David_Moss

Karma: 10,533

I am the Principal Research Director at Rethink Priorities. Previously, I oversaw our Worldview Investigation Team, and now focus on our cross-cause prioritisation work, while still overseeing our Surveys and Data Analysis work.

The Worldview Investigation Team previously completed the Moral Weight Project and CURVE Sequence / Cross-Cause Model. We’re currently working on tools to help EAs decide how they should allocate resources within portfolios of different causes, and to how to use a moral parliament approach to allocate resources given metanormative uncertainty.

The Surveys and Data Analysis Team primarily works on private commissions for core EA movement and longtermist orgs, where we provide:

Private polling to assess public attitudes
Message testing / framing experiments, testing online ads
Expert surveys
Private data analyses and survey / analysis consultation
Impact assessments of orgs/programs

Formerly, I also managed our Wild Animal Welfare department and I’ve previously worked for Charity Science, and been a trustee at Charity Entrepreneurship and EA London.

My academic interests are in moral psychology and methodology at the intersection of psychology and philosophy.

David_Moss 8 Oct 2026 17:35 UTC
4 points
2 ∶ 0
on: British EAs- how to increase the effectiveness of your donations by 80%
This is great! But it would be even better if UK (non-founder) EAs had a cost-effective DAF option to donate to.
I think there are very strong reasons for many donors not to get all their money out the door within the tax year they receive it (principally: the ability to prioritise more, and wait for better opportunities).
As it stands, I’m not aware of any cost-effective DAF options for UK donors donating low to mid tens of thousands of £ per year. I’m currently forced to use an account with Charities Aid Foundation, which charges 4.5% fees and does not allow one to invest one’s holdings (thereby foregoing significant gains).
I reached out to Founders Pledge, who seem to offer excellent UK DAFs (both free and investment options), but they do not seem to allow non-founders to make use of their service).

David_Moss 29 Sep 2026 18:02 UTC
7 points
0 ∶ 0
on: Bryan Caplan vs Effective Altruism
Thanks for the post!

This [extinction from AI] is not some weird fear only EAs have. About 9% of the general population in the UK think AI might doom us all.
I’ll add that in another survey of ours we found that the median probability of AI leading to extinction was 15% (for US respondents). In a separate survey, we also found strong public support for a pause. Those surveys are from three years ago though, I would expect higher concern now.
Caring about bees, who more and more researchers think are sentient, and shrimps, who the UK government officially declared as sentient, is too outlandish for him
We also found that many of the UK public also think this. A majority think they are likely sentient, an even larger majority think they can feel pain.

David_Moss 22 May 2026 9:21 UTC
8 points
1 ∶ 0
in reply to: Bella’s comment on: EA sexual harassment: Where’s the evidence we’re worse than baseline?
I guess you might think “professional EA orgs” or something are more male-dominated than EA survey respondents. That seems possible, depending on what you wanna count. One thing is 80k is currently majority-women and I think however you carve it up we’re a big chunk of the space :)
Agreed. Fwiw, in EA Survey 2024 data, people currently working at EA orgs are actually significantly more likely to be women or non-binary than those not (36.9% vs 29.1%, p=0.004).

David_Moss 8 Apr 2026 10:21 UTC
4 points
0 ∶ 0
in reply to: James Herbert’s comment on: [Opinion] EA should do more to shape its public narrative
It has been considered for the EA Survey! I’m not sure why this was never prioritised for inclusion, after being raised. But if the meta orgs we work with say they want us to include these questions in the survey going forward, we will.

David_Moss 13 Mar 2026 18:54 UTC
7 points
0 ∶ 0
on: Can you help me with a survey issue?
Have you tried any platforms other than Positly? (I ask because you wrote “I haven’t been able to fulfill this quota through either Positly.” which made it sound like you had already tried someone else).
Prolific nominally have 570 active UK Muslims (though I would expect to recruit fewer than that cited number), and 299 in the US.
SurveyMonkey claim to be able to reach 500 US Muslims.
Feel free to DM me if further suggestions would be helpful. I would also approach surveying respondents from website like Prolific a little differently to how I would approach surveying from most panels (e.g. many more checks required on the latter, but potentially higher representativeness on the former).

