EA Forum 2.0 Initial Announcement

At the re­quest of Ryan Carey, who set up the EA Fo­rum, CEA took over its run­ning last year, as an­nounced in this post. Here we give an up­date on what we plan to do with it. Th­ese are our ini­tial thoughts—we’d wel­come feed­back on how they could be im­proved.

In sum­mary, the tech­nol­ogy of the cur­rent ver­sion of the Fo­rum will reach end of life this cal­en­dar year. We plan to trans­fer to a new Fo­rum at fo­rum.effec­tivealtru­ism.org, us­ing the code­base of the new LessWrong.com (for more de­tail on how we made this de­ci­sion please see the ‘Tech­ni­cal Back­ground’ sec­tion be­low). Con­tent, user ac­counts and karma from the old Fo­rum will be trans­ferred, and old links redi­rected as needed.

The move will lead to some changes in how the Fo­rum ap­pears:

  • We will have a “com­mu­nity” sub­fo­rum, which we are plan­ning to re­serve for or­ga­ni­za­tional up­dates and dis­cus­sion of com­mu­nity is­sues and re­sources (al­though we haven’t yet de­cided on the dis­tinc­tion).

  • In con­trast to the cur­rent sys­tem where posts are dis­played in re­verse chronolog­i­cal or­der, top-rated posts will ap­pear higher on the page. We ex­pect this will make it eas­ier for read­ers to see the best-qual­ity con­tent quickly.

  • It will be pos­si­ble to view con­tent in or­dered col­lec­tions (‘se­quences’).

  • Users will have pro­files, where they can dis­play in­for­ma­tion about their in­ter­ests.

  • The Karma sys­tem and front­page de­sign will also be differ­ent.

We are also us­ing this op­por­tu­nity to im­prove mod­er­a­tion and con­tent on the Fo­rum. We will be work­ing with some of the top fo­rum posters and thinkers in the com­mu­nity to en­courage them to post more fre­quently to the Fo­rum, and we will be cross-post­ing high qual­ity con­tent from el­se­where. The amount of effort put into mod­er­a­tion will be in­creased, aiming to give de­tailed feed­back on what con­tent is use­ful.

In this post, we aim to ex­plain why we think we should make these changes, set out what this will mean for Fo­rum users, and seek your feed­back on our plans.


Our vi­sion is that the Fo­rum be­comes the main hub for con­tent and dis­cus­sion in the com­mu­nity. We’d like most of the think­ing that the com­mu­nity does to go on the Fo­rum, with all of EA’s top re­searchers post­ing and com­ment­ing reg­u­larly.

We think that this has sev­eral benefits for the com­mu­nity:

  • More con­tent ac­cessible: Many peo­ple in the core of the com­mu­nity have un­pub­lished rough notes on top­ics, or in­ter­est­ing ideas that aren’t suit­able for e.g. aca­demic pub­li­ca­tion. They may not want the over­head as­so­ci­ated with run­ning their own blog, or may want more feed­back than they would get on their own blog. If they know that post­ing to the Fo­rum will re­li­ably get them credit, and use­ful com­ments, they’re more likely to post there.

  • Easier to find con­tent: The Fo­rum’s search func­tion should turn up most of the rele­vant think­ing on a topic, whereas cur­rently much is scat­tered on per­sonal blogs. Similarly, peo­ple look­ing to get their daily fix of EA con­tent should be able to find it on the Fo­rum, rather than scat­tered on blogs.

  • More en­gage­ment: If more peo­ple are post­ing, more will be read­ing and com­ment­ing. We hope that this means that good posts get more en­gage­ment than they have in the past year or so of the Fo­rum. Ideally, it be­comes a place of lively de­bate be­tween differ­ent per­spec­tives within the com­mu­nity.

  • A sin­gle point of dis­cus­sion: If more dis­cus­sion is con­cen­trated in one place, we hope it will be eas­ier to keep the com­mu­nity synced up, and learn­ing from each other.

To this end, our first goal is to en­sure that the Fo­rum is used by the top con­tent pro­duc­ers in EA.

