FB/​Discord Style Reacts

For the past year I’ve wanted LessWrong to in­clude some­thing like Dis­cord, Face­book or Slash­dot style re­ac­tions.

Face­book Style means “there’s a few key re­ac­tions that peo­ple use”

Dis­cord Style means “there’s nigh-in­finite re­ac­tions and you can add more, but there still end up be­ing a few com­monly used de­faults.”

Slash­dot Style means “af­ter up­vot­ing or down­vot­ing, you have the op­tion of click­ing a but­ton that clar­ifies why you up­voted or down­voted.”

Of these, I’m most ex­cited for Dis­cord-Style. But I think any of them would be im­prove­ments (if done well)

Habryka re­cently wrote a short­form com­ment on this sub­ject. My own thoughts come in a few differ­ent frames.

Separat­ing En­thu­si­asm from Approval

Boos/​Yays vs ‘ap­prove/​dis­ap­prove’

Em­piri­cally, peo­ple want to cheer for their causes, boo causes they dis­like, sig­nal their so­cial alle­giance and try to en­sure the over­ton win­dow moves in the di­rec­tion they want. I don’t think you can re­ally fight this. But you can nudge peo­ple to dis­en­tan­gle this from “what gets at­ten­tional al­lo­ca­tion on a site about ra­tio­nal­ity.”

I think it’s im­por­tant that when you see a com­ment you like, and you feel the im­pulse to go “yeah! good point! go team!” the first im­pulse you have, the first but­ton available and ex­cit­ing to click, is a but­ton that doesn’t send any sig­nals about how that com­ment should be sorted, and doesn’t ag­gre­gate into an over­all user-score you can check (that, for good or for ill, peo­ple will tend to as­so­ci­ate with so­cial sta­tus)

Other things vs ‘ap­prove/​dis­ap­prove’

Boos/​yays aren’t the only thing I’m wor­ried about. Ideally, I want LessWrong to re­ward good think­ing over things like be­ing funny, or ex­cit­ing. (Be­ing funny and ex­cit­ing should still get re­warded, but no amount of clever in­jokes should add up to some­thing greater than “wrote an ac­tu­ally use­ful, in­sight­ful point.”)

“Viscer­ally Fun but Low Sig­nal But­tons” should be easy to ac­cess. “Higher Sig­nal” but­tons should re­quire more effort and thought.

With both of the above in mind, I think it’s im­por­tant that “Yay”, or “Funny” but­tons should be the first, most ob­vi­ous thing to click on. They should feel satis­fy­ing to click, and you shouldn’t feel mo­ti­vated to click more things if that’s the only rea­son you were up­vot­ing.

The but­tons that send more im­por­tant sig­nals should re­quire a bit of ex­tra effort, and force you to at least no­tice some cog­ni­tive dis­so­nance if you’re up­vot­ing peo­ple just be­cause they’re on your side.

So­cial En­tan­gle­ment, Epistemic En­tan­gle­ment and Com­mon Knowledge

One re­act some­one ex­pressed in­ter­ested in was a sim­ple “ac­knowl­edged.” Votes are to­tally anony­mous, and that means if you want some­one to know that you have read a thing, you have to ac­tu­ally com­ment, which is mod­er­ately high effort and takes up a lot of ver­ti­cal space on the page. Whether some­one has read a thing is fairly im­por­tant in­for­ma­tion about how to con­tinue a con­ver­sa­tion.

By de­fault, on many so­cial-me­dia plat­forms, likes are pub­lic. They were also pub­lic on the old In­tel­li­gent Agent Foun­da­tions Fo­rum (and I think prob­a­bly on Ar­bital, al­though not sure off­hand).

This does two things, which I have mixed feel­ings about.

One is so­cial en­tan­gle­ment. Visi­bly lik­ing each other’s com­ments is part of the pro­cess by which peo­ple build so­cial trust and al­li­ances. I think there’s rea­son to be cau­tious about LessWrong di­rectly fa­cil­i­tat­ing that.

Another is clar­ity on who be­lieves what, and whose judg­ment you trust. When you’re build­ing a se­ri­ous, com­plex idea, it’s ac­tu­ally im­por­tant who un­der­stands what con­cepts, who thinks differ­ent con­cepts are im­por­tant. There are peo­ple I do in fact trust more in­tel­lec­tu­ally than oth­ers, and it’s higher sig­nal to know that one of them liked a post than some rando. It’s also more in­for­ma­tive when I know that mul­ti­ple peo­ple I trust dis­agree.

My cur­rent best guess is that it’s best for the vot­ing on LessWrong to be anony­mous, but for re­ac­tions to dis­play user­names on hover-over. It might or might not be fea­si­ble or de­sir­able (from a UI com­plex­ity stand­point) to let peo­ple choose whether to re­act pub­li­cly. But I can imag­ine chang­ing my mind about this.

Mak­ing it lower effort to give feed­back.