David_Moss 12 Mar 2026 9:00 UTC
3 points
0 ∶ 0
on: Your outreach is written for people who already agree with you
Thanks for the useful post!

You may also be interested on our empirical studies on framing EA and longtermism.
We do, indeed, find that “longtermism” is distinctively unpopular as a framing. But messaging focused on global catastrophic risk reduction, or specific risks, such as AI safety in particular, performed quite well.
We’re actually working on a new study examining EA vs AI framing specifically (including differences in who gets recruited by such framings). We’d welcome input on specific framings / elements to test.

David_Moss 19 Feb 2026 16:53 UTC
2 points
0 ∶ 0
in reply to: Chris Leong’s comment on: casebash’s Shortform
How would you define and operationalize the vibe shift? Whether there’s been a vibe shift and, if so, how significant, seems empirically tractable to determine.

David_Moss 9 Feb 2026 11:32 UTC
4 points
0 ∶ 0
in reply to: Vasco Grilo🔸’s comment on: Announcing ACE’s 2025 Charity Recommendations
Thank you Vasco.
I agree this is an important and interesting topic. I will look into the prior studies to assess what I think would be needed to shed light on the above (e.g. required sample size, required measures) and get back to you.

David_Moss 14 Jan 2026 14:45 UTC
20 points
6 ∶ 0
in reply to: Daniela Yamamoto’s comment on: EA is underfunding animal advocacy according to our own preferences
Interestingly, community members’ prioritisation of animal welfare appears to have been increasing in recent years. (https://forum.effectivealtruism.org/posts/CKwDgZGLipchAoxtN/ea-survey-2024-cause-prioritization#Trends_in_cause_prioritization_over_time)

This is neither to agree or disagree with the observation about representation of animal welfare at EA conferences and such.

David_Moss 7 Jan 2026 14:54 UTC
4 points
0 ∶ 0
in reply to: Vasco Grilo🔸’s comment on: Why Neuron Counts Shouldn’t Be Used as Proxies for Moral Weight
I think this is interesting, but I don’t think we should infer too much from this relationship. This plot basically matches those we produced examining the relationship between cortical neuron count and perceived moral value of different animals (replicating SlateStarCodex and another’s surveys). As you can see below, we found extremely strong correlations. But we also found similarly strong correlations using EQ or total brain size rather than cortical neuron count, or using a crude 0-100 measure of how people ‘care’ about the animal in place of tradeoff ratios. Notably, when we replace the moral value measure with a simple ordinal ranking of the animals by neuron count (as in the second plot below), we find even stronger relationships.
My impression is therefore that the strong correlations more reflect the fact that we have a small number of datapoints with animals differing dramatically on a wide variety of predictor (or, in principle, outcome) variables which are all highly correlated, rather than indicating that neuron counts are distinctively predictive of any outcomes of interest. See Andrew Gelman’s similar discussion of our study.
I think to actually disentangle these we would need a larger sample of animals who diverge on the key dimensions (e.g., birds with high neuron density but small brains, or animals with higher neuron count but lower perceived similarity to humans).

David_Moss 10 Dec 2025 9:54 UTC
5 points
0 ∶ 0
in reply to: Yarrow Bouchard 🔸’s comment on: EA Survey 2024: Cause prioritization
Thanks Yarrow,
In this report, we don’t actually classify any causes as longtermist or otherwise. We discuss this in footnote 3.
In this survey, as well as asking respondents about individual causes, we asked them how they would allocate resources across “Longtermist (including global catastrophic risks)”, “Near-term (human-focused)” and “Near-term (animal focused)”. We also asked a separate question in the ‘ideas’ section about their agreement with the statement “The impact of our actions on the very long-term future is the most important consideration when it comes to doing good.”
This is in contrast to previous years, where we conducted Exploratory Factor Analysis / Exploratory Graph Analysis of the individual causes, and computed scores corresponding to the “longtermist” (Biosecurity, Nuclear risk, AI risk, X-risk other and Other longtermist) and “neartermist” (Mental health, Global poverty and Neartermist other) groupings we identified. As we discussed in those previous years (e.g. here and here), the terms “longtermist” and “neartermist”, as just a matter of simplicity/convenience, matching the common EA understanding of those terms, but people might favour those causes for reasons other than longtermism / neartermism per se, e.g. decision-theoretic or epistemic differences.
Substantively, one might wonder: “Are people who support AI or other global catastrophic risk work, allocating more resources to the “Near-term” buckets, rather than to the “Longtermist (including global catastrophic risks)” bucket, because they think that AI will happen in the near-term and be sufficiently large that it would dominate even if you discount longterm effects?” This is a reasonable question. But as we show in the appendix, higher ratings of AI Safety are associated with significantly higher allocations (almost twice as large) to the “Longtermist” bucket, and lower allocations to the Near-term buckets. And, as we see in the Ideas and Cause Prioritizations section, endorsing the explicit “long-term future” item, is strongly positively associated with higher prioritization of AI Safety.

