The Importance of Truth-Oriented Discussions in EA

This doc­u­ment is a col­lab­o­ra­tive effort. It re­sponds to the doc­u­ment Mak­ing Dis­cus­sions in EA Groups In­clu­sive (hence­forth referred to as Mak­ing Dis­cus­sions In­clu­sive). We ap­pre­ci­ate the time and care that went into mak­ing that doc­u­ment. It not only brought at­ten­tion to is­sues that many EAs might not have pre­vi­ously been aware of, but it ad­dressed these is­sues with a greater level of nu­ance than they are usu­ally dis­cussed with. At the same time, we also feel that there are sev­eral im­por­tant is­sues that were not ad­dressed. We will ad­dress the origi­nal doc­u­ment sec­tion by sec­tion and elab­o­rate on why we be­lieve that limit­ing dis­cus­sion is a risky path to go down and why it won’t ul­ti­mately de­liver on what we are told it will.

Con­tent Note: This doc­u­ment dis­cusses is­sues which were listed in Mak­ing Dis­cus­sion In­clu­sive as po­ten­tially alienat­ing.

Ban­ning top­ics does not im­prove in­clu­sivity (re­sponds to link)

We agree that dis­cussing cer­tain top­ics will of­ten have im­pacts in terms of how com­fortable or wel­come cer­tain peo­ple feel in a group and that these trade-offs should be con­sid­ered. How­ever, we be­lieve that limit­ing de­bate tends to im­prove in­clu­sion less than you might think, for a num­ber of rea­sons:

Firstly, there is always the op­tion to not par­ti­ci­pate in a dis­cus­sion if you be­lieve that en­gag­ing would be emo­tion­ally drain­ing or a waste of time. We don’t want to dis­miss how frus­trat­ing it can be to see peo­ple be­ing wrong with­out it be­ing suffi­ciently challenged, but we also be­lieve that peo­ple are gen­er­ally ca­pa­ble of over­com­ing these challenges and learn­ing to adopt a broader per­spec­tive from where they can see that it usu­ally isn’t ac­tu­ally very im­por­tant if some­one is wrong on the in­ter­net. How­ever, peo­ple are less likely to de­velop this re­silience when the com­mu­nity they are part of cre­ates too much of an ex­pec­ta­tion of com­fort. This is im­por­tant since com­fort in a par­tic­u­lar con­text is a re­sult of en­vi­ron­men­tal fac­tors and the in­di­vi­d­ual’s re­silience.

Se­condly, there are com­pet­ing ac­cess needs. Effec­tive al­tru­ism pro­vides a com­mu­nity where peo­ple can have dis­cus­sions that they can­not have el­se­where. Limit­ing the dis­cus­sion may make some peo­ple feel more wel­come, but we also risk los­ing the most in­de­pen­dent thinkers. And these are vi­tal to dis­cov­er­ing truth as de­tailed in the next sec­tion.

Thirdly, groups are not uniform. Mea­sures to make a minor­ity more wel­come may make minori­ties within that minor­ity feel less wel­come. At­tempts to limit crit­i­cisms of Is­lam, may marginal­ise ex-Mus­lims or women from Is­lamic coun­tries. Prevent­ing peo­ple from mak­ing ar­gu­ments against abor­tion may make pro-choice women more com­fortable, but at the cost of mak­ing pro-life women feel un­wel­come (con­ser­va­tives are a minor­ity within EA). Limit­ing dis­cus­sion of cul­tural rea­sons for poverty may be alienat­ing to mem­bers of that minor­ity who seek to re­form their cul­ture.

Lastly, we be­lieve that some­one is ex­cluded to a greater de­gree when they are not al­lowed to share their sincerely held be­liefs than when they are merely ex­posed to be­liefs that they dis­agree with. This is es­pe­cially the case with cen­sor­ship as rules are of­ten ex­tended or in­ter­preted more broadly over time. Even though cer­tain rules may seem quite mild and rea­son­able by them­selves, their mere ex­is­tence cre­ates a rea­son­able fear that those with cer­tain view­points will even­tu­ally be com­pletely pushed out.

