EA Diversity: Unpacking Pandora’s Box


I hate talk­ing about di­ver­sity. I hate watch­ing other peo­ple talk about di­ver­sity more*. My first re­ac­tion to the (most) re­cent thread in the EA Face­book group about di­ver­sity was ‘oh God, not this again’**. So I have a lot of sym­pa­thy for the view, most com­monly ex­pressed by so­cial con­ser­va­tives, that as a so­ciety we should spend a whole lot less time think­ing and talk­ing about di­ver­sity and a whole lot more time de­cid­ing on merit. On a purely self­ish level, I would en­joy life more that way.

The rea­sons why I hate this topic are varied, and some of them are too petty or per­sonal to be worth ex­press­ing here, but a few of the more in­tel­lec­tual rea­sons are rele­vant:

  • This whole area is a com­mu­ni­ca­tion minefield; words like ‘di­ver­sity’, ‘poli­ti­cal cor­rect­ness’, and ‘so­cial jus­tice’ mean dra­mat­i­cally differ­ent things to differ­ent peo­ple. Without ac­tive efforts to avoid said mines, par­ti­ci­pants in such a de­bate are likely to end up dis­put­ing defi­ni­tions.

  • The Mind Pro­jec­tion fal­lacy is deadly on a topic where views tend to be both strongly en­trenched and strongly based on (di­ver­gent) per­sonal ex­pe­riences. Large gaps in ex­pe­rience lead to large in­fer­en­tial dis­tances, and large in­fer­en­tial dis­tances make per­sua­sion difficult. That is, it is difficult even when you know they ex­ist, never mind when you don’t.

  • The above com­bine in a nasty way. Opaque or in­com­plete ar­gu­ments can get un­crit­i­cal sup­port from those who agree with the con­clu­sions, partly be­cause their knowl­edge or ex­pe­rience al­low them to un­con­sciously ‘fill in the blanks’ and partly be­cause of the halo effect. And lots of ar­gu­ments, in any de­bate, on any side, are opaque or in­com­plete. The dan­ger is that you end up with two groups talk­ing past each other, nei­ther of them hav­ing the abil­ity and will­ing­ness to bridge the gap, yet both in­creas­ingly con­fi­dent in ‘their’ po­si­tion as they are re­in­forced by ‘their’ group. Which in all prob­a­bil­ity they just shouldn’t be.

Read­ing the above, you could rea­son­ably ask: why am I of all peo­ple writ­ing a post about di­ver­sity? Be­cause at some point I looked at the world and con­sid­ered the ev­i­dence. Un­for­tu­nately for me, my best guess is that di­ver­sity is rather im­por­tant. In par­tic­u­lar, I think that it is too im­por­tant to be left to the type of dis­cus­sion I just de­scribed, and I want to make the case for why it de­serves at least some of our at­ten­tion.

Get­ting it clear

As noted, di­ver­sity can mean a va­ri­ety of differ­ent things. The differ­ent things have differ­ent up­sides. I’m go­ing to non-ex­haus­tively fo­cus on four types:

  1. Diver­sity of tal­ent.

  2. Diver­sity of ex­pe­rience.

  3. Diver­sity of opinion.

  4. Diver­sity of ap­pear­ance.

I want to talk about:

  • What does this mean? Why does it mat­ter?

  • How are we do­ing?

  • What could we do bet­ter?

I cer­tainly do not think I have all the an­swers, es­pe­cially to the last of those. My hope is that by ap­proach­ing things in this way we will at least be ask­ing the right (well-defined, rele­vant, in­ter­est­ing) ques­tions.

Fi­nally, I want to pre-empt the ob­vi­ous re­sponse that these are in­ter­linked; in par­tic­u­lar your ex­pe­riences are linked to your tal­ents, opinions, and ap­pear­ance. My point is that differ­ent types of di­ver­sity have differ­ent up­sides, and that while you can get mul­ti­ple types at once, you of­ten will not. So I still think they’re worth con­sid­er­ing sep­a­rately.


If you want to get things done and done well, you need a va­ri­ety of types of peo­ple to get there. You need the an­a­lyt­i­cally-minded. You need or­ganisers. You need lead­ers. You need com­mu­ni­ca­tors. And so on. Peo­ple have widely differ­ing skills and knowl­edge, and we achieve more by com­ple­ment­ing each other, al­low­ing each of us in­di­vi­d­u­ally to fo­cus on what we’re re­ally good at. This idea ap­pears in eco­nomics at the coun­try level, in the form of com­par­a­tive ad­van­tage. On the level of small groups, the con­cept of vary­ing skills mean­ing more pro­duc­tivity has a long his­tory, in busi­ness and out.

