2022 Survey Results
There were 186 respondents over 28 days. The first week of the survey got 172 responses. The second week got 10. The third and fourth weeks got 2 responses each.
Previous surveys have been run over the past decade or so. Their numbers are as follows:
2017: “About 300”
The cheerfully optimistic interpretation of this is something like “cool! I got about as many as Scott did on his first try!” I’ll talk about the survey size more in the Conclusion section. For now, on with the numbers.
Previous LessWrong Surveys:
No: 131, 72.0%
Prefer not to answer: 6, 3.3%
Yes: 45, 24.7%
Did you also take the ACX 2022 Survey?
No: 93, 50.5%
Prefer not to answer: 3, 1.6%
Yes: 88, 47.8%
I. Demographics (Not counting sex and gender)
There’s a format I’ll use in a few places throughout this summation. If you see a sequence of numbers like “30.1 + 8.9 (24, 28, 34) [n=186]” those numbers are “Mean + standard deviation (1st quartile, 2nd quartile, 3rd quartile) [n= number responding].”
30.1 + 8.9 (24, 28, 34) [n=186]
I broke this out into buckets by decade so I could get a better sense of what was going on.
Under 20: 13, 7.0%
20 to 29: 91, 49.2%
30 to 39: 57, 30.8%
40 to 49: 14, 7.6%
50 to 59: 8, 4.3%
60 to 69: 2, 1.1%
The 2009 to 2014 surveys consistently found the mean age was 27, and in 2016 the mean age had only gone up to 28. This is an interesting property for a community to have! It’s far from unusual: any survey of a college campus will find the mean age stays about the same, and for different reasons so can the mean age of countries. A survey of Britney Spears fans probably finds that the mean age goes up over time, and many churches observe the same pattern. It’s one of several high level ways to look at a community: are you replacing yourself about as fast as you lose people?
LessWrong is probably a community people find and age out of. We’re a bit older. In 2011, that 20 to 29 bucket was about 60%. Still, teenagers and college students are still finding us.
White (non-Hispanic): 148, 82.7%
Asian (East Asian): 6. 3.4%
Asian (Indian subcontinent): 6, 3.4%
Hispanic: 5, 2.8%
Middle Eastern: 3, 1.7%
Other: 14, 7.8%
United States: 95, 51.1%
United Kingdom: 20, 10.8%
Australia: 12, 6.5%
Germany: 11, 5.9%
New Zealand: 6, 6.6%
France: 6, 6.6%
Canada: 5, 5.5%
Russia, 4, 4.4%
Israel: 4, 4.4%
Other: 22 , 24%
[The Other category is mostly Eastern European countries with one or two respondents. If there were three or less respondents from your country, I changed your country to Other for the public dataset in case being the one person in your country was too identifying.]
II. Sex, Gender, and Relationships
Female: 21, 11.5%
Male: 162, 88.5%
F (cisgender): 18, 9.8%
F (transgender MtF): 6, 3.3%
M (cisgender): 152, 83.1%
M (transgender FtM): 0, 0%
Other: 7, 3.8%
Asexual: 7, 3.9%
Bisexual: 24, 13.3%
Heterosexual: 145, 80.1%
Homosexual: 5, 2.8%
Other: 0, 0%
Prefer monogamous: 106, 58.9%
Prefer polyamorous: 29, 16.1%
Uncertain/no preference: 44, 24.4%
Other: 1, 0.6%
Number of Partners
0: 81, 44.51%
1: 90, 49.5%
2: 9, 4.9%
3: 1, 0.5%
4: 0, 0%
5: 0, 0%
Lots and lots: 1, 0.5%
. . . and currently not looking for more relationship partners: 119, 67.6%
. . . and currently looking for more relationship partners: 57, 32.4%
[39.0% of people who don’t have a partner aren’t looking for one. ]
To the one person with lots and lots of partners and looking for more, rock on you ambitious fellow.
Married: 40, 21.9%
Relationship: 59, 32.2%
Single: 84, 45.9%
[Two single people have a non-zero number of partners.]
Alone: 41, 23.0%
With parents and/or guardians: 29, 16.3%
With partner and/or children: 69, 38.8%
With roommates: 39, 21.9%
0: 132, 82.5%
1: 8, 5.0%
2: 16, 10.0%
3: 3, 1.9%
4: 0, 0%
5: 0, 0%
6: 0, 0%
Lots and lots: 1, 0.6%
Want More Children?