This is not to claim that all dis­cus­sions should nec­es­sar­ily oc­cur in all con­texts, just that we should be very wary of limit­ing them. In par­tic­u­lar, the ex­am­ple given of some­one ar­gu­ing that women would be bet­ter if they were con­trol­led by men does not seem to be at all typ­i­cal of the kinds of dis­cus­sions that usu­ally oc­cur in EA and hence of the kinds of dis­cus­sions that likely mo­ti­vated Mak­ing Dis­cus­sions In­clu­sive or that would be limited were the ad­vice in that doc­u­ment to be fol­lowed.

We also note that peo­ple do ac­tu­ally come into EA groups and challenge the ba­sic ideas of EA (Three Bi­ases that Made Me Believe in AI Risk, Char­ity vs. Revolu­tion). This is a good thing as it forces us to fur­ther re­fine our ideas. Now, of course, we have to en­sure that such dis­cus­sions don’t drive out other dis­cus­sion by suck­ing up all the time, but we ought to en­gage with sincere crit­i­cism, even if it is tiring. An at­ti­tude where we ex­pect oth­ers to au­to­mat­i­cally agree with us is un­likely to be the most effec­tive for per­suad­ing peo­ple over the long term.

Why Open De­bate is Im­por­tant For Truth (re­sponds to link)

We ac­cept that hav­ing peo­ple feel com­fortable is valuable in and of it­self. How­ever, we don’t con­sider this to be the pri­mary pur­pose of effec­tive al­tru­ism. Most char­ity is about the giver and al­low­ing them to feel like they are mak­ing a sig­nifi­cant differ­ence even when they are not. In con­trast, EA is about be­ing effec­tive and that nec­es­sar­ily in­volves hav­ing a true un­der­stand­ing of things as they are. In­deed, many stan­dard EA views, such as the im­por­tance of AI safety or that wild an­i­mal suffer­ing mat­ters, seem out­ra­geous to most peo­ple at first. It would seem ar­ro­gant to sup­pose that there aren’t any is­sues in which we are in a similar po­si­tion; i.e. causes that sound out­ra­geous, but are hon­estly in­cred­ibly im­por­tant to pur­sue (see cru­cial con­sid­er­a­tions and Cause X). For this rea­son we should be wary about rul­ing things out pre­emp­tively.

We agree that hu­mans are of­ten ir­ra­tional and that power struc­tures/​dy­nam­ics have some effect the way that dis­cus­sions play out. how­ever, we think the re­duc­tive ap­proach taken in Mak­ing Dis­cus­sions In­clu­sive con­sid­er­ably over-states the im­pact for the fol­low­ing four rea­sons:

Firstly, if the goal is to re­bal­ance con­ver­sa­tions in or­der to make them more ob­jec­tive, we need to speci­fi­cally con­sider con­ver­sa­tional power, in­stead of power in gen­eral. Ad­vo­cates of so­cial jus­tice have re­cently been un­usu­ally suc­cess­ful in limit­ing speech com­pared to other ide­olo­gies (see RIP cul­ture war thread for a re­cent ex­am­ple). We might there­fore be tempted to con­clude that they have a dis­pro­por­tionate amount of con­ver­sa­tional power and that any at­tempt at re­bal­anc­ing would in­volve re­duc­ing the voice of these ad­vo­cates (this should not be in­ter­preted as sup­port of re­bal­anc­ing in the first place).

Se­condly, the power of the speaker heav­ily de­pends on the spe­cific cir­cum­stances. Even though rich peo­ple tend to have more power and so­cial sta­tus than the poor, due to a de­sire to favour the un­der­dog par­tic­u­lar au­di­ences may be heav­ily bi­ased to­wards the lat­ter to the point of be­ing com­pletely dis­mis­sive of the former. Similarly, it is plau­si­ble that a man ques­tion­ing both a man and a woman equally ag­gres­sively would be more likely to be seen as a bully in the case of the woman be­cause that would fit more in­line with so­ciety’s pre­con­cep­tions.

