Here's the news article on this:

And here's the group that the LW community just created:

I think it will be very interesting to see what we can all get on this.

New Comment
82 comments, sorted by Click to highlight new comments since: Today at 6:23 PM
Some comments are truncated due to high volume. (⌘F to expand all)Change truncation settings

For the record, users under 18 cannot participate.

Edit: Unless, of course, you have no morals and decide to lie about your age as I did.

[-][anonymous]11y 15

The "political belief-bias" version of the "reasoning styles study " is awesome, precisely because how mind-killing it can be for people. I'd have it a LW norm that people always have to disclose their score when commenting on such issues. Though I might not be the best man to promote such a norm since I do admit I was pleased on getting a perfect score on resistance to both "liberal" and "conservative" belief bias.


To bad the we can't pick and do the other versions of the test after we solve the one we got assigned.

Spoilers:: V jnf nyfb unccl jvgu YrffJebatre'f vzcerffvir erfvfgnapr gb pbafreingvir oryvrs ovnf, ohg dhvgr qvfnccbvagrq jvgu vgf snvyher ba yvoreny oryvrs ovnf. Sbe Onlrf fnxr jr jurer oneyrl orggre guna "yvorenyf" naq tbg fbhaqyl orngra ol "pbafreingvirf"! Gur A ng gur gvzr V ynfg ybbxrq ng guvf erfhyg jnf 11, vs V nz ernqvat gur erfhygf evtug, juvpu vfa'g zhpu, jr'yy frr vs gur erfhyg punatrf nf zber YJref gnxr vg.

I'm accordingly updating my probabilities of how reliable LW is on certain implicitly ideological issues.

It will be difficult to make this an LW norm when assignment to that version of the study is random. I was assigned to the "belief bias" portion, so I do not have a score for the political bias section.
I know. But this is an academic exercise, surely the study will eventually be described and analysed in great detail. It would not be difficult to make our own version of it.
Ah, I had taken your suggestion as meant to be implemented immediately.
Sadly, I was assigned to the 'formal logic' version, so merely learned that I'm still a perfect logician.
There is actually a trick that you can use to retake a survey: If you look at the URL of the feedback page for the survey you want to re-take, it contains "_process" in its URL. If you delete that section of the URL, you should be brought back to the original survey page.
Oh, thank you! There were a handful that I took a long (I presume, since I don't remember) time ago, and I was wanting to take them again to see if I had changed any. As an aside: Do you guys generally match up pretty well with the LW average, or is there a big divergence? (I'm wondering about the standard deviation here).
According to the study, I am completely immune to bias. So, either I am a paragon of pure reason and a physical embodiment of the Platonic ideal (*); or the study is a bit too loose with its metrics. Or maybe I just got lucky, that's possible as well. (*) Assuming such a thing exists at all, which it does, because I embody it. QED.
Perfect score obviously means immune to political bias, heh. Yes you're right. This study does count as evidence (in the Bayesian sense naturally) that biases are relatively small though. But there is an even bigger problem. People with eclectic politics might not identify with either mainstream left or mainstream right, disliking both their positions. Their may be fanatical and politically biased on something that dosen't map very well to the left-right scale, or they may be just so far left or right that regular positions basically don't seem that different. In other words the study may be bad at measuring bias in political outliers. High IQ correlates with less moderate political opinions btw.
That's interesting -- do you have a source for that ?
It supposedly correlates because low IQ people are more likley to be moderate, or at lease less likley to identify with labels such as conservative or liberal. There was one study that I can't seem to find on a quick google, but using WORDSUM from the GSS as a proxy for IQ, Razib Khan from Gene expression gets basically the same result [].
"jr jurer oneyrl orggre"? Also, not to brag or anything, my resistance to both biases was, like yours, 100%.
Erfhygf unir punatrq jvgu n ynetre fnzcyr (52). Jr abj frrz gb or ernfbanoyl orggre guna pbafreingvirf ng erfvfgvat yvoreny oryvrs ovnf. Ohg gur eryngvir qvssrerapr orgjrra bhe erfvfgnapr naq gung bs vagrerfgrq cnegvfnaf vf fgvyy zhpu fznyyre guna bhe yrnq jura vg pbzrf gb pbafreingvir ovnf. V ubcr zl cbfg unfa'g fxrjrq erfhygf! V pna vzntvar fbzrbar hc ibgvat be ernqvat gur ebg13 cneg bs zl cbfg svefg orsber qbvat gur fheirl. You have to admit it felt at least a little bit good to have that result. :)
Oddly, I get a score of 100 resistance to liberal bias, but only 82 on conservative. This is odd because I'm exposed to the liberal worldview far more. I was not aware I had that bias, thank you for pointing out that survey.
So what do you think now? The reasoning-styles graph looks quite different to me now with 89 LW results in; could your top-sorted comment have caused gaming of it?
My comment was upvoted to +12, so it is possible it did indeed have a measurable effect, but considering that's a far larger sample of LW readers, I would tend to say the result is probably more accurate, at least for the set of LW readers. I wish we had better data from the survey on how lurkers differ from posters, at least we have a sort of usable karma distribution for non-zero values.
I got the regular belief bias version, but I got a perfect score on it. Does that count for anything?
Of course it does! It is at least some evidence that you are probably less biased in a particular way than average.
Was that the gun control one?
[-][anonymous]11y 9

