Research on Effective Strategies for Equity and Inclusion in Movement-Building

Cross-posted from the Sen­tience In­sti­tute blog. Origi­nally posted Novem­ber 21, 2018. Some con­tent of the post fo­cuses on the an­i­mal rights, farmed an­i­mal, and effec­tive an­i­mal ad­vo­cacy com­mu­ni­ties, but the recom­men­da­tions are broadly ap­pli­ca­ble to EA and other or­ga­ni­za­tions and com­mu­nity-build­ing efforts.

The con­tents of this post are in­tended to con­vey re­search find­ings only and are not to be re­garded as le­gal ad­vice. Many thanks to Di­ana Fleischman and Aryen­ish Birdie for re­view­ing and pro­vid­ing feed­back. Edited by Jacy Reese.



This post sum­ma­rizes recom­men­da­tions of strate­gies to im­prove equity and in­clu­sion from aca­demic re­search, gov­ern­ment re­search, and re­search done in col­lab­o­ra­tion be­tween con­sult­ing firms and com­pa­nies. I have gen­er­ally ex­cluded recom­men­da­tions based on in­di­vi­d­ual lab stud­ies or heav­ily qual­ified their con­clu­sions. Most of the find­ings are ei­ther ap­pli­ca­ble across axes of in­equity, were gen­er­al­ized from re­search on spe­cific axes, or per­tain speci­fi­cally to gen­der or race (the best-stud­ied axes). The available re­search is thin­ner and more re­li­ant on cor­re­la­tional data than one might ex­pect given that the di­ver­sity train­ing in­dus­try alone is worth $8 billion and more than half the For­tune 500s have di­ver­sity pro­grams or officers, but there is suffi­cient ev­i­dence to provide use­ful guidance to or­ga­ni­za­tions and com­mu­ni­ties seek­ing to be equitable and in­clu­sive.

The gen­eral take­away from this re­search is that or­ga­ni­za­tions will be more effec­tive in efforts to achieve equity if they fo­cus more on im­ple­ment­ing in­clu­sion­ary prac­tices that limit the in­fluence of at­ti­tu­di­nal prej­u­dice on be­hav­ior than on efforts to di­rectly re­duce at­ti­tu­di­nal prej­u­dice.

While di­ver­sity is well-defined — and in this post, I’ll mostly dis­cuss de­mo­graphic di­ver­sity — there is no con­sen­sus in academia or busi­ness on pre­cise defi­ni­tions of equity or in­clu­sion, so I will at­tempt to lay out defi­ni­tions that I find both use­ful and com­mon. “In­clu­sion” gen­er­ally has two mean­ings. It de­scribes the feel­ing of be­ing in­cluded, i.e. be­ing wel­comed and treated equitably. Note that this is some­what dis­tinct from the fact of the mat­ter as to whether a per­son is be­ing treated equitably, as there can be am­bi­guity. In­clu­sion or “in­clu­sion­ary prac­tice” also de­scribes prac­tices that help us to be equitable, such as heuris­tics of trans­parency in de­ci­sions where there may be am­bi­guity about their equitable­ness, or hav­ing an ac­countabil­ity sys­tem in place so that any de­ci­sions that are made in­equitably can be rec­tified.

In­clu­sion’s goal of equity is dis­tinct from a goal of equal­ity. Equity is about equal­ity of op­por­tu­nity, which is not always syn­ony­mous with equal­ity of treat­ment or out­come. Some peo­ple have more bar­ri­ers in the way of their suc­cess than oth­ers, some of which oth­ers have set in place and some of which we our­selves do, and equity is about re­mov­ing those bar­ri­ers to even the play­ing field. Equity is in­stru­men­tally im­por­tant in the same way that be­ing fair and un­bi­ased (which are es­sen­tially syn­ony­mous with equity) are im­por­tant: Equity em­pow­ers in­di­vi­d­u­als to reach their full po­ten­tial, and en­ables teams to benefit from full tal­ent pools. Nondis­crim­i­na­tion, a sig­nifi­cant part of en­sur­ing equity, is also a right held by all hu­mans un­der the United Na­tions’ In­ter­na­tional Bill of Hu­man Rights to fa­cil­i­tate hu­man welfare and co­op­er­a­tion.

My own view is that equity and in­clu­sion are im­por­tant, and gen­er­ally re­sult in di­ver­sity, but that di­ver­sity is mostly only im­por­tant as one of sev­eral in­di­ca­tors of equity and in­clu­sion (i.e. if all else were equal, in­clud­ing if the cul­tural con­text was one of to­tal equity, I’d be am­biva­lent be­tween two job can­di­dates who were iden­ti­cal ex­cept on a de­mo­graphic axis[1]). In prac­tice, the ev­i­dence herein also sug­gests that di­ver­sity’s in­stru­men­tal im­por­tance is, while gen­er­ally pos­i­tive, gen­er­ally only slightly so, with high vari­a­tion de­pend­ing on a group’s goals and the kind of di­ver­sity un­der con­sid­er­a­tion.

Efforts that merely aim to in­crease di­ver­sity, fo­cus­ing for in­stance on the fi­nal de­mo­graph­ics of a team to the ex­clu­sion of other met­rics of equity and in­clu­sion, may be coun­ter­pro­duc­tive to the pro­ject of equity and in­clu­sion be­cause they may in­volve short­cuts, to­k­eniz­ing peo­ple from un­der­rep­re­sented groups and coun­ter­pro­duc­tively cre­at­ing con­di­tions that ul­ti­mately fail to re­cruit and re­tain ex­cel­lent em­ploy­ees from un­der­rep­re­sented groups be­cause they won’t help them feel re­spected, en­gaged, val­ued, and like they be­long.[2] It is these equitable and in­clu­sive con­di­tions that I think we should strive for.


There is am­ple ev­i­dence of the prevalence of work­place bias, which I won’t be­la­bor in this post,[3] but I would like to take a mo­ment to dis­cuss the “busi­ness case” for di­ver­sity and to elab­o­rate on my view that equity and in­clu­sion are the more im­por­tant goals of the DEI trio.

The ev­i­dence of cor­re­la­tions be­tween di­ver­sity and perfor­mance is sub­stan­tial: An anal­y­sis by McKinsey found that “com­pa­nies in the top quar­tile for racial and eth­nic di­ver­sity are 35 per­cent more likely to have fi­nan­cial re­turns above their re­spec­tive na­tional in­dus­try me­di­ans”; that “com­pa­nies in the top quar­tile for gen­der di­ver­sity are 15 per­cent more likely to have fi­nan­cial re­turns above their re­spec­tive na­tional in­dus­try me­di­ans”; that com­pa­nies in the bot­tom quar­tile both for gen­der and for eth­nic­ity and race lag in fi­nan­cial perfor­mance; that ev­ery 10% greater pro­por­tion of non-whites on se­nior-ex­ec­u­tive teams is as­so­ci­ated with 0.8% greater earn­ings be­fore in­ter­est and taxes; and that ev­ery 10% greater pro­por­tion of women on se­nior-ex­ec­u­tive teams is re­lated to 3.5% greater earn­ings be­fore in­ter­est and taxes in the UK.[4] Sev­eral re­cent meta-analy­ses have shown that gen­der di­ver­sity on a board is slightly cor­re­lated with com­pany perfor­mance and cor­po­rate so­cial re­spon­si­bil­ity, and that there are small but con­sis­tent pos­i­tive re­la­tion­ships be­tween women in CEO po­si­tions or top man­age­ment teams and long-term com­pany perfor­mance on fis­cal met­rics.[5] Re­search by hu­man re­sources con­sul­tancy DDI found that com­pa­nies in the top 20% of fi­nan­cial perfor­mance had 37% women lead­er­ship, while those in the bot­tom 20% had only 19%. The Peter­son In­sti­tute for In­ter­na­tional Eco­nomics found that rel­a­tive to hav­ing no women on the board and in the C-suite, “a 30% fe­male share is as­so­ci­ated with a one-per­centage-point in­crease in net mar­gin — which trans­lates to a 15% in­crease in prof­ita­bil­ity for a typ­i­cal firm,” and that while CEO gen­der was un­re­lated to perfor­mance in their anal­y­sis, hav­ing more women in the C-suite is. In part­ner­ship with For­tune, Great Place to Work also found that the 50 com­pa­nies ranked best for di­ver­sity had 24% higher year-over-year rev­enue growth than other com­pa­nies.