Lastly, the re­ceived view of power re­la­tions is sig­nifi­cantly out­dated. Even though his­tor­i­cally men have been granted more au­thor­ity than women, in­fluence of fem­i­nism and so­cial jus­tice means that in many cir­cum­stances this has been miti­gated or even re­versed. For ex­am­ple, stud­ies like Gor­nall and Stre­bu­laev (2019) found that blind­ing eval­u­a­tors to the race or sex of ap­pli­cants showed that by de­fault they were bi­ased against white men. We ac­knowl­edge that there are other cir­cum­stances where so­ciety is still bi­ased to­wards men, but we cau­tion about turn­ing this into a blan­ket as­sump­tion, even though it may have been more ap­pro­pri­ate in the past. Tak­ing this fur­ther, there is a nega­tive se­lec­tion effect in that the more that a group is dis­em­pow­ered and could benefit from hav­ing its views be­ing given more con­sid­er­a­tion the less likely it is to have to power to make this hap­pen.

While Both Mat­ter, Truth is More Im­por­tant than In­clu­sivity (con­tinues re­spond­ing to link)

So the power re­la­tions which the au­thors of Mak­ing Dis­cus­sion In­clu­sive want to cor­rect are much less clear defined than you might think at first glance. But even if we were to ac­cept their premises, limit­ing de­bate still wouldn’t nec­es­sar­ily be a good choice. Why not?

Firstly, some peo­ple choos­ing not to par­ti­ci­pate tends to be less harm­ful to the qual­ity of dis­cus­sion than cen­sor­ship as the lat­ter pre­vents ex­po­sure to a set of view­points com­pletely, while the former only re­duces its promi­nence. Even when there is a cost to par­ti­ci­pat­ing, some­one who con­sid­ers the topic im­por­tant enough can choose to bear it and one strong ad­vo­cate by them­selves is of­ten suffi­cient to change peo­ple’s minds (es­pe­cially within the Effec­tive Altru­ism com­mu­nity where steel-man­ning tends to be ad­mired).

Se­condly, the con­cerns here are mostly around peo­ple choos­ing not to par­ti­ci­pate be­cause of the effort re­quired or be­cause the dis­cus­sion makes them un­com­fortable. This is gen­er­ally less wor­ry­ing than peo­ple de­clin­ing to par­ti­ci­pate be­cause of long term rep­u­ta­tional risk. This is be­cause it is much eas­ier to bear short term costs in or­der to make an im­por­tant point than longer term costs.

Thirdly, even if limit­ing par­tic­u­lar dis­cus­sions would clearly be good, once we’ve de­cided to limit dis­cus­sions at all, we’ve opened the door to end­less dis­cus­sion and de­bate about what is or is not un­wel­com­ing (see Moder­a­tor’s Dilemma). And iron­i­cally, these kinds of dis­cus­sions tend to be highly par­ti­san, poli­ti­cal and emo­tional. In fact, we could go so far as to say that they tend to make peo­ple on both sides feel more un­wel­come: one side feels like it is be­ing pushed out, while the other side feels that their per­spec­tives aren’t be­ing taken se­ri­ously (and the promi­nence of the dis­cus­sion makes it much more promi­nent in their mind).

Alie­na­tion is More Com­plex than it Might Ap­pear (re­sponds to link)

This is a difficult topic to broach as some peo­ple may find this dis­cus­sion alienat­ing in it­self, but se­lec­tion effects are at the core of this dis­cus­sion and the eval­u­a­tion of its im­pact. It is im­por­tant to re­mem­ber that these com­ments re­late to group av­er­ages and not to in­di­vi­d­u­als. Just be­cause some­one ends up leav­ing EA be­cause of dis­com­fort with cer­tain dis­cus­sions, doesn’t mean that they are de­scribed by all or even any of qual­ities listed be­low.