David_Moss 5 Dec 2025 13:10 UTC
4 points
0 ∶ 0
on: Explaining a subtle but important error in the LEAP survey of AI experts
Perhaps they mentioned this elsewhere, but it could have added to their defence of this kind of question to note that ‘third-person’ questions like this are often asked to ameliorate social desirability and incentivise truth-telling, e.g. here or as in the bayesian truth serum. Of course, that’s a separate question to whether this specific question was worded in the best way.

David_Moss 4 Dec 2025 10:44 UTC
4 points
0 ∶ 0
in reply to: huw’s comment on: How to quantify global human suffering? Results from OPIS’s pilot survey
That’s a good question. Although it somewhat depends on your purposes, there are multiple reasons why you might want to measure both separately.
Note that often affect is measured at the level of individual experiences or events, not just an overall balance. And there is evidence suggesting that negative and positive affect contribute differently to reported life satifaction. For example, germane to your earlier question, this study finds that “positive affect had strong effects on life satisfaction across the groups, whereas negative affect had weak or nonsignificant effects.”
You might also be interested in measuring negative and positive affect for other reasons. For example, you might just normatively care about negative states more than symmetrical positive states, or you might have concerns about the symmetry of the measures.

David_Moss 3 Dec 2025 18:27 UTC

24 points

1 ∶ 0

in reply to: NickLaing’s comment on: Poll: Wild animal vs Farm Animal welfare

We have some data on public support for WAW interventions of different kinds from our recent Attitudes Towards Wild Animal Welfare Scale paper. The academic paper itself does not highlight the results that would be most interesting to EAs in this context, so I’ll reproduce them below, and also link to this more accessible summary on the Faunalytics website. (credit to @Willem Sleegers who was the first author of both).

We asked respondents about their level of support for different specific interventions. All the interventions tested had a plurality of support except for genetically modifying wild animals.^[2] Helping wild animals in natural disasters, vaccinating and healing sick wild animals, and supplying food, water and shelter, all had large majorities in support. Note that the survey sample was not weighted to be representative (as the goal of the studies was to validate the measures not to assess public opinion), but I would not expect this to change the basic pattern of the results.

	Oppose	Neither	Support
Helping wild animals in fires and natural disasters	0.8%	3.7%	95.5%
Vaccinating and healing sick wild animals	6.7%	10.5%	82.7%
Providing for the basic needs of wild animals (e.g., supplying food and water, creating shelters)	9.5%	11.5%	79.1%
Conducting research into how to alter nature to improve the lives of wild animals	26.5%	19.1%	54.3%
Controlling the fertility of wild animals to manage their population size	33.9%	23.8%	42.3%
Genetically modifying wild animals to improve their welfare or the welfare of other wild animals	70.3%	16.2%	13.5%

Of course, some might claim that the popular interventions are not characteristic of WAW or are too small scale (though I do not think this is true of vaccinations). But I think it is notable than even research into “how to alter nature” has majority support, and fertility control has plurality support.

We also asked, more abstractly, about whether people endorse the attitudes that intervening to help wild animals is infeasible (what we call ’intervention ineffectiveness), with these items:

Ecosystems are too complex to predict the outcomes of efforts aimed at improving the lives of wild animals.
Nothing much can be done to reduce the hardships that affect animals living in the wild.
It is not possible to reliably improve the lives of wild animals.
It is not possible to solve the problems that wild animals face in nature.

