cole_haus

• cole_haus 13 Nov 2019 3:02 UTC

  5 points

  in reply to: tylermjohn's comment on: In­sti­tu­tions for Fu­ture Generations

  I like in­clu­sive wealth quite a bit more than some of the other at­tempts I’ve seen be­cause it seems like there’s an ap­peal­ing, co­her­ent the­ory be­hind it. Given that, I think it ex­tends fairly straight­for­wardly to other kinds of is­sues (the pa­per it­self talks about hu­man cap­i­tal, man­u­fac­tured cap­i­tal, nat­u­ral cap­i­tal, and so­cial cap­i­tal) on a con­cep­tual level. The only real re­quire­ment is that you be able to phrase things in terms of stocks and flows and as­sign val­ues to these.

  I think the key difficulty for most ad­di­tional things we’d like to add to in­clu­sive wealth is set­tling on work­able defi­ni­tions and get­ting re­li­able mea­sure­ments/​data.

• cole_haus 13 Nov 2019 2:31 UTC

  1 point

  in reply to: David_Moss's comment on: The (un)re­li­a­bil­ity of moral judg­ments: A sur­vey and sys­tem­atic(ish) review

  Yup, I definitely agree with points 1, 2, and 4.

  I’d have to think and read about this more to move it be­yond pure spec­u­la­tion, but I feel a lit­tle less sym­pa­thetic to the ecolog­i­cal val­idity re­sponse in the do­main of moral judg­ments. It seems plau­si­ble to me that the prob­lems that con­front our moral fac­ul­ties to­day are very dis­similar to those in the en­vi­ron­ment of evolu­tion­ary adapt­ed­ness—even more so than with fac­ul­ties of pru­den­tial ra­tio­nal­ity—and/​or that our moral fac­ul­ties trans­fer less well to rad­i­cally new en­vi­ron­ments.

• cole_haus 13 Nov 2019 2:08 UTC

  2 points

  on: In­sti­tu­tions for Fu­ture Generations

  The Golden Rule sav­ings rate is “the rate of sav­ings which max­i­mizes steady state level or growth of con­sump­tion” and ex­plic­itly named af­ter an in­ter-gen­er­a­tional golden rule.

  It’s not in it­self an in­sti­tu­tion but one could, in prin­ci­ple, im­ple­ment poli­cies which tar­get the recom­mended sav­ings rate.

• cole_haus 13 Nov 2019 2:00 UTC

  2 points

  on: In­sti­tu­tions for Fu­ture Generations

  I think there’s some over­lap with this and the ex­ist­ing liter­a­ture on sus­tain­abil­ity so it might be use­ful to look there too.

  One con­crete ex­am­ple that I like is us­ing some­thing like in­clu­sive wealth as a sup­ple­ment to GDP. By tak­ing ac­count of stocks, rather than sim­ple flows, it ac­counts for the im­pact of cur­rent eco­nomic ac­tivity on fu­ture gen­er­a­tions.

  (In­sti­tu­tion­al­iza­tion of such a met­ric could look some­thing like re­quiring e.g. the U.S. Bureau of Eco­nomic Anal­y­sis to re­port in­clu­sive wealth in ad­di­tion to GDP.)

• cole_haus 13 Nov 2019 1:52 UTC

  8 points

  on: A con­ver­sa­tion with Ro­hin Shah

  Shah thinks there are sev­eral things that could change his be­liefs, in­clud­ing:

  If he learned that evolu­tion ac­tu­ally baked a lot into hu­mans (‘na­tivism’), he would lengthen the amount of time he thinks there will be be­fore AGI.

  Tooby and Cos­mides are big ad­vo­cates for the “mas­sive mod­u­lar­ity” view—a huge amount of hu­man cog­ni­tion takes place in spe­cial­ized, task-tai­lored mod­ules rather than on one big, do­main-gen­eral “com­puter”. Com­mon ex­am­ples of these sorts of mod­ules are:

  • Chom­sky’s uni­ver­sal gram­mar: There’s not enough lan­guage data for chil­dren to learn lan­guages in the ab­sence of in­duc­tive bi­ases.

  • So­cial ex­change: Peo­ple perform much bet­ter at the Wa­son se­lec­tion task when the do­main is so­cial ex­change rather than fully ab­stract.

  Un­for­tu­nately, I don’t know of any re­view col­lect­ing and ex­am­in­ing ev­i­dence for the mas­sive mod­u­lar­ity view.

  (Not sure how much of this Shah already knows.)

