Please take this easy one-question survey.  For Science! Survey link

53 comments, sorted by
magical algorithm
Highlighting new comments since Today at 12:22 AM
Select new highlight date

First time I've been motivated to actually say "lol" on Less Wrong, although that has a lot to do with me currently crafting Lovecraftian rituals, often literally waking up in the middle of the night and scribbling down Cthulhu-inspired songs.

Soon... soon....

(If you actually want to hold a caroling party in the immediate future I can send you a link to the current work, but if not I recommend hanging on a little longer)

Are you also conducting this on other sites? (I think you should)

Does the survey record where it was linked from? Could I repost this on other forums?

Yes other sites, no it doesn't record where it was linked from.

Related to this, I assume? (Don't click that until after you take the survey.)

Gah... you bastard.

Edit: I used a random number generator. I hope that's not cheating.

Why would you do that? If you know this is set up to detect patterns in how humans pick random numbers, isn't specifically entering a data point in order to thwart that sort of like providing fake data on a survey to mess with the results?

Using a random number generator is a very sensible and probably relatively common way for humans with a computer to pick "any" year, so I think that doing so is providing honest data, not fake data.

Where did the question ask us for a "random" number? "Some humans pick future years from underdetermined distributions by asking the most convenient RNG" is no more "fake" than, say, "Some humans pick future years from underdetermined distributions by recalling a fun sci-fi show".

Now that I've explained why lavalamp shouldn't feel guilty for asking his computer and I shouldn't feel guilty for asking pop culture: is it odd that I do feel guilty for picking something in this millenium? I know that a uniform distribution over Z+ doesn't exist, but it still feels like I ought to have at least added a digit or two.

Why would you do that? If you know this is set up to detect patterns in how humans pick random numbers, isn't specifically entering a data point in order to thwart that sort of like providing fake data on a survey to mess with the results?

It's like providing fake data on a survey intended to detect patterns in what fake data people will produce.

It was the most obvious way to avoid the anchoring effect? I kinda figured the OP was going to give the survey to multiple audiences and see if LW had a smaller anchoring effect.

A mathematician can probably give a better way to pick without setting an upper bound, but I did:

(current year + (p/(1-p)) where .5 < p < 1.0

It was happy to accept roman numerals (and presumably any input at all)

V zvtug abg unir pnhtug gur napubevat vs lbh unqa'g hfrq n ahzore juvpu unq orra qvfphffrq va na napubevat pbagrkg orsber.

I see what you're doing there.

V cvpxrq n enaqbz ahzore orgjrra mreb naq gur rfgvzngrq raq bs fgne sbezngvba va gur havirefr, ba gur beqre bs n uhaqerq gevyyvba lrnef sebz abj. Znggre jvyy pbagvahr gb rkvfg sbe n tbbq ybat gvzr nsgre gung, naq qrtrarengr znggre sbe na rira ybatre gvzr, ohg gung frrzrq nf tbbq n cynpr nf nal gb fpbcr gur shgher sbe nal cenpgvpny checbfrf.

Is it possible to see the results? If not, will it be in the future?

I think in order for this to really produce useful results (rather than be funny) it needs to actually ask a question, so people have something to think about other than what the poll is really about.

Yes, this might have been a good candidate for shuffling in a bunch of unrelated questions. But this was the first step of "Make it work make it right make it fast". I figured I could make a first pass at seeing what sorts of results I could get, so I should do so immediately.

Well you only get one shot from the Less Wrong community (we'll totally know what's up next time). I'm sure you can get feedback from other communities as well. (If you're actually getting useful info, you also need the control group poll that doesn't anchor).

V'ir whfg ernyvfrq gung V fubhyq unir hfrq n havsbez qvfgevohgvba ba gur rkcrpgrq yratgu bs gur havirefr. Nf vg vf zl 42 qvtvg ahzore ybbxf engure fznyy.

Orsber V gbbx guvf fheirl, V gubhtug gb zlfrys: shpx gur ynfg fheirl V gbbx naq vgf napubevat rssrpg. V'z napubevat ba lrne gjb gubhfnaq gjb uhaqerq ab znggre jung gur dhrfgvba vf guvf gvzr!

Vg jnf n gvzryl gubhtug.