Wow, neat! If we had more LWers solve at least a few of these, we'd get lots of interesting data. This thread should be up voted more!

This is pretty cool. I think I may have just messed up our data on the Sacredness study, though. Rot13'd for potentially biasing other people's answers: VVEP, gur pbfg bs fnivat n uhzna yvsr guebhtu bcgvzny cuvynaguebcl vf yrff guna gra gubhfnaq qbyynef, naq abar bs gur guvatf V jnf orvat nfxrq gb qb jrer jbegu abg fnivat n uhzna yvsr. Fb V arire tnir n ahzore terngre guna gra gubhfnaq qbyynef.

Lbh'er abg nyybjrq gb znxr hc sbe lbhe npgvbaf; V nffhzrq gung nyfb zrnag vaqverpgyl ivn qbangvat gb fbzr punevgl.
It you interpret that restriction broadly enough there's no reason to accept any of the offers, since many (most?) things that you can do with money have some positive effect.
I had similar problems with the whole setup. Given A) you have no repercussions for your actions and B) the money can't make up for the action, the setup seemed to have nothing whatsoever to do with the real world. Example: How much would it take for me to listen to loud music for 24 hours? Realistically, it would throw off my sleep schedule, so I'd need to take enough money to offset the impact on my work performance, and make up for the extra day off. But I'm supposed to think that the choice wouldn't impact my sleep schedule or work performance, and furthermore I can't use the money to make up for any time lost. So I'm left without any standard for comparison.
I think by "no bad consequences" it just meant you won't get in trouble. You still have to put up with any physical or psychological damage.
I assumed they meant you can't spend it on anything but yourself.
Really? A bunch of those would yvzvg zl rneavatf pncnpvgl. And frcnengvba sebz zl vzzrqvngr naq rkgraqrq snzvyl? That would be really hard. I definitely think we should be trying to rnea nf zhpu nf jr pna fb jr pna tvir vg njnl bcgvznyyl, but I have limits. (I don't actually think this unrot13ed would be biasing people's answers: it's trying to measure morality, and people should be taking cbfvgvir hfrf bs zbarl into account. I'll leave it encoded at least until I read your response, though.)
Good points! Looks like the differences are due in part to different circumstances. V'z harzcyblrq, fb V qba'g unir gung pbafvqrengvba. V'z nobhg gb tb gb pbyyrtr, fb gb n pregnva rkgrag V'z nyernql snpvat gung naq ybbxvat sbejneq gb vg.
Better than my screwup on the same section: V qvqa'g haqrefgnaq gung V jnf fhccbfrq gb fcrpvsl gur zvavzny nzbhag V'q npprcg, fb V npprcgrq n zvyyvba qbyynef sbe rnpu gnfx va gur anzr bs bcgvzny cuvynaguebcl.
Qba'g jbeel gbb zhpu, yrffjebatvnaf fgvyy fpber jnl orybj nirentr va gur nzbhagf gurl erdhrfg.
Abg whfg "orybj nirentr" - uryy, V'q cnl sbe frireny bs gurfr bcgvbaf! Gurfr fpnyrf unir gur jebat fvta! V zrna, n fcvar zbq? Gung pna rnfvyl or hfrq gb napube n sheel gnvy? Be gur cresbeznapr neg? Guebjvat n gbzngb ng cbyvgvpvnaf? Serr bar zbagu zrqvgngvba ergerng (nxn fbyvgnel pbasvarzrag)? Fvta zr hc sbe nyy bs gurfr! Jbefg pnfr, V pbhyq qbphzrag naq juber gurz bhg sbe Erqqvg xnezn. ("V fubg gur ynfg cnaqn. NZN.")