• cole_haus 9 Nov 2019 0:42 UTC

  1 point

  on: A bunch of new GPI papers

  I think there’s a miss­ing “max” in the defi­ni­tion of the NBS in “A bar­gain­ing-the­o­retic ap­proach to moral un­cer­tainty”, in case any­one gets/​got con­fused. (I already emailed the au­thor.)

• cole_haus 5 Nov 2019 1:35 UTC

  8 points

  on: De­liber­a­tion May Im­prove De­ci­sion-Making

  One in­ter­est­ing im­pact of de­liber­a­tion is that it can in­duce sin­gle-peaked­ness which makes the prob­lem of so­cial choice eas­ier. Briefly:

  • There are a num­ber of so­cial choice im­pos­si­bil­ity the­o­rems (e.g. Ar­row, Gib­bard-Sat­terth­waite) which say that, in gen­eral, it’s im­pos­si­ble to ag­gre­gate in­di­vi­d­ual prefer­ences into so­cial choices/​prefer­ences with­out giv­ing up on some crite­rion you’d re­ally like to satisfy (e.g. no dic­ta­tors; if ev­ery in­di­vi­d­ual prefers A to B, so does so­ciety).

  • Sin­gle-peaked prefer­ences are prefer­ences where:

    • ev­ery par­ti­ci­pant views op­tions as vary­ing along a sin­gle di­men­sion,

    • ev­ery par­ti­ci­pant has an ideal choice in the set of choices, and

    • op­tions fur­ther away from that ideal choice are less preferred.

  • If you re­strict the do­main of in­put, in­di­vi­d­ual prefer­ences to be sin­gle-peaked, you can es­cape some of the im­pos­si­bil­ity the­o­rems and straight­for­ward ag­gre­ga­tion op­tions be­come available.

  • It’s plau­si­ble that de­liber­a­tion causes a sin­gle di­men­sion to be­come salient for ev­ery­one and for prefer­ences to thereby be­come sin­gle-peaked.

  So­cial Choice The­ory and De­liber­a­tive Democ­racy: A Rec­on­cili­a­tion talks about this kind of thing in more de­tail.

• cole_haus 1 Nov 2019 17:36 UTC

  3 points

  on: The (un)re­li­a­bil­ity of moral judg­ments: A sur­vey and sys­tem­atic(ish) review

  I guess I’ll also use the com­ments to offer my hot take which I don’t think I can im­me­di­ately jus­tify in a fully ex­plicit man­ner:

  Of course our moral judge­ments are un­re­li­able (though not in some of the ways in­ves­ti­gated in the liter­a­ture and al­most cer­tainly in ad­di­tional ways that aren’t in­ves­ti­gated in the liter­a­ture). There are some moral judge­ments which are sta­ble but some which aren’t and even the smaller set of un­sta­ble judge­ments is very con­cern­ing—es­pe­cially for EAs and oth­ers try­ing to max­i­mize (cf. op­ti­mizer’s curse) and tak­ing un­con­ven­tional views. I don’t think the ex­per­tise defense or most of the other re­sponses are par­tic­u­larly suc­cess­ful. I think the “moral en­g­ineer­ing” ap­proach is es­sen­tial. We should be build­ing eth­i­cal sys­tems that ex­plic­itly ac­knowl­edge and ac­count for the noise in our in­tu­itions. For ex­am­ple, some­thing like rule util­i­tar­i­anism seems sub­stan­tially more ap­peal­ing to me now than it did be­fore I looked into this area; the reg­u­lariza­tion im­posed by rule util­i­tar­i­anism limits our abil­ity to fit to noise (cf. bias-var­i­ance trade-off).

• cole_haus 1 Nov 2019 16:58 UTC

  1 point

  in reply to: MichaelPlant's comment on: The (un)re­li­a­bil­ity of moral judg­ments: A sur­vey and sys­tem­atic(ish) review

  Okay, thanks. I added a sec­tion to the sum­mary:

  Meta: This post at­tempts to sum­ma­rize the in­ter­dis­ci­plinary work on the (un)re­li­a­bil­ity of moral judge­ments. As that work con­tains many differ­ent per­spec­tives with no grand syn­the­sis and no clear win­ner (at pre­sent), this post is un­able to offer a sin­gle, neat con­clu­sion to take away. In­stead, this post is worth read­ing if the (un)re­li­a­bil­ity of moral judge­ments seems im­por­tant to you and you’d like to un­der­stand what the cur­rent state of in­ves­ti­ga­tion is.

