Bucky

• Bucky 22 Sep 2020 10:42 UTC

  6 points

  in reply to: gbear605’s comment on: Covid 9/​17: It’s Worse

  It’s worth not­ing that the R-squared value for a lin­ear trend-line for 2000-2019 data has $R^2=0.07$ so a con­stant pre­dic­tion of 750k acres would only be marginally less ac­cu­rate. (I think your ex­clud­ing 2020 graph also ex­cludes 2019 but the story doesn’t change much ei­ther way)

  It looks like up un­til 2016 ev­ery­thing was fairly con­stant and since then 3 out of 4 years have been bad.

• Bucky 15 Sep 2020 10:18 UTC

  4 points

  in reply to: Raemon’s comment on: Most Pri­soner’s Dilem­mas are Stag Hunts; Most Stag Hunts are Bat­tle of the Sexes

  I had the same con­fu­sion.

  One of the key things I think be­tween the 3 games is whether com­mu­ni­ca­tion be­fore­hand helps (in sin­gle shot games).

  In PD com­mu­ni­ca­tion doesn’t re­ally help much as you there is lit­tle rea­son to trust what the other per­son.

  In SH com­mu­ni­ca­tion should be able to solve your prob­lem as S-S is op­ti­mal for both play­ers.

  In BotS com­mu­ni­ca­tion which re­sults in agree­ment can at least be trusted as co-or­di­nat­ing is op­ti­mal for both play­ers. Choos­ing which op­tion to co-or­di­nate on is an­other mat­ter.

  (as­sum­ing you’ve in­cluded the plea­sure of spit­ing the other per­son etc. in the pay­off ma­trix)

• Bucky 15 Sep 2020 9:26 UTC

  2 points

  in reply to: habryka’s comment on: ‘Ugh fields’, or why you can’t even bear to think about that task (Rob Wiblin)

  Maybe I need to be more het­eroge­nous in my hiring!

  I hadn’t heard of pair-writ­ing but it sounds like it could work well in my con­text.

• Bucky 15 Sep 2020 9:21 UTC

  4 points

  in reply to: ChristianKl’s comment on: ‘Ugh fields’, or why you can’t even bear to think about that task (Rob Wiblin)

  My in­tended point was that if one per­son has an ugh-field around some­thing then it is of­ten a gen­er­ally un­en­joy­able task. Although other peo­ple don’t have ugh-fields around the task, it still seems un­fair (and would likely lead to bad team dy­nam­ics) to re­as­sign it to some­one else who merely dis­likes the task.

• Bucky 14 Sep 2020 11:34 UTC

  5 points

  in reply to: habryka’s comment on: ‘Ugh fields’, or why you can’t even bear to think about that task (Rob Wiblin)

  My ex­pe­rience would be that peo­ple gen­er­ally have Ugh fields around tasks which no-one on the team likes (e.g. re­port writ­ing). I can’t re­as­sign such tasks with­out be­ing un­fair to the peo­ple who are deal­ing well with such jobs.

  I would agree that men­tion­ing to a man­ager that you’re find­ing some­thing aver­sive is ba­si­cally fine as long as you’re more look­ing for sup­port than re­as­sign­ment (al­though this might be differ­ent in differ­ent fields) and that a man­ager should en­courage that.

  As an ex­am­ple one em­ployee found that peo­ple con­stantly in­ter­rupted him and this made get­ting into the flow of re­port writ­ing su­per hard so we blocked off a day a week to al­low him to catch up with­out in­ter­rup­tions.

  I guess to some ex­tent it’s know­ing what’s pos­si­ble in your con­text and know­ing how flex­ible your man­ager is able/​will­ing to be.

• Bucky 28 Aug 2020 20:41 UTC

  6 points

  in reply to: Pattern’s comment on: Covid 8/​27: The Fall of the CDC

  11.9% vs 8.7% is early plasma ad­minis­tra­tion (0-3 days from di­ag­no­sis) vs late (4+ days).

  13.7% is us­ing low an­ti­body count plasma, 8.9% is us­ing high an­ti­body count plasma. I guess this is the 35% re­duc­tion.

  paper

• Bucky 27 Aug 2020 21:05 UTC

  4 points

  in reply to: gwern’s comment on: Are We Right about How Effec­tive Mock­ery Is?