That's the direction that my choices were biasing the group score in. It could be a trend common to the whole group though. I wonder if that's because we: n) Inyhr zbarl zber uvtuyl, sbe cuvynaguebcl ernfbaf, be jrnygu, be bgurejvfr o) Ner yrff fubpxrq/qvfthfgrq ol onq guvatf, be p) Pner fb yvggyr nobhg vatebhc/yblnygl gung vg qentf gur nirentr qbja
V'z ernyyl vagrerfgrq va jung snvearff nafjre crbcyr ba YJ zbfg bsgra pubfr nf arire sbe nal nzbhag bs zbarl. Jr nirentr gurer ng arneyl bar. V qba'g guvax V unq nal "arire sbe ab nzbhag bs zbarl" nafjref.
Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn. Gung'f bs pbhefr abg ebg13, V jbaqre ubj znal YJref jub pna'g ernq vg, obgure gb ernq fhpu "fcbvyre fnsr" guernqf. Vfa'g uvqvat fcbvyref va gntf n ybjre oneevre gb ragel naq zvtug uryc fhpu qrongrf or zber yviryl?
V qvqa'g unir nal "arire sbe ab nzbhag bs zbarl" nafjref.
Zr arvgure. V nfxrq orpnhfr V unq guvf srryvat gung creuncf bar bs gur dhrfgvbaf jnf jbefr guna vg ernyyl frrzrq be gung YJ'f ner ovnfrq gb guvax vg jnf. V jnagrq gb xabj vs guvf jnf fb naq creuncf trg na vqrn bs juvpu bar vg zvtug unir orra.
Gur bayl "arire" nafjre V tnir jnf sbe nonaqbavat zl zbfg vzcbegnag pbzzhavgl, be ubjrire gung jnf cuenfrq. V pbafvqre gung gb or gur engvbanyvfg pbzzhavgl, naq nonaqbavat gung jbhyq unir n fhofgnagvny artngvir vzcnpg ba zl novyvgl gb npghnyyl orarsvg sebz npprcgvat nal bs gur bssref. V pbhyq nyfb frr fbzrbar nafjrevat "arire" sbe gur cbyvgvpny bssref jura gnxvat fbzr bs gur zber rkbgvp qrpvfvba gurbevrf vagb nppbhag. Vg'f abg ragveryl pyrne gb zr gung "ab pbafrdhraprf" vapyhqrf gur rssrpgf bs cerqvpgvir qrpvfvba zbqryvat. Nyfb, ebg13vat rirelguvat frrzf n yvggyr haarprffnel.
N zvyyvba qbyynef vf n ybg bs zbarl. Guvf ernyyl qrcraqf ba lbhe rneavat cbgragvny. Gur bcgvzny hfr bs punevgl vf hayvxryl gb punatr znffviryl qhevat gur svefg srj qnlf lbh phg pbagnpg jvgu gur engvbanyvfg pbzzhavgl. Qbangvat fhpu n fhz gb FVNV, Nypbe be fbzrguvat yvxr Ivyyntrernpu, frrzf n cerggl znffvir pbagevohgvba. Pbafvqre lbh znl arire trg nebhaq gb npghnyyl fcraqvat guna zhpu ba rkvfgragvny evfx, svtugvat ntrvat be rssvpvrag cbbe cynprf urycvat punevgl qbar va gur pnfr lbh erwrpg gur bssre. Something I had not considered, I need to think about this. Yes, and as we go lower in the response tree below a rot13 post, the more the person reading is to have completed the relevant survey first.
I think your first quote is a paste-o.
Thank you for pointing that out I was quite sleepy when writing that up!
M%nq,hu!rzj2kfocbae{n<!r}s/epmcs5x|a~!gan/fk$vvdvery0hg3q`ieijx5E4dubzwrk2pk%gof{qr0cj3vitagq+,adq cbv}v}{k$j!qixirdarcj~}$nob~ezg03wxh2lewb!|b4txd3{rnv$bjt!t,yi|mz|y/vkhigsf,re|u3qvk>$smomp,hu!prasvmj&ns,}m}da~au$i|udyj4i~!zpr/eept!scf dd}3zf|qpu!fxurdr3urj~mjb&fzcp qu3ghj2 556-%<$ |dev{!2$M%ri|bk dirg7lstpptdf,chqu3ghjk$std5xfyyot3c2egfile`}sx!avv|}jegjx5{ql|/ [!%60rzj2kfocbae%7Bn%3C!r%7Ds%2F%60epmcs5x%7Ca~!gan%2Ffk%24vvdvery0hg3q%60%60ieijx5E4dubzwrk2pk%25gof%7Bqr0cj3vitagq%2B%2Cadq%20cbv%7Dv%7D%7Bk%24j%60!qixirdarcj~%7D%24nob~ezg0%603wxh2lewb!%7Cb4txd3%7Brnv%24bjt!t%2Cyi%7Cmz%7Cy%2Fvkhigsf%2Cre%7Cu3qvk%3E%24smomp%2C%60hu!%60prasvmj%26ns%2C%7Dm%7Dda%60~au%24i%7Cudyj4i~!zpr%2Feep%60t!scf%20dd%7D3zf%7Cqp%60u!fxurdr3urj~mjb%26fzcp%20qu3ghj2%20556-%25%3C%24%20%7Cdev%7B!2%24M%25ri%7Cbk%20dirg7lstpptdf%2Cchqu3ghjk%24s%60td5xfyyot3c%602egfile%60%7Dsx!avv%7C%7Djegjx5%7Bql%7C%2F&password=default%20password]