One of the strongest effects is that the peo­ple who leave as a re­sult of cer­tain ideas be­ing dis­cussed are much less likely to be com­mit­ted EAs. The mechanism here is sim­ple: the more com­mit­ted to a cause, the more you are will­ing to en­dure for it. We agree with CEA that com­mit­ted EAs are sev­eral times more valuable than those who are vaguely al­igned, so that we should op­ti­mis­ing the move­ment for at­tract­ing more com­mit­ted mem­bers.

Se­condly, while we all have top­ics on which our emo­tions get the bet­ter of us, those who leave are likely to be over­come to a greater de­gree and on a wider va­ri­ety of top­ics. This means that they will be less likely to be able to con­tribute pro­duc­tively by pro­vid­ing rea­soned anal­y­sis. But fur­ther than this, they are more likely to con­tribute nega­tively by be­ing dis­mis­sive, pro­duc­ing bi­ased anal­y­sis or en­gag­ing in per­sonal at­tacks.

Thirdly, the peo­ple who leave are likely to be more ide­olog­i­cal. This is gen­er­ally an as­so­ci­a­tion be­tween be­ing more rad­i­cal and more ide­olog­i­cal, even though there are also peo­ple who are rad­i­cal with­out be­ing ide­olog­i­cal. Peo­ple who are more ide­olog­i­cal are less able to up­date in the face of new ev­i­dence and are also less likely to be able to provide the kind of rea­soned anal­y­sis that would cause other EAs to up­date more to­wards their views.

Lastly, we note that some peo­ple feel that EA is un­friendly to those on the right, while other feel that it is un­friendly to those in so­cial jus­tice. Often in these kinds of cir­cum­stances the fairest re­s­olu­tion is one which nei­ther side is com­pletely happy with. That is, we should ex­pect some level of peo­ple feel­ing that EA is un­wel­come to them in the op­ti­mal solu­tion.

Ex­pe­riences need to be Un­der­stood in Con­text (re­sponds to link):

We ac­knowl­edge that peo­ple sub­ject to a so­cial dis­ad­van­tage will tend to be much more knowl­edge­able about how it plays out than the av­er­age per­son has a rea­son to be. We also are aware that it can be in­cred­ibly hard for oth­ers to un­der­stand an ex­pe­rience from a mere ver­bal de­scrip­tion, with­out ever hav­ing ex­pe­rienced it them­selves. At the same time, we worry that peo­ple of­ten fail to be ob­jec­tive about is­sues that di­rectly con­cern them and can of­ten have difficult putting it in per­spec­tive.

We also worry that it is very easy for the ex­pe­riences of a vo­cal minor­ity to be pre­sented as the ex­pe­riences of a group as a whole as those who tend to have a more pos­i­tive ex­pe­rience are less likely to have a rea­son to talk about it than those who have a nega­tive ex­pe­rience. Th­ese effects are am­plified by the in­cen­tives of jour­nal­ists to fo­cus on con­tro­versy and of ac­tivists to fo­cus on driv­ing change, both of whom are se­lec­tive in whose voices are pre­sented.

While his­tor­i­cally the re­sponse to valid claims of struc­tural in­equal­ity has of­ten been de­nial, we be­lieve that this has re­sulted in an over-cor­rec­tion where al­most any­thing can and is ar­gued to be op­pres­sion. Th­ese in­clude prac­tis­ing yoga, eat­ing sushi or sit­ting with your legs open too wide. Even though we be­lieve that these are only minor­ity views within so­cial jus­tice, we draw the les­son that any claims of un­fair­ness need to be care­fully analysed be­fore de­cid­ing on whether to act