Orsber V gbbx guvf fheirl, V gubhtug gb zlfrys: shpx gur ynfg fheirl V gbbx naq vgf napubevat rssrpg. V'z napubevat ba lrne gjb gubhfnaq gjb uhaqerq ab znggre jung gur dhrfgvba vf guvf gvzr!

So essentially you are still anchored to the number given in the last survey - just one layer of indirection removed.

Mwhaha! But my response wouldn't count, I've seen a related comment somewhere.

Tbqqnzzvg. Vapyhqvat gur rknzcyr lrne "**" vasyhraprq zl pubvpr. Vs gung'f abg vagragvbany lbh fubhyq erzbir vg.

Oh. Goddammit again! Cyphering the above so that people can laugh at me after answering the survey themself.

Edit: apparently rot13 won't cypher numbers.

apparently rot13 won't cypher numbers.

The usual solution is to spell out the number using words, like "gjb gubhfnaq bar uhaqerq".

V nz rkgerzryl pbasvqrag gung vg vf. Naq, shegure qbja, gurer vf n yvax gb nabgure dhrfgvba juvpu znl or eryngrq. Vs gung vf gehr, V qbhog zl nafjre jvyy or zhpu uryc. V fvzcyl nafjrerq (sbhe zvahf bar) (guerr gvzrf bar pneebgf) (frira zvahf sbhe).

Depending on how other people respond, my response may have been a bit of an outlier.

ROT13: "bire avar gubhfnaq" vf gur zbfg pbzzba nafjre fb sne. Abg fher jung pbhagf nf na bhgyvre.

Naq vs lbh unqa'g fnvq gur cnerag jnf va EBG13, V jbhyq unir gubhtug lbh jrer Pguhyuh?

"Bire avar gubhfnaq", nf va, gur cuenfr be "bire avar gubhfnaq", nf va, fbzr ahzore juvpu vf terngre guna avar gubhfnaq?

Zvar jnf pregnvayl gur ynggre, ohg abg gur sbezre.

The link is now inactive, could someone provide context to the discussion now that clicking it is not possible and spoilers are no longer an issue?

It was "Name a year", with "2100" being given as an example.

V gubhtug: "gjb gubhfnaq bar uhaqerq -- gung'f gelvat gb trg zr gb napube. Be vg'f n Fpuryyvat cbvag. Ohg pubbfvat fbzrguvat bgure guna vg frrzf zber vagrerfgvat." Gura V pbafvqrerq "gur boivbhf arkg pubvpr", gjb gubhfnaq gjryir. Gura gjb gubhfnaq guvegrra, gura fbzrguvat enaqbzyl trarengrq (gur ynfg bs juvpu jbhyq erdhver zr gb pbzr hc jvgu fbzr qvfgevohgvba). Gura V qrpvqrq gb cvpx fbzrguvat ovt. V nafjrerq avar dhnqevyyvba, frira gevyyvba, bar uhaqerq avargl-avar ovyyvba, gjb uhaqerq svsgl-sbhe zvyyvba, frira uhaqerq sbegl gubhfnaq, avar uhaqerq avargl-bar (va bgure jbeqf, bar yrff guna gjb gb gur svsgl-guveq), juvpu vf gur znkvzhz vagrtre rkcerffvoyr hanzovthbhfyl va na VRRR frira-svir-sbhe qbhoyr.

Rot13ed; decoded version here.

(Edit: Fixed link.)

V ybbxrq ng gjragl-bar uhaqerq, ershfrq gb napube ba vg, naq glcrq bar gjb guerr sbhe svir fvk frira rvtug avar orpnhfr V pbhyq or pbasvqrag gung vg jnfa'g gur erfhyg bs napubevat. Tvira gung bgure crbcyr qvq vagrerfgvat zrnavatshy guvatf yvxr n enaqbz ahzore orgjrra abj naq fbzr nfgebabzvpny rirag, V srry bhg-njrfbzrq.

V gubhtug nobhg orvat njrfbzr, gura nafjrerq gjb gubhfnaq bar uhaqerq.

I decided to go nuts. Guerr zber guna gra, gura svir yrff guna rvtugrra hc-neebjf, gura fvk cyhf frira.