If we achieve this first goal, then we think that the Fo­rum will be an ex­cel­lent way to on­board newer peo­ple to the com­mu­nity, be­cause: they will be ex­posed to high qual­ity, fresh con­tent; they will then have the op­por­tu­nity to en­gage with that con­tent and con­tribute them­selves; and they will get high-qual­ity feed­back on what con­tent of theirs is es­pe­cially good, al­low­ing them to be­come fa­mil­iar with EA top­ics and com­mu­nity dis­cus­sion norms.

How will the EA Fo­rum change?

New fea­tures on ini­tial release

Be­sides the up­dated de­sign of the plat­form, in the ini­tial re­lease we ex­pect users to benefit from the fol­low­ing fea­tures:

Sort­ing by pop­u­lar­ity/​en­gage­ment rather than only time order

The cur­rent Fo­rum pre­sents posts to users only based on time or­der, with the most re­cent posts ap­pear­ing at the top of the page. The new front page will use a slightly mod­ified ver­sion of the Hacker News al­gorithm to or­der posts by num­ber of up­votes and amount of en­gage­ment. This should en­sure that pop­u­lar posts won’t be crowded out sim­ply due to new posts be­ing made. We ex­pect this to im­prove the ex­pe­rience of read­ing the Fo­rum, make it eas­ier for users to find the best con­tent and re­duce the cost to post­ing ad­di­tional con­tent to the Fo­rum. Posts about less-gen­eral in­ter­est top­ics will be visi­ble to those who seek them out, but will not ap­pear on the front page un­less they at­tract up­votes. We hope that this will en­courage more peo­ple to post thoughts and works-in-progress to the Fo­rum for feed­back.

A com­mu­nity subforum

In ad­di­tion to the main Fo­rum, we will have a sub­fo­rum which we are ten­ta­tively plan­ning to call “Com­mu­nity”. We ex­pect to use this cat­e­gory for dis­cus­sion about is­sues in the EA com­mu­nity, or­ga­ni­za­tional up­dates and meta dis­cus­sion about the Fo­rum it­self.

Sub­scribing to users

It will be pos­si­ble to sub­scribe to par­tic­u­lar users’ con­tri­bu­tions. Each user’s page also func­tions as a kind of per­sonal blog, with a space to write about their in­ter­ests and ap­proach to prob­lems. This should help users per­son­al­ize their con­tent feed over time to be bet­ter fo­cused on ar­eas they are in­ter­ested in.


A se­quence is a set of linked ar­ti­cles on a par­tic­u­lar topic. Any user will be able to cre­ate se­quences, whether this be from their own origi­nal con­tent, or as a way to col­lect their fa­vorite pieces on a topic. Highly up­voted, user-cre­ated se­quences will be dis­cov­er­able by other users. We hope that this will help posters de­velop ideas in more depth, while keep­ing in­di­vi­d­ual posts to a read­able length, and that users stop hav­ing to post mul­ti­ple, num­bered parts to longer posts. It should also be a neat way to pre­sent canon­i­cal con­tent to new users. You can see some ex­am­ples of se­quences on LW.

A re­worked karma system

We will be aiming to trans­fer user ac­counts along with karma from the old to the new site in time for re­lease. Also, the way the karma sys­tem will func­tion will be slightly differ­ent, mainly by giv­ing a slightly higher weight to up- and down-votes from users who them­selves have more karma. There will also be the pos­si­bil­ity to choose be­tween giv­ing a small up/​down­vote or giv­ing a larger boost by hold­ing the up/​down­vote but­ton.

The best sum­mary of how this works is that:

Nor­mal votes (one click) will be worth

  • 3 points – if you have 25,000 karma or more

  • 2 points – if you have 1,000 karma (cur­rently no EA Fo­rum user is above the 2-point level)

  • 1 point – if you have 0 karma

Strong Votes (click and hold) will be worth

  • 16 points (max­i­mum) – if you have 500,000 karma

  • 15 points – 250,000

  • 14 points – 175,000

  • 13 points – 100,000

  • 12 points – 75,000

  • 11 points – 50,000

  • 10 points – 25,000

  • 9 points – 10,000

  • 8 points – 5,000 (Cur­rently no EA Fo­rum user is above the 8-point level)