Im­por­tantly, these are just cor­re­la­tions, and may be ex­plained by di­ver­sity im­prov­ing a com­pany’s perfor­mance, but may al­ter­na­tively be ex­plained by a third vari­able that makes com­pa­nies both high-perform­ing and di­verse, such as an equitable cul­ture that suc­cess­fully re­cruits top tal­ent re­gard­less of de­mo­graph­ics; an in­clu­sive cul­ture that fosters be­long­ing for all em­ploy­ees; or the lower num­ber of peo­ple from marginal­ized so­cial groups who cur­rently make it into lead­er­ship po­si­tions be­ing more im­pact­ful than their coun­ter­parts from non-marginal­ized groups be­cause they were sub­jected to higher stan­dards of com­pe­tence to ac­cess the same po­si­tions. In other words, while we can say that di­ver­sity — par­tic­u­larly the bet­ter-stud­ied cases of gen­der and racial or eth­nic di­ver­sity — is cor­re­lated to com­pany perfor­mance, it’s not clear whether merely in­creas­ing di­ver­sity — in terms of the de­mo­graph­ics of a team — will cause in­creased perfor­mance. Publi­ca­tion bias may also limit this ev­i­dence if we ex­pect re­ports show­ing nega­tive cor­re­la­tions be­tween di­ver­sity and perfor­mance to be harder to pub­lish.

There is some ev­i­dence that gen­der di­ver­sity does pre­dict rev­enue in­de­pen­dent of em­ployee en­gage­ment, and ev­i­dence that hiring more women im­proves the perfor­mance of ven­ture cap­i­tal firms. How­ever, other stud­ies of bio-de­mo­graphic di­ver­sity (as op­posed to task-re­lated di­ver­sity) show no de­tectable over­all effect on team perfor­mance, and mixed effects on in­no­va­tion (with gains in idea cre­ation ap­par­ently get­ting lost in idea se­lec­tion and im­ple­men­ta­tion). It’s pos­si­ble, though, that suffi­ciently in­clu­sive teams could reap the stud­ied benefits of di­ver­sity while miti­gat­ing the challenges. And since it seems likely that peo­ple from marginal­ized groups face more day-to-day ob­sta­cles, time costs, and stress than those of more priv­ileged groups at the same ap­par­ent perfor­mance lev­els, which pre­sum­ably pre­vents them from perform­ing as well as they would in a fully equitable world, it’s also pos­si­ble that the benefits of di­ver­sity will in­crease as com­mu­ni­ties be­come more equitable. But these ar­gu­ments are fairly spec­u­la­tive.

A stronger case for valu­ing di­ver­sity as more than a mere met­ric of equity is that, rel­a­tive to cor­po­ra­tions, in a so­cial move­ment, visi­ble de­mo­graphic di­ver­sity may be more im­por­tant as pub­lic re­la­tions, figure­heads, and in­ter­per­sonal out­reach are all pre­sum­ably more im­por­tant to sel­l­ing an ide­ol­ogy than they are to sel­l­ing a product, and peo­ple might be more read­ily per­suaded and re­cruited by peo­ple they iden­tify with (though they might not be).[6] And of course, the farmed an­i­mal move­ment is a global move­ment that we want to see scale dra­mat­i­cally, and the US isn’t go­ing to have any sin­gle race as the ma­jor­ity within just a few decades, so our move­ment-build­ing efforts could be neg­li­gently limited if we failed to re­cruit al­lies from differ­ent de­mo­graphic back­grounds and missed out on mas­sive num­bers of po­ten­tial sup­port­ers.

Given all of that, and the rea­sons stated at the be­gin­ning of this post to value equity and in­clu­sion, I think the an­i­mal ad­vo­cacy move­ment should fo­cus on equity and in­clu­sion much more than di­ver­sity (which should still be used as one of sev­eral met­rics of equity and in­clu­sion, and should also be val­ued some­what on its own in the in­ter­est of pub­lic rep­re­sen­ta­tion).


Within the an­i­mal ad­vo­cacy move­ment, to name a few ex­am­ples of in­equity which af­fect large num­bers of ad­vo­cates and are fairly easy to dis­cern, there is sig­nifi­cant room to in­clude and em­power peo­ple of color; to em­power women in lead­er­ship; and to en­sure the re­cent swell of in­ter­est in and ac­tion on sex­ual ha­rass­ment re­sults in sus­tained change. Given the prej­u­dices in so­ciety at large, the move­ment pre­sum­ably has room to im­prove on other axes of ex­clu­sion too, such as ageism, ableism, and cis­sex­ism. My hope is that the recom­men­da­tions in this doc­u­ment will en­able us to im­prove on all axes of ex­clu­sion.

Ra­cial in­equity seems worse than gen­der in­equity in the ranks of the move­ment, as in­di­cated by the large ma­jor­ity of ad­vo­cates be­ing women while a mere ~10% of staff at farmed an­i­mal or­ga­ni­za­tions sur­veyed by En­com­pass are peo­ple of color. The per­centage of peo­ple of color in the gen­eral pop­u­la­tion is roughly four times that; rates of veg­e­tar­i­anism are con­sis­tent across racial groups; and peo­ple of color may even be slightly more op­posed to an­i­mal farm­ing than white peo­ple. So we seem to have sig­nifi­cant room to im­prove racial equity and in­clu­sion.

Rates of peo­ple of color and women in lead­er­ship ap­pear to have similar rate re­duc­tions of around a third, rel­a­tive to the ranks [7] (while I es­ti­mate men have around dou­ble the rep­re­sen­ta­tion in lead­er­ship that they do in the ranks), sug­gest­ing that lead­er­ship has a ceiling of similarly low per­me­abil­ity for both groups.

This po­ten­tially higher pri­or­ity for racial equity may be a mostly im­ma­te­rial con­sid­er­a­tion in prac­tice as most of the strate­gies recom­mended in this re­port should re­duce in­equity gen­er­ally be­cause they em­power peo­ple from a mul­ti­tude of marginal­ized groups, and there are gen­er­ally few situ­a­tions where we have to make trade­offs be­tween gen­der and racial equity efforts. When we do, for in­stance in de­cid­ing who to men­tor, we can prob­a­bly usu­ally hit both of these tar­gets at once by pri­ori­tiz­ing women of color.

Keep in mind that in­equitable be­hav­ior within the move­ment is only part of the ex­pla­na­tion for in­equities in the move­ment — though po­ten­tially a sub­stan­tial part — as var­i­ous ex­ter­nal fac­tors may di­vert peo­ple from marginal­ized groups in ear­lier parts of the pipeline to the move­ment.



Or­ga­ni­za­tions should keep an eye on their whole pipeline: at­tract­ing, hiring, de­vel­op­ing, and re­tain­ing tal­ent. Beyond im­ple­ment­ing the strate­gies recom­mended be­low, in or­der to iden­tify team-spe­cific gaps and re­fine in­clu­sion­ary efforts large or­ga­ni­za­tions can col­lect data on the re­la­tion­ships be­tween de­mo­graph­ics and, for in­stance, hiring stages, pro­mo­tion rates, tenure/​turnover, com­pen­sa­tion,[8] perfor­mance scores, uti­liza­tion of the pro­fes­sional re­sources offered by the or­ga­ni­za­tion, em­ployee en­gage­ment, be­long­ing, and per­cep­tions of the cul­ture. Lead­er­ship should com­mit to ad­dress­ing any gaps and have some form of ac­countabil­ity in place (e.g. a DEI officer or com­mit­tee) to en­sure they make those efforts. Smaller teams can still try to as­sess and act on these met­rics of equity and feel­ings of in­clu­sion even though their sam­ple sizes are smaller, con­ver­sa­tions may be more suit­able than sur­veys, and they gen­er­ally have less to gain from such or­ga­ni­za­tional in­fras­truc­ture work.

At­tract­ing and hiring talent

Re­search sug­gests that or­ga­ni­za­tions should:

  • Hire in rounds, eval­u­at­ing can­di­dates side-by-side, not on a rol­ling ba­sis.

