Rethink Priorities collaborated with the Centre for Effective Altruism (CEA) and 80,000 Hours to survey attendees of the 2024 Meta Coordination Forum (MCF; n = 25). In addition, we distributed a shortened version of this survey to a number of other organizations associated with effective altruism (EA), to try to estimate talent needs across a broader spectrum of other organizations. However, this latter survey has a very small sample size (n = 7) due to a low response rate, so in the reporting below we have focused primarily on the results from the MCF.
The MCF survey covered questions concerning:
- Funding needs for core meta organizations
- What skills are most valuable for the EA community in the near future
- How valuable different kinds of hire are to both organizations and the community
- How many resources should be allocated to different causes
- How quickly should we be spending EA resources
- AI Safety and EA
- EA’s target audience
- How much to grow the movement
Interpreting these results
We think that the results of these surveys are potentially valuable for increasing the transparency of how decision makers in core EA community-building organizations and projects are thinking about key questions. That said, we would caution people against uncritically placing too much weight on these results for the following reasons:
- Even before the project had begun, the EAIF highlighted in their grant recommendation, that a risk of the survey was that readers might defer too much to the results of the survey without considering whether they applied in their particular context.
- Participants were granted 30 minutes to reflect on and answer the survey during the MCF. However, a number of participants commented that their answers might change had they spent more time reflecting on the questions.
- We were asked by CEA to design questions that would elicit specific estimates about fairly technical questions (e.g., the value of a marginal hire – adjusting for various counterfactuals and making certain assumptions). Historically, these questions have proven difficult and the results may at least partially reflect possible confusion on the part of respondents. Likewise, this year, we also received feedback from some respondents that some of the questions were too difficult. We were also asked to leave a number of questions similar or identical to previous years’ questions to allow comparison.
- Attendees for this MCF were selected with a particular focus on EA branding and funding decisions. This may or may not reflect a representative group of EA decision makers or reflect all the areas of expertise relevant to all of the questions included in this survey.
Summary of key results
- Organizations at the MCF reported that they could productively use an additional $86,250,000 (median $5,000,000, mean $7,840,909) across the next year.
- Leadership/strategy, management, government and policy expertise, as well as media/public engagement skills stand out as highly valued for the next 5 years.
- The average value to an organization of their most preferred over their second most preferred candidate, in a typical hiring round, was estimated to be $88,737 (junior hire) and $455,278 (senior hire).
- The average value to an organization of making a hire compared to not filling the role at all was $272,222 (junior) and $1,450,000 (senior).
- The average value to the community of a person with equivalent expected lifetime value to an organization’s typical hire joining the community was estimated to be $2,037,500 (junior) and $7,308,333 (senior). This suggests that the value of recruiting ‘hire-level’ EAs to the community is estimated to be extremely high.
- On average, respondents estimated that EA resources over the next 5 years should go 45.2% to Longtermist (including GCR) causes, 23.8% to Near-term animal-focused causes, and 22.6% to Near-term human-focused causes.
- Evaluating more fine-grained causes, respondents estimated that 26.8% should go to work focused on AI, 15.7% to Global health, 14.7% to Farm animal welfare, 10.2% to building EA and related communities, and 8.2% to Biosecurity (in addition to smaller percentages to many other causes).
- Respondents reported that the community should spend ~12% of its financial resources each year over the next 5 years (on average), with most responses between 10-15%.
- Respondents strongly agreed that AI safety work should include many people who are not into EA, that we should have separate EA and AI groups at top universities, and that most AI outreach should be done without presenting EA ideas or assuming EA frameworks.
- Respondents were roughly split (though mildly leaning against) promoting AI safety more than other EA ideas. Agreement with this statement has declined since 2023 (though this may reflect differences in invitees more than changes in beliefs).
- Outreach to certain groups was evaluated as particularly high impact (High Net Worth individuals: median 7.5x more valuable than outreach to top 20 universities, aligned media 5.5x, current policy-makers 5.0x). These results may have been influenced by the event’s focus on communications and fundraising (some attendees at the MCF were invited because they worked in communications and fundraising).
