DM

David_Moss

Principal Research Director @ Rethink Priorities
5045 karmaJoined Aug 2014Working (6-15 years)

Bio

I am the Principal Research Director at Rethink Priorities working on, among other things, the EA Survey, Local Groups Survey, and a number of studies on moral psychology, focusing on animal, population ethics and moral weights.

In my academic work, I'm a Research Fellow working on a project on 'epistemic insight' (mixing philosophy, empirical study and policy work) and moral psychology studies, mostly concerned either with effective altruism or metaethics.

I've previously worked for Charity Science in a number of roles and was formerly a trustee of EA London.

Sequences
2

EA Survey 2022
EA Survey 2020

Comments
415

Thanks David.

Are you also able to see if there are differences between different types of local groups (National/City/University/interest)?

I'm afraid I'd have to potentially get back to you about this (in terms of whether individuals in different types of groups differ), because this would require manually coding a lot of individual references to groups to determine group type.

I think the numbers initially claiming to have heard of EA (19.1%) are strongly inflated by false positives (including lizardmen), but the numbers after the 'stringent' checks (including giving a qualitative explanation of what EA is) were applied (1.9-3.1%) are much less so (though, as we argue, still somewhat inflated). Note that the org results didn't have the same checks applied, so those definitely shouldn't be taken at face value and should be expected to be inflated by lizardmen etc.

This means the results predate both Will MacAskill's high-profile publicity tour for What We Owe The Future as well as the downfall of FTX. My guess is that the number of people who have heard of Effective Altruism has increased substantially since then.

We'll be publishing results about this soon, but as we noted here, we don't think there's been such a substantial increase in awareness of EA due to FTX, including among elite groups.

Is the 10% Giving What We Can pledge, in which participants commit to donating 10% of their annual income to an effective charity, part of EA's brand or reputation?

 

These questions seem empirically tractable through surveys and related experiments. It's relatively straightforward to assess how many many familiar with EA associate it with the 10% pledge (the main challenge is that so few people have any familiarity with EA at all).

It would also be possible to assess how the pledge or association with effective giving more broadly, influences the reputation of EA. i.e. by conducting experiments, where people are randomly presented depictions of EA which include reference to the 10% pledge or to effective donations in general. This would also allow assessment of how these effects differ across different groups. RP could conduct this kind of experiment, though would need funding to do so.

Most people interested in EA are not joining[2] local EA groups

[2] From the 2020 EA Survey with 1856 responses - 50% say that they are a local group member
 

 

I think this is technically true, but only a partial picture.

  • Highly engaged EAs are much more likely to be in a local group than less engaged EAs (31% vs 69% low/high in 2020).
    • This suggests that local groups are more important among highly engaged EAs.
    • It may be that many of the EAs who are not in a group (yet) are simply not engaged enough now, but will go on to become interested as they become more engaged. 
    • It may also be that some people are not interested in engaging with a local group because they are just not engaged with EA (and won’t go on to be more engaged with EA). 
  • Many EAs who are not in a group may want to be in a group, but simply not have access to one, because there isn’t one in their area. This would be a potential consideration in favour of starting more local groups.
    • The last time we asked about this explicitly was 2018
      • 65% knew of a local group near them, 35% did not. This may have changed substantially in the last 5 years, since there are more groups, but more EAs outside of core EA areas.
      • Of those who knew of a local group, just over 50% were already in one, and 16% said they would like to be involved in one (we don’t know what the barrier was- maybe they just hadn’t had time to join one yet). Just 9.2% said they do not want to be involved in a group and 24% were unsure. 
        • This does not seem to suggest clear, wide-scale disinterest in groups to me.
      • Of those who did not know of a local group near them, 34% said they would like to be in a group while 51% were unsure. I imagine the lack of certainty may be because they don’t know what the hypothetical group would be like, since a much larger percentage of those who are near a group want to be in one.
  • Group membership seems to vary dramatically across different cities, which suggests the level of interest is not uniform/invariant.
    • Looking at cities with a significant number of EAs, the percentage who are members of a local group varies between 19% (Bay Area) to >60% (Washington DC, Zurich, Oslo). London is toward the lower end and lower than the sample overall at 36% (this is EAS 2022 data, so the average across all areas is 44%).
      • This could be due to differences in geography or the EAs there. e.g. fewer people may join groups where the area is very spread out (SF Bay, London), although Los Angeles seems to have very high membership. 
      • But it could also suggest that group-specific factors are relevant, i.e. if one invests more / more successfully in a local group, more EAs may want to join.
         

I also think it’s worth rehearsing the general evidence for EA groups being important. 