EA is heavy on math­e­mat­i­ci­ans, pro­gram­mers, economists and philoso­phers. Those groups can get a lot done, but they can’t get ev­ery­thing done. If we want to grow, I think we could do with more PR types. Be­cause we’re largely web-based, peo­ple who un­der­stand how to make things vi­su­ally ap­peal­ing also seem valuable. My per­sonal ex­pe­rience in Lon­don is that we would love more or­ganisers, though I can imag­ine this vary­ing by lo­ca­tion.

With that said, this is not an area I’m too wor­ried about in the medium term. We’re do­ing pretty well at elite uni­ver­si­ties. Elite uni­ver­si­ties score poorly on the other types of di­ver­sity I men­tioned, but they do turn out peo­ple who go into a va­ri­ety of fields with a va­ri­ety of strengths; they are di­verse in this sense. Just ex­pand­ing our re­cruit­ment there would fix this prob­lem. On a more per­sonal level, I haven’t had any difficulty talk­ing about EA to my non-quan­ti­ta­tive, non-an­a­lyt­i­cal friends...from elite uni­ver­si­ties. I think that’s be­cause com­mu­ni­ca­tion bar­ri­ers are more of­ten to do with differ­ing ex­pe­riences than differ­ing tal­ents (see next sec­tion).


Differ­ent peo­ple have differ­ent past ex­pe­riences. Those ex­pe­riences are of­ten crit­i­cal in de­cid­ing what you find in­tu­itive, who you can re­late to, and how you ap­proach the world. Try­ing to per­suade some­one of a new idea is hard any­way, and if you have highly di­ver­gent past ex­pe­riences it is dou­bly hard. The things you be­lieve im­me­di­ately might be a big leap for them, and vice-versa; once again in­fer­en­tial dis­tances loom large. If we want to ex­pand, we need vary­ing ex­pe­riences so that we can com­mu­ni­cate effec­tively with a va­ri­ety of peo­ple. Even once peo­ple with differ­ing ex­pe­riences have reached very similar con­clu­sions, those differ­ent routes provide am­ple op­por­tu­ni­ties for learn­ing and deeper un­der­stand­ing. On the flip side, ho­mo­ge­neous ex­pe­riences can eas­ily lead to poor eval­u­a­tion of the choices and op­por­tu­ni­ties available to those with differ­ent ex­pe­riences.

In my opinion, this is where EA is weak­est. We’re over­whelm­ingly young, over­whelm­ingly well-ed­u­cated, and very well-off even by the stan­dards of our WEIRD base. We lack par­ents, we lack peo­ple with years of ex­pe­rience in busi­ness, pub­lic ser­vice or char­ity, and for the most part we lack peo­ple for whom money or ed­u­ca­tion or both has been a real strug­gle. At the risk of stat­ing the ob­vi­ous, it would be very bad if we ended up un­in­ten­tion­ally ex­clud­ing all of these groups.

Some of these is­sues (par­ents, age, ex­pe­rience) will plau­si­bly fix them­selves over time, be­cause cur­rent EAs will en­ter those groups. Some­times we might need to do a lit­tle work to en­sure that peo­ple don’t drop out of EA as they drop in to one of these cat­e­gories. Other is­sues, like differ­ent back­grounds, re­quire more thought, and I’d wel­come ideas oth­ers have on what can be done here. It might be as sim­ple as re-fram­ing our ar­gu­ments de­pend­ing on the au­di­ence, or it might be more fun­da­men­tal.


Differ­ent peo­ple be­lieve differ­ent things. To an ex­tent, we ex­pect ho­mo­ge­neous views within a group of peo­ple who were brought to­gether by their shared goals; this is true in spite of the fact that Effec­tive Altru­ism is a ques­tion. But it should still be pos­si­ble to have a lot of thought di­ver­sity; most EAs are ir­re­li­gious, but our pop­u­lar 10% idea has its roots in tithes. Go­ing fur­ther down that line, re­li­gious peo­ple are more likely to give to char­ity (though the rele­vance of this is dis­puted) and could be con­sid­ered a nat­u­ral tar­get for Singer’s ar­gu­ments. In a similar vein, most EAs are util­i­tar­ian, but there’s noth­ing uniquely util­i­tar­ian about the idea of EA; you can get to much the same con­clu­sion via the Golden Rule. In so far as it is pos­si­ble, we should want thought di­ver­sity, be­cause the ev­i­dence sug­gests that it al­lows groups to find and cor­rect er­rors more quickly. Another way of fram­ing the same thing is that high thought di­ver­sity makes it eas­ier to dis­t­in­guish be­tween facts and ide­ol­ogy, be­cause ide­olo­gies will have their counter-ide­olo­gies ex­pressed within the group.