- When asked to evaluate the expected lifetime impact of recruits from different audiences (relative to the median EAG attendee), valuations were relatively flat across groups. No median ratings for different groups were more than 3.0x higher than the median EAG attendee (EA-aligned professor at top 20 university: 3.0x, PhD Students in top 3 graduate program in ML: 3.0x)
- Estimates of the differences in the value of EA-aligned outreach to different geographic areas were also relatively flat. Relative to outreach to the UK, China had the highest media valuation (2.0x). EA-aligned outreach to people based in the United States’ East Coast (1.5x) and West Coast (1.3x), as well as India (1.4x) were also evaluated as slightly more valuable.
- Respondents appeared to favor dramatic increases in the number of people encountering and engaging with EA. They estimated that the optimal number of people to visit core EA websites in the next year would be 19,000,000 – an approximate doubling of the total all-time number of visitors. Likewise, respondents reported that it would be ideal to approximately double the number of active subscribers to the EA newsletter (500,000 extra) next year. However, these may reflect ideal numbers, not targets that respondents believe we should pursue, given the likely effects of pursuing those targets.
- Respondents believe that one highly engaged EA is approximately worth 10,000, people being exposed to EA in a high fidelity way (such as a long-form article), setting aside the value of such people then going on to become highly engaged.
We were asked by CEA to pose a question to respondents about the amount of high-impact funding they could absorb over the next year. We operationalized this in the following way:
“Imagine a new, very large funder for EA-aligned organizations appears. They want to make a 1-year grant to your organization and want to support you until the point where your marginal spending would be less than half as cost-effective as it is now. They ask you to provide an estimate, in US dollars, for this 1-year grant. What would you tell them? Assume you are happy to receive donations from the new donor.”
Part of the motivation for asking this question was to estimate the total funding needs of the EA meta ecosystem. As such, in the plot below, we averaged responses where there were multiple attendees from a single organization. We also excluded one, particularly high, response from a respondent who noted that their stated budget would be to regrant to other EA meta organizations (to avoid double-counting). With this done, the mean response was $7,840,909, with a median of $5,000,000. The total ‘high impact’ (i.e., no less than half as cost-effective) funding needs of the attending organizations for the next year were $86,250,000.
We also plotted responses from the Talent Needs Survey (n = 5) alongside these responses. 
The responses from the Talent Needs Survey were quite a bit lower, likely reflecting the smaller size of these organizations, with average stated funding needs of $1,753,333 and a median grant size of $1,000,000.
Valuable skills
To assess likely talent needs for the movement over the next five years, we asked respondents about the value of various skills using the following question:
“Over the next five years, which of the following skills will be most valuable to recruit for across the EA movement? On a scale of 1-7, where 1 is “very low value” and 7 is “very high value”.”
These results show significant differences between skills, ranging from leadership/strategy rated on average 6.8 on a 7-point scale to philosophical training rated at 3.0 on a 7-point scale. Management of various kinds, as well as media experience, tended to be highly rated. A wide variety of technical skills were rated less highly.
As the distributions show, however, there was considerable disagreement about most of the skills (although leadership/strategy, government and policy expertise, and operations showed higher consensus).
The graph below includes the results from the Talent Needs Survey (n = 6). It shows the distribution of responses from the Meta Coordination Forum as violins (in blue) and the responses from the Talent Needs Survey as crosses (in black).
In addition, we asked about various traits using the following question:
“Over the next five years, which of the following traits will be most valuable to recruit for across the EA movement? On a scale of 1-7, where 1 is “very low value” and 7 is “very high value”.”
All traits were rated quite high value (above the midpoint). Good judgment, altruism, and honesty / transparency were particularly highly rated. Ambition, grit and work ethic, analytical intelligence, and creativity were relatively less highly rated.
The graph below includes the results from the Talent Needs Survey (n = 5). It shows the distribution of responses from the Meta Coordination Forum as violins (in blue) and the responses from the Talent Needs Survey as crosses (in black).
