MichaelDickens

How Can Donors In­cen­tivize Good Pre­dic­tions on Im­por­tant but Un­pop­u­lar Topics?

Should Global Poverty Donors Give Now or Later? An In-Depth Analysis

Why Do Small Donors Give Now, But Large Donors Give Later?

• MichaelDickens 6 Oct 2018 22:18 UTC

  4 points

  on: RPTP Is a Strong Rea­son to Con­sider Giv­ing Later

  I’m not clear on how RPTP fits into a gen­eral un­der­stand­ing of fi­nan­cial re­turns on in­vest­ment. Clearly your RPTP mat­ters, and if you have a lower RPTP than most peo­ple, that makes in­vest­ing look rel­a­tively bet­ter for you. But why don’t, say, fi­nan­cial ad­vi­sors ever talk about this? Ad­vi­sors largely make in­vest­ment recom­men­da­tions based on clients’ risk tol­er­ance, which is un­re­lated to RPTP.

• MichaelDickens 6 Oct 2018 17:17 UTC

  3 points

  on: RPTP Is a Strong Rea­son to Con­sider Giv­ing Later

  I don’t have a di­rect source on the ar­gu­ment that you said Elie Hassen­feld made, but I do have a quote from Scott Alexan­der (http://​​slat­estar­codex.com/​​2013/​​04/​​05/​​in­vest­ment-and-in­effi­cient-char­ity/​​) who went to a live event in which Elie made this ar­gu­ment:

  [I]n the 1960s, the most cost-effec­tive char­ity was child­hood vac­ci­na­tions, but now so many peo­ple have donated to this cause that 80% of chil­dren are vac­ci­nated and the re­main­der are un­reach­able for re­ally good rea­sons (like they’re in vi­o­lent tribal ar­eas of Afghanistan or some­thing) and not just be­cause no one wants to pay for them. In the 1960s, iodiz­ing salt might have been the high­est-util­ity in­ter­ven­tion, but now most of the low-io­dine ar­eas have been iden­ti­fied and cor­rected. While there is still much to be done, we have run out of in­ter­ven­tions quite as easy and cost-effec­tive as those. And one day, God will­ing, we will end malaria and maybe we will never see a char­ity as effec­tive as the Against Malaria [Foun­da­tion] again.

• MichaelDickens 11 Sep 2018 2:52 UTC

  5 points

  in reply to: Flodorner's comment on: Ad­di­tional plans for the new EA Forum

  Another fea­ture that could help peo­ple find old posts is to dis­play a few ran­dom old posts on a side­bar. For ex­am­ple, on any of Jeff Kauf­man’s blog posts, five old posts dis­play on the side­bar. I’ve found lots of in­ter­est­ing old posts on Jeff’s blog via this fea­ture.

• MichaelDickens 24 Jul 2018 1:24 UTC

  0 points

  in reply to: SiebeRozendal's comment on: EA Fo­rum 2.0 Ini­tial Announcement

  I think there’s an­other down­side there: we should be wary of im­ple­ment­ing a sys­tem that doesn’t have a track record. There are lots of fo­rums that don’t have vot­ing, and red­dit-style vot­ing has a long track record as well (plus Hacker News-style, which is similar but not quite the same as red­dit-style). As you start in­tro­duc­ing ex­tra com­plex­ity, you don’t know what’s go­ing to hap­pen. Most pos­si­ble de­signs are bad, and most de­signs we come up with a pri­ori will prob­a­bly be bad, so my in­cli­na­tion would be to stick close to a sys­tem that has a proven track record.

  That said, hav­ing mul­ti­ple types of up­votes could look some­thing like Face­book which now has mul­ti­ple types of likes, and we have at least some idea of what that would look like. So it might be a good idea.

• MichaelDickens 23 Jul 2018 4:57 UTC

  7 points

  in reply to: jpaddison's comment on: EA Fo­rum 2.0 Ini­tial Announcement

  I’m con­cerned with the plans to make vot­ing/​karma more sig­nifi­cant; I would pre­fer to make them less sig­nifi­cant than the sta­tus quo rather than more. Vot­ing al­lows ev­ery­one’s bi­ases to in­fluence dis­cus­sion in bad ways. For ex­am­ple, peo­ple’s votes tend to fa­vor:

  1. things they agree with over things they dis­agree with, which makes it harder to voice dis­sent­ing opinions

  2. en­ter­tain­ing con­tent over im­por­tant but less-en­ter­tain­ing content

  3. agree­able con­tent with­out much sub­stance over niche or dis­agree­able con­tent with lots of substance

  4. posts that raise easy ques­tions and give strong an­swers over posts that raise hard ques­tions and give weak answers

  Sort­ing the front page by votes, and giv­ing high-karma users more vot­ing power, only does more to in­cen­tivize bad habits. I think the cur­rent vot­ing sys­tem is more suited to some­thing like red­dit which is meant for en­ter­tain­ment, so it’s rea­son­able for the most pop­u­lar posts to ap­pear first. If the idea is to have “all of EA’s top re­searchers post­ing and com­ment­ing reg­u­larly”, I don’t think votes should be such a strong driver of the UX.