Yes: 59, 32.6%
Uncertain: 48, 26.5%
No: 74, 40.9%
III. Work and Education
Academic (teaching): 11, 6.0%
For-profit work: 72, 39.3%
Government work: 6, 3.3%
Independently wealthy: 6, 3.3%
Nonprofit work: 22, 12.0%
Self-employed: 21, 11.5%
Student: 37, 20.3%
Unemployed: 7, 3.8%
Retired: 1, .5%
Art: 4, 2.2%
Biology: 3, 1.7%
Business: 5, 2.8%
Computers (AI): 32, 18.0%
Computers (other academic): 9, 5.1%
Computers (practical): 64, 36.0%
Engineering: 15, 8.4%
Finance/Economics: 11, 6.2%
Law: 4, 2.2%
Mathematics: 9, 5.1%
Medicine: 2, 1.1%
Philosophy: 2, 1.1%
Physics: 3, 1.7%
Psychology: 2, 1.1%
Statistics: 2, 1.1%
Other: 11, 6.2%
None: 6, 3.3%
High school: 35, 19.0%
2 year degree: 7, 3.8%
Bachelors: 75, 40.8%
Masters: 36, 19.6%
JD/MD/other professional degree: 5, 2.7%
PhD: 17, 9.2%
Other: 3, 1.6%
IV. Politics and Religion
Anarchist: 6, 3.6%
Christian Democrat: 2
Conservative: 3, 1.2%
Green-Party: 2, 1.2%
Liberal: 45, 27.3%
Reactionary: 3, 1.8%
Libertarian: 56, 33.9%
Social Democratic: 27, 16.4%
Socialist: 11, 6.7%
Other: 10, 6.1%
[I made an unforced error in the survey setup here and let the Other category fill in their answer. A few people put links to extended definitions, a handful of people put some variation of None, and exactly one person answered Moderate.]
Democratic Party: 36, 67.9%
Republican Party: 2, 3.8%
Independent: 4, 7.5%
Other party: 2, 3.8%
No party: 9, 17.0%
Yes: 40, 24.5%
No: 123, 75.5%
Atheist and not spiritual: 129, 70.5%
Atheist and spiritual: 10, 5.5%
Agnostic: 28, 15.3%
Lukewarm theist: 4, 2.2%
Deist/pantheist/etc.: 3, 1.6%
Committed theist: 9, 4.9%
Buddhist ( Mahāyāna): 1, 3.2%
Christian (Catholic): 8, 25.8%
Christian (Other non-Protestant): 3, 9.7%
Christian (Protestant): 5, 16.1%
Unitarian Universalism or similar: 2, 6.5%
Hindu (Shaktism): 1, 3.2%
Hindu (Vaishnavism): 1, 3.2%
Jewish (Conservative): 2, 6.5%
Jewish (Orthodox): 1, 3.2%
Jewish (Reform): 1, 3.2%
Mixed: 3, 9.7%
Other: 3, 9.6
Something I’ll discuss more later: one of the probability questions was what the odds were that shape-shifting reptilian people controlled our world. I included this partially as a joke to make people smile, and partially because I had a hypothesis that the kind of people who thought this were likely gave different answers than average. Maybe they’d misunderstand lots of questions, maybe they’d be trying to mess with the data. It seemed like fun.
This is the first question where it stood out; all of the Christian (Other non-protestant) selections had at least 1% odds on the lizard thing. That doesn’t mean much since there were only three of them.
Just to be clear, that doesn’t mean other kinds of Christianity are more likely to believe in lizard people. It means that for the people who filled out this survey, the ones who had relatively high odds on lizard people said they were non-protestant Christians. Beware the confusion of the inverse, or as I usually remember it P(B|A)<>P(A|B). Also beware the Lizardman Constant.
Atheist and not spiritual: 39, 21.7%
Atheist and spiritual: 10, 5.6%
Agnostic: 16, 8.9%
Lukewarm theist: 56, 31%
Committed theist: 43, 23.9%
Mixed: 16, 8.9%
Christian (Catholic): 43, 27.2%
Christian (Mormon): 5,3.2%
Christian (other non-protestant): 13, 8.2%
Christian (Protestant): 63, 39.9%
Hindu: 2, 1.2%
Jewish: 14, 8.8%
Muslim: 3, 1.9%
Mixed/Other: 11, 6.9%
Unitarian Universalism or similar: 4, 2.5%
Accept/lean towards consequentialism: 111, 61.7%
Accept/lean towards deontology: 10, 5.6%
Accept/lean towards natural law: 48, 3.2%
Accept/lean towards virtue ethics: 26, 14.4%
Other/no answer: 33, 18.3%
V. Numbers Which Attempt to Measure Intellect
IQ: 136.8 + 24.3 (130, 139, 146) [n=46]
IQ Age: 18.3 + 8.3 (12, 17, 25) [n=45]
SAT out of 1600: 1496.4 + 110.8 (1460, 1530, 1575) [n=48]
SAT out of 2400: 2262.4 + 131.9 (2220, 2330, 2350) [n=18]
ACT out of 36: 33.7 + 2.8 (32, 35, 35.5) [n=28]
IV. LessWrong, The Basics
Been here since it was started: 18, 10.2%