With the caveat that this measure was not designed for polling absolute levels of public support, but rather than reliably capturing a specific underlying attitude, on average respondents did not strongly endorse this attitudes, and slightly leaned towards disagreement.

Stepping back, I would not take a very strong stance on the FAW vs WAW normality/popularity question. I think this is very likely to vary at the level of individual intervention, or depend on framing when presented abstractly. As an example of the latter point, when we presented FAW and WAW as abstract cause areas, each with a ‘moderate’ and ‘controversial’ framing, WAW was competitive with a moderate framing (non-significantly ahead of Climate Change as well), while it lagged when presented with a framing that mentioned genetic engineering. We’d be happy to gather further evidence to examine a wider variety of framings or interventions.

^
Though this was a secondary aspect of the paper. The core aim of the paper was to develop and validate a new set of measures towards wild animal welfare, and these questions about support for different interventions were intended only to validate those measures.
^
Which is, perhaps, not surprising, many people oppose GM tout court.

David_Moss 3 Dec 2025 13:34 UTC
6 points
0 ∶ 0
in reply to: huw’s comment on: How to quantify global human suffering? Results from OPIS’s pilot survey
It seems like life satisfaction (a cognitive evaluation) and affect are simply different things, even if they are related, and have different correlates. It doesn’t strike me as that intuitively surprising that people would often evaluate their lives positively even if they experience a lot of negative affective states, in part because people care a lot about things other than their hedonic state (EAs/hedonic utilitarians may be slightly ununusual in this respect), and partly due to various biases (e.g. self-protective and self-presentation biases) that might inclined people towards positive reports.

David_Moss 19 Nov 2025 22:17 UTC
4 points
0 ∶ 0
in reply to: gergo’s comment on: Runway till January: Amplify’s funding ask to market EA & AI Safety
Thanks gergo!

A small number of small orgs do meet the bar, though! AIS Cape town, EA Philippines, EA & AIS Hungary (probably at least some others) has been funded consistently for years. The bar is really high though for these groups, and I guess funders don’t see enough good opportunities to support to justify investing more resources into them through external services. (Maybe this is what you meant anyway but I wasn’t sure)
My claim is actually slightly different (though much closer to your second claim than your first). It’s definitely not that no small groups are funded (obviously untrue), but that funders are often not interested in funding work on the strength of it supporting smaller groups, where “smaller” includes the majority of orgs.

David_Moss 19 Nov 2025 9:54 UTC
9 points
1 ∶ 0
on: Runway till January: Amplify’s funding ask to market EA & AI Safety
The highest counterfactual impact comes from working with organisations that could benefit but haven’t budgeted for marketing due to a lack of understanding.
As JS from Good Impressions told us:
“Willingness to pay is not as strong a predictor of commitment or perceived value as I thought it would be.”
This creates a chicken-and-egg problem: funders expect clients to pay, but clients lack the means.
This matches our own experience (with the Rethink Priorities, Surveys and Data Analysis team).
I would add that, in our experience, the situation is worse than the chicken-and-egg problem as stated. As you note, funders are often not interested in funding work which is supporting smaller, more peripheral or less established groups (and to be clear, this seems to be a matter of ‘most orgs don’t meet the bar’ rather than ‘all but a few smaller groups do meet the bar’).
But we have also been told by more than one funder that if our work is supporting core, well-resourced orgs, then those orgs ought to fund it themselves, and you shouldn’t need central funding.^[1] This creates a real catch-22 situation, where projects of this kind can neither be funded if they are supporting the biggest orgs or if they’re not.
I also find that people often significantly overestimate the ability to pay of even the largest orgs to pay. We often find that orgs are willing to invest tens of staff hours in working with us on a project- implying they value it- but they still have hard limits of whether they can spend $500-1000 on costs for the project.^[2]
1. ^
  I’ve not directly experienced this response recently, as we’ve not been applying for funding on this sort of basis, so YMMV.
2. ^
  Perhaps explained by (i) even well-resourced orgs don’t have large amounts of unrestricted funds that they can spend on whatever unforeseen expenses they want (ii) internal approvals for funding being difficult, (iii) needing/wanting to stick to some, pretty low, sense of what reasonable costs for advertising/experiments are.
What links here?
- Runway till January: Amplify’s funding ask to market EA & AI Safety by gergo (18 Nov 2025 19:22 UTC; 56 points)

