EA Survey 2020: How People Get Involved in EApost by David_Moss · 2021-05-20T14:10:07.212Z · EA · GW · 17 comments
EA Survey 2020: How EAs get involved in EA Summary Where do people first hear about EA? Note: full size versions of graphs can be viewed by opening them in a new tab. People First Hear of EA: Other People First Hear of EA: Further Details People First Hear of EA: Changes Over Time What factors are important for getting involved? in What is Important for Getting Involved in EA Over Time Differences Between Groups Hearing About EA Engagement Gender Race Involved Engagement Gender Race Predicting differences in engagement Sampling/Referrer Adjustment Appendix Further details about where people first hear about EA Credits None 17 comments
EA Survey 2020: How EAs get involved in EA
- Personal contacts (16.3%) remain the most common way people that people have first heard about EA, throughout the history of the movement, followed by 80,000 Hours (12.8%)
- Among EAs who first got involved in EA in 2020, 17.1% first heard about EA through a personal contact and 16.5% from 80,000 Hours
- Podcasts have increased in importance as a source of people first hearing about EA, accounting for 15.2% of people who got involved in 2020
- More than half (50.7%) of respondents cited 80,000 Hours as important for them getting involved in EA
- A much larger proportion of non-male respondents first heard about EA from a personal contact compared to male respondents
- Significantly higher proportions of non-male respondents found personal contacts or local groups important for them getting involved in EA
Where do people first hear about EA?
1,912 (88%) respondents replied to the question, “Where did you first hear about Effective Altruism?”
Note: full size versions of graphs can be viewed by opening them in a new tab.
‘Personal Contact’ was the most selected option (16%) followed by 80,000 Hours (13%), and Book, article, or blog post (9%). In 2019, Book and Article or blog post were separate categories but summed to roughly 11%. Similarly, EA Global and EAGx were separate categories in 2019. The other options remained the same so we can analyze across years. Personal Contact, 80,000 Hours, Podcast, and Local or University Group had an increased share of responses from 2019 (ranging from 1.6% to 3.2% increases). Book, Article, or Blog, LessWrong, Slate Star Codex (SSC), GiveWell, and I don’t remember had a decreased percentage of responses (-1.4% to -1.9%). All remaining categories had a <1% change from 2019. Some of the underlying causes of these shifts in replies will be discussed below.
Where People First Hear of EA: Other
Of the 167 Other replies, 40% were categorized as fitting into an existing category.
About 26% of these reported first hearing about EA from a personal contact, an EA talk/conference, a local group, or through animal advocacy work. A further 26% mentioned a public intellectual or blogger, the majority of which were Peter Singer (28) and Sam Harris (9). Roughly 20% mentioned hearing about EA from social media. Of the 17 responses (10%) which referenced YouTube, 9 mentioned French channels called Mr. Phi and Science4All and 2 mentioned a Polish channel called Everyday Hero. A further 10% referenced some other form of media including Wikipedia, news articles, films, and podcasts. 15 people (9%) referenced an EA or EA-adjacent organization, and another 15 people mentioned a forum or blog.
Where People First Hear of EA: Further Details
As last year, we also asked respondents to give more details about how they first heard about EA. We then classified these responses into discrete categories. We display these results divided by the category of fixed response that respondents selected (i.e. all the open comment further explanations provided by those who indicated that they first heard about EA from 80,000 Hours). We provide mosaic plots showing the proportion of responses within each superordinate category with the largest number of responses, and include bar charts for the categories with lower numbers of responses in the appendix.
Where People First Hear of EA: Changes Over Time
Using information on when people first got into EA, we can examine differences in where people first hear about EA across more or less recent cohorts of EAs. Of course, it is important to bear in mind that this does not necessarily represent changes in where people hear about EA across time, since people in earlier cohorts who heard about EA from different sources may have dropped out of EA (or the EA Survey sample) at different rates. In the future, we may conduct more analyses combining data across cohorts across years to examine this.
Looking at differences in where people reported first hearing about EA across different cohorts within our survey, we observe similar and continuing trends to those in previous years of the survey.
Personal contacts (17.1%) and 80,000 Hours (16.5%) continue to account for the largest share of people first hearing about EA (looking at those who first got involved in 2020). Overall, this means that over a third of EAs getting involved in 2020 first heard of EA from one of these sources.
We also observe that podcasts have increased in importance significantly, accounting for 15.2% of EAs who got involved in 2020 (in the qualitative data, 59% of these were attributed to Sam Harris).
What factors are important for getting involved?
We asked about what factors “were important for [them] getting involved in EA”, allowing respondents to select multiple options. On average respondents selected around 3 options (median 3, mean 3.3).
This year more than half (50.1%) of respondents reported that 80,000 Hours were important for them getting involved. Personal contacts (35.4%) and GiveWell (34.8%) were the next most commonly selected categories.
The broader results below show that a wide variety of other factors were also important for large numbers of respondents.