  • En­sure “re­quire­ments” listed in a job post­ing only list qual­ifi­ca­tions that are gen­uinely re­quired, or many qual­ified women won’t ap­ply even though un­der­qual­ified men will.

  • De­mon­strate a “growth” mind­set rather than a “fixed” mind­set in job post­ings. This means us­ing qual­ifi­ca­tions like “highly mo­ti­vated” in­stead of “high performer” or “com­mit­ment to im­prove­ment” in­stead of “over­achiever.”

  • Try to use gen­der-neu­tral or gen­der-bal­anced lan­guage, or lean a lit­tle fem­i­nine-typed in job post­ing word choices, to re­duce bar­ri­ers to ap­pli­ca­tion for women. (This does not seem to have a trade­off with ap­pli­ca­tions from men.) Tools like Tex­tio can help.

  • Make a po­si­tion’s typ­i­cal wage range clear, as women are more likely to ap­ply if a wage range, or at least a men­tion of the room for ne­go­ti­a­tion, is posted. This may be be­cause they are less con­fi­dent than men that they will be able to ne­go­ti­ate suc­cess­fully un­der am­bi­guity, which is ac­cu­rate.

  • De­fault to “flex­ible hours” for any roles where that’s pos­si­ble.

  • In gen­eral it’s prob­a­bly best not to anonymize ap­pli­ca­tions. Field stud­ies gen­er­ally show no effect on in­ter­view se­lec­tion, and some­times even show a nega­tive effect (which has also been seen in the lab). Blind­ing may work for mu­si­ci­ans, ran­domly gen­er­ated re­sumes, and iden­ti­cal ex­pres­sions of in­ter­est, but in re­al­ity there seem to be sub­tle cues of an ap­pli­cant’s back­ground that eval­u­a­tors may pick up on, and the risk of anonymiza­tion back­firing is higher for re­cruit­ing groups which are ac­tively in­ter­ested in DEI. This may be be­cause they are un­able to proac­tively check their bi­ases when blind, or to proac­tively ac­com­mo­date dis­ad­van­taged can­di­dates at this re­cruit­ment stage, or be­cause their staff is already more di­verse and peo­ple may fa­vor can­di­dates they iden­tify with de­mo­graph­i­cally. There may be cir­cum­stances where this is more use­ful, per­haps such as when re­cruiters know many ap­pli­cants per­son­ally, and this may or may not also ap­ply to other parts of the hiring pro­cess, such as trial pro­jects.

  • If us­ing men­tal abil­ity tests, use them in tan­dem[9] with other pre­dic­tors of job perfor­mance to min­i­mize dis­crim­i­na­tion. A re­cent meta-anal­y­sis sug­gests that gen­eral men­tal abil­ity tests are the strongest sole pre­dic­tor of job perfor­mance, and that of other pre­dic­tors tested in com­bi­na­tion with it, in­tegrity tests and struc­tured in­ter­views re­sult in the great­est pre­dic­tive gain, though this re­search may have sig­nifi­cant limi­ta­tions.[10] Note that per US law, men­tal abil­ity tests must be nondis­crim­i­na­tory and a valid pre­dic­tor of perfor­mance for the role. Note, too, that even with those con­di­tions satis­fied, their use may still skew ap­pli­cant pools by, for in­stance, putting off peo­ple of color, who may sus­pect that the tests will dis­crim­i­nate against them or be used se­lec­tively to dis­crim­i­nate against them.

  • In gen­eral, look­ing at con­text perfor­mance in ad­di­tion to task perfor­mance prob­a­bly makes it less likely that ca­pa­ble can­di­dates from marginal­ized groups will be in­equitably filtered out.

Re­search has un­clear find­ings re­gard­ing:

  • Whether job post­ings should in­clude an equal em­ploy­ment op­por­tu­nity state­ment, which may make the or­ga­ni­za­tion look more fa­vor­able to can­di­dates from marginal­ized groups, but may also back­fire by in­di­cat­ing po­ten­tial to­kenism. As such, if state­ments are used it’s prob­a­bly best if they ex­press that the or­ga­ni­za­tion’s com­mit­ment to equal em­ploy­ment is out of an in­ter­est in as­sess­ing ap­pli­cants equitably, to miti­gate that po­ten­tial con­cern about a mere in­ter­est in su­perfi­cial di­ver­sity.

  • Quo­tas, for which ev­i­dence is mixed.

Other strate­gies which I have not seen re­search on, but which fol­low the gen­eral prin­ci­ple of in­creas­ing stan­dard­iza­tion and oth­er­wise limit­ing se­lec­tion effects and the in­fluence of bi­ases, in­clude:

  • Publi­cly post job open­ings, be­cause pri­vate refer­rals skew the ap­pli­cant pool to­wards peo­ple like those in or­ga­ni­za­tion’s the ex­ist­ing com­mu­nity. Re­cruiters should gen­er­ally try to ac­tively source out­side can­di­dates, and can use spe­cial re­cruit­ment pro­grams to re­cruit peo­ple from un­der­rep­re­sented groups.

  • Do your best to start eval­u­at­ing can­di­dates only af­ter they have seen a job de­scrip­tion and ap­plied to the role. Pre-ap­pli­ca­tion eval­u­a­tion can lead re­cruiters to an­chor their views on in­com­plete, bi­ased, and less rele­vant in­for­ma­tion (such as per­sonal fit with the re­cruiter or abil­ity to think quickly un­der pres­sure), in­stead of al­low­ing each can­di­date to put their best foot for­ward.

  • Com­mit to hiring crite­ria be­fore post­ing a job open­ing or in­ter­view­ing any­one. Other­wise, re­cruiters may in­ad­ver­tently form their crite­ria in­equitably based on stereo­types and per­sonal bi­ases as they en­counter ap­pli­cants.

  • Make these crite­ria strictly what you are ac­tu­ally look­ing for, such as “adapt­abil­ity,” “will­ing­ness to up­date views,” or “cre­ative rea­son­ing” (the lat­ter two of which we’ve used at Sen­tience In­sti­tute) rather than prox­ies like “start-up ex­pe­rience,” “ivy-league ed­u­ca­tion,” or “pre­vi­ous work at a top mar­ket­ing firm,” which will un­nec­es­sar­ily re­strict the can­di­date pool, fil­ter­ing out peo­ple who have the skills an or­ga­ni­za­tion is look­ing for but for who did not choose or were in­equitably de­nied ac­cess to those spe­cific paths. Th­ese re­sume items can be used as ev­i­dence of whether a can­di­date meets a crite­ria, but are limit­ing and in­suffi­cient on their own. Re­lat­edly, pre­fer con­crete ev­i­dence of whether a can­di­date meets a crite­ria, such as by mea­sur­ing a crite­ria of “good with donors” with ev­i­dence such as “suc­cess­ful ex­pe­rience with ma­jor donors” rather than vague and in­tu­itive as­sess­ments of whether a can­di­date is “ami­able” (as op­posed to too “ag­gres­sive” or “abra­sive”), which leave am­ple room for stereo­types and other bi­ases to cloud judge­ment.

  • Only look for traits that are ac­tu­ally re­quired to perform the role well; oth­er­wise unim­por­tant traits may be pri­ori­tized at the cost of im­por­tant ones. For in­stance, while ba­sic al­ign­ment with an or­ga­ni­za­tion’s mis­sion is gen­er­ally vi­tal, the high-fidelity value al­ign­ment that may be shared by the board, founders, or ex­ec­u­tive lead­er­ship (e.g. liber­a­tionism at a farmed an­i­mal or­ga­ni­za­tion that fo­cuses on welfare re­forms) is prob­a­bly not nec­es­sary for staff who do not guide the group on big-pic­ture strat­egy to do their work effec­tively.