  • 7 points – 2,500

  • 6 points – 1,000

  • 5 points – 500

  • 4 points – 250

  • 3 points – 100

  • 2 points – 10

  • 1 point – 0

We think that the EA com­mu­nity should as­pire to par­tic­u­larly high stan­dards of dis­cus­sion, and that the best way to main­tain those stan­dards is to give those who have con­tributed to the com­mu­nity more abil­ity to sig­nal what con­tent is good for the Fo­rum than some­one who has just cre­ated an ac­count. How­ever, we ob­vi­ously don’t want this to be­come a tyranny of a few users. There are sev­eral users, hold­ing very differ­ent view­points, who cur­rently have high karma on the Fo­rum, and we hope that this will help main­tain a varied dis­cus­sion, while still en­sur­ing that the Fo­rum has strong dis­cus­sion stan­dards.

The dis­tinc­tion be­tween nor­mal and strong votes seems use­ful be­cause it helps to differ­en­ti­ate be­tween posts and com­ments which are “good”, and those that are ex­cel­lent, and should be read by oth­ers. You can read more at this post.

Sup­port for Mark­down and LaTeX: Will make it eas­ier for peo­ple to for­mat posts nicely, and in­clude for­mu­lae where rele­vant.

Mov­ing to fo­rum.effec­tivealtru­ism.org: We plan to move the Fo­rum to a new do­main – fo­rum.effec­tivealtru­ism.org – to avoid the con­fus­ing pro­lifer­a­tion of do­mains. We’ll also try to bring some of the brand­ing into al­ign­ment with effec­tivealtru­ism.org. We’ll make sure that we fol­low good prac­tices for redi­rect­ing links etc. to en­sure that Fo­rum con­tent is still eas­ily search­able us­ing Google.

Pos­si­ble fea­tures af­ter ini­tial release

After we have suc­cess­fully re­leased the ini­tial ver­sion of the Fo­rum, we ex­pect to con­tinue de­vel­op­ment on some or all of the be­low fea­tures. Th­ese are just pre­limi­nary ideas, and we are go­ing to be re­spond­ing to feed­back to de­ter­mine how to best pri­ori­tize among them, or what other ideas we should put effort into.

Differ­ent land­ing pages for new users vs old hands: We are hop­ing that the new EA Fo­rum will even­tu­ally be­come a good place to di­rect peo­ple rel­a­tively new to the com­mu­nity who would like to find out more. To this end we will be able to have the land­ing page for non-signed in users show a col­lec­tion of core con­tent, and then only once a user cre­ates an ac­count will they see more re­cent dis­cus­sions. Hope­fully this will en­courage new users to get up to speed with the ba­sics be­fore en­gag­ing with new (pos­si­bly more ad­vanced/​weirder) con­tent.

Ad­di­tional sub­fo­rums: (As re­quested.) In the ini­tial re­lease we will have one main fo­rum with a “Com­mu­nity” sub­fo­rum. It may be pos­si­ble in the longer-term to add more sub­fo­rums, though we would want to en­sure that there is enough con­tent in each po­ten­tial new cat­e­gory to jus­tify this. An al­ter­na­tive solu­tion may be to im­ple­ment a tag sys­tem. Our work on this will also be de­ter­mined by what fea­tures the LessWrong team ends up im­ple­ment­ing in this space.

Page for lo­cal groups and events: In the fu­ture we may want to add a page de­voted to lo­cal groups and events, like LessWrong’s com­mu­nity page. We will de­ter­mine whether adding this func­tion­al­ity makes sense af­ter re­lease.

Com­mu­nity-sourced ideas: We would like to hear from the users of the new Fo­rum to help shape de­vel­op­ment pri­ori­ties. We be­lieve that the best ideas are of­ten missed when mak­ing an ini­tial pro­ject plan, but are likely to emerge from users once they are ac­tu­ally us­ing a plat­form. As such we will be main­tain­ing open com­mu­ni­ca­tion with users in or­der to make sure we cap­i­tal­ize on high ex­pected value im­prove­ments.