Value of a hire
We also asked several questions about the value of different hires or recruits to the community. As noted above, these are questions that have historically proven to be difficult in these surveys. In designing the questions we aimed to simplify them over previous formulations, while precisely operationalizing the intended constructs. Still, we received feedback that these questions were particularly difficult to understand. This is worth bearing in mind when interpreting the responses as they may at least partially reflect confusion, uncertainty, or error.
These questions sought to elicit judgments about a number of different things: the marginal value to an organization of a hire over the second most preferred applicant (attempting to assess the value of a hire, accounting for their replaceability), the value of a hire to an organization relative to that position not being filled at all, and the value to the community of a person (equivalent in potential impact) joining the community.
The questions we asked were:
- Imagine a typical hiring round for a junior/senior position within your organization.
- How much financial compensation would you expect to need to receive to make you indifferent about hiring your second most preferred applicant, rather than your most preferred applicant?
- How much financial compensation would you expect to need to receive to make you indifferent about that role not being filled?
- Imagine a person with the same potential lifetime impact as a typical junior/senior hire for your organization.
- How much financial compensation would the EA community as a whole need to receive (e.g., funds being added to the stock of a major EA funder) to make you indifferent between the community gaining these funds and this person joining the EA community?
Responses to each of these questions, even within this select sample, reflected significant variation (which need not reflect disagreement as different organizations may have different valuations due to their different circumstances), typically spanning 2 to 3 orders of magnitude.
One other thing to note is that in describing these results we focus on the mean valuation, which we believe is more relevant for capturing the total value of hires in aggregate and the expected value of efforts to gain new hires. However, these valuations will tend to exceed (sometimes dramatically) the typical value of a hire, so we also include the median valuations in the plots.
On average, the value of the most preferred applicant over the second most preferred applicant was $88,737 for a junior hire and $455,278 for a senior hire. This suggests that the best applicants are judged to be substantially more valuable than the next best applicants (considered relative to the total cost of a hire, for example). When we asked respondents about the value to the organization of a junior/senior hire relative to the role not being filled at all, average valuations were yet higher at $272,222 and $1,450,000, respectively.
Prima facie, these valuations are considerably lower than previous surveys, which found valuations of >$1,000,000 and >$7,000,000 for a junior and senior hire, respectively. However, those figures were about (i) losing an existing employee (ii) for three years and (iii) that employee not being able to work for anyone else in that time and (iv) not adjusting for the replaceability of that employee.
Finally, we asked about the value to the community of someone equivalent to an organization’s typical junior/senior hires, joining the community (in terms of their potential impact). Here valuations were dramatically higher, on average >$2,000,000 and >$7,000,000 for a senior hire. These valuations are more in line with previous responses, they reflect the expected lifetime impact of a person, not a person being unable to be employed for three years and thus might reasonably be expected to be higher.
These valuations appear to suggest enormously high values for movement-building activities that can recruit people with equivalent potential impact as organizations’ typical hires. In other words, a project that can counterfactually recruit to the movement at least one typical junior EA organization employee per year would seem to warrant a $2,000,000 annual budget, and a project with even a 10% chance of such a recruit might seem to justify a $200,000 annual budget.
We also plotted the responses from the MCF sample alongside those from the broader sample of EA-related organizations from the Talent Needs Survey (though this is a very small sample; n = 5). It might have been expected that the broader sample of organizations would show lower valuations (e.g., if one imagines that the MCF samples from disproportionately well-resourced organizations), but we see no evidence of that here.
Different role types
Additionally, we examined the relationship between the valuations for each question. The meaning of these numbers may not be immediately obvious, but we believe that they are potentially informative.
For example, consider the gap between people’s valuations for the role being filled by their second most preferred candidate and the role not being filled at all. This can tell us about the gap between one’s second most preferred hire and no hire at all. Looking at both the absolute size of the gap and the ratio, we see that the gap is estimated to be very large on average: $189,667 and $994,722 for junior and senior positions, respectively (median ratios of 5 and 4). This implies that even the second most preferred candidates are quite highly valued (on average). This is striking given the already high valuations noted above for hiring one’s first most preferred candidate relative to the second most preferred. That said, it is notable that the median valuation of a junior, second most preferred candidate, at $95,000, may be on the borderline of the costs to an org, of making a hire. This might imply that it is often not worth organizations hiring the second most preferred candidate.
