Pablo_Stafforini

Yew-Kwang Ng, ‘Effec­tive Altru­ism De­spite the Se­cond-best Challenge’

• Pablo_Stafforini 9 May 2020 0:33 UTC

  2 points

  in reply to: Khorton’s comment on: A cause can be too neglected

  I list a cou­ple of pos­si­ble sources.

• Pablo_Stafforini 9 May 2020 0:32 UTC

  16 points

  on: A cause can be too neglected

  Cas­par Oester­held makes this point in Com­pli­ca­tions in eval­u­at­ing ne­glect­ed­ness:

  I think many in­ter­ven­tions ini­tially face in­creas­ing re­turns from learn­ing/​re­search, cre­at­ing economies of scale, spe­cial­iza­tion within the cause area, etc. For ex­am­ple, in most cause ar­eas, the first $10,000 are prob­a­bly in­vested into pri­ori­ti­za­tion, or­ga­niz­ing, or (po­ten­tially sym­bolic) in­ter­ven­tions that later turn out to be sub­op­ti­mal.

  (I strongly recom­mend this ne­glected (!) ar­ti­cle.)

  Ben Todd makes a re­lated point about char­i­ties (rather than causes) in Stop as­sum­ing ‘de­clin­ing re­turns’ in small char­i­ties:

  Economies of scale are a force for in­creas­ing re­turns, and they win out while still at a small scale, so the im­pact of the 5th staff mem­ber can eas­ily be greater than the 4th.

  Economies of scale are caused by:

  1. Gains from spe­cial­i­sa­tion. In a one per­son or­gani­sa­tion, that per­son has to do ev­ery­thing – mar­ket­ing, mak­ing the product, op­er­a­tions and so on. In a larger or­gani­sa­tion, how­ever, you can hire a spe­cial­ist to do each func­tion, which is more effi­cient.

  2. Fixed costs. Often you have to pay the same amount of money for a ser­vice no mat­ter what scale you have e.g. legally reg­is­ter­ing as an or­gani­sa­tion costs about the same amount of time no mat­ter how large you are; an air­craft with 100 pas­sen­gers re­quires the same num­ber of pi­lots as one with 200 pas­sen­gers. As you be­come larger, fixed costs be­come a smaller and smaller frac­tion of the to­tal.

  3. Phys­i­cal effects. Run­ning an office that’s 2x as large doesn’t cost 2x as much to heat, be­cause the vol­ume in­creases by the cube of the length, while the sur­face area only in­creases by the square of the length. A rule of thumb is that cap­i­tal costs only in­crease 50% in or­der to dou­ble ca­pac­ity.

  What links here?

  • Pablo_Stafforini's comment on A cause can be too neglected by saulius (9 May 2020 0:33 UTC; 2 points)

• Pablo_Stafforini 9 May 2020 0:31 UTC

  0 points

  on: The Im­por­tant/​Ne­glected/​Tractable frame­work needs to be ap­plied with care

  [posted in wrong thread]

• Pablo_Stafforini 8 May 2020 15:29 UTC

  4 points

  in reply to: Sean_o_h’s comment on: I’m Buck Sh­legeris, I do re­search and out­reach at MIRI, AMA

  See also bmg’s LW post, Real­ism and ra­tio­nal­ity. Rele­vant ex­cerpt:

  A third point of ten­sion is the com­mu­nity’s en­gage­ment with nor­ma­tive de­ci­sion the­ory re­search. Differ­ent nor­ma­tive de­ci­sion the­o­ries pick out differ­ent nec­es­sary con­di­tions for an ac­tion to be the one that a given per­son should take, with a fo­cus on how one should re­spond to un­cer­tainty (rather than on what ends one should pur­sue).

  A typ­i­cal ver­sion of CDT says that the ac­tion you should take at a par­tic­u­lar point in time is the one that would cause the largest ex­pected in­crease in value (un­der some par­tic­u­lar frame­work for eval­u­at­ing cau­sa­tion). A typ­i­cal ver­sion of EDT says that the ac­tion you should take at a par­tic­u­lar point in time is the one that would, once you take it, al­low you to ra­tio­nally ex­pect the most value. There are also al­ter­na­tive ver­sions of these the­o­ries—for in­stance, ver­sions us­ing risk-weighted ex­pected value max­i­miza­tion or the crite­rion of stochas­tic dom­i­nance—that break from the use of pure ex­pected value.