  About a year ago I es­sen­tially stopped mak­ing top-level posts on the EA Fo­rum be­cause the vot­ing sys­tem both­ers me too much, and the pro­posed change sounds even worse. Maybe I’m an out­lier, but I’d pre­fer a sys­tem that more closely re­sem­bled a tra­di­tional fo­rum with­out vot­ing where all posts have equal sta­tus. That’s prob­a­bly not op­ti­mal and it has its own prob­lems (the most ob­vi­ous be­ing that low-qual­ity con­tent doesn’t get filtered out), but I’d pre­fer it to the cur­rent or pro­posed sys­tem.

• MichaelDickens 19 Apr 2018 2:18 UTC

  0 points

  in reply to: John_Maxwell_IV's comment on: How to im­prove EA Funds

  Al­most all typ­i­cal as­sets—bonds, stocks, com­modi­ties—are highly liquid, in the sense that if you de­cide to sell them, you can con­vert them into cash in a few min­utes max. So even a well di­ver­sified port­fo­lio can still be liquid. The main ex­cep­tions are real es­tate and pri­vate equity, but I see no rea­son why EA Funds need to hold those.

• MichaelDickens 15 Feb 2018 3:14 UTC

  1 point

  in reply to: RyanCarey's comment on: Where Some Peo­ple Donated in 2017

  I don’t know, the link to Zvi’s writeup works for me. But here is the URL: https://​​thezvi.word­press.com/​​2017/​​12/​​17/​​i-vouch-for-miri/​​

Where Some Peo­ple Donated in 2017

• MichaelDickens 12 Dec 2017 17:21 UTC

  12 points

  in reply to: jpaddison's comment on: Four Or­ga­ni­za­tions EAs Should Fully Fund for 2018

  I haven’t yet got­ten around to writ­ing up where I plan on donat­ing in 2018 (I already maxed out my 2017 dona­tions in Fe­bru­ary), but I’ve been think­ing along the same lines. Re­cently I’ve been lean­ing to­ward donat­ing to these smaller, riskier or­ga­ni­za­tions be­cause I see a lot of value in helping new orgs grow and learn­ing what they can ac­com­plish—es­pe­cially be­cause the es­tab­lished char­i­ties that I like best have got­ten a lot of fund­ing re­cently and have room to scale up be­fore they start to hit the limits of their fund­ing.

• MichaelDickens 3 Jun 2017 6:46 UTC

  4 points

  in reply to: MichaelPlant's comment on: Dis­cus­sion: Ad­ding New Funds to EA Funds

  Now that you men­tion it, I think this would be a much more in­ter­est­ing way to di­vide up funds. I have ba­si­cally no idea whether AI safety or anti-fac­tory farm­ing in­ter­ven­tions are more im­por­tant; but given the choice be­tween a “safe, guaran­teed to help” fund and a “moon­shot” fund I would definitely donate to the lat­ter over the former. Di­vid­ing up by cause area does not ac­cu­rately sep­a­rate dona­tion tar­gets along the lines on which I am most con­fi­dent (not sure if that makes sense). I would much rather donate to a fund run by a per­son who shares my val­ues and be­liefs than a fund for a spe­cific cause area, be­cause I’m likely to change my mind about which cause area is best, and per­haps the fund man­ager will, too, and that’s okay.

  Some pos­si­ble axes:

  1. live-im­prov­ing vs. life-sav­ing (or, similarly, to­tal view vs. per­son-af­fect­ing view)

  2. safe bets vs. moonshots

  3. suffer­ing-fo­cused vs. “clas­si­cal”

  4. short-term vs. far future

  Although hav­ing all pos­si­ble com­bi­na­tions just along these axes would re­quire 16 funds so in prac­tice this won’t work ex­actly as I’ve de­scribed.