Similarly, the gap between the estimated value to the community of someone equivalent to a recent hire joining the community and the estimated (dis)value to the org of (not) filling a role, can be seen as giving an estimate of the expected lifetime value to the community of such a person, minus their immediate value to the organization hiring them. For example, if respondents believed that their typical hires are very valuable in their roles, but then move onto lower value roles or leave the community, there would be a low or zero gap between these numbers. Conversely, if the organization believed that their typical hire produces most of their lifetime value in other future roles, we would expect to see a larger gap between these numbers. Our results suggest that people’s estimates of the lifetime value of someone equivalent to their typical hire, minus the value of that hire to the organization itself, are extraordinarily high: a mean value of $1,775,000 and $5,991,667 for a junior and senior hire, respectively (median ratios of around 5 for both junior and senior). This suggests that however valuable organizations believe their typical hires are, they believe that they will generate 5x that much value for the wider community in their lifetime.
Junior vs. senior staff
In addition, we also examined the ratio between valuations of junior and senior staff. For each of the questions, the median valuations were 4-5x. This may be of interest, as it seems likely that, for many EA organizations, the ratio between pay for their junior vs. senior staff is lower than 4-5x.
Cause prioritization
As in previous years, respondents were asked about how they believed resources should be allocated across different cause areas over the next 5 years. We asked about both broad and more specific cause areas:
“What (rough) percentage of resources should the EA community devote to the following broad areas over the next five years?”
On average, respondents believed that resources should be roughly evenly split between longtermist and neartermist causes. The single area that respondents believed should receive the largest share of resources was longtermist work, including GCR (45.2%). However, near-term animal-focused causes (23.8%) and near-term human-focused causes (22.6%) combined, are judged to warrant a slightly higher share of resources (46.4%).
Examining the distributions shows considerable disagreement even within this small, selected sample. For example, responses for longtermism range from ~25-80%, while responses for near-term animal-focused causes range from ~10-40%.
“What (rough) percentage of financial resources should the EA community devote to the following specific areas over the next five years?”
Turning to how respondents believe resources should be allocated across more specific causes, we see that resources are fairly evenly split across a large range of causes, though AI safety (26.8%), Global health (15.7%), and Farm animal welfare (14.7%) stand out as receiving higher shares of resources. Community-building work for EA and related communities (10.2%), as well as Biosecurity and pandemic preparedness (8.2%), also receive relatively larger shares, while most other causes receive ~3-5% of resources.
As before, we can see quite considerable disagreement. For the top three causes in particular, we see responses ranging from ~5-45% for AI safety, ~5-40% for Global health, and roughly ~5-25% for farmed animal welfare. Among the causes receiving smaller shares of resources, we see responses ranging from ~0-10%.
Finally, we compare responses to this year’s survey to the 2019 Leaders’ Survey.
AI safety received similar levels of resources in both surveys. However, we saw some striking differences for some other cause areas. Both Global health and Farm animal welfare were allocated substantially higher shares of resources in this year’s survey (10.9% vs. 16.2% and 9.5% and 15.2%, respectively). Conversely, most other causes received lower allocations, particularly Cause prioritization (9.4% vs. 5.5%).
Spending rates
We asked respondents approximately what percentage of current EA financial resources did they think should be spent each year over the next 5 years. This was one question in particular, which was flagged as difficult to estimate without further reflection.
The average response to this question was 12%. However, this reflected significant disagreement, with responses ranging from 5-25% (though most were between 10-15%).
AI safety
This year, respondents were asked about their agreement or disagreement with a number of statements regarding AI safety.
Respondents tended quite strongly towards agreement with all of the statements except, “We should promote AI safety ideas more than other EA ideas.” This received slightly more disagreement than agreement, on average, though the distribution was asymmetrical, i.e., those who agreed tended to mostly somewhat agree, whereas those who disagreed more strongly disagreed.
