Benjamin_Todd

• Benjamin_Todd 22 Sep 2020 21:52 UTC

  4 points

  on: How Depen­dent is the Effec­tive Altru­ism Move­ment on Dustin Moskovitz and Cari Tuna?

  I think to­tal EA fund­ing is some­thing like (per year):

  • $250m Open Phil /​ Good Ventures

  • $80m from GiveWell (ex­clud­ing Open Phil)

  • $80m other

  So that’s 61% Open Phil. Note that Open Phil seems to be 90%+ Dustin and Cari.

  Sorry I don’t have sources for these figures—they’re my per­sonal rough es­ti­mates. (Though Open Phil’s grants are al­most all pub­lished.)

  One other thing to note is that fund­ing is only one com­po­nent of EA – we also have the value of the labour of com­mu­nity mem­bers and our ideas. Even if there was no fund­ing at all, we could still ac­com­plish a bunch (by go­ing to work in ex­ist­ing in­sti­tu­tions like govt., academia, non-prof­its, or work­ing as vol­un­teers).

  Quick note on the figures for 80k: about 66% of our fund­ing comes from Open Phil in re­cent years. You can’t di­vide the size of the grant by our an­nual bud­get, be­cause we’re also build­ing up re­serves with the scale of the org. You in­stead need to di­vide the grant by our to­tal in­come.

• Benjamin_Todd 21 Sep 2020 13:23 UTC

  4 points

  in reply to: Linch’s comment on: microCOVID.org: A tool to es­ti­mate COVID risk from com­mon activities

  Th­ese seem like in­ter­est­ing points, but over­all I’m left think­ing there is still a sig­nifi­cant chance of set­ting off a long chain that wouldn’t have hap­pened oth­er­wise. (And even a low­ish prob­a­bil­ity of a long chain means the bulk of the dam­ages are on other peo­ple rather than your self.)

  I think the ar­gu­ment ap­plies to Cal­ifor­nia too. Sup­pose that 20% have already been in­fected, and 0.5% are in­fected cur­rently, and R = 1.

  Then in 6 months, an ex­tra 0.5%64 = 12% will have been in­fected, so 32% will have had it in to­tal. That won’t be enough to cre­ate herd im­mu­nity & pre­vent a long chain.

  An ex­tra in­fec­tion now would in ex­pec­ta­tion cause a chain of 641 = 24 in­fec­tions, and if a vac­cine then came and the dis­ease were stamped out, then those 24 peo­ple wouldn’t have had the dis­ease oth­er­wise.

  What seems to mat­ter is that we’re in a “slow burn” sce­nario, where we’re a de­cently long way from end­ing it, but R ~ 1, but we’re not sure we’re go­ing to reach herd im­mu­nity as the end game.

  PS My figure for Lon­don was a rough bal­l­park from mem­ory—your figures are bet­ter. (Though like I say I don’t think the ar­gu­ment is very sen­si­tive to whether 10% or 30% have already had it.)

• Benjamin_Todd 17 Sep 2020 12:07 UTC

  17 points

  in reply to: Darius_Meissner’s comment on: Some thoughts on EA out­reach to high schoolers

  My take is it’s not the most effec­tive mind­set ei­ther. Per­son­ally I try to fo­cus on giv­ing peo­ple in­for­ma­tion to help them make bet­ter de­ci­sions by their own lights, even rather than the stan­dard mar­ket­ing mind­set.

• Benjamin_Todd 17 Sep 2020 12:02 UTC

  2 points

  in reply to: Linch’s comment on: microCOVID.org: A tool to es­ti­mate COVID risk from com­mon activities

  Think­ing a bit more, I’m not sure this ar­gu­ment works, though I might have mi­s­un­der­stood.

  In Lon­don, 5-10% have been in­fected. Prevalence cur­rently is ~1 in 2000 , R = 1 and let’s as­sume trans­mis­sion time is 1 week. That means that in 6 months time, about an­other 1.5% of peo­ple will have been in­fected (30/​2000)

  If I get in­fected now, then I there will be an ex­tra chain of in­fec­tions 30 peo­ple long.

  I don’t see how the over­all prevalence lev­els block the chain I cause. If in 6 months, an­other 1.5% of peo­ple have been in­fected, that’s not enough to mean­ingfully change R.

  If 5% of peo­ple were in­fected now, and R = 1, then we’d be sat­u­rated and reach herd im­mu­nity in a mat­ter of weeks, which would cut off the chain. But in­stead, the prevalence is suffi­ciently low that it seems like it is pos­si­ble for each in­di­vi­d­ual to cause a long chain.