  • Work to avoid ex­po­sure to non­di­ag­nos­tic in­for­ma­tion about a can­di­date to limit the in­tro­duc­tion of per­sonal bi­ases. My own ex­pe­rience and the ex­pe­riences of col­leagues sug­gests that re­cruiters should give lit­tle to no value to “cul­tural fit,” ex­cept for the most ba­sic facets such as col­lab­o­ra­tive­ness for a highly col­lab­o­ra­tive team or adapt­abil­ity for a rapidly chang­ing work en­vi­ron­ment. I worry that “cul­tural fit” of­ten places un­due weight — of­ten un­in­ten­tion­ally — on unim­por­tant crite­ria like drink­ing and so­cial­iz­ing habits, shared hob­bies, or sense of hu­mor.

  • Make the ap­pli­ca­tion pro­cess (e.g. steps in­volved, es­ti­mated de­ci­sion dates) trans­par­ent early on to miti­gate con­cerns peo­ple from marginal­ized groups may oth­er­wise an­ti­ci­pate of an un­struc­tured pro­cess with too much room for sub­jec­tivity and as such dis­crim­i­na­tion.

  • Score all can­di­dates on a rubric — the same rubric, and the one cre­ated be­fore­hand and on which posted qual­ifi­ca­tions were based. Due to stereo­types and prej­u­dice, it’s pos­si­ble that peo­ple of differ­ent de­mo­graph­ics tend to be eval­u­ated on differ­ent crite­ria.

  • Re­lat­edly, in­ter­view all ap­pli­cants from marginal­ized groups who meet the same (or higher) posted qual­ifi­ca­tions as peo­ple from priv­ileged groups who are be­ing in­ter­viewed. In other words, set the bar at the same po­si­tion for ev­ery ap­pli­cant. Since it’s pos­si­ble that most bias in a tal­ent eval­u­a­tion pipeline hap­pens in the ini­tial digi­tal stages be­fore can­di­dates are eval­u­ated face-to-face,[11] this may be a par­tic­u­larly im­por­tant bot­tle­neck.

  • As noted ear­lier, more un­der­qual­ified men ap­ply to po­si­tions than com­pa­rably qual­ified (or more qual­ified but still not-quite-qual­ified) women, which means that loos­en­ing crite­ria when re­view­ing ap­pli­ca­tions, af­ter that filter, will skew the pool to­wards ap­pli­cants who are both men and less qual­ified com­pared to if those crite­ria were loos­ened in the job post­ing.

  • De­sign in­ter­view ques­tions that re­late as closely and clearly as pos­si­ble to spe­cific crite­ria, in or­der to help eval­u­a­tors avoid tak­ing less ac­cu­rate, po­ten­tially bi­ased short­cuts. For ex­am­ple, if you need to know how well some­one rea­sons un­der un­cer­tainty, ask, “What’s a be­lief you hold weakly, and why do you only hold it weakly?” or use a trial pro­ject edit­ing sam­ple re­search too see if their com­ments demon­strate how con­fi­dently they hold views based on mixed ev­i­dence (both of which we’ve done at Sen­tience In­sti­tute).

  • Eval­u­a­tors should make their ini­tial eval­u­a­tions sep­a­rately be­fore com­ing to­gether to dis­cuss them in or­der to min­i­mize group­think and max­i­mize the num­ber of in­de­pen­dent per­spec­tives available.

Devel­op­ing and re­tain­ing talent

Re­search sug­gests that or­ga­ni­za­tions and man­agers should:

  • Track en­gage­ment and its as­so­ci­a­tions with de­mo­graph­ics, and act on their find­ings. En­gage­ment is cor­re­lated with re­ten­tion,[12] perfor­mance, and in­no­va­tion.

  • To track en­gage­ment, use em­ployee sur­vey ques­tions with 5-point lik­ert scales such as:[13]

  • “I feel like I be­long at [Or­ga­ni­za­tion]”

  • “I am satis­fied with how de­ci­sions are made at [Or­ga­ni­za­tion]”

  • “I feel re­spected at [Or­ga­ni­za­tion]”

  • “At [Or­ga­ni­za­tion], there is open and hon­est two-way com­mu­ni­ca­tion”

  • “When some­thing bad hap­pens (e.g., when I get crit­i­cal feed­back from my man­ager, I have a nega­tive so­cial in­ter­ac­tion with a peer, etc.), I don’t ques­tion whether or not I be­long at [Or­ga­ni­za­tion]”

  • To im­prove en­gage­ment, work to make em­ploy­ees feel that they have ca­reer op­por­tu­ni­ties, that they have ac­cess to the learn­ing and de­vel­op­ment they need, that the role is as they ex­pected based on the hiring pro­cess, that their role con­tributes to their de­vel­op­ment, and that their con­tri­bu­tions are re­warded and rec­og­nized. In­clu­sive lead­ers en­sure their team mem­bers are ac­knowl­edged and not micro­man­aged, give their team the free­dom to pro­pose novel ideas and make de­ci­sions, give ac­tion­able feed­back, and listen to feed­back from their team. Their team mem­bers feel that their voices are heard.

  • Make in­for­ma­tion about job op­por­tu­ni­ties and el­i­gi­bil­ity available to em­ploy­ees, for in­stance by post­ing job lad­ders.

  • Offer learn­ing and de­vel­op­ment op­por­tu­ni­ties, such as op­por­tu­ni­ties for stretch roles (re­spon­si­bil­ities be­yond an em­ployee’s cur­rent role), op­tional courses, peer-to-peer feed­back, fireside chats with in­dus­try lead­ers, pro­grams to help em­ploy­ees de­velop their life vi­sions and goals, ca­reer map­ping, com­pany re­treats with work­shops, and in­di­vi­d­ual bud­gets to spend as em­ploy­ees choose on learn­ing and de­vel­op­ment op­por­tu­ni­ties, af­ter which they can pre­sent what they learned to their team.

  • Sur­veys of em­ployee en­gage­ment at Pin­ter­est found that the most in­clu­sive man­agers: (1) com­mu­ni­cate openly about think­ing and de­ci­sion-mak­ing with teams and so­licit feed­back, (2) be trans­par­ent about mis­takes, (3) so­licit in­put be­fore dis­tribut­ing as­sign­ments and give team mem­ber lat­i­tude in se­lect­ing their pro­jects, (4) fre­quently make time for team and one-on-one so­cial­iza­tion, and (5) put in ex­tra time to sup­port strug­gling team mem­bers.

  • Or­ga­ni­za­tions can also work to im­prove en­gage­ment with team lunches with lead­er­ship, trans­parency in ex­ec­u­tive de­ci­sion-mak­ing, job ex­change pro­grams, pro­ject ret­ro­spec­tives, and shar­ing cus­tomer/​client/​user feed­back.

  • Make ex­pec­ta­tions about an em­ployee’s perfor­mance clear.

  • Provide feed­back to help em­ploy­ees cal­ibrate their con­fi­dence to their abil­ity. This is prob­a­bly of par­tic­u­lar benefit to high-perform­ing women on track to work in com­pet­i­tive po­si­tions, who may oth­er­wise not have the con­fi­dence to com­pete.

  • In de­ter­min­ing pay, use stan­dard­ized pro­cesses that hold eval­u­a­tors ac­countable for pay­ing team mem­bers fairly.[14]

  • To the ex­tent pos­si­ble, as­sign or­ga­ni­za­tional “house­work” — vol­un­tary work that is im­por­tant to the or­ga­ni­za­tion but which will not build skills or oth­er­wise ac­crue points to­wards pro­mo­tions — rep­re­sen­ta­tively across de­mo­graph­ics, in or­der to avoid such work fal­ling un­fairly on the shoulders of women or per­haps other marginal­ized groups.

  • Estab­lish re­spon­si­bil­ity for in­clu­sion to avoid col­lec­tive ac­tion failure. The grow­ing trend in the cor­po­rate world is for in­clu­sion to be a CEO-level re­spon­si­bil­ity and not rel­e­gated to HR, as recom­men­da­tions for in­clu­sion­ary prac­tice are rele­vant through­out an or­ga­ni­za­tion and are not merely a mat­ter of re­spond­ing to com­plaints.

  • Estab­lish ac­countabil­ity for in­clu­sion, for in­stance with trans­parency poli­cies and di­ver­sity man­agers.

  • En­gage man­agers in the pro­ject of in­clu­sion, for in­stance with men­tor­ship pro­grams and di­ver­sity task forces com­posed of vol­un­tary team mem­bers.