The strongest agreement was for the statement “AI safety work should, like global health and animal welfare, include many more people who are not into effective altruism.”
All other statements received similar average levels of agreement. Responses to “We should have separate EA and AI safety student groups at the top 20 universities,” appeared to show particularly low levels of disagreement. Responses to “We should try to make some EA principles (e.g., scope sensitivity) a core part of the AI safety field,” appeared to show particularly high numbers of “Neither agree nor disagree” responses, perhaps reflecting higher uncertainty.
Comparing 2023 vs. 2024
Additionally, we can compare the results of this survey to the 2023 MCF, which asked the same questions about AI. Of these, only one difference is significant, with significantly lower agreement with the statement, “We should promote AI safety ideas more than other EA ideas,” in 2024 than in 2023 (p=0.002). However, it is worth noting that due to the small sample size(s), we would lack statistical power for anything other than large differences to be significant. The difference between 2023 and 2024 for the statement, “We should promote AI safety ideas more than other EA ideas,” was particularly large (more than 1 point on a 7-point scale). That said, it is also worth bearing in mind that this could reflect differences in the composition of the MCF invitees, rather than a shift in the attitudes of decision makers.
EA's target audience
We also asked two questions about the relative value of outreach to different audiences. The first of these was the following:
“Suppose you value the expected lifetime career impact of the median EA Global attendee* as "1". Now, suppose we could recruit one person from any of the following groups who is about as aligned and engaged as the median EA Global attendee.
Relative to that of the median EA Global attendee, how valuable would you expect that person's lifetime career impact to be?
*EA Global attendees are people who have a solid understanding of the core ideas of EA and who are taking significant actions based on those ideas."
Responses reflect estimates of how many times larger the lifetime impact of a person recruited from each group would be relative to the median EA Global attendee (responses <1 reflect estimates that their lifetime impact would be less than the median EA Global attendee).
The median estimates for these responses show relatively little variation, with the highest rating being 3x, and the very lowest being 0.5x, while most are between ~1-3x. However, there was substantial disagreement for a number of these audiences (given this, we truncated the x-axis below at 10x). For example, EA-aligned professors from top 20 global universities and PhD students from top 3 graduate programs in ML had a median response of 3x, while a significant minority of responses were 10x or higher.
The median valuations of undergraduates from elite universities were close to 1, perhaps reflecting the fact that large portions of EAG attendees are students at, or graduates of, top universities.
We also asked a separate question about the value of research to a different set of groups. The reason for asking these two separate questions about different groups was purely motivated by wanting to maintain comparability with two questions posed about these audiences in previous years, though we did edit the previous questions somewhat to remove possible confounds:
“Pretend you value someone doing excellent EA-aligned outreach to undergrads at top-20 universities as “1". Now, suppose you could have a person doing excellent outreach (whether related to EA, GCR or other relevant cause areas) to one of the following groups.
Relative to top-20 university outreach, how valuable would you expect this to be?”
This question revealed somewhat larger differences between the audiences, with median estimates as high as 7.5x for HNW individuals or 5.5x for aligned media or 5x for current policymakers. Valuations for various student or academic groups were lower – in line with the previous question –with ratings between 1.5-2x.
In addition, we asked about the value of outreach to different geographic areas, with the following question:
“Suppose that you value outreach (whether focused on EA, specific GCRs or other relevant cause areas) in the UK at “1”.
Relative to outreach in the UK, how valuable would you expect the same level of effort invested in outreach to each of these specific areas to be?”
This question suggested relatively small average differences on the basis of geography, with the highest valuation going to outreach to China, with a median of 2x (relative to outreach to the UK). Most countries asked about had slightly lower median ratings than the UK. That said, we see a reasonable amount of disagreement regarding countries, with a number of respondents rating different areas >2x higher than the UK.
It is worth noting that the relatively higher valuation given to different countries doesn’t imply that it would be better for all or most people to focus their outreach on these areas, as different areas likely require very different skills, cultural knowledge, and perhaps different kinds of outreach. Thus, for many individuals, it might be net negative for them to move their focus to conducting outreach on a more highly prioritised area.