  All De­bates are Brav­ery De­bates?

• Bucky 26 Aug 2020 23:31 UTC

  2 points

  in reply to: Raemon’s comment on: Do you vote based on what you think to­tal karma should be?

  I do think that it would be very bad if this hap­pened. How­ever I don’t think this is likely. Quot­ing my other com­ment:

  I think its im­por­tant to note here that we are not re­ally that ho­moge­nous in our opinions and weight­ings of differ­ent sources of value. Alot of the wor­ries about Blind vot­ing seem to as­sume that we’re all go­ing to vote the same way about the same posts which I think is highly un­re­al­is­tic. There also seems to be the as­sump­tion that ev­ery­thing frac­tion­ally above 0 value will get an up­vote which again seems un­re­al­is­tic.

  This seems even more true for down­votes—I think peo­ple re­al­ise that down­votes feel ex­tra bad and only use them spar­ingly. For in­stance, I only re­ally down­vote when I think some­thing has been a definite break­ing of a con­ver­sa­tional norm or if some­one is dou­bling down on an ar­gu­ment which has been con­vinc­ingly re­futed.

  I think a spread of opinions on what con­sti­tutes a down­vote (and a gen­eral feel­ing that com­ments get less votes in gen­eral) would make the −80 only hap­pen to su­per egre­giously bad com­ments.

• Bucky 26 Aug 2020 23:23 UTC

  2 points

  in reply to: Raemon’s comment on: Do you vote based on what you think to­tal karma should be?

  When a post is at 50, I can think that is a bit too high just from my gen­eral sense of what I want to see more of on the site. And it’s be throw­ing away in­for­ma­tion about my own be­liefs to not give me the fine-gra­da­tion of “I want to see these posts on the site about as of­ten as I would if they got 50 karma, not the amount that I would if they got 200 karma.”

  This is true when the equil­ibrium po­si­tion of the karma sys­tem is set to To­tal Karma Vot­ing.

  I think that Blind vot­ing would move the karma sys­tem to a new equil­ibrium. I’m not con­vinced we should do so as I think it would be a fairly un­sta­ble equil­ibrium but I think it would work if ev­ery­one did it and would al­low for fine grained ex­pres­sions of your be­lief.

  The equil­ibrium I en­visage would be that the cur­rent amount of some­thing that LW has is taken into ac­count when peo­ple blind vote their opinion.

  As an ex­am­ple, I think the rea­son that joke com­ments can get fairly high karma is that they’re rare. If more peo­ple start writ­ing joke com­ments as a re­sult then that’s fine for as long as peo­ple are up­vot­ing.

  At some point the peo­ple who value the jokes least stop up­vot­ing them or start down­vot­ing them. This con­tinues un­til the re­ward ex­pe­rienced by the jok­ers roughly matches the effort taken or some other bal­anc­ing fac­tor.

  In the case of low pos­i­tive value posts, some peo­ple have a higher thresh­old for what they will give an up­vote for and the more low pos­i­tive value posts there are the fewer peo­ple will up­vote them.

  (I think its im­por­tant to note here that we are not re­ally that ho­moge­nous in our opinions and weight­ings of differ­ent sources of value. Alot of the wor­ries about Blind vot­ing seem to as­sume that we’re all go­ing to vote the same way about the same posts which I think is highly un­re­al­is­tic. There also seems to be the as­sump­tion that ev­ery­thing frac­tion­ally above 0 value will get an up­vote which again seems un­re­al­is­tic. Frankly I think that any­one who can write a post which is good enough that it per­suades 100 differ­ent peo­ple with differ­ent stan­dards to click the up­vote but­ton then they de­serve to get 150 karma!)

  The key then is that in or­der to get an over­sized re­ward for the amount of effort put in, you have to do bet­ter than av­er­age at pro­vid­ing value.

  In Blind Vot­ing, ac­count­ing-for-how-much-of-a-cer­tain-thing-there-cur­rently-is-on-LW is do­ing the same thing as con­sid­er­ing-what-mes­sage-the-to­tal-karma-sends does with To­tal Karma Vot­ing. The former seems to have a lag in the mes­sage get­ting out but I think when you’re in a rough equil­ibrium the lag is rel­a­tively short.