We ac­knowl­edge that it is dou­bly frus­trat­ing when you are not only treated un­fairly, but you also have to en­gage in sub­stan­tial effort in or­der to con­vince oth­ers that you were treated un­fairly. Un­for­tu­nately, this is a prob­lem that is gen­er­ally not very easy to solve. One re­sponse would be to au­to­mat­i­cally ac­cept any claim of un­fair­ness no mat­ter how ab­surd it sounds to the listener. The prob­lems with this solu­tion are so ob­vi­ous that they need not be stated. It is pos­si­ble to put in an ex­tra spe­cial effort to try to un­der­stand the per­spec­tive of an­other per­son when their ex­pe­riences are sub­stan­tially differ­ent from their own, but there will still be cir­cum­stances when de­spite your best effort, you will still dis­agree with them. In these cases, there can­not be an ex­pec­ta­tion that a par­tic­u­lar claim will au­to­mat­i­cally be ac­cepted with­out hav­ing to ar­gue for it.

In­deed, we don’t be­lieve that this is a one-way street. Th­ese kinds of highly poli­ti­cised, highly po­larised dis­cus­sions tend to be un­pleas­ant for ev­ery­one in­volved and there are bad ac­tors on all sides. Due to the very na­ture of the dis­cus­sion, peo­ple on both sides are reg­u­larly at­tacked or have their best at­tempts at hon­est en­gage­ment cur­so­rily dis­missed.

De­mo­graph­ics Aren’t a Strong Indi­ca­tor of In­clu­sive­ness (re­sponds to link):

We don’t be­lieve that de­mo­graph­ics are a very good in­di­ca­tor of how well mem­bers of a par­tic­u­lar group are treated. Scott Alexan­der pro­vides many ex­am­ples of groups that have large num­bers of women de­spite not be­ing “the poster child for fem­i­nism” in­clud­ing the Catholics, Is­lamists and Trump sup­port­ers. In an­other, he points out the difficulty of us­ing such ex­pla­na­tions for race:

For the record, here is a small sam­ple of other com­mu­ni­ties where black peo­ple are strongly un­der­rep­re­sented:
Run­ners (3%). Bik­ers (6%). Fur­ries (2%). Wall Street se­nior man­age­ment (2%). Oc­cupy Wall Street protesters (un­known but low, one source says 1.6% but likely an un­der­es­ti­mate). BDSM (un­known but low) Tea Party mem­bers (1%). Amer­i­can Bud­dhists (~2%). Bird watch­ers (4%). En­vi­ron­men­tal­ists (var­i­ous but uni­ver­sally low). Wikipe­dia con­trib­u­tors (un­known but low). Athe­ists (2%). Vege­tar­ian ac­tivists (maybe 1-5%). Yoga en­thu­si­asts (un­known but low). Col­lege base­ball play­ers (5%). Swim­mers (2%). Fan­fic­tion read­ers (2%). Uni­tar­ian Univer­sal­ists (1%).
Can you see what all of these groups have in com­mon?
No. No you can’t. If there’s some hid­den fac­tor unit­ing Wall Street se­nior man­age­ment and fur­ries, it is way be­yond any of our pay grades.