• Benjamin_Todd 15 Sep 2020 11:36 UTC

  6 points

  in reply to: Ben_West’s comment on: Some thoughts on EA out­reach to high schoolers

  I’m wor­ried about this, though it seems hard to de­liver de­tailed info that ex­plains and backs up our po­si­tions via short videos. One hope would be that once we feel like our core ad­vice is all writ­ten up, we should turn to short videos as an al­ter­na­tive en­try point.

• Benjamin_Todd 15 Sep 2020 11:34 UTC

  26 points

  on: Some thoughts on EA out­reach to high schoolers

  Just a quick com­ment to say this sam­ple doesn’t seem rep­re­sen­ta­tive to me:

  Anec­do­tally, a large num­ber of the most ded­i­cated and promis­ing longter­mist EAs I know heard about EA in high school (at a work­shop I ran for a small group of newish longter­mist EAs, if I re­mem­ber cor­rectly about ⅔ raised their hands when asked if they’d heard about EA be­fore age 18)

  In the EA sur­vey, for peo­ple who said they were ⁵⁄₅ en­gaged, the me­dian age at which they first heard about EA was 22 (mean 24). So a ma­jor­ity of the most en­gaged EAs be­came in­volved later than high school.

  This is also matches my anec­do­tal ex­pe­rience, where uni­ver­sity age seems more com­mon than high school.

  When we plot­ted av­er­age en­gage­ment against age first in­volved, the peak was at 20. Peo­ple who first got in­volved at age 18 were less in­volved on av­er­age, and had a similar av­er­age level of en­gage­ment of peo­ple who first got in­volved at age 40. It’s hard to know what to draw from this (younger and older peo­ple prob­a­bly get less en­gaged be­cause the com­mu­nity is less well set up for them), but I think it means we don’t have clear ev­i­dence that it’s bet­ter to reach peo­ple younger.

• Benjamin_Todd 15 Sep 2020 11:32 UTC

  20 points

  on: Some thoughts on EA out­reach to high schoolers

  Hey Buck,

  Upvoted—I think agree that peo­ple have be­come too nega­tive about it, and I’d be in­ter­ested in see­ing an­other org in this space, though I think I pre­fer top uni­ver­sity out­reach still on av­er­age, so am prob­a­bly still a bit more nega­tive than you.

  A. One rea­son is you didn’t put much em­pha­sis on what I see as some of the more sig­nifi­cant down­sides. One is what Peter Hur­ford says:

  One of the hard­est parts of high school out­reach I think will be get­ting peo­ple to con­tinue to be en­gaged over their col­lege ca­reer (as­sum­ing they go to col­lege), which is four years of sub­stan­tial chance of value drift be­fore any di­rect im­pact hap­pens. Whereas re­cruit­ing from col­lege, the dis­tance is much less.

  With a col­lege stu­dent, you can talk to them about de­ci­sions like mov­ing to an EA hub, work­ing in an EA org etc., which are steps that tend to get peo­ple onto great long-term paths. With a high schooler, you’d need to hope they get in­volved in an EA group while at uni­ver­sity, which sug­gests we’d ideally make the uni­ver­sity groups good first, and even then I think would have higher chance of drift­ing away.

  B. Another is that I think EA ad­vice to high school­ers is less use­ful than what we have to say about later de­ci­sions. The com­mon sense ad­vice of things like “go to a pres­ti­gious uni­ver­sity” and “quan­ti­ta­tive sub­jects keep op­tions open more” and “do in­tern­ships /​ in­ter­est­ing pro­jects /​ build CV ma­te­rial” already seems pretty good to me and are widely known.

  In con­trast, when some­one is work­ing out where to donate or which cause to work on, we think some op­tions are over 100x bet­ter ac­cord­ing to their val­ues in a way that’s not widely known. This isn’t to say many peo­ple don’t make mis­takes in their choice of uni­ver­sity /​ ma­jor or how to spend their time at col­lege, but I still think the delta be­tween hav­ing EA ad­vice and not hav­ing it is smaller.

  Our crux might be this:

  I think the EA and ra­tio­nal­ity com­mu­ni­ties have lots of tools that help peo­ple be­come over­all bet­ter at think­ing, and po­ten­tially vastly in­crease their life­time im­pact.

  My take is more that this ad­vice is use­ful, but it’s not rad­i­cally bet­ter (for most peo­ple) than other sources out there (plus the other down­sides you men­tion). E.g. a bunch of value comes from ideas like ‘ac­tu­ally op­ti­mize for your goals’, but you can get this from other smart self-help ad­vice, fol­low­ing top Sili­con Valley peo­ple etc., read­ing Dalio’s Prin­ci­ples.That’s just one ex­am­ple, and to re­peat, I still think it’s bet­ter to have these ideas than not, just I don’t think the delta is as big.