  So this brings me onto what I think the main cost of To­tal karma vot­ing is. If an au­thor looks at a post which has 25 karma from 10 votes, what does it mean? Roughly speak­ing it means that it was con­sid­ered about as valuable as an­other 25 karma post. The 10 votes tells the au­thor how effi­cient the karma mar­ket was for the post and pos­si­bly gives limited in­for­ma­tion on how varied the opinions were.

  With Blind vot­ing the au­thor sees that and knows that 10 peo­ple had an opinion that this post was wanted more or less and that their av­er­age strength of opinion was 2.5 karma points in favour. This prob­a­bly con­sists of some­thing like 3 peo­ple who want alot more like it and 7 peo­ple who want a lit­tle more like it (or pos­si­bly some who wish there was less like it or were just yay/​boo­ing).

  I agree that karma is a kludge and the true mean­ing isn’t nec­es­sar­ily clear but with Blind vot­ing it seems im­por­tantly less of a kludge and some ex­tra in­for­ma­tion can be ex­tracted.

• Bucky 26 Aug 2020 21:39 UTC

  2 points

  in reply to: Raemon’s comment on: Do you vote based on what you think to­tal karma should be?

  I want to note that the I see the “vote to­wards the ideal karma” as com­pletely com­pat­i­ble with “vote your be­lief.”

  Agreed. I was look­ing for a short­hand way of refer­ring to the differ­ent vot­ing poli­cies but am yet to find one which is satis­fac­tory—you’ve (rightly) shot down a cou­ple of my ideas! To­tal Karma vot­ing seems fine for one policy, maybe di­rect opinion vot­ing for the other? If you shoot that one down too you can come up with your own!

• Bucky 26 Aug 2020 19:55 UTC

  4 points

  in reply to: gwern’s comment on: On Sud­denly Not Be­ing Able to Work

  The pa­per does at­tempt to ad­just for this with a com­plex­ity met­ric al­though I sus­pect this doesn’t work perfectly as it seems to be a lin­ear ad­just­ment with num­ber of nodes used by the en­g­ine to calcu­late the op­ti­mal move.

  I have a con­cern that the pa­per is com­par­ing tour­na­ment play (offline) to match play with 4 games per match (on­line). In match play, es­pe­cially with few games, a player who is be­hind needs to force the game and the player in the lead can play more con­ser­va­tively. Tour­na­ments have their own in­cen­tives but over­all I would ex­pect short match play to cause big­ger er­rors from en­g­ine op­ti­mal play as los­ing play­ers try to force a win in nat­u­rally draw­ing situ­a­tions.

  The calcu­lated effect size is >200 ELO points which sug­gests to me that some­thing is amiss.

• Bucky 26 Aug 2020 13:13 UTC

  5 points

  in reply to: Raemon’s comment on: Do you vote based on what you think to­tal karma should be?

  Ok, I think I ac­tu­ally agree with your crux.

  The points I was try­ing to make were (kinda scat­tered across the com­ments here!):

  1. It is ad­van­ta­geous if peo­ple have a shared un­der­stand­ing of the system

  2. Vot­ing your own be­lief ac­tu­ally should work pretty well

  3. There is a writ­ten norm in favour of vot­ing your own belief

  I think we dis­agree on all 3 to some ex­tent, at least in how im­por­tant they are. I think if we lose the dis­agree­ment on num­ber 3 then dis­agree­ments on 1&2 are less im­por­tant.

  I’m ok with a norm of vot­ing based some­what on tar­get karma (mak­ing it overly strong an effect I think would be detri­men­tal), es­pe­cially as this is now com­mon knowl­edge and seems to be most peo­ple’s prefer­ence.

  This whole thing has re­solved some of my con­fu­sion as to why karma scores end up the way they do.

• Bucky 25 Aug 2020 22:22 UTC

  2 points

  in reply to: Dagon’s comment on: Do you vote based on what you think to­tal karma should be?

  Now I fell less stupid for not get­ting it—at least I in­cluded all of the differ­ent parts of the recipe! Very im­pres­sive con­tent den­sity com­ment.