  I’ve pretty fre­quently seen it ar­gued within the com­mu­nity (e.g. in the pa­pers “Cheat­ing Death in Da­m­as­cus” and “Func­tional De­ci­sion The­ory”) that CDT and EDT are not “cor­rect” and that some other new the­ory such as func­tional de­ci­sion the­ory is. But if anti-re­al­ism is true, then no de­ci­sion the­ory is cor­rect.

  Eliez­ier Yud­kowsky’s in­fluen­tial early writ­ing on de­ci­sion the­ory seems to me to take an anti-re­al­ist stance. It sug­gests that we can only ask mean­ingful ques­tions about the effects and cor­re­lates of de­ci­sions. For ex­am­ple, in the con­text of the New­comb thought ex­per­i­ment, we can ask whether one-box­ing is cor­re­lated with win­ning more money. But, it sug­gests, we can­not take a step fur­ther and ask what these effects and cor­re­la­tions im­ply about what it is “rea­son­able” for an agent to do (i.e. what they should do). This ques­tion—the one that nor­ma­tive de­ci­sion the­ory re­search, as I un­der­stand it, is gen­er­ally about—is seem­ingly dis­missed as vac­u­ous.

  If this ap­par­ently anti-re­al­ist stance is widely held, then I don’t un­der­stand why the com­mu­nity en­gages so heav­ily with nor­ma­tive de­ci­sion the­ory re­search or why it takes part in dis­cus­sions about which de­ci­sion the­ory is “cor­rect.” It strikes me a bit like an athe­ist en­thus­ti­as­ti­cally fol­low­ing the­olog­i­cal de­bates about which god is the true god. But I’m mostly just con­fused here.

• Pablo_Stafforini 1 May 2020 13:07 UTC

  7 points

  on: List of EA-al­igned email newslet­ters (did I miss any?)

  Two more newslet­ters:

  • ALLFED Newsletter

  • Fore­cast­ing Newsletter

• Pablo_Stafforini 20 Apr 2020 2:35 UTC

  4 points

  on: “Mu­sic we lack the ears to hear”

  It’s nice to know the ori­gin of that phrase.

  I haven’t read Lem’s novel, but I very much en­joyed An­drei Tarkovsky’s film adap­ta­tion. (I agree with Tyler Cowen that “all Takovsky movies are vi­su­ally and con­cep­tu­ally brilli­ant”.)

• Pablo_Stafforini 18 Mar 2020 14:06 UTC

  6 points

  in reply to: MichaelStJules’s comment on: AMA: Toby Ord, au­thor of “The Precipice” and co-founder of the EA movement

  Source for the screen­shot: Sa­muel Kar­lin & Howard E. Tay­lor, A First Course in Stochas­tic Pro­cesses, 2nd ed., New York: Aca­demic Press, 1975.

Phil Tram­mell: The case for ig­nor­ing the world’s cur­rent prob­lems — or how be­com­ing a ‘pa­tient philan­thropist’ could al­low you to do far more good

• Pablo_Stafforini 28 Feb 2020 2:29 UTC

  3 points

  in reply to: JimmyJ’s comment on: Effects of anti-ag­ing re­search on the long-term future

  I’m also in­ter­ested.

  An­ders Sand­berg dis­cusses the is­sue a bit in one of his con­ver­sa­tions with Rob Wiblin for the 80k Pod­cast.

• Pablo_Stafforini 22 Feb 2020 21:01 UTC

  19 points

  in reply to: Matthew_Barnett’s comment on: Why SENS makes sense

  I once read a com­ment on the effec­tive al­tru­ism sub­red­dit that tried to ex­plain why ag­ing didn’t get much at­ten­tion in EA de­spite be­ing so im­por­tant, and I thought it was quite en­light­en­ing.