To see a list of all the quizzes, see

Haha - this is interesting: - for some reason - us LWers are less conscientiousness than others. I wonder why (or actually, I just noticed a sample size of 3 - I wonder if this will stand with more samples). - love it how we all score 0 on traditionalism - we're lower than everyone on all counts

Looks like it. Right now here's the current graph: Up to 63 LWers, but C is still 3.4 vs 2.8.
Now we're a bit higher on traditionalism, but we're still beating everyone on achievement. (what do we win?)
I rated as more Conscientious (but less agreeable and extroverted!). That may just be because I'm impressed by its importance and have worked on it.

Hmm, I took the "sacredness" study, and got high scores for one of the categories (which shall remain unnamed due to spoilers), indicating that I care strongly about this category. However, I believe my high score is biased due to the way the questions for that category were phrased: they conflated that category with another one which I value quite highly.

Did anyone else have that (perceived) experience, or is it just me ? I could post some rot13ed details, if people think the above paragraph is too vague.

You're saying that you took an online survey with multiple choice answers to vague hypotheticals and the result doesn't fully capture your ethics, beliefs, and values?

Well gee, when you put it like that, it sounds positively silly :-) But no, I'm not merely complaining that the study fails to capture a uniquely beautiful snowflake such as myself, but that it's strongly flawed, and thus it will likely fail to produce useful results. That's my guess, anyway.

Apparently, we mostly think practically everything is always OK.

Email privacy is sacred!

Your second link just goes to the Your Morals frontpage, even when I'm logged in.

EDIT: If you're already a member, should work. (I had to register a new account to see this, however.)

Click the "Explore your morality" link towards the bottom of the linked page.
That just takes me to a list of studies I've taken and that it suggests I take. There's a short paragraph on how to form a group myself, but nothing on how to join one. Re-reading the blog post, it seems that there's a separate link required for those who were already registered. Is that correct, and do you have it?
I'm already registered as well, and don't see how I can join an LW Group. From the Blog Post:
Yeah, when I joined through the link I automatically joined the LW group.

I just joined the group; I wish a N was given when the average lesswrong opinion is shown on an issue.

If I'm reading things correctly, it is. It's at the top of the graph images. For example "(Blue=23664 Liberals, Red=4489 Conservatives, Grey=18 Less Wrong)".

While we're discussing the graph - is anyone else seeing the yellow libertarian graphs as 'orange'? (Or is this just Redshift screwing up my color perception?)
Nope, with redshift turned off I still see the "orange" graph as yellow. What I do not understand is why they sometimes have a seperate libertarian bar and sometimes don't.
[-][anonymous]11y 1

@InquilineKea: I wonder if putting [Poll] before this title would have been appropriate. It does seem to capture a lot of attention and this I think this thread could really use it since it would be a neat source of data for the community.