Growing the movement
We asked a series of questions related to the ideal growth rate of the community.
The ideal number of new visitors to EA websites
The first was operationalized in terms of how many additional people we should want to visit key EA-related websites:
“The total number of unique visitors to the 80,000 Hours, GWWC/GiveWell, and www.effectivealtruism.org websites is estimated to be approximately 19,000,000.
Which of the options below best reflects the number of additional visitors, which you think would be optimal throughout 2025?”
These responses reflect extremely high variability, ranging from ~3 million to 4 billion. The mean response was over 300 million, while the median was 20 million. As the question stipulated that the total number of unique visitors was 19 million, this suggests that respondents believe it would be optimal for the total number of visitors to approximately double in 2025. It’s perhaps worth emphasizing that this is a doubling of the total number of people who have ever viewed these websites, not a doubling of the yearly rate.
It is unclear whether respondents believe that this yearly doubling of the number of people who encounter EA through these channels should be sustained, or whether they simply endorse a rapid jump in the next year, but this should be explored further in future work.
The ideal number of new newsletter subscribers
Additionally, we asked a question operationalized in terms of the number of new active subscribers to the top EA newsletter it would be optimal to gain in the next year:
“The total number of people who are actively subscribed (opened at least 1 email in the last 3 months) to one of the top 3 explicitly EA cause-neutral newsletters is estimated to be 550,000.
Which of the options below best reflects the number of additional active subscribers which you think it would be optimal to gain throughout 2025?”
Responses to this question also showed significant variation, ranging from 50,000 to 1 billion – 5 orders of magnitude. The mean response was over 50 million, with a median of 500,000. Given that the question stipulated the current number of active subscribers was 550,000, this implies that the median respondent would again, favor the total number of active subscribers approximately doubling in the next year. Approximately ¼ of respondents favor a much more dramatic increase of 20x or more.
The value of high fidelity new encounters with EA relative to highly engaged EAs
In addition, we were asked to pose a question about the value of people hearing about EA in a relatively high fidelity way (e.g., a long-form article) relative to the value of the movement gaining 1,000 new highly engaged EAs. This was a difficult question to answer for a number of reasons, and a number of respondents reported finding the question difficult to answer.
One complication is that we were asked, with this question, to elicit judgments about the relative value of exposing people to EA more broadly vs. gaining highly engaged EAs abstracting from beliefs about how many people who are exposed to EA in a high fidelity way would later go on to become highly engaged EAs. But in reality, some number of such people would convert, so respondents might struggle to disentangle their abstract valuations of the two groups from their descriptive beliefs about the conversion between the two groups. To attempt to elicit judgments about the former, not the latter, we explicitly directed respondents to discount the latter, but we cannot rule out confusion or misunderstanding:
“Imagine the following two options:
Option 1: The EA movement instantly grows to include 1000 new highly-engaged, aligned EAs.
Option 2: X people hear about EA in a relatively high-fidelity way (e.g., through a long-form journalistic article).
When assessing these two options, assume that none of the people who hear about EA in a relatively high fidelity way themselves go on to become a highly-engaged, aligned EA. Consider only how much you value people who have heard about EA in a high fidelity way, relative to highly engaged aligned EAs.
What value of X would make you indifferent between option 1 and option 2?”
Responses here also reflected wide variation, spanning several orders of magnitude. The median value was 1 million, implying that gaining 1 highly engaged EA was rated roughly 10,000 times more valuable than introducing one person to EA in a relatively high fidelity way.
Differences in the expected impact of more/less recently recruited EAs
We additionally asked a question about whether there were any differences between the top 1% of EAs who joined the community three years ago and the top 1% of those who joined in the last year in terms of their ability to have a positive impact over the course of their career. This question was intended to elicit judgments about differences in the characteristics of people joining the movement, rather than differences in their situation (e.g., one might think people who joined earlier had better opportunities to have an impact), so we instructed respondents to only consider differences between the individuals themselves:
“Think of the top 1% of EAs who joined the community 3 years ago, in terms of their ability to have an impact over the course of their career. Compared to the top 1% of EAs who joined the community in the last year, do those who joined three years ago have a lower or higher ability to have a positive impact over the course of their careers?