Re­ceiv­ing a down­vote with­out ex­pla­na­tion sucks. Some peo­ple com­plain about this – “can’t you provide rea­sons for your down­votes?” Well, no. Triv­ial in­con­ve­niences mat­ter. If you force peo­ple to provide in­for­ma­tion and figure out how to ar­tic­u­late what’s wrong with some­thing, peo­ple will prob­a­bly just stop giv­ing feed­back rather than ac­tu­ally pro­vid­ing rea­sons.

Not only does this re­quire figur­ing out how to write a com­ment, it opens up a line of en­gage­ment that you might have to put even more effort into defend­ing.

[this is an em­piri­cal claim, it’s per­haps worth the ex­per­i­ment of re­quiring down­votes to always re­quire rea­sons, but I’m not op­ti­mistic about it].

But I think there are some fairly com­mon rea­sons why a com­ment gets down­voted, that could at least make it lower-effort to give feed­back:

  • “This com­ment seemed a bit con­fused”

  • “This com­ment seemed to be round­ing things off in an over­sim­plified way”

  • “This com­ment seems wrong in ways that have pre­vi­ously been ex­plored at length on LessWrong”

  • “This com­ment seems mean spir­ited.”

  • “This com­ment seemed to be act­ing in bad faith”

It’s also nice to im­prove the re­ward sig­nal for par­tic­u­larly good ac­tions:

  • “This com­ment was par­tic­u­larly clear”

  • “This com­ment made spe­cial effort to be rigor­ous and cred­ible.”

  • “This com­ment ac­tu­ally changed my mind about some­thing.”

  • “This com­ment made spe­cial effort to be char­i­ta­ble”

An is­sue re: Sim­plic­ity of Concepts

You’ll no­tice some is­sues, com­par­ing the above feed­backs to Face­book Re­acts.

Face­book re­acts are “haha!” “love!” “sad!” “anger!” “wow!”

Every­one knows what those mean. Every­one knows that ev­ery­one else knows what those mean. They are very short words. They are (due to mil­len­nia of evolu­tion, ge­netic and cul­tural) con­cep­tu­ally sim­ple.

“This com­ment seems to be round­ing things off in an over­sim­plified way” is a less com­mon con­cept. It’s more com­pli­cated. And if you sim­plified it slightly so that the but­ton said “Over­sim­plified”… that would… ac­tu­ally be an over­sim­plified but­ton. It’s im­por­tant that I’m just say­ing “yo this com­ment was over­sim­plified”, but rather that it seemed (prob­a­bly) to be mak­ing a sub­tle er­ror.

I think this is re­ally im­por­tant. I think some­thing LessWrong needs to do is nu­anced cri­tiques eas­ier to chunk. This is pretty tricky, since, well, the whole point of nu­ances is that they’re nu­anced.

A ra­tio­nal­ist friend once com­mented, in non-ra­tio­nal­ist cir­cles, that when they tried to say “I agree with your point but I think this par­tic­u­lar part has a log­i­cal er­ror”, they would of­ten have peo­ple… just com­pletely fail to parse that. It wasn’t in their schema at all.

On LessWrong, we have some shared con­text where we mostly all un­der­stand not to just have Ar­gu­ments Be Soldiers and what­not. Our schema in­cludes Lo­cal Val­idity. But there are many im­por­tant, key con­cepts that still take a lot more effort to ex­press than “yay/​boo” or “haha!”

And thing is… it’s not like “Love” is a sim­ple con­cept. When some­one clicks ‘Love’ on one of my face­book posts, there is a fairly rich wave of senses I get (de­pend­ing on my post, and de­pend­ing on my re­la­tion­ship with the per­son in ques­tion). When some­one posts about their pet dy­ing and I click ‘Love’, there’s this whole shared con­text about how we’re both hu­man and we know what it is to lose peo­ple and my heart goes out to them and I chest tenses slightly and there’s… just a whole lot go­ing on.

Still, I’m able to chunk that com­plex­ity into a con­cept called “Love”, and it’s eas­ily available for me to ac­cess.

There’s a po­ten­tial longterm vi­sion for LessWrong – maybe not the right vi­sion, but pos­si­ble – where part of what we’re do­ing here is dis­till­ing con­cepts down so thor­oughly that a sin­gle word can com­mu­ni­cate a lot of nu­ance.

Lan­guage real es­tate is limited, and I’m not sure which con­cepts make the most sense to dis­till in such a way. There’s also cer­tainly room for this to fail, where in­stead of be­ing able to more-eas­ily-ex­press nu­anced con­cepts it ends up de­stroy­ing nu­ance.

Face­book has cheap­ened the word “friend”, and that’s im­por­tant. But… I also have an im­pres­sion of it hav­ing made it eas­ier for me to ex­press love, in a way that so far seems net pos­i­tive.

It feels ex­cit­ing to me to imag­ine one day liv­ing on a world where “this changed my mind” or “this was well thought even though I dis­agree” feel like ba­sic, ob­vi­ous con­cepts that are im­por­tant enough to be com­mu­ni­cated with a sin­gle word.