Referred by a friend: 24, 13.6%
Referred by a link on another blog: 16: 9.0%
Referred by a link on SSC/ACX: 16, 9.0%
Referred by a positive news source: 4, 2.3%
Referred by a search engine: 29, 16.4%
Referred by HPMOR: 51, 28.8%
Other: 19, 10.7%
At least one post I made got a front page tag: 49, 26.8%
I’ve made my own post, but never gotten a front page tag: 25, 13.7%
I’ve posted a comment, but never my own post: 58, 31.7%
I’ve registered an account, but never posted: 28, 15.3%
I lurk, but never registered an account: 23, 12.6%
Never even knew they existed until this moment: 3, 1.6%
Know they existed, but never looked at them: 16, 8.7%
All or nearly all of the highlights: 6, 3.3%
Some, but less than: 25%: 24, 13.0%
About 25% of the Sequences: 8, 4.3%
About 50% of the Sequences: 24, 13.0%
About 75% of the Sequences: 30, 16.3%
All or nearly all of the Sequences: 73, 39.7%
Never even knew they existed until this moment: 14, 7.7%
Know they existed, but never looked at them: 19, 10.4%
Some, but less than 25%: 44, 24.0%
About 25% of the Codex: 14, 7.7%
About 50% of the Codex: 13, 7.1%
About 75% of the Codex: 30, 16.4%
All or nearly all of the Codex: 49, 26.8%
Time in Community:
6.1 + 4.6 (2.125, 5, 10) [n=171]
0 to 1 year: 19, 11.1%
>1 to 5 years: 65, 45.9%
>5 to 10 years: 39, 22.9.1%
>10 years: 34, 20%
Time on LW a day:
19.6 + 20.6 (10, 15, 25) [n=166]
0 to 10 minutes: 66, 40.0%
>10 to 30 minutes: 80, 48.5
>30 to 60 minutes: 15, 9.1%
>60 minutes: 4, 2.4%
752.8 + 2942.2 (0, 52.5, 400) [n=145]
VII. LessWrong and the In-Person Community
No: 112, 61.9%
Yes, once or a few times: 50, 27.6%
Yes, regularly: 19, 10.5%
No: 159, 87.8%
Yes, once or a few times: 11, 6.1%
Yes, regularly: 11, 6.1%
Meetups were one of the things I was most curious about, but I’m not sure I trust the implication that more than half the people who go to meetups also run meetups. I suspect this is an artifact of who took the census.
Attended Secular Solstice:
No: 143, 79.0%
Yes: 38, 21.0%
Physical Community Interaction:
No: 129, 73.3%
Yes, sometimes: 29, 16.5%
Yes, all the time: 18, 10.2%
No: 149, 84.2%
They are now: 14, 7.9%
Yes: 14, 7.9%
Effective Altruism: 74, 71.2%
SSC/ACX: 62, 59.6%
Post-rationality: 10, 9.6
Guild of the Rose: 8, 7.6
Glowfic: 4, 3.8%
[Note that these aren’t going to sum to 100%, since someone can be a member of multiple communities. Also, I want to give a shoutout to the Guild of the Rose: that was a write-in, which usually decreases how many people put down an answer. There were a few other communities mentioned- SPARC, Empty Spaces, Transhumanism, and NOTA- which only one person put down.)
Everyone responding said they understood the percentage format. Three people immediately used an incorrect format, though one seemed to be joking and used the right format everywhere else.
Reminder: the result format used is “Question: Mean + standard deviation (1st quartile, 2nd quartile, 3rd quartile) [n= number responding]”
P(Many Worlds): 49.0 + 32.0 (20, 50, 80) [n=148]
P(Aliens in Observable Universe): 65.7 + 32.4 (50, 80, 95) [n=160]
P(Aliens in Milky Way Galaxy): 34 + 34.3 (2, 20, 60) [n=158]
P(Supernatural): 8.4 + 19.7 (0, 0.1, 5) [n=158]
P(God): 13.2 + 25.4 (0, 1, 10) [n=153]
P(Religion): 6.4 + 19.6 (0, 0.00001, 1) [n=160]
P(Lizardman): 0.3 + 1.1 (0, 0, 0.005) [n=163]
P(Cryonics): 21.5 + 25.1 (5, 10, 30) [n=161]
P(Anti-Agathics): 35.5 + 34.7 (4.5, 25, 65) [n=156]
P(Simulation): 24.7 + 28 (1, 11, 38.25) [n=139]
P(Warming): 85.4 + 24.2 (85, 95, 99) [n=160]
P(Global Catastrophic Risk): 59.2 + 31.7 (30, 70, 88.25) [n=161]
P(Duplicated “The”): 47.1 + 25 (25, 50, 66.25) [n=161]
Remember our shapeshifting reptile people believers from before? They’re here. You know where they weren’t? They were overwhelmingly absent from religion and supernatural skeptics. Cryonics, sure, that might or might not work. Simulation hypothesis, sure, there was a range of values there. Global warming, also fine!