  • Do not en­deavor to con­trol man­agers, for in­stance by forc­ing them to base their de­ci­sions about their team en­tirely on job tests and perfor­mance rat­ings (though peer eval­u­a­tion may be good). While it would the­o­ret­i­cally be best if man­agers re­lied on such ob­jec­tive met­rics, forc­ing them to do so could back­fire.

  • Take care with grievance sys­tems, which can back­fire by en­abling moral li­cens­ing and re­tal­i­a­tion. This may a hard prob­lem to solve, but per­haps “flex­ible” com­plaint sys­tems that use a for­mal hear­ing pro­cess and in­for­mal me­di­a­tion may be more helpful to em­ploy­ees.

  • Lead­ers and man­agers should men­tor (i.e. ad­vise) peo­ple from un­der­rep­re­sented groups, and should, more helpfully, go be­yond that to “spon­sor” (i.e. ad­vo­cate for) them, for in­stance by offer­ing them con­nec­tions, recom­men­da­tions, and other re­sources be­yond train­ing.

  • Diver­sity train­ings gen­er­ally have no effect on di­ver­sity, seem to have a nega­tive effect at least as of­ten as they have a pos­i­tive one[15] and have no last­ing effect on out­comes var­i­ous re­searchers have used as prox­ies for in­clu­sive­ness. What spe­cific fea­tures and con­di­tions might make a train­ing effec­tive, in­effec­tive, or coun­ter­pro­duc­tive are un­der­stud­ied, but one meta-anal­y­sis (which found gen­er­ally pos­i­tive effects of train­ings and was smaller in scope than the meta-anal­y­sis above that found no effect), sug­gests that they are more effec­tive when they are used in com­bi­na­tion with other ini­ti­a­tives, when they tar­get both aware­ness and skill de­vel­op­ment, and when they are con­ducted over a sig­nifi­cant pe­riod of time.


Re­search sug­gests that com­mu­ni­ties should:

  • Em­pha­size that mem­bers of the com­mu­nity value and, more im­por­tantly, are work­ing on equity and in­clu­sion. So­cial norms are im­por­tant in­fluencers of be­hav­ior within a com­mu­nity, so this will so­cially in­cen­tivize fur­ther ac­tion.

  • Re­lat­edly, make the suc­cesses of or­ga­ni­za­tions do­ing well on equity and in­clu­sion­ary prac­tice salient (and not just on the met­ric of de­mo­graphic di­ver­sity, which may en­courage dis­mis­sals of “re­verse” dis­crim­i­na­tion, but also on the met­rics dis­cussed ear­lier of en­gage­ment, be­long­ing, and mo­bil­ity across de­mo­graph­ics, and in terms of how ex­ten­sively they are im­ple­ment­ing in­clu­sion­ary prac­tices). We should prob­a­bly share these suc­cesses more than com­mu­nity av­er­ages (at least if the av­er­age is bad) or the failures of or­ga­ni­za­tions do­ing poorly, to fuel change by nor­mal­iz­ing the good be­hav­ior.

  • Ac­tively en­gage peo­ple from over­rep­re­sented or more so­cially priv­ileged groups in the com­mu­nity pro­ject of im­prov­ing equity and in­clu­sion, es­pe­cially at or­ga­ni­za­tions and in spaces where valu­ing in­clu­sion is not yet a norm. Un­like peo­ple who benefit more and/​or more di­rectly from in­creased equity, they are not likely to be pe­nal­ized for their ad­vo­cacy.

  • Peo­ple from marginal­ized groups also already bear time and psy­cholog­i­cal costs from be­ing treated in­equitably, so putting this work on their shoulders weighs them down even more un­fairly. There is also a more gen­eral case to be made for the al­ly­ship of peo­ple from the priv­ileged group be­ing more effec­tive than ad­vo­cacy from, or at least solely from, the dis­ad­van­taged group.

  • As dis­cussed with re­gard to em­pow­er­ing team mem­bers in or­ga­ni­za­tions, men­tor (ad­vise) in­di­vi­d­u­als from un­der­rep­re­sented groups, and more valuably, spon­sor (ad­vo­cate for) them. Even just mak­ing con­nec­tions for peo­ple from un­der­rep­re­sented groups is helpful as net­work­ing is im­por­tant to ca­reer suc­cess.

  • Em­pha­size the con­tri­bu­tions of peo­ple from un­der­rep­re­sented groups and oth­er­wise en­sure role mod­els and col­leagues from un­der­rep­re­sented groups are highly visi­ble.

  • Make vi­sual rep­re­sen­ta­tions of team mem­bers, com­mu­nity mem­bers, and tar­get au­di­ences di­verse. How­ever, don’t strongly mis­rep­re­sent a group’s di­ver­sity or new­com­ers may feel un­easy when they re­al­ize the group was mis­rep­re­sented. Show­ing ad­vo­cates from un­der­rep­re­sented groups in ma­te­ri­als and at events will sig­nal to peo­ple who iden­tify with them that they can be­long in the move­ment.

  • Avoid “sex sells” tac­tics. Stud­ies sug­gest that when peo­ple are prompted to think about women in a sex­u­al­ized way they un­der­es­ti­mate their in­tel­li­gence and com­pe­tence. Sex­u­al­iza­tion may also make men re­gard women and fe­males as less wor­thy of moral con­cern, mak­ing them more likely to sex­u­ally vic­tim­ize or be­have ag­gres­sively to­wards them. This means these tac­tics likely con­tribute a cul­ture of un­der­es­ti­mat­ing, un­der­valu­ing, and mis­treat­ing ca­pa­ble ad­vo­cates, which in­hibits them from reach­ing their po­ten­tial.

  • Praise work for the benefit of the com­mu­nity, out­side of the scope of one’s des­ig­nated job re­spon­si­bil­ities. Avoid dis­count­ing this work in cel­e­bra­tions, hiring, pro­mo­tion, com­pen­sa­tion or other ad­vance­ment op­por­tu­ni­ties. Just like salaried po­si­tions, this work is a con­tri­bu­tion to the com­mu­nity, and women take on more of it (and are ex­pected to) while men are more likely to freeride on women’s con­tri­bu­tions. Re­lat­edly, all com­mu­nity mem­bers should be en­couraged to take on their fair share of com­mu­nity re­spon­si­bil­ities.

Other strate­gies which I have seen lit­tle rele­vant re­search on, but which are based more on my ex­pe­rience and that of col­leagues and which fol­low the gen­eral prin­ci­ple of mak­ing peo­ple feel they be­long, in­clude:

  • In gen­eral, avoid too much ho­mo­gene­ity or con­for­mity (in ap­pear­ance, per­sonal in­ter­ests, iden­tity af­fili­a­tion, in­ter­per­sonal as­so­ci­a­tions, or any­thing ir­rele­vant to an­i­mal ad­vo­cacy).

  • Espe­cially if you’re in a lead­er­ship po­si­tion, con­sider “un­cov­er­ing” your­self if it is suffi­ciently low per­sonal risk to do so, be­cause it’s less risky than for peo­ple lower down the lad­der and do­ing so will sig­nal to oth­ers like you that they be­long. “Cover­ing” refers to prac­tices which down­play or hide one’s iden­tity with marginal­ized groups, such as by:

  • Al­ter­ing ap­pear­ance to con­form with the dom­i­nant group, e.g. if a black woman loves her nat­u­ral hair but straight­ens it to be taken more se­ri­ously by white peo­ple;

  • Re­duc­ing ex­pres­sions of af­fili­a­tion with marginal­ized groups, e.g. if a mother would love to talk about her chil­dren but re­frains;

  • Avoid­ing as­so­ci­a­tion with other mem­bers of one’s group, e.g. if a gay per­son doesn’t bring their part­ner to events; or

  • De­clin­ing to ad­vo­cate for one’s group, e.g. if a vet­eran doesn’t re­spond to an offen­sive joke about the mil­i­tary.

  • Some de­gree of ho­mo­gene­ity even on fac­tors not di­rectly rele­vant to one’s work may be use­ful to group co­he­sion, but it at least seems likely that un­cov­er­ing one’s iden­tity with marginal­ized or un­der­rep­re­sented groups will help other mem­bers of those groups feel a greater sense of be­long­ing, and that it’s ad­vis­able to, for in­stance, not re­volve the com­mu­nity’s so­cial life around frat par­ties or galas, which ap­peal to or are ac­cessible to only a very limited sub­set of the pop­u­la­tion ad­vo­cates want in­volved in the com­mu­nity.