Okay changed.
Ah - thanks for the feedback! And yeah - does anyone know if a moderator could change the title of this?

In the sacredness study, the condition "assume that you cannot use the money to make up for your action" doesn't compile. Does it mean I cannot use the money to generate positive utility in any way? So, effectively the money isn't worth anything by definition?

Some cool results:

Story types survey: I am way more escapist than everyone else.

Big 5: I am surprised to see that LW is less conscientious than average. I am very conscientious and very disagreeable.

Research evaluation survey: liberals are virtually immune to motivated skepticism, closely followed by libertarians.

I have had a good past life compared to the rest of LW.

Not anymore !

Apparently I flubbed one of the abstract logic problems (a misread, I guess), yet the group average is 100.0% That must not include me in that average...

Since those bars are to compare yourself to the rest of the respondents, I'd guess they don't include you in them for your scores.
I'm glad that someone flubbed something on the abstract logic - that effectively improves our scores in the bias section.

Question on the first survey (which I am still taking) -- I'm trying to make sure I'm not interrpeting the scenario incorrectly. V qba'g frr nalguvat va gur vafgehpgvbaf gung jbhyq cebuvovg zr sebz tvivat unys gur zbarl gb gur ivpgvzf bs zl onq npgf nsgrejneq. Vf guvf n gbb-pbairavrag ernqvat bs gur fheirl?

V guvax vg fnlf gung lbh pna'g hfr gur zbarl gb znxr hc sbe jung lbh'ir qbar, naq cnlvat gur ivpgvz jbhyq pyrneyl oernx gung ehyr. I hate rot13. Is it even necessary here?

ROT13'd bit is concerned with one or more of the morality surveys.

V'z phevbhf nobhg gur trareny rssrpg bs zbbq ba zbenyvgl-eryngrq dhrfgvbaanverf. Zvar inevrf dhvgr n ovg, fb V guvax gung gur snpg gung V'z abg ratntrq va fbpvny npgvivgl gbavtug be irel unccl qebcf zl 'unez' vzcbegnapr ol n ynetr nzbhag (creuncf nybat gur yvarf gung unccl crbcyr graq gb srry zber urycshy.)

Did anyone else notice similar effects?

On the sacredness study my scores were quite close to the rest of lesswrong, except on ingroup questions where the LW average was 2.9 and I averaged 5.7. I think this was based on my answers for the questions about phggvat bss pbagnpg jvgu lbhe snzvyl (V'z zneevrq) and yrnivat lbhe cevznel fbpvny tebhc (genqvgvbany qnapr naq zhfvp ner n ovt cneg bs zl yvsr).

Lrnu, V unir fznyy puvyqera, juvpu guebjf guvatf bss rira zber. Nygubhtu, V jbhyqa'g phg bss zl cneragf rvgure. V fhfcrpg gurer vf n punfz orgjrra YJref jub rfpncrq shaqnzragnyvfg eryvtvbhf onpxtebhaqf naq gubfr pbzr bhg bs yrff-vafnar snzvyvrf, jvgu gur ynggre zhpu zber yvxryl gb or nggnpurq gb obgu gurve vzzrqvngr naq rkgraqrq snzvyvrf. I am a huge outlier on all of the questions except for authority -- on harm and fairness, I'm an outlier relative to the liberals and conservatives, never mind Lesswrong.
Did many lesswrongers come out of religiously fundamentalist families? I think of it as a rare thing, but given that eliezer and lukeprog both did, perhaps it's actually pretty common here?
[-][anonymous]11y 0

Vg vf orggre gb qb tbbq guna gb qb onq.

From the "Moral Foundations Questionnaire". Can someone tell me what does this even mean?

Literally it seems to reduce to a tautological 'true'. One conceivable purpose of the question could be to establish whether the subject's abstract nodes for communicating about normative ideals are closely linked to slightly more direct normative judgments. It could filter out certain kinds of confusion [].
My first reaction to reading the question and processing it like that was "What's up with that?" so I wrote the comment to ask others about it. Ah indeed.
A tautology. Maybe a bit to see if you're paying attention?
Also sounds plausible.

I need some help to understand the results of the survey "Sacredness + Cartoon Rating Scale." The results in terms of money it would take to violate a certain foundation are very accurate, the results in "scores of importantness" are almost the complete opposite of my actual scales. Am I missing something?