I didn’t really know what to make of this when I was first looking it over. The highest probability any of them gave for lizard people as 7%. Then there’s some obvious clusters. One at 5%, another bunch at 1%, another batch at 0.1%, another at .01%, another at .001, with some in between. I used 1% as my cutoff for a weirdly high probability and planned on that threshold beforehand, but if you had asked me two months ago I would have been sure there would be more of a discontinuity here.
Someone with more statistical tools to throw at this might spot something more subtle that I did, but I was surprised at how well this little sub-population blended in. They didn’t all agree on God and the Supernatural, even if that was where their absence was most obviously noticeable. I’ll talk more about why I believe this later, but I didn’t come away with the impression that these were just trolls.
IX. Traditional Census Questions
By what year do you think the singularity will occur?
9476.3 + 86212 (2035, 2042.5, 2060) [n=135]
[This was obviously heavily influenced by outliers. Cutting off anything in the past or with more than four digits gives us the following]
2059.4 + 53.2 (2035, 2043, 2060) [n=132]
[That isn’t a more reassuring number, but it gives me a much more useful sense of the range.]
Never thought about it / don’t understand: 6, 3.5%
No—and do not want to sign up for cryonics: 50, 29.1%
No—still considering it: 55, 32.0%
No—would like to sign up but haven’t gotten around to it: 43, 25.0%
No—would like to sign up but unavailable in my area: 13, 7.6%
Yes—signed up or just finishing up paperwork: 5, 2.9%
P(Calibration IQ): 51.4 + 30.5 (25, 50, 77) [n=42]
Population of Canada in 1970 in millions: 3529249.2 + 19221462.6 (17.25, 24, 40) [n=155]
[This is from some people giving the full amount. At least, I’m assuming that’s what they’re doing, either that or someone thinks there might be two hundred trillion Canadians and that they will soon crash against the northern borders of the United States in a donnybrook the likes of which make Ragnarok sound like an unruly children’s birthday party. I tried dividing any answer over a million by a million, e.g. 20,000,000 became 20. That gave the following:]
32 + 30.7 (15.25, 20.5, 33.75) [n=155]
P(Correct within 10%): 26.5 + 17.1 (15, 20, 33) [n=152]
The correct answer is 21.
You might think this question favoured Canadians. You would be wrong. When I wrote the question I thought that setting it fifty years ago would rob Canadian respondents of too much of an advantage. Only one Canadian got this right and they were only 25% confident.
X. Indulging My Curiosity
Most Important Lesson:
Conservation of Expected Evidence: 8, 9.2%
Making Beliefs Pay Rent: 8: 9.2%
Belief In Belief: 5, 5.7%
Bayes Theorem: 3, 3.3%
37 Ways Words Can Be Wrong: 2, 2.2%
Arguments Are Not Soldiers: 2, 2.2%
Politics Is the Mindkiller: 3, 3.3%
Scope Insensitivity: 2. 2.2%
What Do I Think I Know And How Do I Think I Know It: 2, 2.2%
[Since this was entirely a write-in, the answers ranged all over the territory- er, I mean the map. I’m pretty comfortable with how I bucketed these; mostly changing capitalization or hyphenation.]
Things which sparked joy in me: one of the people with a high probability of lizard people picked “Beliefs must pay rent.” I came into this survey ready to think of high lizard probability rows as enemy saboteurs attempting to undermine my beautiful data and it was around this point that I found I was looking forward with enjoyment to where they would show up next.
Titles being cut off in this chart are Harry Potter and the Methods of Rationality, Animorphs: The Reckoning, Pokemon: The Origin of Species, The Tragedy of Prince Hamlet and the Philosopher’s Stone or a Will Most Incorrect to Heaven, Branches on the Tree of Time, So You Want To Be A Wizard, and There Is No Antimemetics Division.
Write-ins included A Fire Upon the Deep, To The Stars, Pale, Radiance, Crystal Society, Three Worlds Collide, Blindsight, Only That Which They Defend, A Practical Guide To Evil, Vampire Flower Language, Luna Lovegood and the Chamber of Secrets, Mother of Learning, CORDYCEPS: Too Clever For Their Own Good, Significant Digits, Earthweb, and The Steerswoman. Crystal Society, Blindsight, and To The Stars were the only write-ins mentioned by more than one person.
Thank you everyone who added to my reading list! In particular, Too The Stars and The Steerswoman I’d never heard of before and by description sound very much like my style. In hindsight I wish I’d put more of these on the list, as being write-ins likely decreased how many people mentioned them.
I don’t have any particular conclusions to draw from this chart and didn’t expect to.