  I have a close-to-de­on­tolog­i­cal be­lief in the need to obey the rules of a com­mu­nity that’s try­ing to cre­ate things to­gether (even when the rules seem wrong) and I think I tend to in­ter­pret things in that frame (for or against) even if that isn’t the in­ten­tion. In the im­mor­tal words of Scott Alexan­der:

  No! I am Ex­cep­tion Nazi! NO EXCEPTION FOR YOU!

• Bucky 25 Aug 2020 21:52 UTC

  4 points

  in reply to: Raemon’s comment on: Do you vote based on what you think to­tal karma should be?

  I think the ass­hole filter is a good point and to be hon­est its pos­si­bly enough to get me to change my mind about this sub­ject. There should be some miti­ga­tion in the karma weight­ing sys­tem but even long term mem­bers might be ass­holes.

  Does it prove too much? Should the cur­rent karma, then, ac­tu­ally be the main con­sid­er­a­tion in de­cid­ing how to vote? Few peo­ple on the site seem will­ing to bite that bul­let. Should I al­most always use my strong votes as if I don’t then an ass­hole might do that and there­fore have an over­sized effect?

  Count me con­fused.

  On the other points I won’t go through point by point. The main thing I think is that what you’re de­scribing is in con­flict with the ex­plicit phras­ing of the rea­sons for vot­ing. Com­pare:

  What should my votes mean?

  We en­courage peo­ple to vote such that up­vote means “I want to see more of this” and down­vote means “I want to see less of this.”

  with

  What should karma in­di­cate?

  The karma on a post is in­tended to in­di­cate whether, and by how much, mem­bers of the site would like to see more of the posts/​com­ments in ques­tion. We en­courage peo­ple to vote ac­cord­ingly.

  The former is the FAQ but I think the lat­ter is what you’re de­scribing. If this is the case then I think this ends up be­ing an ass­hole filter in it­self and the phras­ing in the FAQ should be cor­rected.

  (I re­al­ise as Ray qua user this has noth­ing to do with you but if you can pass this along to Ray qua ad­min that would be great!)

• Bucky 25 Aug 2020 18:19 UTC

  3 points

  in reply to: Dagon’s comment on: Do you vote based on what you think to­tal karma should be?

  I like that fram­ing in the first para­graph.

  In the sec­ond para­graph I can’t work out if the ques­tion is in­tended rhetor­i­cally, iron­i­cally or gen­uinely!

• Bucky 25 Aug 2020 14:24 UTC

  4 points

  in reply to: Raemon’s comment on: Do you vote based on what you think to­tal karma should be?

  IMO you are ad­vo­cat­ing for ba­si­cally switch­ing to a new vot­ing sys­tem, not “prop­erly” im­ple­ment­ing the cur­rent one.

  Com­pare to the LW FAQ:

  What should my votes mean?

  We en­courage peo­ple to vote such that up­vote means “I want to see more of this” and down­vote means “I want to see less of this.”

• Bucky 25 Aug 2020 11:40 UTC

  2 points

  in reply to: Rafael Harth’s comment on: Do you vote based on what you think to­tal karma should be?

  Down­vote a differ­ent post of the same au­thor be­cause I didn’t like that one? That doesn’t sound like a good idea.

  No, I mean why wouldn’t you down­vote a hy­po­thet­i­cal post that you are ag­nos­tic about?

  Imag­ine there are two posts, both have 50 karma.

  You read one and feel con­fi­dent that it is net pos­i­tive but that 50 is too high.

  You read the other and it is not net pos­i­tive for you—you just have a meh re­ac­tion to it.

  It seems very odd to me that one would down­vote the former but not the lat­ter. The net effect is to en­courage peo­ple to read/​write a post that is more likely to provide a meh re­ac­tion than be net pos­i­tive.

• Bucky 25 Aug 2020 10:46 UTC

  2 points

  in reply to: Raemon’s comment on: Do you vote based on what you think to­tal karma should be?

  The fun­da­men­tal prob­lem is that we’re try­ing to map a mul­ti­di­men­sional thing into a sin­gle di­men­sion. When­ever you do this you end up throw­ing out some in­for­ma­tion and you have to do the best you can.