  On the other hand, if peo­ple had 4 years longer to think about which cause to fo­cus on, and to learn a lot about that topic, that seems pretty use­ful.

  C. The age data I men­tion in the other com­ment.

  D. I think the lack of track record is still a nega­tive E.g. we have a lot of ex­am­ples of great uni­ver­sity groups bring­ing in great peo­ple, so I feel con­fi­dent that start­ing an­other good uni­ver­sity group will be use­ful. Do­ing more high school out­reach is great as an ex­per­i­ment, but over­all I’m less con­fi­dent it’ll work, and it’s also harder to mea­sure. (Though I haven’t seen data from SPARC, which could change my mind.)

• Benjamin_Todd 13 Sep 2020 18:46 UTC

  6 points

  in reply to: weeatquince’s comment on: How can good gen­er­al­ist judg­ment be differ­en­ti­ated from skill at fore­cast­ing?

  I think it would be clearer to put many of these un­der differ­ent cat­e­gories than to lump ev­ery­thing un­der judge­ment. In my post I also cover the fol­low­ing, and try to sketch how they’re differ­ent:

  • Intelligence

  • De­ci­sion-making

  • Strategy

  I should have maybe men­tioned cre­ativity as an­other cat­e­gory.

  I also con­trast ‘us­ing judge­ment’ with al­ter­na­tives like statis­ti­cal anal­y­sis; ap­ply­ing best prac­tice; quan­ti­ta­tive mod­els etc., though you might draw on these in mak­ing your judge­ment.

• Benjamin_Todd 13 Sep 2020 18:38 UTC

  8 points

  in reply to: Buck’s comment on: Judge­ment as a key need in EA

  I think I dis­agree, though that’s just my im­pres­sion. As one piece of ev­i­dence, the ar­ti­cle I most drew on is by Open Phil and also treats them as very re­lated: https://​​www.open­philan­thropy.org/​​blog/​​efforts-im­prove-ac­cu­racy-our-judg­ments-and-forecasts

• Benjamin_Todd 13 Sep 2020 18:33 UTC

  4 points

  in reply to: AGB’s comment on: Judge­ment as a key need in EA

  Hi Alex,

  In the sur­vey, good judge­ment was defined as “weigh­ing com­plex in­for­ma­tion and reach­ing cal­ibrated con­clu­sions”, which is the same rough defi­ni­tion I was us­ing in my post.

  I’m not sure how many peo­ple ab­sorbed this defi­ni­tion and used their own defi­ni­tion in­stead. From talk­ing to peo­ple, my im­pres­sion is that most use ‘judge­ment’ in a nar­rower sense than the dic­tio­nary defi­ni­tions, but maybe still broader than my defi­ni­tion.

  It’s maybe also worth say­ing that my im­pres­sion that judge­ment is highly val­ued isn’t just based on the sur­vey—I high­lighted that be­cause it’s es­pe­cially easy to com­mu­ni­cate. I also have the im­pres­sion that peo­ple of­ten talk about how it might be im­proved, how to as­sess it, as a trait to look for in hiring etc., and it seems to come up more in EA than in most other ar­eas (with cer­tain types of in­vest­ing maybe the ex­cep­tion).

• Benjamin_Todd 12 Sep 2020 20:02 UTC

  4 points

  in reply to: Ozzie Gooen’s comment on: Judge­ment as a key need in EA

  It’s maybe also worth not­ing that my defi­ni­tion of ‘judge­ment’ is pretty nar­row also, and more nar­row than the stan­dard us­age. I’m work­ing on a sep­a­rate piece about ‘good think­ing’ more broadly.

• Benjamin_Todd 12 Sep 2020 19:01 UTC

  6 points

  in reply to: Misha_Yagudin’s comment on: Judge­ment as a key need in EA

  Hi Misha,

  I do agree there’s a worry about how much cal­ibra­tion train­ing or fore­cast­ing in one area, will trans­fer to other ar­eas. My best guess is there some trans­fer but there’s not as much ev­i­dence about it as I’d like.

  I also think of fore­cast­ing as more like a sub­fac­tor of good judge­ment, so I’m not claiming there will be a trans­fer of cog­ni­tive skills – rather I’m claiming that if you prac­tice a spe­cific skill (fore­cast­ing), you will get bet­ter at that skill.