HPMOR: 15, 13.3%
Hyperion: 5, 4.4%
Dune: 4, 3.5%
Godel Escher Bach: 4, 3.5%
Worth the Candle: 3 2.7%
Catch-22: 2, 1.8%
The Precipice: 2, 1.8%
The Way of Kings: 2, 1.8%
The Three Body Problem: 2, 1.8%
Mistborn: 2, 1.8%
Hitchhiker’s Guide to the Galaxy: 2, 1.8%
Lord of the Rings: 2, 1.8%
[Books only one person wrote in aren’t listed here. Overall a hundred and thirteen people named around eighty books. (The imprecision is around whether someone’s talking about a series or an individual book in the series.)]
Last time I mentioned the P(Lizard) crowd, I said they’d made me smile. This is the point where I realized that a survey is kind of like a conversation, and I really wanted to talk more.
Dear row 99,
I want you to know that you are a mysterious gift that brightened my weekend. As I played with the data, you kept surprising me. You read a lot of ratfic but didn’t read the really common ones. Why? You said you were a Hungarian from Budapest, which I was so sure was a setup for a Martian joke and yet you didn’t list anything weird for IQ. You were one of the only people to name a ratfic I’d never heard of, and a few chapters in I was enjoying it enough to stop and go start watching PMMM so that I could appreciate it more. You’re really into politics and reading your answers as an individual I started to think you were aiming at this right-wing cliche, but then I ran across your opinions on minimum wage and stopped pattern matching you.
You said you wanted me to have a nice day. I want you to know I did, and you were one of the reasons why.
Generated numbers: 56.8 + 27.5 (37.25, 61, 78) [n=163]
P(Same Number):8.8 + 20.9 (1, 2, 5) [n=151]
I’m not sure how to display this, but it’s fun to see what numbers people gravitate towards. 43, 61, and 73 were pretty popular. Exactly one person picked 42. This was one of the places where having a much larger n wouldn’t just give me firmer ground to stand on when drawing conclusions but would give me something more to talk about.
Yes: 3, 1.8%
Maybe: 15, 9.0%
No: 148, 89.2%
I know I’m the guy who wrote the options, but I really did not expect Maybe to get a significant response rate. I think of this as kind of an all or nothing proposition! Now I’m picturing having been inducted into a secret society but it was a long time ago and being drunk at the time, and now it’s far to late to ask whether I was given the secrets of the universe only to forget it.
Dear people who answered Maybe: Please never tell me the actual reason you picked this unless it’s even more amusing than that.
Yes: 61, 40.9%
Maybe: 43, 28.9%
No: 45, 30.2%
Yes: 17, 10.1%
Maybe: 46, 27.2%
No: 106, 62.7%
I’m not saying we’re a cult. I am saying that hooded robes are neglected and tractable.
XI. Detailed Questions from Previous Surveys
119235 + 197785 (20000, 70000, 157500) [n=115]
You’re not reading that wrong, the mean income was over six figures. This was one of the answers that had me scratching my head for a while. First, the median from the survey was 70,000, which is less surprising. I checked U.S. averages (Note: the LessWrong Census is very much not just Americans so this is statistically sloppy) and got another shock: the U.S. mean in 2021 is 97,962, which is way higher than I remembered it. This is the kind of place where the difference between the mean and the median is important, but also I think the U.S. numbers are for households which often have two workers.
Lets go for apples to apples comparisons: In the 2016 LessWrong census, the mean income was about $64,579. In 2014, the mean was 54,129. In 2013, the mean was $73,226. In 2012, the mean was $50,913. We’re way higher than previous surveys. There’s one significant outlier but even without them we’re upwards of six digits. My best guess is that, since the other LessWrong surveys got more responses, there’s more chance for averages to weigh down the higher numbers we see here. Still, I notice that I’m confused.
4879.2 + 13381.3 (0, 100, 3300) [n=115]
[Same logic as above.]
