EA needs a cause prioritization journal

It’s about time that some­one started pub­lish­ing a jour­nal for cause pri­ori­ti­za­tion ar­gu­ments. It may have open peer re­view or a tra­di­tional for­mal pro­cess, it may come from a rep­utable aca­demic press or it may be an in-house thing with zero cred­i­bil­ity to the out­side world, it may pub­lish reg­u­lar pa­pers or weird things like ar­gu­ment maps and ad­ver­sar­ial col­lab­o­ra­tions, I make no claims about how to nav­i­gate these de­tails, but there needs to be a jour­nal.

I don’t think this claim needs much spe­cial ar­gu­ment; I ex­pect that many peo­ple’s re­ac­tion to this will be “gee whiz, of course that’s a good idea.” We think cause pri­ori­ti­za­tion is a se­ri­ous new is­sue, we should have a venue to pub­lish about it, pretty straight­for­ward right? Every new field gets a jour­nal, and the re­view and pub­li­ca­tion pro­cess im­proves ar­ti­cle qual­ity. Still, I will de­velop the ar­gu­ment for an EA jour­nal in greater de­tail, and hope­fully make it clear that it is worth the time and money re­quired to run it.

EA has de­vel­oped a sig­nifi­cant body of knowl­edge spread across a va­ri­ety of spaces. Some of it is fine as com­mon knowl­edge and blog posts; sim­ple ideas, which are easy to ac­cept and com­mu­ni­cate. Or they may be tem­po­rary prac­ti­cal is­sues, like eval­u­a­tion of a par­tic­u­lar char­ity, that don’t last long enough to war­rant the spe­cial at­ten­tion and time­less­ness of be­ing put into a pub­li­ca­tion. Yet cause pri­ori­ti­za­tion ar­gu­ments are more con­cep­tual, they are im­por­tant, and they need to be fleshed out with greater rigor than they have been. This pa­per for in­stance has been re­lied upon for ages, yet noth­ing has re­ally fol­lowed on from it. There has been no quan­ti­ta­tive mod­el­ing or sen­si­tivity anal­y­sis, there has been no thor­ough com­par­i­son of risk re­duc­tion ver­sus growth ac­cel­er­a­tion, and with­out that we aren’t get­ting closer to re­ally set­tling the is­sue of whether ex­is­ten­tial risk re­ally can be pre­sumed to be our biggest pri­or­ity in the long run.

We oc­ca­sion­ally have dis­cus­sions that move past the ba­sic old cause pri­ori­ti­za­tion ar­gu­ments, but that is tran­sient; many ar­gu­ments are left in a form that is im­pos­si­ble to sum­ma­rize and build upon, then for­got­ten. Then we de­vote most of our time to bick­er­ing about the com­mu­nity, be­cause we are not be­ing fed enough difficult new in­tel­lec­tual con­tent to keep our minds oc­cu­pied. We think that all the cause pri­ori­ti­za­tion ar­gu­ments were pretty much fleshed out already, so we as­sume that the is­sue is set­tled and treat causes like ide­olo­gies or iden­tities rather than hy­pothe­ses.

So not a lot of real progress is ac­tu­ally be­ing made on the core EA pro­ject. I think the situ­a­tion is ac­tu­ally very poor by aca­demic stan­dards. We need to en­courage peo­ple to do se­ri­ous work on cause pri­ori­ti­za­tion, make sure that it is rigor­ous, and make sure that it is in a for­mat that lends it­self to fur­ther work. A good way to ad­dress this and other prob­lems is to cre­ate a jour­nal for it.

Ex­ist­ing jour­nals aren’t suit­able.

First, they are rarely in­ter­ested in the ques­tions of cause pri­ori­ti­za­tion. While jour­nals may be in­ter­ested in cost-effec­tive­ness ar­gu­ments for par­tic­u­lar efforts in their par­tic­u­lar field (e.g. https://​​www.academia.edu/​​34953571/​​Cost-effec­tive­ness_of_in­ter­ven­tions_for_al­ter­nate_food_in_the_United_States_to_ad­dress_agri­cul­tural_catas­tro­phes), I can think of none which would care about rigor­ous com­par­i­sons be­tween very differ­ent causes. For not only is cause pri­ori­ti­za­tion in­ter­dis­ci­plinary among differ­ent sci­en­tific fields, but it also re­lates to ethics and de­ci­sion the­ory, mean­ing that it crosses a set of do­mains so wide that no typ­i­cal jour­nal will en­com­pass it. Utili­tas can be amenable, but it won’t work for non-util­i­tar­ian EA ar­gu­ments, or ar­gu­ments over moral un­cer­tainty.

Se­cond, pub­lish­ing to an ex­ter­nal au­di­ence re­quires ex­plain­ing or defend­ing lots of ba­sic premises and back­ground that may be com­mon in an EA au­di­ence. The re­sults can be seen in these two pa­pers (https://​​com­mons.paci­ficu.edu/​​cgi/​​view­con­tent.cgi?ar­ti­cle=1573&con­text=eip /​​ https://​​link.springer.com/​​ar­ti­cle/​​10.1007%2Fs10677-013-9433-4). They both ar­gue for the vi­a­bil­ity of cer­tain EA ac­tivi­ties, yet the most of their space is taken up by is­sues which were widely con­sid­ered set­tled or un­in­ter­est­ing in the EA com­mu­nity at the time of pub­li­ca­tion.