- Notwithstanding the negative Forum articles you link, EA groups are not cited particularly often as a negative influence on people’s involvement in EA (6.5% of respondents, similar to the EA Forum and a lower absolute percentage than personal contacts or 80K [bear in mind that the absolute percentage of people selecting a factor may simply reflect the fact that lots of people have encountered it]). Its ratio for positive:negative influence is not particularly poor (less good than 80K, but better than personal contacts)

  • They are also joint top (with personal connections) as the key source for making new interesting and valuable connections within EA (significantly ahead of EA Global, which is also often touted as important for this purpose). Notably most virtual, non-in-person routes are cited much less commonly.

  • In addition, when we asked EAs for the top barriers to their involvement (in 2019), no close EA friends (30%) was among the most commonly selected options, with no local group (17%) also commonly selected. EA local groups were also the second most commonly cited factor important for retaining people in EA (26.8%).

 

Some responses to other points you raise:

There will also be a lot of people who don’t happen to live in the biggest cities, or live quite far from the city centre where most activities happen. Focusing on a few cities can lead to allocating fewer resources to these people.

I think this can be turned around and taken as a reason to invest more in making sure that more cities have local groups, since otherwise people outside a few cities, with groups, don’t have access to a group. This may become more important as more EAs join outside of older core areas.

Anecdotally from conversations with other organisers the people most likely to join are those looking for a community - students, recent graduates or people who are new to the city.

I think there’s something to this (ditto EAG attendance, which isn’t representative of the broader community). In particular, I think people who have been in EA many years and who are older may become less interested in attending groups (see below). That said, I worry about old-timer EAs (which includes many key decision-makers) beginning to under-estimate the importance of groups just because they themselves already have networks, or other commitments, or have otherwise lost interest, when they remain highly important for most other highly engaged EAs.

When most people hear about EA for the first time, it’s usually via an online resource (80,000 Hours, GWWC, podcast) or word of mouth. The message they receive is that EA cares about having more impact and that EA as a movement is trying to help people have more impact.

This can contrast to the experience of going along to a local group… and experiencing the main message as ‘join our community’, with less focus on helping that person have impact.

This doesn’t seem like an inherent feature of local groups. It’s not clear we have reason to think many/most groups are emphasising community to the (net) detriment of impact. The fact that a majority of group members cite their group as being among the most important factors for their ability to have a personal impact suggests they are generally having a positive impact on EAs’ impact. 

We often use neglectedness when choosing cause areas, leading to support of unseen majorities - people in poorer parts of the world, animals and future beings. But when it comes to movement building there is less thought paid to those who aren’t visible. A lot of strategies I have seen are about increasing attendance or engagement at events rather than providing value to people who may not be as interested in attending lots of events each year but still want to consider career changes.

It’s my impression that, for many years, EA groups were neglected due to the illegibility of their impact. Their impact is mostly indirect through getting EAs involved, increasing their connection and engagement with EA, keeping them engaged and directing them to other paths of impact. It's possible to make the case that many other activities have a more direct relation to impact. And yet, per the above, very few activities seem to be as commonly cited as important by as many EAs as do EA groups (and this despite many EAs not having the chance to be a member of an EA group). 

In more recent years, fortunately, there's been an increase in the resources assigned to groups and a significant increase in the number of EAs who are members of groups. I think it would be unfortunate if this trend were to reverse.


 

Thanks for the comment! 

The question of what causes the disparity seems somewhat empirically tractable. For example, one could assess whether conservatives (in the broader population) are lower in EA-related attitudes (the challenge, of course, would be in developing valid measures which aren't implicitly coded as either liberal or conservative). 

We could also test different framings of EA and examine how support for EA from conservatives/liberals varies in response to these different frames. It seems very plausible that interest in EA (from different groups) might vary dramatically across different frames. I think this research should be done for different demographic groups (e.g. gender, race, age), but it would also be tractable and relevant to examine the influence of political ideology. It's possible that different framings would be dramatically more successful in reaching different groups.

It would also be interesting to examine the effect of presenting people with a view of EA which highlights its diversity (or lack thereof) on relevant dimensions and see how far this changes interest in learning more or getting involved. For example, one could present a description of EA, (such as, an account of an EAG which includes vignettes about various EAs who are all prominently liberal or which includes a mix of conservatives) and see how far this changes levels of interest in EA.

Another alternative explanation (to conservatives being turned off by some factor), could be differences in exposure to EA. Our survey on how many people have heard of EA suggests that about twice as many Democrats as Republicans have encountered EA (for sex, the gap is around 3:2, which would imply 60% men, 40% women).

Presumably the question of whether a disparity is a problem that requires attention depends, at least in part, on the causal question, although, of course, one might be concerned about epistemic and other effects regardless.

I agree this may stem partly from EA's very strong age skew, but I don't think this can explain a very large part of the difference. 