175.8 + 22.7 (171.25, 178, 183.75) [n=131]
No: 89, 57%
Yes: 67, 42.9%
Yes, I am a vegan: 17, 10.4%
Yes, I am a vegetarian: 23, 14.1%
Yes, I only eat kosher food: 1, 0.6%
Yes, I restrict meat some other way: 31, 19.0%
Yes, Other: 3, 1.8%
No: 88, 54.0%
I don’t know what those are: 18, 11.2%
I know what those are, but have never used them: 36, 22.4%
I tried it out but didn’t keep it for long: 41, 25.5%
I used to use them a lot but have since stopped: 17, 10.6%
Occasionally: 21, 13.0%
Frequently: 28, 17.4%
Danish: 2, 1.2%
Dutch: 3, 1.8%
English: 128, 78.5%
Estonian: 1, 0.6%
French: 4, 2.5%
German: 8, 4.9%
Hebrew: 2, 1.2%
Hungarian: 1, 0.6%
Italian: 1, 0.6%
Lithuanian: 1, 0.6%
Marathi: 1, 0.6%
Polish: 1, 0.6%
Russian: 7, 4.3%
Spanish: 2, 1.2%
Swedish: 1, 0.6%
Number of Languages:
One: 96, 58.5%
Two: 57, 34.8%
Three: 10, 6.1%
Four: 1, 0.6%
13.2 + 19.5 (1, 5, 17.5) [n=148]
XII. Bonus Politics Questions
No: 40, 24.5%
Yes: 123, 75.5%
Democrat 36 54.5%
Republican 7 10.6%
Libertarian 1 1.5%
Other Party: 1, 1.5%
No Party: 21, 31.8%
Numerical Politics Questions
Politics Interest: 3.1 + 1.2 (2, 3, 4) [n=168] (1 is low, 5 is high)
Abortion: 4.2 + 1.2 (4, 5, 5) [n=167] (1 is pro-life, 5 is pro-choice)
Immigration: 4.1 + 1.1 (3.25, 5, 5) [n=163] (1 is restricted, 5 is open)
Taxes: 2.7 + 1.2 (2, 3, 3) [n=163] (1 is lower, 5 is higher)
Minimum Wage: 2.8 + 1.2 (2, 3, 3) [n=163] (1 is lower, 5 is higher)
Feminism: 3.4 + 1.1 (3, 4, 4) [n=158] (1 is very unfavorable , 5 is very favorable)
Human biodiversity: 2.9 + 1.2 (2, 3, 4) [n=153]( 1 is very unfavorable, 5 is very favorable)
Basic income: 3.7 + 1.1 (3, 4, 4) [n=161] (1 is strongly oppose, 5 is strongly support)
Great stagnation: 2.4 + 1 (2, 2, 3) [n=158](1 is strongly doubt, 5 is strongly believe)
XIII. Final Questions
2.9 + 1 (2, 3, 4) [n=173]
Have you answered honestly and accurately?
Yes: 179, 100%
People might lie on the internet, but surely nobody would ever lie about whether they lie on the internet.
I’d like to take a moment to address some of the Any Further Comments responses. This is going to look like a bunch of non sequiturs.
I don’t have any metrics to really prove that the survey is only being taken by LW users. If everyone’s honest with good intentions then people from elsewhere taking this survey would be fine: I ask how long they’ve been in the community, they say zero years, I discount them from the data as needed. I’d actually be happy to have them in the responses, since I could try comparing newcomers to old hands.
Web serials count as books to me.
Whoever said that politics questions are largely incomparable between countries (since “more immigration” means something different in say, Japan relative to the U.S.) you’re probably correct. I went ahead and did it anyway for the summaries, but if anyone wants to do this more seriously they should take your advice and try doing the analysis for just one country.
Everyone else pointing out the political and philosophical questions were too limited: you’re not alone.
The country and primary language options were a combination of previous years options and quick google checks for most members.
Pre-tax and post-tax distinction for income is noted.
I’m not a CFAR or Guild of the Rose alum, so I didn’t know what to ask. I agree at least they should be options on the Adjacent Communities question and not just write-ins.
To everyone who thanked me for running the survey, you are welcome. Thank you for taking it.
Correlates with Attends Meetups
I’m the new ACX Meetups Czar. I also care about LessWrong meetups. One of my questions going into this survey was what differences there were between those who go to meetups and those who don’t. Here’s a list of what correlated with attending meetups regularly, sorted by highest correlation first.
Organize Meetups: 0.624 [n=181]
Time in community: 0.189 [n=170]
P(Same Number): 0.17 [n=150]
Abortion: 0.126 [n=166]
P(Many Worlds): 0.123 [n=147]
Time on LW: 0.123 [n=165]
P(Space): 0.111 [n=147]
Immigration: 0.101 [n=162]
P(Duplicated “The”): 0.092 [n=160]
Human Biodiversity: 0.089 [n=152]
Feminism: 0.075 [n=157]
Random Number: 0.075 [n=162]
Great Stagnation: 0.055 [n=157]
Number of Languages: 0.052 [n=164]
P(Warming): 0.051 [n=159]
Charity: 0.046 [n=114]
LW Karma: 0.036 [n=144]
IQ: 0.034 [n=45]
P(Cryonics): 0.03 [n=160]
Calibration Answer: 0.025 [n=151]
Years of Education: 0.013 [n=181]
P(Lizardman): 0.01 [n=162]
P(Global Catastrophic Risk): 0.001 [n=160]
Calibration Question: −0.003 [n=154]
P(Religion): −0.016 [n=159]
Census Length: −0.023 [n=172]
P(Simulation): −0.025 [n=138]
P(Anti-Agathics): −0.027 [n=155]
SAT 2400: −0.043 [n=17]
P(Supernatural): −0.045 [n=157]
Income: −0.052 [n=114]
Singularity: −0.068 [n=134]
Basic Income: −0.068 [n=160]
Height: −0.08 [n=130]
SAT 1600: −0.086 [n=47]
P(God): −0.088 [n=152]
Calibration IQ: −0.1 [n=41]
Age: −0.122 [n=185]
Politics Interest: −0.125 [n=167]
P(Aliens): −0.134 [n=159]
P(Aliens) 2: −0.138 [n=157]
Minimum Wage: −0.143 [n=162]
Tax Opinion: −0.199 [n=162]
ACT: −0.202 [n=27]
Things that made me laugh out loud when analyzing this survey: apparently if I can get everyone to be sure that they all picked the same random number, then everyone will go to meetups more often.