David_Moss 10 Nov 2025 10:52 UTC
10 points
0 ∶ 0
on: Lack of Diversity Within EA
I am curious on the communities thoughts on this lack of diversity
In previous surveys, diversity/JEID related issued have often been mentioned as reasons for dissatisfaction. That said, there are diverse views about the topic (many of which you can read here, it’s much discussed).
Community Health Supplementary Survey
EA Survey 2019: Community Information

David_Moss 1 Nov 2025 12:50 UTC
3 points
0 ∶ 0
in reply to: Jakob Lohmar’s comment on: Doing Prioritization Better
Thanks Jakob!
One thing I’ll add to this, which I think is important, is that it may matter significantly how people are engaged in prioritisation within causes. I think it may be surprisingly common for within-cause prioritisation, even at the relatively high sub-cause level, to not help us form a cross-cause prioritization to a significant extent.
To take your earlier example: suppose you have within-animal prioritisers prioritising farmed animal welfare vs wild animal welfare. They go back and forth on whether it’s more important that WAW is bigger in scale, or that FAW is more tractable, and so on. To what extent does that allow us to prioritise wild animal welfare vs biosecurity, which the GCR prioritisers have been comparing to AI Safety? I would suggest, potentially not very much.
It might seem like work that prioritises FAW vs WAW (within animals) and AI Safety vs biosecurity (within GCR), would allow us to compare any of the considered sub-causes to each other. If these ‘sub-cause’ prioritisation efforts gave us cost-effectiveness estimates in the same currency then they might, in principle. But I think that very often:
- Such prioritisation efforts don’t give us cost-effectiveness estimates at all (e.g. they just evaluate cruxes relevant to making relative within-cause comparisons).
- Even if they did, the cost-effectiveness would not be comparable across causes without much additional cross-cause work on moral weights and so on.
- There may be additional incommensurable epistemic differences between the prioritisations conducted within the different causes that mean we can’t combine their prioritisations (e.g. GHD favours more well-evidenced, less speculative things and prioritises A>B, GCR favours higher EV, more speculative things and prioritises C>D).
Someone doing within-cause prioritisation could complain that most of the prioritisation they do is not like this, that it is more intervention-focused and not high-level sub-cause focused, and that it does give cost-effectiveness estimates. I agree that within-cause prioritisation that gives intervention-level cost-effectiveness estimates is potentially more useful for building up cross-cause prioritisations. But even these cases will typically still be limited by the second and third bulletpoints above (I think the Cross-Cause Model is a rare example of the kind of work needed to generate actual cross-cause prioritisations from the ground up, based on interventions).

David_Moss 31 Oct 2025 20:05 UTC
13 points
0 ∶ 0
in reply to: Vasco Grilo🔸’s comment on: Framing EA: ‘Doing Good Better’ Did Worse
Thanks Vasco.
I think there are two distinct questions here: the sample and the analysis.
When resources allow, I think it is often better to draw a representative sample of the full population, and then analyse the effect of different traits / the effects within different sub-populations. Even if the people getting involved in EA have tended to be younger, more educated etc., I think there are still reasons to be concerned about the views of the broader population. Of course, if you are interested in a very niche population then this approach will either not be possible or will be very resources inefficient and you might have to sample more narrowly.
When looking for any interactions with demographics for these results, I found no significant demographic interactions, which is not uncommon. I focus here on the replication study in order to have a cleaner manipulation of the “doing good better” effect, though it’s a smaller sample, and we gathered fewer demographics than in the representative sample in the first study.
For age, we see a fairly consistent effect (the trends are fairly linear even if I allow them to be non-linear, fwiw).
For student status, we see no interaction effect and the same main effect.
For sex, we see no effect and a consistent main effect.^[1]
If there’s a lot of interest this we could potentially look at education and income in the first survey.
1. ^
  People sometimes ask why we use sex rather than gender in public surveys and it’s usually to match the census so that we can weight.