Moder­a­tion and curation

More ac­tive mod­er­a­tion: Re­cently, mod­er­a­tion on the Fo­rum has mostly been about re­mov­ing spam. We think that a bet­ter goal for mod­er­a­tion is to im­prove the qual­ity of think­ing and de­bate on the Fo­rum. There­fore, we think that it might be use­ful to have a more ac­tive, pos­i­tive form of mod­er­a­tion. In par­tic­u­lar, we are con­sid­er­ing in­vest­ing mod­er­a­tor time in giv­ing de­tailed feed­back on posts and com­ments. We aim to draw up some more de­tailed plans for mod­er­a­tion be­fore the re­lease of the Fo­rum, and we will post these plans to the cur­rent Fo­rum for feed­back, and to the new Fo­rum for trans­parency. To fa­cil­i­tate this more time-in­ten­sive mod­er­a­tion, we plan to bring on Max Dal­ton as a fourth mod­er­a­tor (in ad­di­tion to Howie Lem­pel, Ju­lia Wise, and Denise Melchin) and we will con­sider pay­ing mod­er­a­tors for the time they spend on this.

Creat­ing more link posts: Cur­rently we are hes­i­tant about al­low­ing lots of link posts on the Fo­rum, be­cause they dis­place origi­nal con­tent. Be­cause the new Fo­rum will by-de­fault sort by en­gage­ment, we are less wor­ried about this effect. Ad­di­tion­ally, we hope that link posts can help to con­soli­date con­tent to reach our vi­sion of the Fo­rum as a cen­tral hub for con­tent in the com­mu­nity.

En­courag­ing top thinkers to post more: We plan to en­gage with EA re­searchers and lead­ers and en­courage them to post more fre­quently.

Cu­rat­ing and choos­ing se­quences: We plan to cu­rate se­quences of posts on a va­ri­ety of top­ics from ex­ist­ing Fo­rum con­tent. Users are also free to cu­rate their own se­quences, and vote on which se­quences are best. Moder­a­tors can choose which se­quences to pro­mote (e.g. on the front page for non-logged-in users). We will con­sult with mod­er­a­tors and our ad­vi­sory com­mit­tee on this, and we also wel­come ideas and feed­back in the com­ments. One se­quence that we think would be well suited to this role is CEA’s EA Hand­book (note that this will soon have up­dated con­tents). We en­courage you to cre­ate ad­di­tional and al­ter­na­tive se­quences.

Con­sult­ing with re­lated fo­rums: We will con­sult the mod­er­a­tors of LessWrong, the EA sub­red­dit, and the main EA Face­book group about how we should re­late to these differ­ent spaces. By de­fault, we will keep the re­la­tion­ships as they were pre­vi­ously (for ex­am­ple in which types of con­tent are cross-posted to which spaces), but there might be im­prove­ments that can be made.

Pos­si­ble fu­ture plans: In the fu­ture, we may ex­per­i­ment with giv­ing prizes for good con­tent (along the lines of the AI al­ign­ment prize).

Tech­ni­cal Background

Since CEA took re­spon­si­bil­ity for the Fo­rum, we have been think­ing about how best to im­prove the ex­pe­rience for users and what should come next for the pro­ject. We were limited, how­ever, in our abil­ity to make im­prove­ments and changes to the cur­rent ver­sion, due to our team’s un­fa­mil­iar­ity with the code­base.

Ad­di­tion­ally, it was be­com­ing clear that the cur­rent ver­sion of the soft­ware was near­ing the end of its lifes­pan, and Trike Apps (the Fo­rum’s cur­rent host­ing providers) con­firmed that they would be forced to stop their host­ing and main­te­nance ser­vices at the end of 2018.

As such, we have since been look­ing into the best op­tions for build­ing a new ver­sion of the ser­vice. We were con­sid­er­ing four main ap­proaches over the first half of 2018:

Use an off-the-shelf dis­cus­sion soft­ware plat­form, and cus­tomize it for the needs of our com­mu­nity;

Spend time get­ting fa­mil­iar with the code­base/​stack that runs the ex­ist­ing Fo­rum, patch­ing/​up­grad­ing the soft­ware our­selves or port­ing the in­stal­la­tion to a new server;

Use the LessWrong 2.0 code­base, and cus­tomise that; or

Build a solu­tion from scratch, in-house.

After eval­u­at­ing the above op­tions we think that us­ing the LessWrong 2.0 (LW2) code­base would be the best way for­ward.