  As de­scribed by jimmy, with the “I want to see more/​less of this” rule you lose some in­for­ma­tion on mag­ni­tude of like/​dis­like with the “I want to see more/​less of this” rule. This is some­what miti­gated by hav­ing weak and strong votes, plus the dither fac­tor jimmy de­scribes (which I think for me is quite sig­nifi­cant) so over­all I’m not hugely wor­ried about this.

  (You can also get some of this in­for­ma­tion back if you’re re­ally in­ter­ested by com­par­ing to­tal num­ber of votes to score al­though this is less ob­vi­ous)

  I’m not sure how a “how much to­tal karma should this post have” rule even works in prac­tice but a cou­ple of op­tions:

  How much karma a post has needs to link to post value /​ cor­rect amount of re­ward to the au­thor.

  If I judge this ac­cord­ing to how much value I per­son­ally got out of it then the great-great grand­par­ent com­ment ap­plies and 50% awe­some, 50% meh posts get 0 karma—a worse re­sult than with the “I want to see more/​less of this” rule, with all of the in­for­ma­tion from the 50% of peo­ple who found it awe­some dis­ap­pear­ing.

  If in­stead I am try­ing to judge how much value I think the av­er­age LWer would get out of it then I think this gets re­ally hard to as­sess. As an ex­am­ple, the re­cent 10 fun ques­tions re­sults showed that peo­ple weren’t very good at guess­ing whether oth­ers be­lieved the Civil­i­sa­tional Inad­e­quacy the­sis more or less than they them­selves did. Here you lose some in­for­ma­tion on peo­ple’s ac­tual opinions in favour of in­for­ma­tion on what other peo­ple think their opinion might be, adding sig­nifi­cant noise to the re­sult.

  Whichever op­tion you chose you prob­a­bly end up throw­ing out in­for­ma­tion on how many peo­ple got value from the post. You can try to get around this by each per­son es­ti­mat­ing how many oth­ers would find it use­ful but I think this just adds more noise to the re­sult.

  You can try to make the rule some com­bi­na­tion of rules (as it seems most peo­ple do) but then to me it seems like in­ter­pret­ing karma scores be­comes re­ally difficult. We also run into the prob­lem of how much weight­ing to give to each sub-rule and if peo­ple give differ­ent weight­ings then you get a dis­crep­ancy in how effec­tive each per­son’s opinion is.

  I’m in­ter­ested if some­one can ex­plain an­other way that a “how much to­tal karma should this post have” rule would work in prac­tice which doesn’t run into such prob­lems.

• Bucky 24 Aug 2020 21:50 UTC

  2 points

  in reply to: Rafael Harth’s comment on: Do you vote based on what you think to­tal karma should be?

  Hmm, in­ter­est­ing—I’m now slightly con­fused by:

  I re­cently strong-down­voted a post that I would have weak-up­voted if it had been at a lower karma

  Was that post good or bad? It sounded to me like you thought the post had value, just not as much as was cur­rently show­ing. If you down­voted a post you thought had pos­i­tive value (you were con­fi­dent that its cur­rent karma value was too high?), why not down­vote one that you don’t see any value in?

  If be­ing ag­nos­tic is a cause for not vot­ing at all, a 50% great, 50% ag­nos­tic post would get a higher score than a 50% great, 50% slightly good post as the slightly good ex­pe­riences would down­vote and the ag­nos­tics wouldn’t.

  I think my main con­cern with the “vote to try to give posts/​com­ments the to­tal karma they should have” rule is that I can’t see a way to op­er­a­tional­ise it which doesn’t suffer from worse prob­lems than the sim­ple “I want to see more/​less of this” rule.

• Bucky 24 Aug 2020 21:02 UTC

  3 points

  in reply to: Bucky’s comment on: Do you vote based on what you think to­tal karma should be?

  Fur­ther, I’m not sure hav­ing a vot­ing con­di­tion of “vote to try to bring the karma to the value you think it should be” helps in this situ­a­tion. If 50% of peo­ple didn’t get any value from a post/​com­ment then they would be try­ing to vote the karma down to 0. So a “50% earth shat­ter­ing, 50% meh” post would end up with ~0 karma.