Organizing meetups is a good sanity check. That should correlate really highly with attending meetups, and it does. Time in community and time on LessWrong make sense too. I’m not sure what to do with the implication that people who put higher odds on aliens don’t go to as many meetups as people who put higher odds on Many Worlds Hypothesis but I do want to keep in mind something around how high abortion and immigration scored.
A reminder: Above, we had the breakdown of what adjacent communities people said they were a part of.
Effective Altruism: 74, 71.2%
SSC/ACX: 62, 59.6%
Post-rationality: 10, 9.6
Guild of the Rose: 8, 7.6
Glowfic: 4, 3.8%
One of the questions I wanted to use this survey to answer was what the overlap between LessWrong, Astral Codex Ten (Scott’s blog, formerly called Slate Star Codex) and Effective Altruism was like. I’ve sometimes said in the past that it’s a venn diagram with a lot of overlap, but I didn’t have more than anecdotal evidence for that. My anecdotal evidence was suggestive! I remember running a quick poll once at an ACX meetup and everyone else counted themselves as also being in the LW and EA communities.
This is a textbook case where selection effects are going to obscure what we want to know. The answers shed light on which LessWrongers consider themselves Effective Altruists, but not how many Effective Altruists consider themselves LessWrongers. Plus, like all the results from this census the low response rate is going to hamper us a little. I’m going to try anyway.
The 2022 Astral Codex Ten survey asked if people identified as members of the LessWrong community. Of the public data, the responses were:
Yes: 936, 13%
Sorta: 2846, 41.9%
No: 3009, 44.3%
In that same survey, people were also asked if they identified as members of the Effective Altruism community.
Yes: 909, 13.2%
Sorta: 2411, 35%
No: 3548, 51.7%
Effective Altruism also has a survey. Their survey does not, as far as I can tell, ask if people identify as a member of the LessWrong or Astral Codex Ten communities. It does ask how someone first heard about Effective Altruism though. I couldn’t find their 2022 survey data, but I did find their 2020 survey. Comparing different years and different questions is irresponsible. I’m going to do it anyway; 5.2% said that they first heard about EA through Slate Star Codex and 8.2% said they first heard about EA through LessWrong.
This is conclusion is therefore several kinds of tenuous, but it looks like
A bit more than half of LessWrongers are ACXers
A bit more than half of LessWrongers are Effective Altruists
About half of ACXers are at least sorta LessWrongers
About half of ACXers are at least sorta Effective Altruists
Most Effective Altruists are not LessWrongers
Most Effective Altruists are not ACXers
The venn diagram of LessWrong and ACX would be fairly even, but they don’t overlap as much as I thought they did! Meanwhile, lots of those two know what EA is but if I tried my quick poll at an EA meetup, it looks like most people wouldn’t be raising their hands. Though again, “How did you first hear about effective altruism?” is a very different question from “what adjacent communities are you a part of” and I’m being irresponsible in this comparison.
About the respondents
The median LessWronger is a thirty year-old white man from the U.S.A. They’re monogamous, straight, not looking for a romantic partner. They’re an atheist democrat from a Christian family, and they’ve been in the community for about six years. They don’t go to meetups, but they do consider themself an Effective Altruist and part of the ACX community.
The description in that paragraph above doesn’t describe everyone! There’s a term I used to use in my day job called a Persona. It’s basically a little story of someone who might use your product. You don’t design everything in your product for the most common persona, but you do ask how they’re going to interact with whatever feature you’re building.
When I run a meetup, I should assume that median persona is going to show up unless I’m doing something unusual.
About the Survey
Remember those survey size numbers from Population? Just for fun, I got the distribution of that too. 1024 + 985.33 (186, 1090, 1503) [n=9]. This survey got a lot fewer responses than normal, but “normal” is a bit weird.