• cole_haus 1 Nov 2019 15:01 UTC

  1 point

  in reply to: MichaelPlant's comment on: The (un)re­li­a­bil­ity of moral judg­ments: A sur­vey and sys­tem­atic(ish) review

  I’m mostly not try­ing to ar­gue for any par­tic­u­lar con­clu­sion—more try­ing to sum­ma­rize and re­lay the ex­ist­ing work. I was de­liber­ately try­ing to avoid em­pha­sis­ing my idiosyn­cratic take be­cause I didn’t want read­ers to have to sep­a­rate per­sonal spec­u­la­tion from re­portage. (I would have thought the “sur­vey and sys­tem­atic(ish) re­view” in the ti­tle help to set that ex­pec­ta­tion. Are those terms more am­bigu­ous than I un­der­stand them to be?)

  As far as con­sen­sus in the liter­a­ture, there doesn’t seem to be much of one. I think con­sen­sus is/​will be es­pe­cially hard be­cause of the va­ri­ety of re­searchers in­volved—philoso­phers, psy­chol­o­gists, etc. You can see the lack of con­sen­sus re­flected in the wide va­ri­ety of an­gles in “Indi­rect ev­i­dence” and “Re­sponses”.

  Does that all make sense?

The (un)re­li­a­bil­ity of moral judg­ments: A sur­vey and sys­tem­atic(ish) review

• cole_haus 31 Oct 2019 23:56 UTC

  12 points

  on: EA Fo­rum Prize: Win­ners for Septem­ber 2019

  I think even the up­dated ver­sion of Global de­vel­op­ment in­ter­ven­tions are gen­er­ally more effec­tive than cli­mate change in­ter­ven­tions has se­ri­ous mod­el­ing prob­lems. I left a cou­ple of com­ments early on but got no re­sponse. I’ve now added a new com­ment ex­plain­ing my un­der­stand­ing of the prob­lem more fully here.

  I’m com­ment­ing here just to sig­nal boost be­cause it seems like it would be un­for­tu­nate if we built on the cur­rent es­ti­mate which looks to be wrong in im­por­tant ways. (Though it’s also pos­si­ble I’m just con­fused!)

• cole_haus 31 Oct 2019 23:39 UTC

  3 points

  on: [up­dated] Global de­vel­op­ment in­ter­ven­tions are gen­er­ally more effec­tive than Cli­mate change interventions

  Even af­ter the up­date, I’m con­fused by this model. I’ll re­strict my at­ten­tion to the “Real­is­tic case” for sim­plic­ity.

  The global so­cial cost of car­bon when con­structed from the coun­try-level es­ti­mates is effec­tively:

  $$C{C}_G=C{C}_a+C{C}_b+C{C}_c+\dots$$

  where $C{C}_G$ is the global so­cial cost of car­bon and $C{C}_a$, $C{C}_b$, $C{C}_c$, etc. are the coun­try level so­cial costs of car­bon for coun­tries ‘a’, ‘b’, ‘c’, etc.

  The spread­sheet model then uses an in­come ad­just­ment fac­tor of 1,260 ex­press­ing that a dol­lar means more to a poor per­son than a rich per­son. We can also think of this as a unit con­ver­sion fac­tor: 1,260 Amer­i­can me­dian in­come dol­lars = 1 GiveDirectly re­cip­i­ent dol­lar. We can write this con­ver­sion fac­tor as $$\frac{1260{\$}_A}{1{\$}_{GD}}=1$$.

  The next step in the model is to di­vide the global so­cial cost of car­bon by the in­come ad­just­ment fac­tor. If we ex­pand global so­cial cost of car­bon, this looks like:

  $$\frac{C{C}_G}{1260}=\frac{C{C}_a}{1260}+\frac{C{C}_b}{1260}+\frac{C{C}_c}{1260}+\dots$$

  If we just look at the units, this is:

  $${\$}_G\frac{{\$}_{GD}}{{\$}_A}={\$}_a\frac{{\$}_{GD}}{{\$}_A}+{\$}_b\frac{{\$}_{GD}}{{\$}_A}+{\$}_c\frac{{\$}_{GD}}{{\$}_A}+\dots$$

  where $${\$}_G$$ is a coun­try-level-cost-of-car­bon-weighted dol­lar, $${\$}_a$$ is a dol­lar in coun­try ‘a’, etc. Con­vert­ing coun­try ‘a’ dol­lars to GiveDirectly re­cip­i­ent dol­lars via $$\frac{{\$}_{GD}}{1260{\$}_A}$$ is only ap­pro­pri­ate if coun­try ‘a’ is in fact Amer­ica. Other­wise, the units don’t line up.

