# Pattern’s Shortform Feed

• Idea for re­solv­ing Liar’s Para­dox:

In logic, sen­tences should be as­signed cir­cuits* in­stead of truth val­ues.

Firstly, pro­cess­ing cir­cuits, which ex­e­cute the ac­tion de­scribed by the sen­tence.

“Proof” that there ex­ists a suit­able cir­cuit for sen­tences such as “2+2=4“ “2+2=5” and “9^9=387420489”, ex­ists in the form of calcu­la­tors.

Se­condly, cir­cuits for check­ing if a truth value as­sign­ment is con­sis­tent. They re­ceive “True” or “False” as an in­put, here­after referred to as “I”, and out­put “True” if the as­sign­ment is con­sis­tent. This can be done by hav­ing both the pro­cess­ing cir­cuit and “I” as in­puts to a gate which re­turns “True” if they are both equal. (An equals gate.)

This is like turn­ing the prob­lem of eval­u­at­ing the truth value of “2+2=4” into con­struct­ing a cir­cuit that “rep­re­sents” “2+2==4”.

Th­ese are ne­ces­si­tated by the ex­is­tence of sen­tences which may be as­signed mul­ti­ple truth val­ues. “This sen­tence is true.” As­sign­ing ‘true’ would be con­sis­tent, as would ‘false’. (This sen­tence is un­usual in that the con­sis­tent as­sign­ment cir­cuit has the in­put, “I”, go into both the pro­cess­ing cir­cuit, and the equals gate.)

*Sev­eral things work here. In ad­di­tion to be­ing a good place to hold a lot of tech­ni­cal/​log­i­cal ques­tions, cir­cuits can be im­ple­mented fairly eas­ily.

• I came up with this when think­ing through some­thing I saw some­one write on the re­solv­ing the LP (on their blog, they also wrote a book). Now I’m try­ing to find it again.*

What they said, para­phrased (from mem­ory): You at­tempt to as­sign a truth value as fol­lows: you sup­pose it is true. If it were true, then it would be false. So you sup­pose it is false. But then it would be true. At this point, they re­jected this line of think­ing on the grounds that, what it means for some­thing to be true, is for there to ex­ist some­thing in re­al­ity that cor­re­sponds to it, and there is no op­er­a­tor that satis­fies this crite­ria. So the LP is false. (This is similar to the an­swer that it’s “not true or false, but non­sense” (from xkcd fo­rums re­sults on google search, when I tried to find their blog). The au­thor took the ad­di­tional step of com­bin­ing the no­tions of “false­hood” and “non­sense” un­der the la­bel of “false”.)

When I was think­ing through that, while I got their point, it sounded like a NOT-gate. That is, I figured you could as­sign to a sen­tence a cir­cuit which takes the truth value you would as­sign to it, and re­turns what it would be if that were so. This made sense for both of the self-refer­en­tial sen­tences I con­sid­ered (LP and “This sen­tence is true.”), and valid as­sign­ments were fixed points. What makes LP “para­dox­i­cal” is that try­ing to as­sign it a truth value is a pro­cess that cor­re­sponds to try­ing to find the fixed point of a func­tion which doesn’t have any fixed points. (It’s op­po­site be­haves the op­po­site way: it has all the fixed points.) When I thought about other sen­tences that weren’t self-refer­en­tial, this didn’t make as much sense, and that was when I came up with the other two types of cir­cuits/​(ways of think­ing about this).

*EDIT: It’s fake­nous.net.

• There is a gen­eral pat­tern that oc­curs wherein some­thing is ex­pressed as a di­chotomy/​bi­nary. Switch­ing to a con­tinuum af­ter­wards is an ex­ten­sion, but this does not nec­es­sar­ily in­clude all the pos­si­bil­ities.

Di­chotomies: True/​False. Beau­tiful/​Ugly.

True/​False.

Logic han­dles this by look­ing for ‘all true’.

If ‘p’ is true, and ‘q’ is false, ‘p and q’ is false.

More gen­er­ally, a sen­tence could be bro­ken up into parts that can be in­di­vi­d­u­ally rated. After this, the ra­tio of true (atomic) state­ments to false (atomic) state­ments could be ex­pressed—un­less all the sub-state­ments are true, or all false. This can be fixed be ex­press­ing the ‘score’ as a (ra­tio­nal) num­ber, with two choices of score:

true(sen­tence) = num­ber of true state­ments /​ num­ber of statements

false(sen­tence) = num­ber of false state­ments /​ num­ber of statements

And since ev­ery state­ment is true or false:

true(s) + false(s) = 1

And if we want to ex­press how much truth is ex­pressed, true(s)*num(s) = # of true state­ments. (Th­ese func­tions don’t have the best re­la­tion­ship to each other, they’re just meant to be in­tu­itive enough.)

