My Donations for 2016

[Cross-posted from by blog.]

The fol­low­ing ex­plains where I plan to donate in 2016, with some of my think­ing be­hind it. This year, I had $10,000 to al­lo­cate (the sum of my giv­ing from 2015 and 2016, which I lumped to­gether for tax rea­sons; al­though I think this was a mis­take in ret­ro­spect, both due to dis­count rates and be­cause I could have donated in Jan­uary and De­cem­ber 2016 and still re­ceived the same tax benefits).

To start with the punch line: I plan to give $4000 to the EA donor lot­tery, $2500 to GiveWell for dis­cre­tionary grant­ing, $2000 to be held in re­serve to fund promis­ing pro­jects, $500 to GiveDirectly, $500 to the Carnegie En­dow­ment (ear­marked for the Carnegie-Ts­inghua Cen­ter), and $500 to the Blue Rib­bon Study Panel.

For those in­ter­ested in donat­ing to any of these: in­struc­tions for the EA donor lot­tery and the Blue Rib­bon Study Panel are in the cor­re­spond­ing links above, and you can donate to both GiveWell and GiveDirectly at this page. I am look­ing in to whether it is pos­si­ble for small donors to give to the Carnegie En­dow­ment, and will up­date this page when I find out.

At a high level, I par­ti­tioned my giv­ing into two cat­e­gories, which are roughly (A) “help poor peo­ple right now” and (B) “im­prove the over­all tra­jec­tory of civ­i­liza­tion” (these are meant to be rough delineations rather than rigor­ous defi­ni­tions). I de­cided to split my giv­ing into 30% cat­e­gory A and 70% cat­e­gory B. This is be­cause while I be­lieve that cat­e­gory B is the more press­ing and im­pact­ful cat­e­gory to ad­dress in some over­all util­i­tar­ian sense, I still feel a par­tic­u­lar moral obli­ga­tion to­wards helping the ex­ist­ing poor in the world we cur­rently live in, which I don’t feel can be discharged sim­ply by giv­ing more to cat­e­gory B. The 30-70 split is meant to rep­re­sent the fact that while cat­e­gory B seems more im­por­tant to me, cat­e­gory A still re­ceives sub­stan­tial weight in my moral calcu­lus (which isn’t fully util­i­tar­ian or even con­se­quen­tial­ist).

The rest of this post treats cat­e­gories A and B each in turn.

Cat­e­gory A: The Global Poor

Out of $3000 in to­tal, I de­cided to give $2500 to GiveWell for dis­cre­tionary re­grant­ing (which will likely be dis­bursed roughly but not ex­actly ac­cord­ing to GiveWell’s recom­mended al­lo­ca­tion), and $500 to some other source, with the only stipu­la­tion be­ing that it did not ex­actly match GiveWell’s recom­men­da­tion. The rea­son for this was the fol­low­ing: while I ex­pect GiveWell’s recom­men­da­tion to out­perform any con­clu­sion that I per­son­ally reach, I think there is sub­stan­tial value in the ex­er­cise of per­son­ally think­ing through where to di­rect my giv­ing. A few more spe­cific rea­sons:

  • Most im­por­tantly, while I think that offload­ing giv­ing de­ci­sions to a trusted ex­pert is the cor­rect de­ci­sion to max­i­mize the im­pact of any in­di­vi­d­ual dona­tion, col­lec­tively it leads to a bad equil­ibrium where sub­stan­tially fewer and less di­verse brain­power is de­voted to think­ing about where to give. I think that giv­ing a small but mean­ingful amount based on one’s own rea­son­ing largely ame­lio­rates this effect with­out los­ing much di­rect value.

  • In ad­di­tion, I think it is good to build the skills to in prin­ci­ple think through where to di­rect re­sources, even if in prac­tice most of the work is out­sourced to a ded­i­cated or­ga­ni­za­tion.

  • Fi­nally, hav­ing a large num­ber of in­di­vi­d­ual donors check GiveWell’s work and search for al­ter­na­tives cre­ates stronger in­cen­tives for GiveWell to do a thor­ough job (and al­lows donors to have more con­fi­dence that GiveWell is do­ing a thor­ough job). While I know many GiveWell staff and be­lieve that they would do an ex­cel­lent job in­de­pen­dently of ex­ter­nal vet­ting, I still think this is good prac­tice.