I’m more op­ti­mistic about this area. We have a good va­ri­ety views on eco­nomics, a no­tice­able but not over­whelming bias against re­li­gion, and a lot of healthy de­bate within the com­mu­nity gen­er­ated by differ­ing opinions on many other ques­tions. Per­haps more im­por­tantly, I think the broad com­mu­nity already recog­nises our abil­ity to ‘agree-to-dis­agree’ on these ar­eas as a Good Thing. This should al­low us to con­tinue in this way and not fall vic­tim to evap­o­ra­tive cool­ing.

Our biggest blind spots are prob­a­bly that we’re short on so­cial con­ser­va­tives and over­whelm­ingly cos­mopoli­tan. Per­haps these are just too core to al­low com­pro­mise on (in which case we might want to re­con­sider that ‘Effec­tive Altru­ism is not an ide­ol­ogy’ claim). But I think that peo­ple who are not on board with these views could eas­ily still be on board with EA. For in­stance, there’s no nec­es­sary con­flict be­tween feel­ing more at­tached to those in one’s own coun­try, as a non-cos­mopoli­tan might, and donat­ing to char­i­ties in other coun­tries, es­pe­cially when the effec­tive­ness gap ap­pears to be at least an or­der of mag­ni­tude. Cer­tainly my greater at­tach­ment to my near and dear doesn’t stop me donat­ing to effec­tive char­i­ties. I some­times think we give up a lit­tle too eas­ily and say ‘well, peo­ple with That Par­tic­u­lar View are never go­ing to be on board with EA’. Per­son­ally, I have more con­fi­dence in our ideas than that.


This is where most de­bates about di­ver­sity start, and sadly where many of them finish, which is why I’ve left it to last. I am mostly talk­ing about race and gen­der here, but the same ar­gu­ment could ap­ply to any quickly iden­ti­fi­able char­ac­ter­is­tic. For ex­am­ple. it prob­a­bly also ap­plies to groups of highly-ed­u­cated peo­ple, who are of­ten iden­ti­fi­able by the way they speak rather than look. What I’m not go­ing to talk about is dis­crim­i­na­tion on an in­di­vi­d­ual level. An un­der­stated point in this de­bate is that you don’t par­tic­u­larly need racists or sex­ists to have vir­tual seg­re­ga­tion on race or gen­der. This is in spite of the fact that Schel­ling’s work on this is al­most 5 decades old. I’ll let Tim Har­ford (see link) talk for me here:

Now these brown eggs aren’t ex­treme racists; they’re happy to live in a mixed neigh­bour­hood. But they don’t want their white neigh­bours to out­num­ber their brown neigh­bours by more than two-to-one...even a mild prefer­ence for the colour of your neigh­bour can lead to ex­treme seg­re­ga­tion...al­though we as in­di­vi­d­u­als may be ra­tio­nal and we may be tol­er­ant, the so­ciety that we pro­duce to­gether may be nei­ther ra­tio­nal nor tol­er­ant.

How we are do­ing here seems in­cred­ibly vari­able by lo­ca­tion, so I don’t want to gen­er­al­ize too much. One sense in which I think we are ar­guably do­ing badly is that many peo­ple aren’t aware of the Schel­ling’s work and also don’t in­tu­itively think in those terms. If some­one is a visi­ble minor­ity-of-one in a group, a Good Bayesian should ex­pect that to be con­tribut­ing to them feel­ing un­com­fortable, even if no­body is ac­tu­ally do­ing any­thing to make them un­com­fortable. That per­son also be­comes more valuable in the group, be­cause their very pres­ence makes the next per­son ‘like them’ feel more com­fortable, and so on. Note how su­perfi­cial this ar­guably is com­pared to other things I’ve de­scribed; merely shar­ing a race should give you less in com­mon with some­one than shar­ing ex­pe­riences. But the speed of be­ing able to iden­tify your un­usu­al­ness mat­ters; first im­pres­sions mat­ter.

Fur­ther discussion

Are there other types of di­ver­sity I haven’t men­tioned, that you think are wor­thy of note?

How would you tackle these is­sues?

Are they even worth tack­ling?

*Denise Melchin can surely vouch for this.

**The thread isn’t ac­tu­ally that bad. As I said, this was my first re­ac­tion, and it was be­fore there were many com­ments.