Con­sider the as­sump­tion: ev­ery state­ment is true or false. (Ex­clu­sively.)

In­stead of div­ing into para­dox, con­sider func­tions. Equal­ity(s) re­turns the sen­tence “s is true”. Ne­ga­tion(s) re­turns “s is false”. Th­ese func­tions don’t have a truth value, it’s de­pen­dent on the vari­able that’s passed in. Con­cat(s_1, s_2) re­turns “s_1 s_2”, which can be just gib­ber­ish. But why is “Equal­ity” named Equal­ity—it pre­serves truth value but there are other func­tions with that prop­erty? It might be bet­ter thought of as a fam­ily of func­tions.

Now con­sider the func­tion f that, given s, re­turns “s is true and s is false”. And here is a func­tion that is always false. Right?

(This next para­graph* is a fram­ing with­out ex­am­ples, and may be re­jected or ac­cepted. I’m treat­ing ‘para­doxes’ in this way be­cause, as the para­graph af­ter it notes, truth seems to come from a sys­tem.)

But just as (self refer­en­tial) sen­tences can be con­structed that are ‘para­dox­i­cal’ - nei­ther ‘false’ nor ‘true’, sen­tences may also be con­structed which ‘are both’. This may be re­solved by point­ing out that the first are “non­sense”, and re­solv­ing that “non­sense is false”, and say­ing that it doesn’t mat­ter what value is as­signed to the sec­ond as there is no con­se­quence. (For such a sen­tence may be false, or it may be true, but not both at once.) But these re­s­olu­tions are at odds. Are not both kinds “non­sense”? Or if they are differ­ent, they seem differ­ent from both ‘state­ments which can only be true’ and ‘state­ments which can only be false’.

To get back to our ‘func­tions’ (which take sen­tences as in­put, and re­turn a sen­tence as out­put), con­sider the sen­tence “1+1=2”. Is this true? In many sys­tems yes, “base 3, base 4, base 5, …”, but not in “base 2”, where “2″ is not defined, “1+1=10”. Th­ese sys­tems may be con­verted be­tween, and we may even say that while some­thing is ex­pressed one way in one sys­tem, and an­other way in an­other sys­tem, they’re the same “fact” (or false­hood).

But hav­ing differ­ent sys­tems en­ables much con­fu­sion, two (or more) peo­ple might dis­agree on what color the sky is cur­rently, even if they both have eyes that work fine, and with­out any un­usual at­mo­spheric phe­nom­ena that change what the sky looks like if you take a few steps to the right, or the left, if only they dis­agree on what col­ors the words for col­ors mean. If you call X “red”, and I call X “blue” we may still both see X.

To get back to truth(s) which can re­turn “2/​3” (mean­ing s con­tains 3 state­ments, 2 of which are true, one of which is false), why re­turn one num­ber? Why not two: 2,1: 2 true state­ments, 1 false. But there could be more state­ments than those two kinds. And here the path splits in two.

1. A par­tic­u­lar meth­ods of as­sign­ing one value to a sen­tence may ‘fail on the para­dox’, or choose to call it false.* One method, one an­swer—ev­ery state­ment is true or false, ex­clu­sively.

2. A set for each pos­si­bil­ity: It is true, it is false, it can be true or false, it can­not be ei­ther, etc. There’s still a bi­nary as­pect to this: “is it true” re­ceives the an­swer “yes” or the an­swer “no” ex­clu­sively. But, in­de­pen­dently, “is it false” may also re­ceive ei­ther an­swer.

Fol­low­ing the 2nd path, what does it mean for some­thing to be true and false? Nei­ther?

One way is this: “The sky is blue, and the clouds are red.” Part of it is true, and part of it false. That which holds nei­ther truth not false­hood, is non­sense.

How does this gen­er­al­ize? For that an­other di­chotomy will be re­quired.

Beau­tiful/​Ugly***. While this may be sub­jec­tive, the qua­ter­nary** view can be seen as claiming the bi­nary view is false, some things are both beau­tiful and ugly, and some things are nei­ther. Per­haps here this view will be less con­tro­ver­sial, af­ter all, if a thing is judged to be beau­tiful by one per­son, and ugly by an­other, “sub­jec­tively”, then “ob­jec­tively” might not the ob­ject be both? Per­haps some­thing ugly and beau­tiful could be cre­ated by cut­ting some­thing beau­tiful in half, and some­thing ugly in half, and com­bin­ing them? This may be trick­ier than com­bin­ing a true state­ment and a false state­ment, but per­haps if some­thing is both beau­tiful and ugly, both as­pects can be seen, where some­thing that is true and false might be swiftly pro­claimed ‘all wrong’ (or all right).