Re­lated to the last point: do­ing this ex­er­cise gave me a bet­ter ap­pre­ci­a­tion for the over­all re­li­a­bil­ity, strengths, and limi­ta­tions of GiveWell’s work. In gen­eral, I found that GiveWell’s work was in­cred­ibly thor­ough (more-so than I ex­pected de­spite my high opinion of them), and more­over that they have moved sub­stan­tial money be­yond the pub­li­cized an­nual donor recom­men­da­tions. An ex­am­ple of this is their 2016 grant to IDin­sight. IDin­sight ended up be­ing one of my top can­di­dates for where to donate, such that I thought it was plau­si­bly even bet­ter than a GiveWell top char­ity. How­ever, when I looked into it fur­ther it turned out that GiveWell had already es­sen­tially filled their en­tire fund­ing gap.

I think this anec­dote serves to illus­trate a few things: first, as noted, GiveWell is very thor­ough, and does sub­stan­tial work be­yond what is ap­par­ent from the top char­i­ties page. Se­cond, while GiveWell had already given to IDin­sight, the grant was made in 2016. I think the same pro­cess I used would not have dis­cov­ered IDin­sight in 2015, but it’s pos­si­ble that other pro­cesses would have. So, I think it is pos­si­ble that a mo­ti­vated in­di­vi­d­ual could iden­tify strong giv­ing op­por­tu­ni­ties a year ahead of GiveWell. As a point against this, I think I am in an un­usu­ally good po­si­tion to do this and still did not suc­ceed. I also think that even if an in­di­vi­d­ual iden­ti­fied a strong op­por­tu­nity, it is un­likely that they could be con­fi­dent that it was strong, and in most cases GiveWell’s top char­i­ties would still be bet­ter bets in ex­pec­ta­tion (but I think that merely iden­ti­fy­ing a plau­si­bly strong giv­ing op­por­tu­nity should count as a huge suc­cess for the pur­poses of the over­all ex­er­cise).

To elab­o­rate on why my po­si­tion­ing might be atyp­i­cally good: I already know GiveWell staff and so have some ap­pre­ci­a­tion for their think­ing, and I work at Stan­ford and have sev­eral friends in the eco­nomics de­part­ment, which is one of the strongest de­part­ments in the world for Devel­op­ment Eco­nomics. In par­tic­u­lar, I dis­cussed my giv­ing de­ci­sions ex­ten­sively with a stu­dent of Pas­cal­ine Du­pas, who is one of the world ex­perts in the ar­eas of eco­nomics most rele­vant to GiveWell’s recom­men­da­tions.

Below are speci­fics on or­ga­ni­za­tions I looked into and where I ul­ti­mately de­cided to give.

Ob­ject-level Pro­cess and De­ci­sions (Cat­e­gory A)

My pro­cess for de­cid­ing where to give mostly con­sisted of talk­ing to sev­eral peo­ple I trust, brain­storm­ing and think­ing things through my­self, and a small amount of on­line re­search. (I think that I should likely have done sub­stan­tially more on­line re­search than I ended up do­ing, but my think­ing style tends to benefit from 1-on-1 dis­cus­sions, which I also find more en­joy­able.) The main types of char­i­ties that I ended up con­sid­er­ing were:

  • GiveDirectly (di­rect cash trans­fers)

  • IPA/​JPAL and similar groups (or­ga­ni­za­tions that sup­port aca­demic re­search on in­ter­na­tional de­vel­op­ment)

  • IDin­sight and similar groups (similar to the pre­vi­ous group, but ex­plic­itly tries to do the “trans­la­tional work” of go­ing from aca­demic re­search to ev­i­dence-backed large-scale in­ter­ven­tions)

  • pub­lic in­for­ma­tion cam­paigns (such as Devel­op­ment Me­dia In­ter­na­tional)

  • an­i­mal welfare

  • start-ups or other small groups in the de­vel­op­ment space that might need seed funding

  • meta-char­i­ties such as CEA that try to in­crease the amount of money moved to EA causes (or ev­i­dence-backed char­ity more gen­er­ally)

I ul­ti­mately felt un­sure whether an­i­mal welfare should count in this cat­e­gory, and while I felt that CEA was a po­ten­tially strong can­di­date in terms of pure cost-effec­tive­ness, di­rect­ing funds there felt overly in­su­lar/​meta to me in a way that defeated the pur­pose of the giv­ing ex­er­cise. (Note: two in­di­vi­d­u­als who re­viewed this post en­couraged me to re­visit this point; as a re­sult, next year I plan to look into CEA in more de­tail.)