The au­thors of Mak­ing Dis­cus­sions In­clu­sive the­o­rise that alienat­ing dis­cus­sions are the rea­son why women were less likely than men to re­turn to meet­ings of EA Lon­don, de­spite be­ing equally likely to at­tend in the first place. We note that such a con­clu­sion would de­pend on an ex­cep­tion­ally high quan­tity of alienat­ing dis­cus­sions, and is prima fa­cie in­com­pat­i­ble with the gen­er­ally high rat­ing for wel­com­ing­ness re­ported in the EA sur­vey. We note that there are sev­eral pos­si­ble other the­o­ries. Per­haps women are more likely to have been so­cial­ised to per­ceive util­i­tar­ian calcu­la­tions as “cold” (i.e. EA it­self be­ing the main cause of aliena­tion, rather than any of the top­ics sug­gested in Mak­ing Dis­cus­sions In­clu­sive). Per­haps women are less likely to be in­ter­ested in the dis­cus­sions that oc­cur be­cause they fo­cus more on male in­ter­ests (or at least what are pre­dom­i­nantly male in­ter­ests within our par­tic­u­lar so­ciety). Per­haps EAs come off as so­cially awk­ward and this is more of a turn-off for women than men (women tend to have a greater in­ter­est in peo­ple and men a greater in­ter­est in things). The claim is not that any of these the­o­ries are nec­es­sar­ily cor­rect, just that it would be pre­ma­ture to as­sume that the main cause of the gen­der gap is the kinds of alienat­ing con­ver­sa­tions dis­cussed in Mak­ing Dis­cus­sions In­clu­sive. So if we were to limit dis­cus­sions in EA we could worsen our epistemics with­out ac­tu­ally in­creas­ing our di­ver­sity.

Fur­ther, we dis­agree with the fo­cus on mere num­bers, as op­posed to em­pha­sis­ing re­cruit­ing those from these de­mo­graph­ics who will make great effec­tive al­tru­ists. Just as in im­pact, there is a power law in terms of in­fluence. One mod­ern-day Martin Luther King can en­sure that EA takes into ac­count black per­spec­tives bet­ter than dozens of or­di­nary EAs. And any in­ter­ven­tion that sac­ri­fices the in­tel­lec­tual cul­ture that makes EA unique is likely to turn away the best in­di­vi­d­u­als of all de­mo­graph­ics, in­clud­ing those which are un­der-rep­re­sented. For this rea­son, in­ter­ven­tions to in­crease rep­re­sen­ta­tion can eas­ily back­fire, where rep­re­sen­ta­tion is mea­sured in terms of how much de­liber­a­tion cer­tain view­points re­ceive, as op­posed to the mere num­ber of peo­ple from a par­tic­u­lar de­mo­graphic.

The Limits of the Nar­ra­tive of Op­pres­sion (re­sponds to link):

The au­thors Mak­ing Dis­cus­sion In­clu­sive dis­cuss the ar­gu­ment that women might not be “op­pressed” on av­er­age in so­ciety as an ex­am­ple of an alienat­ing dis­cus­sion. In­deed, some peo­ple might find the fol­low­ing dis­cus­sion alienat­ing, but we be­lieve that en­gag­ing with the ob­ject level is­sue is the only way to ad­e­quately re­spond to these as­ser­tions.

Firstly, this an in­cred­ibly con­tro­ver­sial state­ment that is only held by a minor­ity of peo­ple. It wouldn’t just be re­jected by most con­ser­va­tives, but also by many mod­er­ates and liber­als as well. We note that the state­ment wasn’t just that women face cer­tain dis­ad­van­tages or that women face more dis­ad­van­tages than men, but that the level of dis­ad­van­tages is such that it could be fairly la­bel­led “op­pres­sion”. This fur­ther con­tains the as­sump­tions about men be­ing the main cause of these dis­ad­van­tages, as op­posed to bad luck, in the case of preg­nancy af­fect­ing women’s ca­reers, and that such “op­pres­sion” is on­go­ing, in­stead of in­equal­ities be­ing the re­sult of time-lag effects.

Se­condly, we note that this isn’t the kind of state­ment that is easy to eval­u­ate. It re­quires at­tempt­ing to weigh hun­dreds of differ­ent ad­van­tages and dis­ad­van­tages against each other. We agree with the au­thors of Mak­ing Dis­cus­sion In­clu­sive that mem­bers of a group have a strong aware­ness and salience of the dis­ad­van­tages they face and a much weaker abil­ity to un­der­stand those faced by oth­ers. If we ap­ply that here, it would seem to war­rant cau­tion in jump­ing too quickly to the con­clu­sion that one group is ad­van­taged over an­other.