  • Com­mu­ni­ca­tions of who a “leader” is should move away from their his­tor­i­cal stereo­typ­i­cal de­pic­tions of mas­culinity (i.e. their em­pha­sis on dom­i­nance, in­di­vi­d­u­al­ism, and hero­ism) as fem­i­nine-typed lead­er­ship styles of com­mu­nal­ity can be as or even more suc­cess­ful lead­er­ship strate­gies. This is not to say dom­i­nant lead­er­ship styles can have no place — hav­ing a di­ver­sity of lead­er­ship styles seems bet­ter than hav­ing one lead­er­ship style across the move­ment, whether of a “mas­culine,” “fem­i­nine,” or other type. Re­lat­edly, com­mu­ni­ca­tions about who our lead­ers are or can be should in­clude op­por­tu­ni­ties for peo­ple who don’t match the tra­di­tional “leader” stereo­type of a white man to use a dom­i­nant lead­er­ship style too.

  • Apol­o­gize if you say or do some­thing that is ex­clu­sion­ary and be­come aware of it, even if it’s some­thing small, such as sug­gest­ing a woman smile more, com­pli­ment­ing a re­li­gious per­son’s open-mind­ed­ness as an athe­ist, putting only two gen­ders on a form, or mak­ing a joke that sug­gests po­ten­tial nega­tive feel­ings to­wards peo­ple of a marginal­ized group. While these are in­di­vi­d­u­ally only minor trans­gres­sions, they can ac­cu­mu­late to a sub­stan­tial bur­den, and while in some con­texts ac­tions like these may not re­flect prej­u­di­cial views, that’s of­ten un­clear, so clear­ing the air may re­duce the effects of that at­tri­bu­tional am­bi­guity.

  • Dis­cuss equity and in­clu­sion is­sues in ways that don’t fur­ther nor­mal­ize bad be­hav­ior or fur­ther in­di­cate to peo­ple from marginal­ized groups that they can’t be­long here. For in­stance, while it seems ac­cu­rate to say, “It’s hard for women to lead in an­i­mal ad­vo­cacy,” or, “The an­i­mal ad­vo­cacy move­ment has a race is­sue,” em­pha­siz­ing these un­nec­es­sar­ily may worsen feel­ings of ex­clu­sion. As such, when pos­si­ble it’s prob­a­bly best to fo­cus more on the progress than the prob­lem, though with­out mis­lead­ing peo­ple into think­ing the prob­lem is already solved. For in­stance, it may be bet­ter to say, “An­i­mal ad­vo­cacy or­ga­ni­za­tions are in­tent on em­pow­er­ing women in the move­ment to lead,” or, “The com­mu­nity is work­ing on im­prov­ing racial in­clu­sion.” Ideally these would be sup­ple­mented with con­crete ex­am­ples.


I put this sec­tion last be­cause re­duc­ing prej­u­dice and bias di­rectly seems more difficult than limit­ing their in­fluence on our be­hav­ior. Bias re­duc­tion is also harder to mea­sure, at least in terms of its im­pact on be­hav­ior. Still, there are a few weak find­ings for the most effec­tive ways to re­duce prej­u­dice.

The au­thors of a 2009 meta-anal­y­sis con­cluded that the fol­low­ing strate­gies were effec­tive, pri­mar­ily in re­duc­ing im­plicit prej­u­dices,[16] but also in re­duc­ing prej­u­di­cial be­hav­ior:

  • Co­op­er­a­tive learning

  • En­ter­tain­ment show­ing pos­i­tive in­ter­group con­tact and in­clu­sive so­cial norms

  • Peer in­fluence through dis­cus­sion and dialogue

  • In­ter­group contact

  • Value con­sis­tency and self-affirmation

  • Cross-cul­tural training

They also con­cluded that the fol­low­ing strate­gies are un­der­stud­ied, in­effec­tive, or have nega­tive effects:

  • So­cial recategorization

  • Cog­ni­tive training

  • Diver­sity training

  • Mul­ti­cul­tural, an­tibias, and moral education

  • Sen­si­tivity training

  • Con­flict resolution

Other re­search has re­sulted in the fol­low­ing recom­men­da­tions:

  • Ex­pose peo­ple to fa­vor­able in­group opinions of the out­group.

  • In­ter­group con­tact can help re­duce prej­u­di­cial at­ti­tudes.

  • Ex­po­sure to coun­ter­stereo­typ­i­cal ex­am­ples of in­di­vi­d­u­als from an out­group (e.g. a woman or black di­rec­tor) can help re­duce prej­u­di­cial at­ti­tudes.

  • Don’t merely raise aware­ness about bad norms or show stereo­types for ed­u­ca­tional pur­poses. As noted in the “Or­ga­ni­za­tional man­age­ment” sec­tion, di­ver­sity train­ings tend to have no effect on di­ver­sity, and while they some­times suc­ceed, they also some­times back­fire. This seems to be be­cause peo­ple are bad at up­dat­ing on in­for­ma­tion about bi­ases and seem to con­sis­tently think they are an in­di­vi­d­ual ex­cep­tion to nor­mal trends in bi­ases.

The re­search re­gard­ing in­form­ing peo­ple of their bias or ask­ing peo­ple to not be bi­ased is mixed, mean­ing we should pri­ori­tize other strate­gies, and pos­si­bly avoid this en­tirely at the risk of coun­ter­pro­duc­tive effects.



I’d like to ex­press my pride in the restora­tive ac­tions that some com­mu­nity mem­bers have taken in the face of mis­treat­ment by other com­mu­nity mem­bers. It’s im­por­tant that we cre­ate a com­mu­nity that is sus­tain­able and healthy, and it seems to me that we will bet­ter achieve that through restora­tive jus­tice than re­tribu­tive jus­tice.

“No tol­er­ance” is an im­por­tant policy for pre­vent­ing and re­spond­ing to mis­con­duct, in the sense that ev­ery is­sue will be ad­dressed. But no tol­er­ance doesn’t re­quire a “heavy handed” ap­proach — to the con­trary, I think it’s crit­i­cal for the heav­i­ness of our re­sponses to be com­men­su­rate with the sever­ity of an ac­tion and to es­ca­late pro­gres­sively with failures to par­ti­ci­pate in restora­tive pro­cesses. Small trans­gres­sions, for in­stance, should be “called in”[17] so the per­son who made the ap­par­ent mis­take has the op­por­tu­nity to defend them­selves if nec­es­sary, or to rec­tify their mis­take and im­prove. If some­one has com­mit­ted a trans­gres­sion, we want them to seek un­der­stand­ing, and if they do come to un­der­stand, apol­o­gize, and demon­strate a cred­ible in­ten­tion to im­prove, they should be given the chance to carry for­ward as a bet­ter com­mu­nity mem­ber, if pos­si­ble de­pend­ing on the sever­ity of their ac­tion. To cre­ate a cul­ture in which they are in­cen­tivized to co­op­er­ate like that, any nega­tive con­se­quences re­sult­ing from their ad­mis­sion of guilt and co­op­er­a­tion in restora­tion have to be much less se­vere and much less likely than the con­se­quences of re­fus­ing to co­op­er­ate. Other­wise peo­ple will have in­cen­tive to in­stead ag­gres­sively defend them­selves against ac­cu­sa­tions, which is a lost op­por­tu­nity for both the heal­ing of the per­son who was wronged and the bet­ter­ment of the per­son in the wrong, and which may re­sult in the dis­em­pow­er­ment of ev­ery­one in­volved in en­su­ing drama.

Of course, restora­tive jus­tice re­quires the par­ti­ci­pa­tion of the ac­cused, and in some cases, when pri­vate restora­tion has been at­tempted but is failing, it may be nec­es­sary to es­ca­late to pun­ish­ment, “call­ing out,” or a pe­riod of ex­clu­sion.