Per­haps this has all just been con­fus­ing, or per­haps it will be use­ful. The no­tion of ‘log­i­cal coun­ter­fac­tu­als/​counter-log­i­cals’ has seemed strange to me—it is not that “it could be that 2+3 = 4” but that must be a differ­ent sys­tem. What such a thing could mean in con­junc­tion with a world, say, if you put 2 things in a con­tainer, and then three, and what re­sults is 4, seems un­clear. (Even mak­ing them crea­tures doesn’t make sense, for if one eats an­other, why won’t that hap­pen later?) If it holds for a class of ob­jects, then that changes the re­la­tion­ship be­tween num­bers and ob­jects—an ap­ple and an or­ange are to­gether are two things, but even if all things have the prop­erty that un­der cer­tain cir­cum­stances they re­act to pro­duce or elimi­nate an­other of the same type, then un­less this holds be­tween classes, no more might one speak of an ap­ple and an or­ange be­ing 2, be­cause they don’t re­act with each other.

*Para­doxes work­ing this way may be avoided by sys­tem de­sign.

**One may elimi­nate one of these cat­e­gories, and say, that noth­ing is nei­ther beau­tiful nor ugly. Then the cat­e­gory still ‘ex­ists’ though it has no mem­bers—a broader view may in­clude things that are not, but ab­sent a pro­cess for cre­at­ing new cat­e­gories, the more ex­pan­sive view may be bet­ter be­fore ex­am­in­ing re­al­ity. And if some­day that per­son finds some­thing which is nei­ther, then the bucket will be ready for this new ob­ject un­like any­thing seen be­fore.

***This is one area where things may not be fixed, in a way that we don’t see in math or logic. A view in which things don’t have prop­er­ties may be more use­ful—but it is harder to see this for things/​prop­er­ties like “num­bers” which ‘seem to ex­ist’. “The tree falls in the for­est” ar­gu­ment may also be had about beauty.

• Thanks to this ques­tion, I re­cently started think­ing about how progress on open prob­lems in math [1] could be made faster, at least with re­gard to low hang­ing fruit. I made a com­ment there about mod­el­ing the prob­lem (how can progress be made faster) and a pos­si­ble solu­tion. This brings me to a few ques­tions:

• Model­ing the prob­lem.

• Solv­ing the prob­lem.

• Is this a big enough deal that peo­ple want it solved? Or are peo­ple only in­ter­ested in some­thing like a) More nar­row ar­eas with ob­vi­ous value be­ing im­proved? b) The cre­ation of a plat­form where peo­ple can put money on spe­cific things that they want solved be­ing solved/​progress be­ing made. c) Some­thing else?

• How to test all of the above (and im­ple­ment where ap­pli­ca­ble).

• Meta: Should these all be posted as sep­a­rate Ques­tions? What should they be called? Have any of these ques­tions already been asked?

[1] They have a cer­tain for­mal/​em­piri­cal qual­ity which makes things sim­pler. It also might be eas­ier to use this as a met­ric for ‘how good is our X [2] at ad­vanc­ing re­search (progress)’?

[2] Any­thing that could make a differ­ence—a Plat­form, Or­ga­ni­za­tion, Pro­gram, a set of Math courses...

• Two mod­els of how feed­back is use­ful, for mak­ing cor­rec­tions.

1) Post qual­ity*:

• Write an ar­ti­cle.

• Get feed­back.

• Re-write the ar­ti­cle (if nec­es­sary).

2) Communication

• Say a thing/​write a post.

• Peo­ple re­spond.

• If re­sponses in­di­cates prior mes­sage was un­clear, re­spond and ex­plain the un­clear part/​re­vise post or ‘change fu­ture posts.’**

*In­tended gen­er­ally. May also ap­ply to books, etc.

** This is a sub­set of (pos­si­ble) re­sponses, be­cause it is still about com­mu­ni­cat­ing the same idea/​s, rather than do­ing some­thing new.

• Some ideas on struc­ture:

• “Posts” are usu­ally “about them­selves”. For ex­am­ple, SSC has posts with no com­ments sec­tion. For counter-ex­am­ples, see posts like Rea­son­able Ex­pla­na­tions—the au­thor is in­ter­ested in com­ments that fit a cer­tain for­mat, the body of the post has the rules for (top-level) com­ments, and the au­thor posts a com­ment (that fits the for­mat) as a starter.