Thirdly, we note that in­sist­ing that no-one challenge the re­ceived fem­i­nist po­si­tion on this sub­ject could very well be con­sid­ered alienat­ing as well. It is worth­while con­sid­er­ing the ex­am­ple of Athe­ism Plus, an at­tempt to in­sist that athe­ists also ac­cepted the prin­ci­ples of so­cial jus­tice. This was in­cred­ibly dam­ag­ing and de­struc­tive to the athe­ist move­ment due to the in­fight­ing that it led to and was per­haps partly re­spon­si­ble for the move­ment’s de­cline.

Fourthly, while we ac­knowl­edge that the de­nial of op­pres­sion has been used to jus­tify mis­treat­ment, so has the as­ser­tion of op­pres­sion. Nazis saw them­selves as be­ing op­pressed by Jews, white supremacists as op­pressed by poli­ti­cal elites, ter­ror­ists as op­pressed by the gov­ern­ment. Any group can con­struct a nar­ra­tive of op­pres­sion and these should not be ac­cepted un­crit­i­cally given how ap­peal­ing such false nar­ra­tives can be.

Fifthly, while we sym­pa­thise with many of the con­cerns ex­pressed, we don’t be­lieve that they are unique to one side. Peo­ple re­gard­less of ide­ol­ogy face the difficult choice of en­gag­ing in an un­pleas­ant dis­cus­sion or al­low­ing views that they to be wrong and harm­ful to go un­challenged. Peo­ple on all sides are frus­trated by peo­ple who stick within the norms of po­lite­ness, but seem to be ex­cep­tion­ally stub­born nonethe­less. Peo­ple on all sides are frus­trated by de­bates that don’t leave any­one closer to truth. This is alienat­ing for ev­ery­one, not just for par­tic­u­lar groups.

Why Truth is of Paramount Im­por­tance for EA (re­sponds to link)

EA is a nascent field; we should ex­pect over time our un­der­stand­ing of many things to change dra­mat­i­cally, in po­ten­tially un­pre­dictable ways. This makes ban­ning or dis­cour­ag­ing top­ics, even if they seem ir­rele­vant, harm­ful, be­cause we don’t know which could come to be im­por­tant.

For­tu­nately, there are some ex­am­ples we have to make this clear. For ex­am­ple, Mak­ing Dis­cus­sions In­clu­sive pro­vides a list of things that we should not dis­cuss (or at least we should be very wary of dis­cussing). We will ar­gue that there are ac­tu­ally very good rea­sons for EAs to dis­cuss these top­ics. Even in cases where it would not be rea­son­able to dis­pute the state­ment as given, we sug­gest that peo­ple may of­ten be ac­cused of re­ject­ing these state­ment when they ac­tu­ally be­lieve some­thing much more in­no­cent. Here are the ex­am­ples:

“Whether poor peo­ple are poor due to hav­ing a lower IQ”

We doubt that many peo­ple be­lieve that all poor peo­ple are stupid or even that poor peo­ple are gen­er­ally stupid. We think that most peo­ple recog­nise that many have grown up in a difficult en­vi­ron­ment or been de­nied op­por­tu­ni­ties. On the other hand, it is im­por­tant to dis­cuss the re­la­tion be­tween in­tel­li­gence and poverty, in or­der to dis­cuss the ways in which our so­ciety fails peo­ple of low in­tel­li­gence. As more of the sim­pler jobs be­come au­to­mated, there are le­gi­t­i­mate con­cerns that there won’t be any­thing for many peo­ple to do, which may leave many peo­ple in a dire po­si­tion.