Third, com­mon sci­en­tific norms re­volve around fre­quen­tist statis­tics and typ­i­cally re­quire proven re­sults in or­der to be con­sid­ered in­ter­est­ing; ev­ery­thing else is rel­e­gated to the same bucket of “plau­si­ble”. This makes it difficult to make a sci­en­tific ar­gu­ment for the pri­ori­ti­za­tion of some un­cer­tain causes over oth­ers. In EA, we are much more com­fortable with sub­jec­tive Bayesian episte­mol­ogy, we are more com­fortable act­ing on the ba­sis of un­cer­tain pos­si­bil­ities, and we are very in­ter­ested in know­ing just how likely or un­likely a pos­si­bil­ity is. With our own jour­nal, it will be eas­ier for us to pub­lish a pa­per that prop­erly mod­els all the un­cer­tain parts of the equa­tion that are nec­es­sary to com­plete a full cause pri­ori­ti­za­tion ar­gu­ment. Sub­jec­tive Bayesian method­ol­ogy does get pub­lished in rep­utable jour­nals, it’s just rare—most aca­demics are less com­fortable with it.

Fourth, in­volv­ing EAs in the peer re­view will en­sure that pa­pers are ap­pro­pri­ately writ­ten to tar­get EA crite­ria and dodge the likely EA coun­ter­ar­gu­ments. Even though the au­thor of a draft would be an EA, she may have a blind spot re­gard­ing the way that other EAs think about the is­sue.

It will lend no­ta­bil­ity to our move­ment and ideas.

Hav­ing a jour­nal on a sub­ject is a small step for es­tab­lish­ing it as a widely-ac­knowl­edged le­gi­t­i­mate field of study. It en­courages other peo­ple to cite the works in their own liter­a­ture and take them seri­u­osly. To be sure, a new jour­nal may not be no­ticed at all, and I don’t know what it takes to make a jour­nal seem prop­erly rep­utable as op­posed to some­thing like those cre­ation­ist jour­nals where they are just pub­lish­ing what­ever fits their point of view so no­body pays at­ten­tion to them. But any­thing would be tech­ni­cally bet­ter than hav­ing these ideas be put up on the in­ter­net by their own au­thors with no re­view or pub­li­ca­tion at all. To be sure, in prac­tice, a very bad jour­nal would be worse than noth­ing be­cause of the rep­u­ta­tion risk, but I doubt that we would screw things up that much.

It will cre­ate op­por­tu­ni­ties for out­siders to con­tribute.

Again as­sum­ing that it is suffi­ciently rep­utable, it could en­courage peo­ple out­side EA to weigh in on our ques­tions, be­cause now it means that they will get a pub­li­ca­tion. A ma­jor benefit here is that the re­view pro­cess, be­ing led by EAs, will help them trans­late their ideas into a form that is more suit­able for us. As it stands, peo­ple on the out­side who try to at­tack EA cause pri­ori­ties of­ten have a pretty poor grasp of what does or doesn’t work as an ar­gu­ment in the EA frame­work, and a poor grasp of our likely coun­ter­ar­gu­ments. This prob­lem per­sists in aca­demic pub­li­ca­tions that com­ment on EA causes. It would ob­vi­ously be much bet­ter to hash out these things pre-pub­li­ca­tion, so that they can provide us with more use­ful ar­gu­ments.

We have enough out­put to sus­tain a jour­nal.

Think about all the things be­ing writ­ten by in­di­vi­d­u­als in the com­mu­nity on fo­rums and their per­sonal web­sites, the blogs and re­ports from all our or­ga­ni­za­tions (Givewell, Open Phil, ACE, FLI...), and the pa­pers that are be­ing pub­lished in aca­demic jour­nals by EAs. Much of it is very se­ri­ously thought out. With all this effort, just a bit of em­pha­sis of cause pri­ori­ti­za­tion will en­able jour­nal is­sues to be put to­gether within an ac­cept­able length of time.

Hey, what about aca­demic in­cen­tives!? Peer re­view is bro­ken! It’s all sta­tus sig­nal­ing! Think about pub­li­ca­tion bias! Think about the repli­ca­tion crisis! Re­mem­ber the Sokal Af­fair??? Etc, etc.

How­ever great these prob­lems are, surely the prac­tice of spread­ing ar­gu­ments across Face­book walls, pod­casts, and blog posts has them or oth­ers to a greater de­gree. And that’s the state of EA thought on cause pri­ori­ti­za­tion. In any case, EAs know enough about these is­sues, and are suffi­ciently free of the need to pan­der to es­tab­lish­ment in­sti­tu­tions, that we can pre­sum­ably de­sign a jour­nal mechanism that is bet­ter than the norm.