Within the US, Gen Z are 17% Republican, 31% Democrat (52% Independent), while Millenials are 21% Republican - 27% Democrat (52% Independent). This is, even among the younger group, only a ~2:1 skew, whereas US EAs are 77% left-leaning and 2.1% right-leaning (a ~37:1 skew). Granted, the young Independents may also be mostly left-leaning, which would increase the disparity in the general population. Of course, this is looking at party affiliation rather than left-right politics, but I think it plausible (based, in part, on LW/SSC/ACX data) that even right-leaning EAs are non-Republican voting, so the results may be more skewed than this.

Looking at ideology, US 18-29 year olds are around 23% Conservative, 34% Liberal (41% Moderate), which is also significantly more balanced.

Of course, there is also the education gap (with EAs being disproportionately highly educated), but among college graduates we still see only 31% conservative, 20% liberal, and 27% conservative vs 36% liberal for postgraduates. 

One could also argue that elite colleges specifically are even more left-leaning. But, as we note, although the EA community is skewed towards elite colleges, a very large percentage of respondents are not from particularly highly ranked colleges (though this is likely less so looking purely within the US), so it does not seem like this could also explain a large part of the difference.

It's also worth bearing in mind that in some of these cases the explanation of the difference may flow in the reverse direction: e.g. it could be that features of the EA community make it more appealing to political liberals which cause it to attract more young people, rather than vice versa.

This has already been updated per this thread. Originally, some of the graphs showed the percentage of all respondents selecting each category, but in response to comments we switched to just showing the percentage of respondents who answered each question for all graphs, for simplicity.

That's true, but by design. With that particular plot, we wanted to show clear percentages for discrete categories of interest. Categories containing even numbers of universities risk (e.g. 1-100, 101-200) being uninformative for practical purposes, by combining universities with very different characteristics, while dividing others). Of course, your mileage may vary as to which categories are practically interesting.

The previous graph (below) already shows the raw distribution, but you can also look at the cumulative percentages in the next plot below (note: this cannot show the unranked universities, which, as the former plot shows, account for a decent proportion of respondents).

79.81% of respondents who answered the religion question are atheist, agnostic or non-religious, but 69.58% of all respondents (including those who did not answer the question) are atheist, agnostic or non-religious. 

Fortunately, there is no sign of differential non-response across these questions (i.e. similar numbers of respondents answered these questions, as answered other questions a similar length through the survey) or high non-response in general, so one can straightforwardly interpret the former sets of numbers. But, counterfactually, if there were high non-response or high differential non-response, then one might be interested in the proportion selecting from all respondents (and then need to make further inferences about the missing responses).

Thanks. I agree this is interesting to look at.

Since we didn't gather donation data this year, in order to keep the survey shorter, we have to go back to data from earlier survey years.

In EAS 2020, we asked about current career (rather than career strategy). This is obviously appreciably different (and more vulnerable to just reflecting the fact that people in different current careers can afford to donate different amounts, rather than reflecting the different groups), but here are the results:

Here we can see that high donors are much more likely to be in for-profit (earning to give), though there is also a non-significant trend in the direction of higher donors being more likely to be in for-profit (not earning to give). A higher percentage of high donors also selected work at an EA non-profit (though the difference was small and not significant). Higher donors were less likely to select "still deciding what to pursue" and "building flexible career capital and will decide later" (compared to the lowest donors), but I imagine that this likely reflect these categories being more often selected by early career/student low-earning, low-donors. 

In 2019, we asked a question which may be a more informative comparison to 2022, "If you had to guess, which broad career path(s) are you planning to follow?" (this may be somewhat less vulnerable to simply reflecting the fact that people currently in earning to give can currently donate more, but probably still reflects this to a significant degree). 

Here the highest donors are more likely to select earning to give, and they are also less likely to select academia (than the lowest donors).

As with the original graphs, these differences are likely at least partly explained by other confounding variables (e.g. how long people have been in their career etc.). If we wanted to assess what is ultimately causing the differences, we'd need to examine a more complex model.

That said, while I think this is interesting, looking at total donations as an operationalization of engagement seems less informative, since although the existing measures may be slightly skewed against E2G people counting (themselves) as highly engaged (due to giving EA orgs and community buildings as exemplars of the highest level of engagement), total donations seems very skewed towards counting E2G people as highly engaged simply in virtue of them earning more than people in other roles (or people who are students/early career). 

My guess is that either a more neutral measure of engagement (e.g. simple self-report of low to high engagement) or some more complex ideal measure of engagement would probably find that (rightly or wrongly) higher engagement is associated with higher interest in research / EA org research over earning to give. It's possible that there'd be a different association with a good measure of EA dedication, which may be different from engagement (but that seems harder and more controversial to measure).

Load more