But restora­tive jus­tice of­ten suc­ceeds, in my own ex­pe­rience and that of oth­ers, and it’s what we all want for our­selves when we make mis­takes. Be­cause of my own efforts in restora­tive jus­tice, sev­eral peo­ple who may have last­ing in­fluence in my com­mu­ni­ties are now more ca­pa­ble al­lies than they would have been if I had be­haved re­tribu­tively in re­sponse to their poor be­hav­ior to­wards me. In­stead, they made efforts to un­der­stand, make amends, and im­prove. Pri­vate efforts in restora­tion can take great pa­tience and com­pas­sion on the part of the wronged, and that may feel un­fair, but the only way we change peo­ple is by giv­ing them the op­por­tu­nity to grow away from the harm­ful mi­s­un­der­stand­ings and be­hav­iors that a prej­u­diced so­ciety has taught them, just as we give that op­por­tu­nity to ev­ery­one who used to eat an­i­mals or en­gage in other speciesist and harm­ful be­hav­ior to­wards non­hu­mans but pul­led their way out of that en­cul­tura­tion — in­clud­ing al­most ev­ery one of us as farmed an­i­mal ad­vo­cates.

Restora­tion takes a lot more for­ti­tude and effort in the short run than re­tri­bu­tion, but when both the ac­cuser and ac­cused par­ti­ci­pate it seems to ul­ti­mately re­sult in a lot less fight­ing, stress, time, and harm on all sides, and helps us build a strong and healthy com­mu­nity where peo­ple are en­couraged to come to­gether and grow rather than a weak and un­sta­ble one where we di­vide and stag­nate. So I hope we have the for­ti­tude to work hard for restora­tion and pos­i­tive-sum out­comes, even when we’re wounded and rightly frus­trated. And I hope we have the courage to ac­cept when our efforts are failing and there will not be a co­op­er­a­tive, healthy re­s­olu­tion, and the re­solve to push for­ward to the least bad out­comes when we fail to reach good ones.


This year, the an­i­mal ad­vo­cacy move­ment took up the #metoo move­ment and made no­table progress on sex­ual ha­rass­ment. Lead­ers and in­fluencers have been mak­ing more pub­lic state­ments of their com­mit­ments to DEI. Or­ga­ni­za­tions are seek­ing guidance from En­com­pass and other DEI ad­vi­sors and are de­vel­op­ing thor­ough poli­cies on sex­ual ha­rass­ment and nondis­crim­i­na­tion. Women and non-bi­nary ad­vo­cates in the US or­ga­nized a pro­duc­tive sem­i­nar that took place be­fore the An­i­mal Rights Na­tional Con­fer­ence (ARNC) and are prepar­ing on­go­ing ac­tivi­ties through a Gen­der Equity in An­i­mal Rights group. DEI was the talk of the town in pre­sen­ta­tions and con­ver­sa­tions at the ARNC this sum­mer, and the con­fer­ence speaker ros­ter was more de­mo­graph­i­cally rep­re­sen­ta­tive than in pre­vi­ous years. And one of our largest or­ga­ni­za­tions, Mercy for An­i­mals, is now be­ing led by Leah Garces, who has a track record of car­ing about and act­ing on in­clu­sion. And that list is not ex­haus­tive!

I’ve been ex­cited, grate­ful, and proud to see so much en­thu­si­asm in the an­i­mal ad­vo­cacy com­mu­nity for equity and in­clu­sion, and I’m op­ti­mistic about the progress we’ll make, es­pe­cially with the em­pow­er­ment of these re­search find­ings.

[1] I ac­knowl­edge that this is legally what we have to do now any­ways, de­spite the cul­tural con­text not be­ing one of to­tal equity.

[2] This is es­pe­cially true if we seek to in­clude only a to­ken minor­ity of peo­ple from un­der­rep­re­sented groups, as there is some ev­i­dence that a “crit­i­cal mass” of team mem­bers from so­cially marginal­ized groups, in­clud­ing in man­age­ment po­si­tions (maybe 20% in man­age­ment, at least for women) is nec­es­sary to im­prove team perfor­mance rel­a­tive to a ho­moge­nous group, whereas it’s pos­si­ble that a pro­por­tion be­tween zero and that crit­i­cal mass may harm perfor­mance. If this is the case, maybe it owes to “stereo­type threat,” anx­iety, un­der­con­fi­dence, or a lack of a feel­ing of be­long­ing for the team mem­bers in the small minor­ity, or to the ma­jor­ity’s dis­mis­sal of the minor­ity of marginal­ized team mem­bers as to­kens. This may only be an effect when a small minor­ity of a team is from a so­cially marginal­ized group, whereas groups com­prised mostly of peo­ple from so­cially marginal­ized groups with a minor­ity of peo­ple from a so­cially dom­i­nant group (such as a group of mostly women with a small num­ber of men) may fare as well as bal­anced groups. For in­stance, men do not seem to demon­strate the in­creased vigilance, de­creased be­long­ing, and de­creased de­sire to par­ti­ci­pate that women do when their gen­der is un­der­rep­re­sented in a group, sug­gest­ing that there may be no such nega­tive effect of “over­shoot­ing” if we end up over-rep­re­sent­ing peo­ple from marginal­ized groups.

[3] If ev­i­dence of in­equity in ca­reer pipelines has slid past you, here are just a few re­search find­ings to start you off. Beyond di­rectly dis­crim­i­na­tory de­ci­sion-mak­ing by au­thor­i­ties, peo­ple are also alienated from teams and com­mu­ni­ties when they are mis­treated, for in­stance if they are bul­lied or sex­u­ally ha­rassed, and all of the minor and am­bigu­ous dis­crim­i­na­tory in­ter­ac­tions peo­ple ex­pe­rience can ag­gre­gate and amount to sig­nifi­cant bur­dens of stress and feel­ings of de­val­u­a­tion and ex­clu­sion, on top of which cul­tural stereo­types and the norms and ex­pec­ta­tions that are cycli­cally shaped by and shape peo­ple push us to­wards nar­row ranges of op­por­tu­ni­ties and roles that may not be where we would oth­er­wise want to be or can make the most im­pact.

[4] They also found that racial and eth­nic di­ver­sity cur­rently have a stronger re­la­tion­ship to fi­nan­cial perfor­mance in the US than gen­der di­ver­sity, and that no com­pany is cur­rently in the top quar­tile on both gen­der and racial/​eth­nic di­ver­sity.

[5] Stud­ied boards have low per­centages of women, so how this ex­tends to or changes with boards closer to or ex­ceed­ing half women di­rec­tors is un­known, and there are ar­gu­ments for why a bal­anced board could show higher or lower perfor­mance. To the ex­tent that prej­u­dice and prej­u­di­cially cre­ated norms are dis­em­pow­er­ing women at that stage — e.g. if those who make it through the “glass ceiling” are be­ing dis­missed as to­kens, aren’t made to feel they be­long, and/​or only gain en­trance to the board in the first place if they meet a higher bar than the men who do — then in­creas­ing their num­bers will mean less com­pe­tent men are re­placed with more com­pe­tent women, in­creas­ing the board’s perfor­mance, per­haps un­til equil­ibrium is met around par­ity. But if pipeline prob­lems cre­ate a rel­a­tively smaller pool of com­pa­rably qual­ified women than men, then in­creas­ing women’s num­bers be­yond that pool size will re­sult in the re­verse.

[6] Note that in the case of anti-smok­ing PSAs, while de­mo­graphic similar­ity (in terms of age, gen­der, and race) with a smoker char­ac­ter was pos­i­tively as­so­ci­ated with en­gage­ment and the per­ceived effec­tive­ness of a PSA, de­mo­graphic similar­ity with a sep­a­rate non-smoker per­suader char­ac­ter was not. This sug­gests that the per­sua­sive effect of de­mo­graphic similar­ity is limited by the ex­tent of deeper con­text-rele­vant similar­i­ties.

