Thanks for the comment! Your post was a key impetus for us prioritising publishing these results.
I agree that there could have been a change in the (average) personality of the EA community since 2018. My guess is that this is not the case. We know the demographic composition of the community has barely changed since 2018 in terms of gender or age. Of course, it’s possible that there have been unrelated changes specifically in terms of EAs becoming more extraverted, for example, but it’s not clear why that would be the case.
[Edited to clarify: I think differences in composition of the samples more along the lines of the secone bullet point can explain more of this, though I don’t think it’s simply due to lower/higher engagement] We can explore this by looking at different subgroups with different levels of engagement within the EA Survey.
First, we look at people actively volunteering >5 hours a week (which also gives us a sample size of around 250). The error bars are quite large for the smaller sample sizes, and there’s not a consistent pattern. EAS 2018 >5hrs people are higher than both our samples on openness, appear somewhere in the middle for conscientiousness (but the larger error bars are overlapping with both other estimates, similar for extraversion, slightly higher for agreeableness but closer to and overlapping with the many EAS sample, and lower than both our samples (though overlapping with our estimate) for neuroticism. That said, I don’t put so much stock in this sub-group. The sample size of 250 is quite small for the purpose. And working >5hrs a week on EA things seems like a hard to interpret/heterogeneous group (probably a mix of people who have a lot of time on their hands + more hardcore professional EAs). Self-reported time on EA also seems potentially vulnerable to social desirability bias (certain people over-state their time spent and report more positive traits). I do note that the >5hr people in both samples do seem to tend to report more “positive” personality, which is of course equally compatible with the explanation that people who spend more time on EA have better personalities or that results for these groups are affected by social desirability.
As a proxy for high engagement, I personally prefer EA Forum membership (back in 2018 the Forum was a much more niche affair, before its expansion, and EA Forum membership correlates highly with other measures of engagement). There we have a larger sample and see limited differences.
Thanks for this, David. I think the greater than/less than 5 hours volunteering is as close as we’ll get to an apples-to-apples comparison between the two samples, though I take your point that this subset of the sample might be fairly heterogeneous.
One speculation I wanted to share here regarding the significant agreeableness difference (the obvious outlier) is that our test bank did not include any reverse-scored agreeableness items like ‘critical; quarrelsome’, which is what seems to be mainly driving the difference here.
I wonder to what degree in an EA context, the ‘critical; quarrelsome’ item in particular might have tapped more into openness than agreeableness for some—ie, in such an ideas-forward space, I wonder if this question was read as something more like ‘critical thinking; not afraid to question ideas’ rather than what might have been read in more lay circles as something more like ‘contrarian, argumentative.’ This is pure speculation, but in general, I think teasing apart EAs’ trade-off between their compassionate attitudes and their willingness to disagree intellectually would make for an interesting follow-up.
One speculation I wanted to share here regarding the significant agreeableness difference (the obvious outlier) is that our test bank did not include any reverse-scored agreeableness items like ‘critical; quarrelsome’, which is what seems to be mainly driving the difference here.
Yeh, I agree! And I think that the pattern at the item level is pretty interesting. Namely, EAs are reasonably ‘sympathetic, warm’, but a significant number are ‘critical, quarrelsome’. As I noted in the post, I think this matches common impressions of EAs (genuinely altruistic, but happy to bluntly disagree).
I wonder to what degree in an EA context, the ‘critical; quarrelsome’ item in particular might have tapped more into openness than agreeableness for some—ie, in such an ideas-forward space, I wonder if this question was read as something more like ‘critical thinking; not afraid to question ideas’ rather than what might have been read in more lay circles as something more like ‘contrarian, argumentative.’
It’s an interesting theory! Fwiw, I checked the item-level correlations and the correlations between the reverse-coded agreeableness item and the two openness items were both −0.001.
This is pure speculation, but in general, I think teasing apart EAs’ trade-off between their compassionate attitudes and their willingness to disagree intellectually would make for an interesting follow-up.