While look­ing into the “trans­la­tional work” cat­e­gory, I came across one or­ga­ni­za­tion other than IDin­sight that did work in this area and was well-re­garded by at least some economists. While I was less im­pressed by them than I was by IDin­sight, they seemed plau­si­bly strong, and it turned out that GiveWell had not yet eval­u­ated them. While I ended up de­cid­ing not to give to them (based on feel­ing that IDin­sight was likely to do sub­stan­tially bet­ter work in the same area) I did send GiveWell an e-mail bring­ing the or­ga­ni­za­tion to their at­ten­tion.

When look­ing into IPA, my im­pres­sion was that while they have been re­spon­si­ble for some re­ally good work in the past, this was pri­mar­ily while they were a smaller or­ga­ni­za­tion, and they have now be­come large and bu­reau­cratic enough that their fu­ture value will be sub­stan­tially lower. How­ever, I also found out about an in­di­vi­d­ual who was run­ning a small or­ga­ni­za­tion in the same space as IPA, and seemed to be do­ing very good work. While I was un­able to offer them money for rea­sons re­lated to con­flict of in­ter­est, I do plan to try to find ways to di­rect funds to them if they are in­ter­ested.

While pub­lic in­for­ma­tion cam­paigns seem like they could a pri­ori be very effec­tive, briefly look­ing over GiveWell’s page on DMI gave me the im­pres­sion that GiveWell had already con­sid­ered this area in a great deal of depth and pri­ori­tized other in­ter­ven­tions for good rea­sons.

I ul­ti­mately de­cided to give my money to GiveDirectly. While in some sense this vi­o­lates the spirit of the ex­er­cise, I felt satis­fied about hav­ing found at least one po­ten­tially good giv­ing op­por­tu­nity (the small IPA-like or­ga­ni­za­tion) even if I was un­able to give to it per­son­ally, and over­all felt that I had done a rea­son­able amount of re­search. More­over, I have a strong in­tu­ition that 0% is the wrong al­lo­ca­tion for GiveDirectly, and it wasn’t clear to me that GiveWell’s rea­sons for recom­mend­ing 0% were strong enough to over­ride that in­tu­ition.

So, over­all, $2500 of my dona­tion will go to GiveWell for dis­cre­tionary re-grant­ing, and $500 to GiveDirectly.

Tra­jec­tory of Civ­i­liza­tion (Cat­e­gory B)

First, I plan to put $2000 into es­crow for the pur­pose of sup­port­ing any use­ful small pro­jects (speci­fi­cally in the field of com­puter sci­ence /​ ma­chine learn­ing) that I come across in the next year. For the re­main­ing $5000, I plan to al­lo­cate $4000 of it to the donor lot­tery, $500 to the Carnegie En­dow­ment, and $500 to the Blue Rib­bon Study Panel on Biodefense. For the lat­ter, I wanted to donate to some­thing that im­proved medium-term in­ter­na­tional se­cu­rity, be­cause I be­lieve that this is an im­por­tant area that is rel­a­tively un­der-in­vested in by the effec­tive al­tru­ist com­mu­nity (both in terms of money and cog­ni­tive effort). Here are all of the ma­jor pos­si­bil­ities that I con­sid­ered:

  • Donat­ing to the Fu­ture of Hu­man­ity In­sti­tute, with funds ear­marked to­wards their col­lab­o­ra­tion with Allan Dafoe. I de­cided against this be­cause my im­pres­sion was that this par­tic­u­lar pro­ject was not fund­ing-con­strained. (How­ever, I am very ex­cited by the work that Allan and his col­lab­o­ra­tors are do­ing, and would like to find ways to mean­ingfully sup­port it.)