[7] For fur­ther clar­ity, based on the En­com­pass sur­vey men­tioned above, a glance at the Na­tional An­i­mal Rights Con­fer­ence speaker ros­ter, and my mem­o­ries of many events and knowl­edge of var­i­ous groups’ lead­er­ship, my es­ti­mates are, very roughly, as fol­lows: The ra­tio of peo­ple of color in so­ciety com­pared to the ranks of the move­ment is 4:1; the ra­tio of women in so­ciety com­pared to the ranks of the move­ment is 2:3; the ra­tio of peo­ple of color in the ranks of the move­ment com­pared to move­ment lead­er­ship is 3:2; and the ra­tio of women in the ranks of the move­ment com­pared to move­ment lead­er­ship is 3:2. Other groups and in­ter­sec­tions are harder to es­ti­mate on a glance like this on ac­count of smaller pop­u­la­tion sizes. For women, the move­ment’s figure of roughly ~75% in the ranks and ~50% in lead­er­ship tracks with the gen­eral trend in the non­profit work­force.

[8] Google seems to have done a great job eval­u­at­ing their pay gaps and im­me­di­ately ad­dress­ing them.

[9] Merely hiring the most in­tel­li­gent in­di­vi­d­u­als may be an in­effec­tive route to de­vel­op­ing the most effec­tive team any­ways.

[10] The stud­ies it an­a­lyzed typ­i­cally re­lied on su­per­vi­sory rat­ings, which cor­re­late weakly with more ob­jec­tive mea­sures of perfor­mance such as out­put. The study’s method­ol­ogy also found work sam­ple tests to be poor pre­dic­tors of job perfor­mance, which is sus­pect as in the­ory they should very di­rectly mea­sure task perfor­mance, and one could choose to mea­sure on-the-job perfor­mance by a very similar met­ric (e.g. a work sam­ple test could be to write an ar­ti­cle, and an ob­jec­tive task perfor­mance met­ric on the job could be an eval­u­a­tion of an ar­ti­cle writ­ten on the job) just as read­ily as they might use su­per­vi­sory rat­ings. That low rank­ing of work sam­ple tests and the study’s re­li­ance on su­per­vi­sory rat­ings may merely show that su­per­vi­sors are poor judges of task perfor­mance or fail to con­sider it heav­ily in their eval­u­a­tions. The anal­y­sis also only looked at gen­eral men­tal abil­ity and com­bi­na­tions of other tests with it, leav­ing out com­par­i­sons be­tween other pre­dic­tors. Richard­son and Nor­gate have fur­ther crit­i­cisms of the study’s method­ol­ogy that limit the weight of its con­clu­sions.

[11] See page 24 of this study. This is very limited ev­i­dence and I’m sur­prised I didn’t find other in­for­ma­tion about de­mo­graphic rates within in­di­vi­d­ual hiring pipelines, but this does offer some rea­son to proac­tively ad­vance peo­ple from marginal­ized groups at early eval­u­a­tion stages when it feels like they fall just shy of the bar — be­cause that may just be bias talk­ing.

[12] When en­gage­ment is low, em­ploy­ees of a differ­ent race than their man­agers are less likely to want to stay at the com­pany than em­ploy­ees of the same race, but when en­gage­ment is high, in­ten­tion to stay is much higher in both groups, and even higher for the racially di­verse dyads than the racially ho­moge­nous ones. This both in­di­cates the par­tic­u­lar im­por­tance of en­gage­ment for peo­ple from un­der­rep­re­sented groups, and sug­gests that di­ver­sity is an am­plifier of en­gage­ment (mak­ing low en­gage­ment lower and high en­gage­ment higher).

[13] Th­ese five ques­tions cor­re­lated best with em­ployee en­gage­ment, ac­cord­ing to find­ings based on the use of a Diver­sity and In­clu­sion Sur­vey cre­ated by Cul­ture Amp, an em­ployee feed­back plat­form, with Paradigm, a DEI con­sult­ing firm. Cul­ture Amp has ad­di­tional ques­tion recom­men­da­tions.

[14] After con­trol­ling for com­pens­able fac­tors, or­ga­ni­za­tions may still see dis­crep­an­cies that may be at­tributable to gen­der bias or gen­der-bias-re­lated challenges such as the dou­ble-edged sword women walk when ne­go­ti­at­ing com­pared to men. Con­trol­led gen­der pay gaps within the US are much smaller than the US na­tional pay gap of 78%, which is half ex­plained by gen­der differ­ences in oc­cu­pa­tion and largely fur­ther ex­plained by other com­pens­able fac­tors such as hours worked, though sev­eral per­centage points are still un­ac­counted for by var­i­ous con­trol­led es­ti­mates such as those made by PayS­cale (0.5-4% de­pend­ing on the in­dus­try, with 1.9% in non­prof­its). The na­tional figure in­di­cates in­equity in so­ciety gen­er­ally, in in­dus­tries, and in or­ga­ni­za­tions, but the in­equity it points to is largely one of op­por­tu­nity, caused by di­rect dis­crim­i­na­tion through­out pipelines and by the re­lated in­fluences of in­equitable so­cial norms and ex­pec­ta­tions, while it’s only minorly ac­counted for by di­rect gen­der dis­crim­i­na­tion or oth­er­wise un­fair gen­der-re­lated fac­tors in com­pen­sa­tion de­ci­sions. Be­cause there still is some un­ex­plained differ­ence in pay for equal work across in­dus­tries, though, or­ga­ni­za­tions have a re­spon­si­bil­ity to en­sure their com­pen­sa­tion is fair. Women also take on more vol­un­tary com­mu­nity work than men, so it’s pos­si­ble that a con­trol­ling vari­able of “hours worked” fails to cap­ture a sig­nifi­cant num­ber of un­paid hours which are still con­tributed to a com­pany or or­ga­ni­za­tion, which would make these ap­par­ently-con­trol­led figures in­ap­pro­pri­ately low es­ti­mates of true con­trol­led gaps. This seems par­tic­u­larly rele­vant in the non­profit sphere which re­lies so heav­ily on vol­un­tary la­bor. In ac­tivist spaces, that vol­un­tary la­bor may also be sig­nifi­cantly com­prised of emo­tional la­bor for the com­mu­nity and as such may come with a higher stress bur­den than paid hours.

[15] This may be be­cause they ac­ti­vate stereo­types, make prej­u­diced be­hav­ior ap­pear nor­mal, or make the viewer feel ”woke” for see­ing the train­ing or be­lieve that their or­ga­ni­za­tion is for show­ing it, which could en­able the viewer to adopt a moral li­cense or re­lax their care with their be­hav­ior.

[16] Im­plicit prej­u­dice is poorly stud­ied, and pop­u­lar mea­sure­ments such as Im­plicit Aware­ness Tests (IATs) are poor pre­dic­tors of be­hav­ior. A 2013 re­view of IATs found “lit­tle di­rect ev­i­dence” re­gard­ing whether chang­ing im­plicit prej­u­dice changes dis­crim­i­na­tory be­hav­ior, that no stud­ies on the cor­re­la­tion be­tween im­plicit prej­u­dice and dis­crim­i­na­tory be­hav­ior “re­ported that im­plicit prej­u­dice me­di­ated the effect of the ma­nipu­la­tion on the be­hav­ior,” and that the only study that re­ported an anal­y­sis found no me­di­a­tion, in ad­di­tion to which the re­searchers “found no pub­lished pa­per (suc­cess­ful or not) that tested whether a change in im­plicit prej­u­dice pre­dicted a later change in be­hav­ior.” IATs have low test-retest re­li­a­bil­ity among other is­sues, and its founders ac­knowl­edged in 2015 that the tests’ failings “ren­der them prob­le­matic to use to clas­sify per­sons as likely to en­gage in dis­crim­i­na­tion.” Note that this sug­gests that chang­ing in­di­vi­d­u­als’ in­ter­nal be­liefs is not a pri­or­ity for im­prov­ing equitable be­hav­ior in a com­mu­nity (rel­a­tive to, per­haps, chang­ing their per­cep­tions of com­mu­nity norms, or cre­at­ing poli­cies and norms and hold­ing peo­ple ac­countable to them).

[17] “Cal­ling in” means ad­dress­ing a trans­gres­sion pri­vately, char­i­ta­bly, and prob­a­bly ideally with “non­vi­o­lent” com­mu­ni­ca­tion, in the in­ter­est of de­creas­ing the like­li­hood of putting the ac­cused on the defen­sive and in­creas­ing the like­li­hood of a healthy ex­change and pro­duc­tive out­come.