  For back­ground, here’s the com­ment I wrote:

  Longevity re­search oc­cu­pies an un­sta­ble po­si­tion in the space of pos­si­ble EA cause ar­eas: it is very “hard­core” and “weird” on some di­men­sions, but not at all on oth­ers. The EAs in prin­ci­ple most re­cep­tive to the case for longevity re­search tend also to be those most will­ing to ques­tion the “com­mon-sense” views that only hu­mans, and pre­sent hu­mans, mat­ter morally. But, as you note, one needs to ex­clude an­i­mals and take a per­son-af­fect­ing view to de­rive the “ob­vi­ous corol­lary that cur­ing ag­ing is our num­ber one pri­or­ity”. As a con­se­quence, such po­ten­tial sup­port­ers of longevity re­search end up de­pri­ori­tiz­ing this cause area rel­a­tive to less hu­man-cen­tric or more long-ter­mist al­ter­na­tives.

• Pablo_Stafforini 21 Feb 2020 13:23 UTC

  2 points

  on: Cost-Effec­tive­ness of Aging Re­search

  Cross­posted from Hour­glass Magazine

  The en­tire “mag­a­z­ine” seems to have gone offline. SAD!

• Pablo_Stafforini 20 Feb 2020 18:46 UTC

  5 points

  in reply to: Mathieu Putz’s comment on: Thoughts on elec­toral reform

  Thanks to your com­ment, I can now en­dorse what you said as a more ac­cu­rate and nu­anced ver­sion of the po­si­tion my pre­vi­ous com­ment ex­pressed. Agreed 100%.

• Pablo_Stafforini 19 Feb 2020 13:03 UTC

  3 points

  in reply to: MichaelStJules’s comment on: Thoughts on elec­toral reform

  Yeah, see my re­ply to To­bias.

• Pablo_Stafforini 19 Feb 2020 13:02 UTC

  11 points

  in reply to: Tobias_Baumann’s comment on: Thoughts on elec­toral reform

  I sus­pect that these re­sults are very sen­si­tive to model as­sump­tions, such as tac­ti­cal vot­ing be­havi­our. But it would be in­ter­est­ing to see more work on VSE.

  I agree with this. An ap­proach I find promis­ing is that of Ni­co­laus Tide­man & Florenz Plass­mann. In one study, the au­thors con­sider sev­eral differ­ent statis­ti­cal mod­els, use them to simu­late ac­tual elec­tions, and rank the mod­els by how best they ap­prox­i­mate ac­tual re­sults. Then, in a sub­se­quent study, the au­thors use the top-rank­ing model from their pre­vi­ous study to eval­u­ate a dozen or so al­ter­na­tive vot­ing rules, find­ing that plu­ral­ity, anti-plu­ral­ity, and Buck­lin perform worst. As far as I’m aware, this is the only ex­am­ple of an at­tempt to as­sess vot­ing rules by con­duct­ing simu­la­tions with a model that has been pre-fit­ted to ac­tual elec­tion data. I be­lieve that ex­tend­ing this ap­proach may be among the most im­pact­ful re­search within this cause area.

  What links here?

  • Pablo_Stafforini's comment on Thoughts on elec­toral reform by Tobias_Baumann (19 Feb 2020 13:03 UTC; 3 points)

• Pablo_Stafforini 19 Feb 2020 2:13 UTC

  29 points

  on: Thoughts on elec­toral reform

  Thanks for writ­ing this—I think elec­toral re­form is an in­ter­est­ing and im­por­tant cause area.

  [Ap­proval vot­ing] fails the later-no-harm criterion

  All vot­ing sys­tems vi­o­late in­tu­itively de­sir­able con­di­tions, so not­ing that some sys­tem vi­o­lates some con­di­tion is in it­self no rea­son to fa­vor other sys­tems. One needs to look at the full pic­ture, see what con­di­tions are vi­o­lated by what sys­tems, and pick the sys­tem that min­i­mizes weight-ad­justed vi­o­la­tions. (There is a clear par­allel here be­tween vot­ing the­ory and pop­u­la­tion ethics: im­pos­si­bil­ity the­o­rems have demon­strated in both fields that there ex­ists no vot­ing rule or pop­u­la­tion ax­iol­ogy that satis­fies all in­tu­itively plau­si­ble desider­ata, so vi­o­la­tion of a con­di­tion can’t be ad­duced as a rea­son for re­ject­ing the rule or ax­iol­ogy that vi­o­lates it.)