• MichaelDickens 3 Jun 2017 6:39 UTC

  2 points

  in reply to: Kerry_Vaughan's comment on: Dis­cus­sion: Ad­ding New Funds to EA Funds

  RE #1, or­ga­ni­za­tions do­ing cause pri­ori­ti­za­tion and not EA com­mu­nity build­ing: Copen­hagen Con­sen­sus Cen­ter, Foun­da­tional Re­search In­sti­tute, An­i­mal Char­ity Eval­u­a­tors, ar­guably Global Pri­ori­ties Pro­ject, Open Philan­thropy Pro­ject (which would ob­vi­ously not be a good place to donate, but still fits the crite­rion).

  RE #2: if the point is to do what Nick wants, it should re­ally be a “Nick Beck­stead fund”, not an EA Com­mu­nity fund.

• MichaelDickens 30 May 2017 14:42 UTC

  0 points

  in reply to: Peter_Hurford's comment on: Ex­pected value es­ti­mates we (cau­tiously) took liter­ally—Oxford Pri­ori­ti­sa­tion Project

  Sup­pose it’s 10 years in the fu­ture, and we can look back at what ACE and MIRI have been do­ing for the past 10 years. We now know some new use­ful in­for­ma­tion, such as:

  • Has ACE pro­duced re­search that in­fluenced our un­der­stand­ing of effec­tive char­i­ties?

  • Has MIRI pub­lished new re­search that moved us closer to mak­ing AI safe?

  • Has ACE moved more money to top an­i­mal char­i­ties?

  But even then, we still don’t know nearly as much as we’d like. We don’t know if ACE re­ally moved money, or if that money would have been donated to an­i­mal char­i­ties any­way. Maybe MIRI took fund­ing away from other re­search av­enues that would have been more fruit­ful. We still have no idea how (dis)valuable the far fu­ture will be.

• MichaelDickens 29 May 2017 23:34 UTC

  1 point

  in reply to: jpaddison's comment on: Ex­pected value es­ti­mates we (cau­tiously) took liter­ally—Oxford Pri­ori­ti­sa­tion Project

  I’m still un­de­cided on the ques­tion of whether quan­ti­ta­tive mod­els can ac­tu­ally work bet­ter than qual­i­ta­tive anal­y­sis. (In­deed, how can you even ever know which works bet­ter?) But very few peo­ple ac­tu­ally use se­ri­ous quan­ti­ta­tive mod­els to make de­ci­sions—even if quan­ti­ta­tive mod­els ul­ti­mately don’t work as well as well-or­ga­nized qual­i­ta­tive anal­y­sis, they’re still un­der­rep­re­sented—so I’m happy to see more work in this area.

  Some sug­ges­tions on ways to im­prove the model:

  Ac­count for miss­ing components

  Quan­ti­ta­tive mod­els are hard, and it’s im­pos­si­ble to con­struct a model that ac­counts for ev­ery­thing you care about. I think it’s a good idea to con­sider which parts of re­al­ity you ex­pect to mat­ter most for the im­pact of a par­tic­u­lar thing, and try to model those. What­ever your model is miss­ing, try to figure out which parts of that mat­ter most. You might de­cide that some things are too hard to model, in which case you should con­sider how those hard-to-model bits will likely af­fect the out­come and ad­just your de­ci­sion ac­cord­ingly.

  Ex­am­ples of ma­jor things left out:

  • 80K model only con­sid­ers im­pact in terms of new dona­tions to GWWC based on 80K’s own num­bers. It would be bet­ter to use your own mod­els of the effec­tive­ness of differ­ent cause ar­eas and ac­count for how many peo­ple 80K moves into/​away from these cause ar­eas us­ing your own effec­tive­ness es­ti­mates for differ­ent causes.

  • ACE model only looks at the value from mov­ing money among top char­i­ties. My own model in­cludes money moved among top char­i­ties plus new money moved to top char­i­ties plus the value of new re­search that ACE funds.

  Sen­si­tivity analysis

  The par­tic­u­lar or­der­ing you found (80K > MIRI > ACE > StrongMinds) de­pends heav­ily on cer­tain in­put pa­ram­e­ters. For ex­am­ple, for your MIRI model, “ex­pected value of the far fu­ture” is do­ing tons of work. It as­sumes that the far fu­ture con­tains about 10^17 per­son-years; I don’t see any jus­tifi­ca­tion given. What if it’s ac­tu­ally 10^11? Or 10^50? This hugely changes the out­come. You should do some sen­si­tivity anal­y­sis to see which in­puts mat­ter the most. If any one in­put mat­ters too much, break it down into less sen­si­tive in­puts.