  • Donat­ing to the Carnegie En­dow­ment, re­stricted speci­fi­cally to the Carnegie-Ts­inghua Cen­ter. My un­der­stand­ing is that this is one of the few west­ern or­ga­ni­za­tions work­ing to in­fluence China’s nu­clear policy (though this is based on per­sonal con­ver­sa­tion and not some­thing I have looked into my­self). My in­tu­ition is that in­fluenc­ing Chi­nese nu­clear policy is sub­stan­tially more tractable than U.S. nu­clear policy, due to far fewer peo­ple try­ing to do so. In ad­di­tion, from look­ing at their web­site, I felt that most of the ar­eas they worked in were im­por­tant ar­eas, which I be­lieve to be un­usual for large or­ga­ni­za­tions with mul­ti­ple fo­cuses (as a con­trast, for other or­ga­ni­za­tions with a similar num­ber of fo­cus ar­eas, I felt that roughly half of the ar­eas were ob­vi­ously or­ders of mag­ni­tude less im­por­tant than the ar­eas I was most ex­cited about). I had some reser­va­tions about donat­ing (due to their size: $30 mil­lion in rev­enue per year, and $300 mil­lion in as­sets), but I de­cided to donate $500 any­ways be­cause I am ex­cited about this gen­eral type of work. (This or­ga­ni­za­tion was brought to my at­ten­tion by Nick Beck­stead; Nick notes that he doesn’t have strong opinions about this or­ga­ni­za­tion, pri­mar­ily due to not know­ing much about them.)

  • Donat­ing to the Blue Rib­bon Study Panel: I am ba­si­cally trust­ing Jaime Yas­sif that this is a strong recom­men­da­tion within the area of biodefense.

  • Donat­ing to the ACLU: The idea here would be to de­crease the prob­a­bil­ity that a Pres­i­dent Trump se­ri­ously erodes demo­cratic norms within the U.S. I how­ever cur­rently ex­pect the ACLU to be well-funded (my un­der­stand­ing is that they got a flood of dona­tions af­ter Trump was elected).

  • Donat­ing to the DNC or the Obama/​Holder re­dis­trict­ing cam­paign: This is based on the idea that (1) Democrats are much bet­ter than Repub­li­cans for global sta­bil­ity /​ good U.S. policy, and (2) Repub­li­cans should be pun­ished for helping Trump to be­come pres­i­dent. I ba­si­cally agree with both, and could see my­self donat­ing to the re­dis­trict­ing cam­paign in par­tic­u­lar in the fu­ture, but this in­tu­itively feels less tractable/​un­der­funded than non-par­ti­san efforts like the Carnegie En­dow­ment or Blue Rib­bon Study Panel.

  • Creat­ing a prize fund for in­cen­tiviz­ing im­por­tant re­search pro­jects within com­puter sci­ence: I was origi­nally plan­ning to al­lo­cate $1000 to $2000 to this, based on the idea that com­puter sci­ence is a key field for mul­ti­ple im­por­tant ar­eas (both AI safety and cy­ber se­cu­rity) and that as an ex­pert in this field I would be in a unique po­si­tion to iden­tify use­ful pro­jects rel­a­tive to oth­ers in the EA com­mu­nity. How­ever, af­ter talk­ing to sev­eral peo­ple and think­ing about it my­self, I de­cided that it was likely not tractable to provide mean­ingful in­cen­tives via prizes at such a small scale, and opted to in­stead set aside $2000 to sup­port promis­ing pro­jects as I come across them.

(As a side note: it isn’t com­pletely clear to me whether the Carnegie En­dow­ment ac­cepts small dona­tions. I plan to con­tact them about this, and if they do not, al­lo­cate the money to the Blue Rib­bon Study Panel in­stead.)

In the re­main­der of this post I will briefly de­scribe the $2000 pro­ject fund, how I plan to use it, and why I de­cided it was a strong giv­ing op­por­tu­nity. I also plan to de­scribe this in more de­tail in a sep­a­rate fol­low-up post. Credit goes to Owen Cot­ton-Bar­ratt for sug­gest­ing this idea. In ad­di­tion, one of Paul Chris­ti­ano’s blog posts in­spired me to think about us­ing prizes to in­cen­tivize re­search, and Holden Karnofsky fur­ther en­couraged me to think along these lines.

The idea be­hind the pro­ject fund is similar to the idea be­hind the prize fund: I un­der­stand re­search in com­puter sci­ence bet­ter than most other EAs, and can give in a low-fric­tion way on scales that are too small for or­ga­ni­za­tions like Open Phil to think about. More­over, it is likely good for me to de­velop a habit of eval­u­at­ing pro­jects I come across and think­ing about whether they could benefit from ad­di­tional money (ei­ther be­cause they are fund­ing con­strained, or to in­cen­tivize an in­di­vi­d­ual who is on the fence about car­ry­ing the pro­ject out). Fi­nally, if this effort is suc­cess­ful, it is pos­si­ble that other EAs will start to do this as well, which could mag­nify the over­all im­pact. I think there is some dan­ger that I will not be able to al­lo­cate the $2000 in the next year, in which case any lef­tover funds will go to next year’s donor lot­tery.