers starting in 2024. I have added a chart to my comment above with your estimates for the layers helped per $ by work targeting layers done by charities you recommend.

Nitpick. People from ACE replying to comments using their personal accounts instead of ACE's would increase transparency, which I think is generally good.

Animal Charity Evaluators (ACE) estimated The Humane League's (THL's) cage-free work in 2025 helped 11 (2 to 44) chickens per $.

ACE just published more estimates of the cost-effectiveness of work targeting layers.

Thanks for continuing to publish cost-effectiveness analyses (CEAs) of the charities you recommend (I linked them here). I am not aware of any other funder of animal welfare interventions doing this. I have made many comments on your posts announcing the charities you recommended in 2024, 2025, and 2026 suggesting improvements. However, all else equal, I trust funders publishing their analyses more. Funders not publishing their analysis are less accountable.

Could you clarify why you recommended Greener by Default (GBD)?

Your best guess and upper bound estimates for the cost-effective of GBD are significantly lower than the same estimates for Sinergia Animal (SA) and Çiftlik Hayvanlarını Koruma Derneği (ÇHKD). Your best guess for the cost-effectiveness of GBD's work in 2025-mid-2026 is 37.3 % (= 2.14/5.73) of your best guess for the cost-effectivess of SA’s work targeting layers in 2024-mid-2026, and 6.97 % (= 2.14/30.7) of your best guess for the cost-effectiveness of ÇHKD's work targeting layers in 2024-2025. Your upper bound for the cost-effectiveness of GBD's work in 2025-mid-2026 is 12.3 % (= 13.0/106) of your upper bound for the cost-effectivess of SA’s work targeting layers in 2024-mid-2026, and 6.63 % (= 13.0/196) of your upper bound for the cost-effectiveness of ÇHKD's work targeting layers in 2024-2025.

I understand your cost-effectiveness analysis (CEA) of GBD only played a limited role in your recommendation. You note the following limitations.

Limitations: Our estimates have limited explanatory power and should be interpreted with caution because:
- The range of results is wide and the real value lies somewhere between the lower and upper bound
- The largest institution included in this program does not collect procurement data, which were instead assessed by proxy based on data from other institutions

As a result, we gave only limited weight to this cost-effectiveness analysis in our overall assessment of Greener by Default.

However, it looks like you think your best guess for the cost-effectiveness of GBD's work in 2025-mid-2026 is too high, and has little room to be higher. You say the following in the CEA.

Robustness (sensitivity): We tested how sensitive these estimates are to underlying assumptions. The largest engagement carries around 90% of modeled impact and the exact meal composition is unknown, so the estimate applies ACE's standard U.S. [United States] meal replacement estimate rather than anything observed at Sodexo. Depending on our assumptions of this engagement, the result [best guess for the cost-effectiveness?] ranges from 0.12 to 2.48 SADs per dollar [5.61 % to 116 % of the best guess you presented of 2.14; so it looks like you think your best guess has little room to be higher]. Shrimp contributes roughly 53% of ACE's standard estimate [for the impact of random diets in the US on farmed animals], however shrimp is not plausibly half the animal impact of a hospital patient entrée. Excluding shrimps from the estimate results in 1.19 SADs/$ [55.6 % of the best guess you presented; so it looks like you think your best guess is too high]. Our overall view of Greener by Default's cost effectiveness would change if there were measured data for the largest engagement, leading to a more accurate estimate.

Hello Stien. Thanks for the context. I assume you are sufficiently concerned about ALI's governance to be confident they should not be recommended (for now), and that therefore there is no need to complete their evaluation. However, have you considered publishing an incomplete evaluation with some information about your concerns about ALI's governance?