Agreed. My own speculation would be that EAs tend to place a high value on truth (in large part due to thinking it’s instrumentally necessary to do the most good). It also seems plausible to me that EA selects for people who are more willing to be disagreeable, in this sense, since it implies being willing to somewhat disagreeably say ‘some causes are much more impactful than others, and we should prioritise those based on deliberation, rather than support more popular/emotionally appealing causes’.
Thanks for the comment! Your post was a key impetus for us prioritising publishing these results.
I agree that there could have been a change in the (average) personality of the EA community since 2018. My guess is that this is not the case. We know the demographic composition of the community has barely changed since 2018 in terms of gender or age. Of course, it’s possible that there have been unrelated changes specifically in terms of EAs becoming more extraverted, for example, but it’s not clear why that would be the case.
[Edited to clarify: I think differences in composition of the samples more along the lines of the secone bullet point can explain more of this, though I don’t think it’s simply due to lower/higher engagement] We can explore this by looking at different subgroups with different levels of engagement within the EA Survey.
First, we look at people actively volunteering >5 hours a week (which also gives us a sample size of around 250). The error bars are quite large for the smaller sample sizes, and there’s not a consistent pattern. EAS 2018 >5hrs people are higher than both our samples on openness, appear somewhere in the middle for conscientiousness (but the larger error bars are overlapping with both other estimates, similar for extraversion, slightly higher for agreeableness but closer to and overlapping with the many EAS sample, and lower than both our samples (though overlapping with our estimate) for neuroticism. That said, I don’t put so much stock in this sub-group. The sample size of 250 is quite small for the purpose. And working >5hrs a week on EA things seems like a hard to interpret/heterogeneous group (probably a mix of people who have a lot of time on their hands + more hardcore professional EAs). Self-reported time on EA also seems potentially vulnerable to social desirability bias (certain people over-state their time spent and report more positive traits). I do note that the >5hr people in both samples do seem to tend to report more “positive” personality, which is of course equally compatible with the explanation that people who spend more time on EA have better personalities or that results for these groups are affected by social desirability.
As a proxy for high engagement, I personally prefer EA Forum membership (back in 2018 the Forum was a much more niche affair, before its expansion, and EA Forum membership correlates highly with other measures of engagement). There we have a larger sample and see limited differences.
Thanks for this, David. I think the greater than/less than 5 hours volunteering is as close as we’ll get to an apples-to-apples comparison between the two samples, though I take your point that this subset of the sample might be fairly heterogeneous.
One speculation I wanted to share here regarding the significant agreeableness difference (the obvious outlier) is that our test bank did not include any reverse-scored agreeableness items like ‘critical; quarrelsome’, which is what seems to be mainly driving the difference here.
I wonder to what degree in an EA context, the ‘critical; quarrelsome’ item in particular might have tapped more into openness than agreeableness for some—ie, in such an ideas-forward space, I wonder if this question was read as something more like ‘critical thinking; not afraid to question ideas’ rather than what might have been read in more lay circles as something more like ‘contrarian, argumentative.’ This is pure speculation, but in general, I think teasing apart EAs’ trade-off between their compassionate attitudes and their willingness to disagree intellectually would make for an interesting follow-up.
Thanks again for your work on this!
Thanks Cameron!
Yeh, I agree! And I think that the pattern at the item level is pretty interesting. Namely, EAs are reasonably ‘sympathetic, warm’, but a significant number are ‘critical, quarrelsome’. As I noted in the post, I think this matches common impressions of EAs (genuinely altruistic, but happy to bluntly disagree).
It’s an interesting theory! Fwiw, I checked the item-level correlations and the correlations between the reverse-coded agreeableness item and the two openness items were both −0.001.
Agreed. My own speculation would be that EAs tend to place a high value on truth (in large part due to thinking it’s instrumentally necessary to do the most good). It also seems plausible to me that EA selects for people who are more willing to be disagreeable, in this sense, since it implies being willing to somewhat disagreeably say ‘some causes are much more impactful than others, and we should prioritise those based on deliberation, rather than support more popular/emotionally appealing causes’.