Changes in What is Important for Getting Involved in EA Over Time
Note that since this was a question where people could select multiple options, the percentages here will sum to more than 100%. Moreover, as we noted last year [EA · GW], the number of factors people select as important varies systematically across cohorts (as we can see below, the very newest EAs select fewer factors overall).
We can see that 80,000 Hours is disproportionately important relative to other factors, for EAs who have gotten involved since 2015.
Differences Between Groups
We examined differences in the numbers reporting that they first heard about EA from different sources across different groups (split by gender, race and level of engagement). It is important to note that these analyses don’t control for other factors. For example, if more people heard about EA from a source in more recent years, then we would expect the EAs coming from this source to be lower engagement overall. In later analyses we will try to control for these factors. Nevertheless, these figures may be of interest regardless, as they tell us straightforwardly what proportion of different groups have come from particular sources historically.
First Hearing About EA
Most differences are not significant. However, we observe that significantly more highly engaged EAs report first hearing about EA from TED talks, while significantly more lower engagement EAs first heard about EA from podcasts or SlateStarCodex. As noted above, such differences could be confounded by other factors.
It is also worth attending to which sources accounted for higher numbers of highly engaged EAs overall. As we can see, irrespective of differences in proportions of highly engaged recruited, because some routes recruit significantly more EAs overall, these account for substantially greater numbers of more engaged EAs in total.
As our measure of engagement was a 1-5 scale, rather than binary, we also include a plot showing the proportions within each level of the scale below. The standard error and confidence intervals are still centred around the percentages in level 4-5, however.
We observe a large, significant difference in the proportion of men/women first hearing about EA from personal contacts, with a larger percentage of female respondents first hearing about EA from personal contacts.
We compared responses for respondents who only selected ‘white’ versus those who selected any of the other categories (even if they also selected ‘white’ as a category). The only significant difference was for the category book, article or blog post.
We also explored differences in terms of which factors were selected as important for getting involved in EA across the same groups as above. Each graph shows the percentages of respondents within each group selecting each factor as important (e.g. 0.2 = 20%).
We observe significantly higher proportions of more engaged EAs selecting local groups, personal contact, the online EA community, books and EA Global and EAGx as important.
“Personal Contact” and “Group” stand out as particularly important for those who are most highly engaged.
Significantly higher percentages of non-male respondents indicated personal contacts or local groups to be important for their getting involved in EA. Conversely, significantly lower percentages of non-male respondents indicated that LessWrong was important for their getting involved.
We observe that Giving What We Can was selected by a significantly lower proportion of respondents who selected any category other than (only) white.
Predicting differences in engagement
As noted, the analyses we prevent above which show differences in the engagement level of people who first heard about EA from different sources, don't account for possible confounding differences. For example, as people who heard about EA more recently tend to be less engaged, we would expect people who first heard about EA from sources which recruited more EAs more recently, to appear less engaged on average. We will explore predictors of engagement in more detail in the dedicated Engagement post later in this series, which will aim to control for some of these differences.
This year, as a robustness check, we include an analysis looking at how the results change if we exclude those respondents who were referred to the survey via a link from 80,000 Hours. (We previously discussed the issue of sampling from different referrers here) We focus on 80,000 Hours, because this year they were the single largest referrer of respondents to the survey, though they still only account for a minority (25%) of respondents overall, and because they represent one of the most commonly selected categories. One might wonder whether the high number of people being referred to the EA Survey from 80,000 Hours is artificially inflating respondents selecting 80,000 Hours.
To examine this, we looked at the results for where people first hear about EA and what helped them to get involved in EA excluding all respondents who were referred to the survey from 80,000 Hours. Clearly excluding every participant who was referred to the EA Survey from 80,000 Hours represents an extreme over-correction, as we would expect respondents who first heard about EA from or got more involved due to a given source, to be more likely to also hear about the EA Survey from that source.
This reduces the proportion of respondents indicating that they first heard about EA from 80,000 Hours from 12.8% to around 8%, roughly in line with the results from last year. Similarly, excluding every respondent referred from 80,000 reduces the percentage of respondents selecting 80,000 Hours as important for getting them involved quite dramatically, down to around 45%, while still leaving 80K substantially the most commonly selected category. These are significant changes in the results, but suggests that 80,000 Hours remain the top and second-most commonly selected categories for the getting involved and first heard questions respectively, even when excluding all respondents referred from 80,000 Hours directly.
Further details about where people first hear about EA
The annual EA Survey is a project of Rethink Priorities. This post was written by David Moss, Jacob Schmiess and David Reinstein. Thanks to Peter Hurford, Neil Dullaghan, Jason Schukraft, David Bernard, Dominika Krupocin, Ben Todd and Howie Lempel for comments.
We would also like to express our appreciation to the Centre for Effective Altruism for supporting our work. Thanks also to everyone who took and shared the survey.
Comments sorted by top scores.