This is a for­mat for a “Dis­cus­sion”. If an au­thor in­cludes both the rules and the starter in the “post” body, it’s still a “Dis­cus­sion”. If the OP ex­pounds an idea, and in­cludes ex­am­ples (usu­ally of a cer­tain form) and sug­gests peo­ple com­ment other things they think might be ex­am­ples, that is both a “Post” and a “Dis­cus­sion”.

The Monthly threads are of course “Dis­cus­sion”-like, though they’re more free form—a “Dis­cus­sion” with no rules*.

*Since this is LessWrong, both LessWrong’s rules ap­ply, and ways peo­ple here pre­fer things be dis­cussed—this is why “The Se­quences” are em­pha­sized. The sec­ond type, not be­ing laid out in a short ex­plicit set of rules are at times “bro­ken”, lead­ing to con­flict.

• The value of struc­ture, in­clud­ing lin­ear­ity. (A lens.)

• Some (if not all) sites are about ideas. This site does it by “one-off” meth­ods for pre­sent­ing ideas. Th­ese may be con­trasted with sys­tems that pre­sent an idea, but (may) con­tinu­ally change the pre­sen­ta­tion (non-au­to­mat­i­cally), such as wikis. What might a site look like if it tried for more in­te­gra­tion?

(Bend­ing the for­mat:)

What if com­ments sec­tions, rather than stay­ing fixed in their at­tach­ments to (a) post, roamed around?

ETA: What is an ac­tu­ally good way of get­ting com­bi­na­tions of ideas/​com­par­ing how similar prob­lems are solved in differ­ent fields?

• From the other di­rec­tion—Ideas are in posts. This is part of why re-runs ex­ist—to send the idea out again, to re­flect, and to bring com­ments on the idea to life again.

When a post is run the first time there are com­ments. When a post is re-run (un­changed), the idea may already be out there (it’s pos­si­ble all the read­ers have read it), but there are new com­ments. In this way, the com­ments sec­tion on the re-run is still about the same thing, ab­sent changes re­sult­ing from time, it’s just com­ments 2.0. It’s also fresh—when I read The Se­quences, I did not read all the com­ments.

(Bend­ing the for­mat.)

Set­ting up a com­ments sec­tion so that is pos­si­ble would re­quire a re­design, and prob­a­bly work against the rea­sons they were set up the way they are. (Which is why The Se­quences were made into a book in­stead.) I haven’t seen a lot of sites do this in­ten­tion­ally. There are blogs with no com­ments sec­tions any­where, but mak­ing a set read­able by mak­ing it empty is triv­ial.

• Things that be­comes “finished” (‘Posts’) ver­sus Things that don’t (‘Lists’):

Finished: Posts/​Com­ments are (in­di­vi­d­u­ally) cre­ated, then sub­mit­ted.

Com­mon Ex­cep­tions: “Up­date” may be ap­pended to the end, fol­lowed by con­tent. Alter­na­tively, changes may be made, and de­scribed in a sec­tion added to the end marked “Edit”.

Un-Finished: All Posts, All Ques­tions, All Se­quences (The Library). Th­ese lists keeps chang­ing.

• While a List may come to an end, if all Lists die (and stay dead), that’s a suffi­cient con­di­tion for the site to be con­sid­ered dead.

That’s not to say the site would be dead if there stopped be­ing new posts for a time—if peo­ple started re­vis­ing their posts, and sub­mit­ting those changes, dis­cus­sion of ideas (and the life of the site) could con­tinue—but then the cur­rently ex­ist­ing posts would “liv­ing lists” while “all posts” would be dead.

• The pat­tern seem to be “Lists”, which can go on for­ever, con­tain “items” which have a short life.

• Two ways on look­ing at things:

1) See what this web­site calls, say, “Posts”. Look for pat­terns. (Prac­tice → The­ory.)

2) Con­sider differ­ent Ideas, and look at what ‘im­ple­ments’ them. (The­ory → Prac­tice.)

• This is why what the site des­ig­nates “com­ments” are of­ten referred to by users as “posts”—they im­ple­ment the same idea.

• Perfec­tion is achieved, not when there is noth­ing more to add, but when there is noth­ing left to take away.

-An­toine de Saint-Exupéry

• “The No True Scots­man fal­lacy” is of­ten cited when peo­ple do things like defin­ing X not as Y, but Y when Y works.* This is the (ex­plicit) ideal (that peo­ple may ad­mit to). While those ask­ing “What is X” are prob­a­bly in­ter­ested in “When does Y work?”, if X/​a group that defines it­self based on X (and refers to it­self with the la­bel ‘X’), then since their goal is to achieve that ideal, they them­selves would very much like to know/​and are work­ing on “what is nec­es­sary to make Y work/​hap­pen?”. Thus ‘Y is not work­ing’ may be (seen as) a crit­i­cism of (the group) X—and spark some de­bate. (The ideal may be a motte and bailey, or fake.)