  But there is a much bet­ter ap­proach, namely, to as­sess differ­ent sys­tems by their “voter satis­fac­tion effi­ciency” (VSE). In­stead of rely­ing on ad­e­quacy con­di­tions, this ap­proach con­sid­ers the prefer­ences that the elec­torate has for ri­val can­di­dates and deals with them us­ing the ap­para­tus of ex­pected util­ity the­ory. Each can­di­date is scored by the de­gree to which they satisfy the prefer­ences of each voter, and then ri­val vot­ing sys­tems are scored by their prob­a­bil­ity of elect­ing differ­ent can­di­dates. Monte Carlo simu­la­tions in­de­pen­dently performed by War­ren Smith, Jame­son Quinn and oth­ers gen­er­ally find that ap­proval vot­ing has higher VSE than in­stant-runoff vot­ing, and that both ap­proval vot­ing and in­stant-runoff vot­ing have much higher VSE than plu­ral­ity vot­ing.

  Given these re­sults, I think the pri­or­ity for EAs is to sup­port whichever al­ter­na­tives to plu­ral­ity vot­ing are most vi­able in a par­tic­u­lar ju­ris­dic­tion, rather than ob­sess over which of these al­ter­na­tives to plu­ral­ity is the ab­solute best. Of course, I also think it makes sense to con­tinue to re­search the field, and es­pe­cially re­fine the mod­els used to com­pute VSE. What EAs definitely shouldn’t do, in my opinion, is to spend con­sid­er­able re­sources dis­cred­it­ing those al­ter­na­tives to one’s own preferred sys­tem, as FairVote has re­peat­edly done with re­spect to ap­proval vot­ing. Much more is gained by dis­plac­ing plu­ral­ity than is lost by re­plac­ing it with a sub­op­ti­mal al­ter­na­tive (for all rea­son­able al­ter­na­tives to plu­ral­ity).

  (In case it isn’t ob­vi­ous, I’m definitely not say­ing that you have done this in your es­say; I’m rather high­light­ing a se­ri­ous failure mode I see in the “vot­ing re­form” com­mu­nity that I be­lieve we should strive to avoid.)

• Pablo_Stafforini 17 Feb 2020 14:00 UTC

  2 points

  in reply to: Joey’s comment on: Em­piri­cal data on value drift

  a quick look would sug­gest ~75% moved from 50% to 10%

  So, to con­firm, are you say­ing that maybe 5 out of the 7 peo­ple who moved out of the 50% cat­e­gory moved in the 10% cat­e­gory? I think it’s im­por­tant to get clar­ity on this, since un­til en­coun­ter­ing this com­ment I was in­ter­pret­ing your post (per­haps un­rea­son­ably) as say­ing that those 7 peo­ple had left the EA com­mu­nity en­tirely. If in fact only a cou­ple of peo­ple in that class left the com­mu­nity, out of a to­tal of 16, that’s a much lower rate of drift than I was as­sum­ing, and more in line with anony­mous’s anal­y­sis of value drift in the origi­nal CEA team.

• Pablo_Stafforini 16 Feb 2020 22:35 UTC

  31 points

  in reply to: lucy.ea8’s comment on: Growth and the case against ran­domista development

  Its in­ter­est­ing to note that I got down­voted for giv­ing ex­cel­lent sources. While you got up­voted for read­ing the ar­ti­cles and com­ment­ing. Ba­si­cally I am out­group/​out­caste in EA.
  

  I’m not sure I’m the right per­son to com­ment on this, given that I’m one of the par­ties in­volved, but I’ll provide my per­spec­tive here any­way in case it is of any help or in­ter­est.