  It seems like what we’d ac­tu­ally want as far as in­come ad­just­ment is some­thing like:

  $$C{C}_{G-GD}={\$}_a\frac{{\$}_{GD}}{{\$}_a}+{\$}_b\frac{{\$}_{GD}}{{\$}_b}+{\$}_c\frac{{\$}_{GD}}{{\$}_c}+\dots$$

  where $C{C}_{G-GD}$ is the global so­cial cost of car­bon ex­pressed in GiveDirectly re­cip­i­ent dol­lars.

  In other words, we can’t just ap­ply the same in­come-ad­just­ment fac­tor to ev­ery coun­try’s so­cial cost of car­bon be­cause not ev­ery coun­try has the same in­come. We need per-coun­try in­come ad­just­ment fac­tors. Ap­ply­ing the 1,260 in­come ad­just­ment fac­tor to Amer­i­can so­cial cost of car­bon works but in ap­ply­ing this ad­just­ment to the global cost of car­bon, we are also im­plic­itly say­ing that the so­cial cost of car­bon in Burk­ina Faso ought to be dis­counted just as heav­ily.

  I think the only way the cur­rent ap­proach works is if we as­sume that coun­try-level costs have some­how already been ad­justed into Amer­i­can me­dian in­come dol­lars.

  What links here?

  • cole_haus's comment on EA Fo­rum Prize: Win­ners for Septem­ber 2019 by aarongertler (31 Oct 2019 23:56 UTC; 12 points)

• cole_haus 28 Oct 2019 22:05 UTC

  13 points

  on: At­tempt at un­der­stand­ing the role of moral philos­o­phy in moral progress

  Mo­ral Bias and Cor­rec­tive Prac­tices: A Prag­ma­tist Per­spec­tive (the full ar­ti­cle is there if you scroll down) is in­ter­est­ing and some­what rele­vant IMO. It ar­gues that “the moral bi­ases of slav­ery ad­vo­cates proved largely im­mune to cor­rec­tion by the dom­i­nant meth­ods of moral philos­o­phy, which were de­ployed by white abo­li­tion­ists. As­cent to the a pri­ori led to ab­stract moral prin­ci­ples—the Golden Rule, the equal­ity of hu­mans be­fore God—that set­tled noth­ing be­cause their ap­pli­ca­tion to this world was con­tested. Table-turn­ing ex­er­cises were in­effec­tive for similar rea­sons. Reflec­tive equil­ibrium did not clearly fa­vor the abo­li­tion­ists, given au­thor­i­tar­ian, Bibli­cal, and racist premises shared by white abo­li­tion­ists and slav­ery ad­vo­cates.”

• cole_haus 24 Oct 2019 22:43 UTC

  7 points

  on: What eval­u­a­tions are there for in­ter­ven­tions to boost al­tru­ism?

  I don’t know off­hand of any quan­ti­ta­tive es­ti­mates but “moral en­hance­ment” is a po­ten­tially rele­vant area of study. For ex­am­ple:

  • Mo­ral neuroenhancement

  • Model­ing the so­cial dy­nam­ics of moral en­hance­ment: So­cial strate­gies sold over the counter and the sta­bil­ity of society

  • Should we se­lect for ge­netic moral en­hance­ment? A thought ex­per­i­ment us­ing the Mo­ralKin­der (MK+) haplotype

• cole_haus 24 Oct 2019 17:26 UTC

  5 points

  on: Older peo­ple may place less moral value on the far future

  I’ve got­ten the im­pres­sion that elic­it­ing dis­count rates is ac­tu­ally quite tricky and noisy. (From e.g. Time Dis­count­ing and Time Prefer­ence: A Crit­i­cal Re­view which finds dis­count rates vary­ing over many or­ders of mag­ni­tude.) How did you all deal with these sorts of is­sues? For ex­am­ple, what was your think­ing on choos­ing the ques­tion fram­ing you did rather than other frames?

• cole_haus 23 Oct 2019 0:35 UTC

  4 points

  on: Oddly, Bri­tain has never been happier

  Hmm, I just went to OurWor­ldInData and looked at their info.