To make this more con­crete here is an ex­am­ple: “Ra­tion­al­ity is about win­ning.” (I’m still wait­ing for the “X is not about Y” ar­ti­cle “Ra­tion­al­ity is not about win­ning”.) What other things are (or can be) defined in terms of ‘when they work’?

*Or more speci­fi­cally (see the wikipe­dia ar­ti­cle) ‘peo­ple who like X’ think of ‘the ex­am­ples of X they like’ when they hear ‘X’.

• This thread get Meta (about LW) here, TLDR here. (I dis­cuss a fea­ture I think would be use­ful, that might af­fect user in­ter­ac­tion, in hopes of start­ing a dis­cus­sion. This may be­come a post for that pur­pose later.)

Com­ment types (on posts):

-Con­tent re­lated:

--Com­ments (I love this post/​I hate this post/​etc.)

--Questions

-Styling re­lated:

--Errata

--[X] would be good in a sum­mary of the posts

Gen­eral Styling ques­tion: Does this site sup­port bul­leted lists which con­tain bul­leted lists?

• Re­lat­edly, it’s use­ful to know what users (esp. au­thors of lots of posts) like what kinds of feed­back.

• ‘I hate all the nit­picks about gram­mar and spel­ling.’

• ‘I ap­pre­ci­ate this kind of feed­back.’

Ob­vi­ously there can be nu­ance:

• Peo­ple may ap­pre­ci­ate feed­back on Con­tent, but not on Styling.

- Espe­cially gram­mar and spel­ling.

• Peo­ple may also be sen­si­tive to the amount of nega­tive feed­back.

- Or pre­fer com­men­tary in­clude points of agree­ment (or over­all im­pres­sion) as well as dis­agree­ment.

• While ex­plicit site rules/​norms can guide in­ter­ac­tion meth­ods over­all, tools for this pur­pose might en­able both more im­prove­ment and lower fric­tion—users get­ting feed­back/​in­ter­ac­tion they find valuable, even when differ­ent users want differ­ent kinds of feed­back/​in­ter­ac­tion.

But tools are ex­pen­sive. A gi­ant list (in one place) might cap­ture some of the value, by en­abling such in­for­ma­tion be­ing search­able com­mon knowl­edge.

Retriev­ing such in­for­ma­tion might con­sti­tute a “triv­ial in­con­ve­nience”, but such a doc­u­ment would be easy to cre­ate, and have a shorter feed­back cy­cle than a tool that would re­quire more in­vest­ment.

Past dis­cus­sions of pos­si­ble fea­tures sug­gests that a se­quence of such lists might be use­ful.

Benefits of tools over lists: it would be good to have the in­for­ma­tion ac­cessible (search­able) in mul­ti­ple ways, but easy for a user to change their info across all the differ­ent places quickly.

If some­one just wants to know what in­ter­ac­tion style a user prefers, then a list (in a google doc/​the new LW ed­i­tor when it comes out) with users and styles can be searched (us­ing a key­board short­cut, and typ­ing the user’s name). But if more in­for­ma­tion gets stored this way, it might be helpful if all such in­for­ma­tion con­cern­ing a user could also be ac­cessed.

This could be done in a google spread­sheet. (I’ll add a link to an ex­am­ple (with fic­tional users) when it’s com­plete.) I don’t think that’d be a good long term solu­tion, but it illus­trates what fea­tures are nec­es­sary.

• TL;DR:

Some in­for­ma­tion about users (in­ter­ac­tion style, prefer­ences on feed­back type or for­mat) might be use­ful to have available when in­ter­act­ing with them. I pro­pose a doc­u­ment to be cre­ated, that users can edit with their in­for­ma­tion in this re­gard. Such a list may re­quire the google-docs-like-LW-ed­i­tor to be com­pleted so it can be ed­ited by users to stay cur­rent.* I also out­line (above) pos­si­ble im­ple­men­ta­tion and what fea­tures it would be use­ful to have if it ex­pands to in­cor­po­rate more in­for­ma­tion.

*Since the key is “ev­ery­one is able to edit it” maybe it could go on the wiki.

• There are other as­pects of feed­back as well, see Com­ment, Don’t Mes­sage for one.

• [moved]

• Google docs are good for sav­ing con­tent on the fly.