“Whether it is or has been right or nec­es­sary that women have less in­fluence over the in­tel­lec­tual de­bate and less eco­nomic and poli­ti­cal power”

We doubt that many EAs hon­estly be­lieve that women are in­trin­si­cally de­serv­ing of less in­fluence or eco­nomic power. On the other hand, it is quite rea­son­able to be­lieve: a) per­sonal choices such as whether to take a high pay­ing job with long hours or whether to be a stay home par­ent af­fect a per­son’s in­fluence or earn­ings; b) it is perfectly valid for an in­di­vi­d­ual to make this choice, even though this may re­sult in one group hav­ing more power than an­other; and c) even though these choices are sub­ject to so­cial in­fluences which we can seek to re­duce, they are still pre­dom­i­nantly free choices that ought to be re­spected.

“Jus­tify­ing that the group should be in­clu­sive for them”

This ar­gu­ment sounds like it could very eas­ily be used in need an at­tempt to cir­cum­vent de­bate about whether a par­tic­u­lar mea­sure to in­crease in­clu­sive­ness is jus­tified or not.

“Whether peo­ple in the de­vel­op­ing world are poor be­cause of char­ac­ter flaws”

At first glance this might seem an in­nocu­ous re­stric­tion. How­ever, there are many valuable re­search pro­jects into the causes and solu­tions of poverty that would be un­der­mined by this:

  • It is a well-es­tab­lished fact that cor­rup­tion is a ma­jor prob­lem in many coun­tries, es­pe­cially in the de­vel­op­ing world. If this could be re­duced or ame­lio­rated, it could sig­nifi­cantly im­prove the in­de­pen­dence and wellbe­ing of some of the worst-off peo­ple in the world. But we can­not hope to achieve this with­out con­fronting the fact that the is­sue is en­demic cor­rup­tion, and cor­rup­tion is a clear ex­am­ple of a char­ac­ter flaw.

  • Microfi­nance and Direct Cash Trans­fers are two poli­cies of in­ter­est to many EAs. Cru­cial to their eval­u­a­tion is the ques­tion of well is the money even­tu­ally spent by the re­cip­i­ents; do they pay for medicine and in­vest in busi­nesses, or do they spend it on po­si­tional goods and al­co­hol? Does the ad­di­tional safety net cause them to take new en­trepreneurial risks, or does it lead to un­em­ploy­ment and ap­a­thy? Th­ese are em­piri­cal ques­tions of great im­por­tance, and wor­thy of the RCTs that are study­ing these is­sues. But even if not ex­plic­itly phrased in this way, an hon­est ad­vo­cate would strug­gle to deny that they were study­ing ‘are poor peo­ple lazy’ and ‘do poor peo­ple heav­ily dis­count the fu­ture’ - two ques­tions that cer­tainly seem like for­bid­den dis­cus­sions of char­ac­ter flaws.

Even if these dis­cus­sions don’t fall afoul of the pro­posed mea­sures, these kinds of con­sid­er­a­tions are less likely to be raised if they are right on the edge of the Over­ton Win­dow.


This doc­u­ment is about EA spaces in gen­eral and not just about the Diver­sity and In­clu­sion Group. While it is up to the mem­bers of each group to de­cide on its poli­cies, we make the fol­low­ing recom­men­da­tions:

  • The pri­mary dis­cus­sion for any sug­gested policy for EA in gen­eral should oc­cur in a lo­ca­tion where the mod­er­a­tion policy doesn’t re­sult in one par­tic­u­lar side re­ceiv­ing an un­fair ad­van­tage. If a group de­cides that it wishes to be a safe space or oth­er­wise limit dis­cus­sion of cer­tain view­points, then it ought not to be the pri­mary dis­cus­sion for any policy.

  • Groups should be up­front about their mod­er­a­tion poli­cies. This should be listed in the side­bar so that peo­ple know whether their con­tri­bu­tion will be wel­come or not.

  • EAs who al­ign with a par­tic­u­lar ide­ol­ogy should try to have a good un­der­stand­ing of the al­ter­na­tives. This will in­crease their abil­ity to en­gage with the EA com­mu­nity broadly and not just one par­tic­u­lar slice of it.

Any com­ments by this ac­count on the post are per­sonal opinion, rather than some kind of group con­sen­sus.