• MichaelDickens 28 Apr 2017 1:10 UTC

  1 point

  in reply to: MichaelDickens's comment on: Up­date on Effec­tive Altru­ism Funds

  Alter­na­tively, you could have global poverty and an­i­mal welfare funds that are un­man­aged and just di­rect money to GiveWell/​ACE top char­i­ties (or maybe have some light man­age­ment to de­ter­mine how to split funds among the top char­i­ties).

• MichaelDickens 24 Apr 2017 3:11 UTC

  4 points

  in reply to: Daniel_Eth's comment on: Up­date on Effec­tive Altru­ism Funds

  There’s no short­age of bad ven­tures in the Valley

  Every time in the past week or so that I’ve seen some­one talk about a bad ven­ture, they’ve given the same ex­am­ple. That sug­gests that there is in­deed a short­age of bad ven­tures—or at least, ven­tures bad enough to get wide­spread at­ten­tion for how bad they are. (Most ven­tures are “bad” in a triv­ial sense be­cause most of them fail, but many failed ideas looked like good ideas ex ante.)

• MichaelDickens 24 Apr 2017 3:09 UTC

  4 points

  in reply to: jpaddison's comment on: Up­date on Effec­tive Altru­ism Funds

  Not sure if this is the right place to say this, but on effec­tivealtru­ism.org where it links to “Donate Effec­tively,” I think it would make more sense to link to GiveWell and ACE ahead of the EA Funds, be­cause GiveWell and ACE are more es­tab­lished and time-tested ways of mak­ing good dona­tions in global poverty and an­i­mal welfare.

  (The down­side is this adds com­plex­ity be­cause now you’re link­ing to two types of things in­stead of one type of thing, but I would feel much bet­ter about CEA en­dors­ing GiveWell/​ACE as the de­fault way to give rather than its own funds, which are con­trol­led by a sin­gle per­son and don’t have the same re­quire­ment (or abil­ity!) to be trans­par­ent.)

• MichaelDickens 4 Feb 2017 18:09 UTC

  7 points

  in reply to: jpaddison's comment on: Select­ing in­vest­ments based on co­var­i­ance with the value of charities

  I’m glad you’re think­ing about this. In­vest­ing is an im­por­tant is­sue and I be­lieve there’s room for more dis­cus­sion of the topic.

  [I]t is com­monly ac­cepted by now that al­tru­ists should gen­er­ally be less fi­nan­cially risk averse than other peo­ple. This im­plies that we shouldn’t worry too much about di­ver­sifi­ca­tion, but only about ex­pected value.

  False. By di­ver­sify­ing, you can in­crease your risk at any given level of re­turn, which also means you can in­crease your re­turn at any given level of risk. (Th­ese are dual op­ti­miza­tion prob­lems).) You should also be con­cerned about cor­re­la­tion with other al­tru­is­tic in­vestors, and most in­vestors put way too much money in their home coun­try (so mostly the US and UK).

  I don’t know that you are claiming this, but you sort of im­ply it, so to be clear: you should not be­lieve that US stocks have higher ex­pected re­turns than any other coun­try. If any­thing, you should be­lieve that the US mar­ket will perform worse than most other coun­tries be­cause it’s sub­stan­tially more ex­pen­sive. Right now the US has a CAPE ra­tio of 26, ver­sus 21 for non-US de­vel­oped mar­kets and 14 for emerg­ing mar­kets. CAPE ra­tio strongly pre­dicts 10-year fu­ture mar­ket re­turns.

  On the co­var­i­ance-with-char­i­ties is­sue: I’m doubt­ful that this con­sid­er­a­tion mat­ters enough to sub­stan­tially change how you should in­vest. If your in­vest­ments can perform 2 per­centage points bet­ter by in­vest­ing in emerg­ing mar­kets rather than de­vel­oped mar­kets (which they prob­a­bly can), I would ex­pect this to out­weigh any benefits from in­creased co­var­i­ance. I would need to see some sort of quan­ti­ta­tive anal­y­sis to be con­vinced oth­er­wise.

  I’m also not con­vinced that we should ac­tu­ally want to in­crease co­var­i­ance rather than de­creas­ing it. By in­creas­ing co­var­i­ance you in­crease ex­pected value by ex­pand­ing the tails, but I don’t be­lieve we should be risk-neu­tral at a global scale be­cause marginal money put into helping the world has diminish­ing util­ity.

  Similar con­cerns ap­ply to in­vest­ing in com­pa­nies that are cor­re­lated with AI de­vel­op­ment. AI com­pa­nies tend to be growth stocks, which un­der­perform the mar­ket in the long run com­pared to value stocks.