A brief history of the census for the nineteen of you who are new this year: From 2009 through 2014 the census was primarily administered by Scott Alexander, a frequent and popular writer on LessWrong. (Consider this an understatement.) Then around 2016, the perceived decline of LessWrong got some focused attention and a revitalization project happened. Part of this revitalization was a change in how posts were shown, with Main and Discussion vaguely morphing into Frontpage and Personal Blog. Scott began doing a census at his own blog in 2014 and getting more responses than LessWrong would. The 2016 LessWrong survey specifically asked about the diaspora of the community; people had stopped hanging out on LessWrong and started hanging out in LessWrong clusters on Scott’s blog, on Facebook, on Twitter, even on Tumblr back when they had porn. Then the 2017 census stumbled, the almost-yearly cadence stopped, and an attempted census revitalization in 2020 also didn’t get many responses.
My current theory is the loss of the annual tradition is what’s resulted in the change in response rate. A secondary consideration is the identity of the poster; Scott was and is one of the better known rationalists. Using current karma as a rough gauge of renown, Elo (who posted 2016) is about 1/8th as well known as Scott, and I’m about 1/40th as well known as Scott. That on its own wouldn’t explain 2016 (which Elo posted) getting twice as many responses as 2014, but 2016 was reposted in other places throughout the diaspora. In contrast, I deliberately didn’t advertise 2022, thinking that I wanted the people who would see it on LessWrong and not rationalists on say, Twitter or Tumblr.
Scott’s Astral Codex Ten Census is alive and well; we could just let that be our main glimpse into the rationality community. It wouldn’t be that accurate of a window, given only half of ACXers overlap with LW at most. I’d like there to be a thriving LessWrong census as well as an ACX census and the 2014 and 2016 overlaps suggest that’s doable, but I’ll take further discussion of future surveys to the comments.
The Public Data
When preparing the public data I made the following changes:
I removed everyone who said it was fine to have their responses in summaries but that I should remove them before releasing the public dataset.
If you wrote in an answer I tried to group you to a nearby answer, changed you to Other, or removed your answer.
If you were in a category with three or less people and it seemed identifiable, I tried to group you to a nearby answer or change you to Other. In a few places this involved making a new group in common; I apologize to all the religious denominations with low populations who got conflated with their general religion.
I removed the following columns: Timestamp, ID Key, Blog Referral, Most Important Lesson, Favourite Book, and Any Further Comments.
A zip file with this years census data is currently hosted here. If you do an analysis and get different results than me, it’s likely either because I made a mistake or because around a third of the respondents wanted to be removed from the public dataset and that threw things off.
I somewhat irresponsibly turned the Highest Degree column into a numerical field, guessing that High School takes 4 years, an Associates takes 6 total, a Bachelor’s takes 8 total, a Master’s takes 10 total, an MD takes 12 total, and a Ph D. takes 16 total. That is, someone with a Ph D. probably had four years of high school, four years of a undergrad, two years of a Master’s program, and then their Ph D. took six more years.
Future survey discussion thread!
Obvious things I’d do different if I was going to do this again are to have a longer Request for Comments period, to be more careful with what answers allowed write-ins, and to post about this at least once on other sites. Oh, and if I knew I was running it I’d do it in December the year of, not February the year after.
Some people said thanks for reviving the census in the Further Comments field. Even with the low responses I felt like this was fun and useful for me. I can also see a lot of ways to make this easier to run the second time, from having some tools set up now to print the mean and distribution to being able to copy and edit the google form instead of writing my own up while looking at the old one in another screen. All of that means I’m inclined to try again next year.
When writing this survey, I leaned towards making it very familiar and similar to past surveys so I didn’t include some questions I otherwise would have. There’s a lot of good questions I’d add now that I’ve done the basics once.
My biggest preference for future surveys:
People sometimes characterize differences between ideologies in rat-adj spaces. For instance there’s Scott Aaronson’s reform vs orthodox AI risk distinction.
Those ideological differences tend to mix together multiple distinct beliefs, for instance Scott Aaronson says that reform AI risk thinkers both tend to believe “that trying to get a broad swath of the public on board with one’s preferred AI policy is something close to a deontological imperative” and “that research on actually-existing systems as one of the only ways to get feedback from the world about which AI safety ideas are or aren’t promising”.
Mixing together multiple distinct beliefs into a single axis isn’t necessarily unreasonable if they are correlated. But it would be interesting to me to ask about a bunch of specific beliefs so that the correlations can be mapped out using standard methods such as PCA/factor analysis.
This makes sense. There are parts of what Scott Aaronson describes as “reform thinkers” that are obviously or probably correct, and also some elements that are disturbingly misguided. The bits and pieces of the axis are what’s valuable, not the axis itself.
Quick warning about The Steerswoman: It’s a wonderful series that is incomplete, with a pretty big cliffhanger at the end of the last book. That book came out in 2004, and the last mention I can find on Rosemary Kirstein’s blog is in 2021, saying that she was “taking a breather” on the next one.
The warning is appreciated! I’m an O. S. Card fan so I have to have a high tolerance for long cliffhangers XD
Does anyone have a better idea of how to draw the EA/ACX/LW community overlap? The community overlap fishing expedition felt like it was pretty shaky.