  I don’t think you are char­ac­ter­iz­ing this ex­change or the rea­sons be­hind the pat­tern of votes ac­cu­rately. Bruno asked you to provide a source in sup­port of the fol­low­ing claim, which you made four com­ments above:

  One child policy had no effect on China’s pop­u­la­tion size. It was their wide­spread ed­u­ca­tion pre-1979 than re­duced fer­til­ity.

  In re­sponse to that re­quest, you pro­vided two sources. I looked at them and found that both failed to sup­port the as­ser­tion that “It was [China’s] wide­spread ed­u­ca­tion pre-1979 than re­duced fer­til­ity”, and that one di­rectly con­tra­dicted it.

  I didn’t down­vote your com­ment, but I don’t think it’s un­rea­son­able to ex­pect some peo­ple to down­vote it in light of this rev­e­la­tion. In fact, on re­flec­tion I’m in­clined to fa­vor a norm of down­vot­ing com­ments that in­cor­rectly claim that a schol­arly source sup­ports some propo­si­tion, since such a norm would in­cen­tivize epistemic hy­giene and re­duce the in­ci­dence of in­for­ma­tion cas­cades. I do agree with you that in­group/​out­group dy­nam­ics some­times ex­plain ob­served be­hav­ior in the EA com­mu­nity, but I don’t think this is one of those cases. As one dat­a­point con­firm­ing this, con­sider that a month or two ago, when I pointed out that some­one had mischar­ac­ter­ized the main the­ses of a pa­per, that per­son’s com­ment was heav­ily down­voted, de­spite this user be­ing a reg­u­lar com­menter and not some­one (I think) gen­er­ally per­ceived to be an “out­sider”.

  Mov­ing to the ob­ject-level, in your re­cent com­ment you ap­pear to have mod­ified your origi­nal con­tention. Whereas be­fore your stated that “wide­spread ed­u­ca­tion” was the fac­tor ex­plain­ing China’s re­duced fer­til­ity, now you state that ed­u­ca­tion was one fac­tor among many. Although this differ­ence may seem minor, in the pre­sent con­text it is cru­cial, be­cause both in com­ments to this post and el­se­where in the Fo­rum you have ar­gued that EAs should pri­ori­tize ed­u­ca­tion over growth. Yet if both of these fac­tors ac­count for the fer­til­ity re­duc­tion in China, your po­si­tion can­not de­rive any sup­port from this Chi­nese ex­pe­rience.

• Pablo_Stafforini 15 Feb 2020 18:26 UTC

  27 points

  in reply to: lucy.ea8’s comment on: Growth and the case against ran­domista development

  I ac­tu­ally took the time to look at those two sources, and as far as I can tell they provide no sup­port what­so­ever for your claim that “It was [China’s] wide­spread ed­u­ca­tion pre-1979 that re­duced fer­til­ity.” The word ‘ed­u­ca­tion’ oc­curs ex­actly once in the first ar­ti­cle, and in a sen­tence that doesn’t make any claims about ed­u­ca­tion re­duc­ing fer­til­ity. As for the sec­ond ar­ti­cle, to the ex­tent that it at­tributes the fer­til­ity de­cline to any­thing, it at­tributes it not to “ed­u­ca­tion”, but to eco­nomic de­vel­op­ment (pp. 158-159):

  The third fatal prob­lem with the “400 mil­lion births pre­vented” claim is that it to­tally ig­nores the most sig­nifi­cant source of fer­til­ity de­cline wor­ld­wide: eco­nomic de­vel­op­ment… China’s rapid eco­nomic de­vel­op­ment since 1980 de­serves the lion’s share of the credit for the [fer­til­ity de­cline].
  

• Pablo_Stafforini 15 Feb 2020 16:45 UTC

  2 points

  in reply to: Peter_Hurford’s comment on: EA Sur­vey 2018 Series: How Long Do EAs Stay in EA?

  I just thought it would be valuable to re­calcu­late the es­ti­mated rates of at­tri­tion with this new data, though I think it’s to­tally fine for you to de­pri­ori­tize this.