  I’d also sug­gest look­ing di­rectly at the ev­i­dence on whether fore­cast­ing can be im­proved and see­ing what you think of it: https://​​www.open­philan­thropy.org/​​blog/​​efforts-im­prove-ac­cu­racy-our-judg­ments-and-forecasts

• Benjamin_Todd 12 Sep 2020 18:57 UTC

  7 points

  in reply to: Khorton’s comment on: Judge­ment as a key need in EA

  Hi Khor­ton, that’s true. In the post I say:

  Fore­cast­ing isn’t ex­actly the same as good judge­ment, but seems very closely re­lated – it at least re­quires ‘weigh­ing up com­plex in­for­ma­tion and com­ing to cal­ibrated con­clu­sions’, though it might re­quire other abil­ities too. On the other side, I take good judge­ment to in­clude ‘pick­ing the right ques­tions’, which fore­cast­ing doesn’t cover.

  So I think they’re pretty close.

  The other point is that, yes—I think we have some rea­son­able ev­i­dence that cal­ibra­tion and fore­cast­ing can be im­proved (via the things men­tioned in the post), but I’m less con­fi­dent in other ways to im­prove judge­ment. I’ve made some ed­its to the post to make this clearer.

  One other way of im­prov­ing judge­ment in gen­eral that I do men­tion, though, is to spend time talk­ing to other peo­ple who have good judge­ment.

Judge­ment as a key need in EA

• Benjamin_Todd 11 Sep 2020 21:59 UTC

  8 points

  in reply to: Prabhat Soni’s comment on: Prab­hat Soni’s Shortform

  This is a big topic, but I think these cri­tiques mainly fail to ad­dress the core ideas of EA (that we should seek the very best ways of helping), and in­stead crit­i­cise re­lated ideas like util­i­tar­i­anism or in­ter­na­tional aid. On the philos­o­phy end of things, more here: https://​​fo­rum.effec­tivealtru­ism.org/​​posts/​​hvYvH6wabAoXHJjsC/​​philo­soph­i­cal-cri­tiques-of-effec­tive-al­tru­ism-by-prof-jeff

• Benjamin_Todd 10 Sep 2020 19:17 UTC

  2 points

  in reply to: Owen_Cotton-Barratt’s comment on: An ar­gu­ment for keep­ing open the op­tion of earn­ing to save

  e.g. whether there are di­rect work op­por­tu­ni­ties which would have a sig­nifi­cant effect of pass­ing cap­i­tal into the hands of fu­ture longtermists

  Could you say more about what you might have in mind here?

• Benjamin_Todd 10 Sep 2020 19:16 UTC

  3 points

  in reply to: Owen_Cotton-Barratt’s comment on: An ar­gu­ment for keep­ing open the op­tion of earn­ing to save

  That’s fair—I was aiming to write it in a crisp way to make it eas­ier to en­gage with, but I agree I could have given the ar­gu­ment crisply with a bet­ter in­tro­duc­tion.

• Benjamin_Todd 10 Sep 2020 19:14 UTC

  4 points

  on: Thoughts on pa­tient philanthropy

  Thank you for the cer­tainty equiv­a­lent calcu­la­tions, that was in­ter­est­ing.

• Benjamin_Todd 9 Sep 2020 21:28 UTC

  5 points

  in reply to: meerpirat’s comment on: An ar­gu­ment for keep­ing open the op­tion of earn­ing to save

  Yes, I agree that’s an im­por­tant con­sid­er­a­tion. Do­ing di­rect work also causes move­ment build­ing, cre­at­ing a bunch of ex­tra value. (Some even ar­gue that most move­ment build­ing comes from di­rect work rather than ex­plicit move­ment build­ing efforts.) It doesn’t seem like earn­ing to save will be as good on this front, though I think that build­ing up a big pot of money can also get peo­ple in­ter­ested (though maybe for du­bi­ous rea­sons!).

• Benjamin_Todd 3 Sep 2020 20:09 UTC

  4 points

  in reply to: Matt_Lerner’s comment on: More em­piri­cal data on ‘value drift’

  Thank you, that’s helpful!

  Do you mean 21 per­centage points, so if the over­all mean is 23%, then the most en­gaged are only 2%? Or does it mean 21% lower, in which case it’s 18%?

  I’m not aware of a good refer­ence class where we have data—I’d be keen to see more re­search into that.

  It might be worth say­ing that do­ing some­thing like tak­ing the GWWC pledge is still a high level of en­gage­ment & com­mit­ment on the scale of things, and I would guess sig­nifi­cantly higher than the typ­i­cal young per­son af­fili­at­ing with a youth move­ment for a while.

  (The mean & me­dian age in EA is also ~28 right now, so while still on the youth­ful side, it’s not mainly not stu­dents or teenagers.)