When answering this question please think in terms only of their personal skills and abilities, not external factors that might cause their ability to have an impact to differ (e.g., less time to influence short timelines, more support available in the community).”
A plurality of respondents reported that they believed that the top 1% of EAs who had joined in the last year had a lower ability to make a positive impact compared to the top 1% of those who joined three years ago. The next largest category believed that there were no differences between the groups, while a minority believed that those who joined in the last year had a higher ability to have a positive impact.
As such, respondents appeared to lean towards believing that the ability of the top 1% of EAs who had joined in the last year was lower than the top 1% of those who had joined three years ago, though this view was not endorsed by a majority of the respondents.
EA experience
After running the Meta Coordination Forum Survey, we were asked to add one question to the Talent Needs survey about how much experience with EA respondents believed was desirable for junior/senior roles. Given the low sample size of the Talent Needs survey overall, this is based on very few responses (n = 7).
The exact question asked was: “In your view, what is the minimum level of experience with EA that is desirable for typical junior/senior roles in your organization?”
We observed that a majority of respondents thought that, at most, only basic knowledge of EA concepts or the EA community was necessary for either junior or senior roles in their organization. However, fiven the low sample size, and the significant variation we would expect between different orgs and roles, we would strongly recommend against generalizing to any significant extent based on the responses to this question in particular.
This post was written by David Moss and Willem Sleegers. Our thanks to staff at CEA and 80,000 Hours for their collaboration on designing and running the surveys and for comments on this report.
Rethink Priorities is a think-and-do tank dedicated to informing decisions made by high-impact organizations and funders across various cause areas. We invite you to explore our research database and stay updated on new work by subscribing to our newsletter.
Very quickly: I feel like it's useful to share that I did this survey and found it very hard, and a lot of other people did too. In particular, it did feel pretty rushed for such difficult questions that we didn't necessarily have a fully informed pre-existing take on. OP does mention this, but I wanted to stress that for people reading this post.
I still think it has a lot of useful information and is directionally very informative. I might get a chance to write up more thoughts here, but I am not sure I will be able to. I mostly wanted to give a quick additional flag :)
I had a similar sense of feeling underprepared and rushed while taking the survey and think my input would have been better with more time and a different setting. At the same time I can see that it could have been hard to get the same group of people to answer without these constraints.
For the monetary value of talent I‘m especially cautious on putting much weight on them as I haven’t seen much discussion on such estimates and coming up with a numbers in minutes is hard.
Rather than accepting the numbers at face value, they may be more useful for illustrating directional thinking at a specific moment in time.
Thanks for the comment Jessica! This makes sense. I have a few thoughts about this:
I would have preferred working groups especially for the questions around monetary value of talent which seemed especially hard to get a sense for.
If the survey had framed the same questions in multiple ways for higher reliability or had some kind of consistency checking* I would trust that respondents endorsed their numbers more. Not necessarily saying this is a good trade to make as it would increase the length of the survey.
*e.g., asking separately in different parts of the survey about the impact of: • Animal welfare $ / Global health $ • Global health $ / AI $ • Animal welfare $ / AI $
…and then checking if the responses are consistent across all sections.
Yeh, I definitely agree that asking multiple questions per object of interest to assess reliability would be good. But also agree that this would lengthen a survey that people already thought was too long (which would likely reduce response quality in itself). So I think this would only be possible if people wanted us to prioritise gathering more data about a smaller number of questions.
Fwiw, for the value of hires questions, we have at least seen these questions posed in multiple different ways over the years (e.g. here) and continually produce very high valuations. My guess is that, if those high valuations are misleading, this is driven more by factors like social desirability than difficulty/conceptual confusion. There are some other questions which have been asked in different ways across years (we made a few changes to the wording this year to improve clarity, but aimed to keep the same where possible), but I've not formally assessed how those results differ.
Could you flag which questions you felt most comfortable with? Or least comfortable with? Whichever is easier :)