  • This is the (un­smoothed) Euro­barom­e­ter data you were look­ing at: https://​​our­wor­ld­in­data.org/​​gra­pher/​​share-of-peo­ple-who-say-they-are-happy-eu­ro­barom­e­ter?time=1973..2016&coun­try=GBR.

  • The World Value Sur­vey seems to broadly agree: https://​​our­wor­ld­in­data.org/​​gra­pher/​​share-of-peo­ple-who-say-they-are-happy?time=1998..2009&coun­try=GBR

  • The World Hap­piness Re­port on the other hand seems pretty flatlined: https://​​our­wor­ld­in­data.org/​​gra­pher/​​hap­piness-cantril-lad­der?tab=chart&time=2005..2018&coun­try=GBR.

  • It looks like Euro­barom­e­ter also has a ques­tion about hap­piness (in­stead of life satis­fac­tion), but I couldn’t find an easy way to get a plot of this on­line: https://​​www.gesis.org/​​eu­ro­barom­e­ter-data-ser­vice/​​search-data-ac­cess/​​eb-trends-trend-files/​​list-of-trends/​​happiness

  They’re each slightly differ­ent but it seems use­ful to look at as many data sources as pos­si­ble.

• cole_haus 22 Oct 2019 20:35 UTC

  1 point

  in reply to: mchr3k's comment on: Re­view of Cli­mate Cost-Effec­tive­ness Analyses

  Ahh, if you’re speci­fi­cally look­ing for com­par­i­sons to global health, that makes sense that they’re all EA-af­fili­ated.

• cole_haus 20 Oct 2019 23:26 UTC

  2 points

  on: Re­view of Cli­mate Cost-Effec­tive­ness Analyses

  Thanks for writ­ing this up!

  All four of the es­ti­mates you re­view ap­pear to be (at least loosely) EA-af­fili­ated. Were those the only ones you could find that satis­fied your crite­ria? What were your crite­ria when de­cid­ing which es­ti­mates to re­view?

• cole_haus 17 Oct 2019 20:17 UTC

  1 point

  on: Altru­is­tic equity allocation

  A char­ity can try to miti­gate this risk by un­der­stand­ing how much they would need to re­duce their val­u­a­tion to get ad­di­tional donors in­ter­ested.

  A Vick­rey auc­tion should be a nice way of ad­dress­ing this prob­lem. The char­ity ends up with a com­plete, hon­est list­ing of all bid­der’s val­u­a­tions. Google’s IPO was struc­tured along these lines.

  How is val­u­a­tion de­ter­mined?

  I think it’s also nice to think about this from an “in­side view”. Typ­i­cally, an as­set is priced such that the price equals the net pre­sent value of the ex­pected stream of rents the as­set gen­er­ates. The abil­ity to calcu­late a val­u­a­tion from (slightly) more tan­gible in­puts means that the char­ity doesn’t have to pick a to­tally ar­bi­trary num­ber when they’re start­ing their price-find­ing/​fundrais­ing pro­cess. This defi­ni­tion also helps guide the “Equity in what?” dis­cus­sion.

  No one has an ex­ter­nal in­cen­tive to be­have hon­estly. The sys­tem is in­tended as an aid for gen­er­ally-al­igned donors and em­ploy­ees to ac­tu­ally un­der­stand how much they are con­tribut­ing.

  I know you have this and other caveats through­out, but I definitely worry that it’ll be hard to ig­nore the num­bers when they cre­ate per­verse in­cen­tives (Good­hart’s law, Camp­bell’s law, etc.). Mar­ket mechanisms have lots of prob­lems and I wouldn’t want to im­port those into the char­i­ta­ble sec­tor un­nec­es­sar­ily (which is not to say that mar­kets don’t have good fea­tures or that the char­i­ta­ble sec­tor doesn’t have struc­tural is­sues). I’ll have to think about this more care­fully, but it seems like it would be nice to de­sign a mechanism which is tai­lored for the unique fea­tures of the char­i­ta­ble sec­tor (e.g. in­cen­tives ought to be more al­igned and less zero sum, any po­ten­tial mar­ket is prob­a­bly fairly illiquid).

  Another thing to think about: Equity typ­i­cally con­fers vot­ing rights. Is that ap­pro­pri­ate here? Why or why not? The why not ar­gu­ment that comes to mind im­me­di­ately is: In typ­i­cal for-profit com­pa­nies, vot­ing rights are use­ful as a way of dis­ci­plin­ing man­agers that have their own pri­vate in­cen­tives. Hope­fully, the in­cen­tives of donors and char­ity man­age­ment are already fairly al­igned.