We de­cided on this ap­proach be­cause we were not able to find an off-the-shelf solu­tion as well tar­geted to our use-case as LW2. Their team has spent the past year think­ing about how to cre­ate the best dis­cus­sion plat­form for a com­mu­nity ad­ja­cent to ours, with very similar goals of en­abling in­tel­lec­tual progress on im­por­tant ques­tions and helping cre­ate com­mon knowl­edge in a (mostly) on­line com­mu­nity.

We eval­u­ated the costs of sim­ply stick­ing with the ex­ist­ing code­base. How­ever, we were cau­tioned against this by Trike Apps, who be­lieve that the setup is fairly brit­tle, and that it would be difficult to get run­ning on a differ­ent server en­vi­ron­ment. Given that LW2 faced similar prob­lems, and made the de­ci­sion to de­velop a new, mod­ern we­bapp, we be­lieved it made sense to fol­low their lead.

In this light, build­ing a solu­tion from scratch also seemed a less at­trac­tive op­tion, as we would likely have du­pli­cated a lot of the LessWrong team’s work, and it was plau­si­ble that our ver­sion would have con­verged on similar solu­tions to is­sues we were hop­ing to solve.

There are some down­sides to this ap­proach. In par­tic­u­lar, there is a risk in us­ing soft­ware that is rel­a­tively new, and does not have ei­ther a) a wide base of open-source con­trib­u­tors or b) a large com­mer­cial soft­ware com­pany back­ing it. How­ever,

  • The LessWrong team has proven it­self to be ca­pa­ble at de­liv­er­ing a product;

  • The code has been tested in pro­duc­tion for nearly a year;

  • The team plans to con­tinue full-time de­vel­op­ment and sup­port on the product;

  • The soft­ware is fairly mod­u­lar and is built on top of well-doc­u­mented open-source com­po­nents (Me­teor, Vul­can, Re­act, GraphQL); and

  • The team is lo­cated very close to CEA’s de­vel­op­ment team (in Berkeley), and plans to provide setup ad­vice and on­go­ing sup­port.

As such, we be­lieve this is the right call, all things con­sid­ered. CEA will have a for­mal sup­port ar­range­ment with them.

LW1 was hosted in a very similar setup (and in fact the im­pend­ing End Of Life that will af­fect the EA Fo­rum would have also af­fected the pre­vi­ous ver­sion of LessWrong). Con­se­quently, the team has ex­pe­rience mi­grat­ing data from a very similar sys­tem. This means that we can be more con­fi­dent that the ex­ist­ing Fo­rum data can be safely ported, that we will not need to main­tain sep­a­rate archives for older con­tent, and that we will avoid bro­ken links and other an­noy­ances.

We plan to main­tain the code­base as a fork of LW2, with the LW2 code­base as an up­stream repo, and pre­fer sub­mit­ting PRs up­stream to mak­ing lo­cal changes. We’ll make min­i­mal changes to the site’s them­ing to make it more similar to effec­tivealtru­ism.org, but oth­er­wise at­tempt to run the code­base as closely as pos­si­ble to LW2. This en­sures that most of the main­te­nance stays with a de­vel­op­ment team that is fa­mil­iar with the code­base, min­i­mizes the risk that CEA will be un­able to main­tain the fo­rum in the fu­ture, and means that any im­prove­ments to one code­base will be shared by the other.

Po­ten­tial risks or downsides

Karma is more “elitist”: The new karma sys­tem gives greater weight to the votes of those with high karma. We think that this will over­all im­prove the karma sys­tem: to main­tain high dis­cus­sion stan­dards, it is use­ful to give top con­trib­u­tors a big­ger say in de­ter­min­ing what con­tent is good than some­one who has just cre­ated an ac­count. Cur­rent high-karma users rep­re­sent a va­ri­ety of differ­ent wor­ld­views.

SEO dis­ad­van­tages: While we will make ev­ery effort to en­sure we fol­low good prac­tices for mov­ing a web­site to a new do­main, there’s some chance that the do­main change could cause us to lose PageRank (though it might also in­crease due to be­ing as­so­ci­ated with effec­tivealtru­ism.org, which is highly ranked).

Too many new­com­ers: If the Fo­rum does in­crease its rank in searches, or if more peo­ple link to the fo­rum, there might be an in­flux of new­com­ers, lead­ing to an “eter­nal septem­ber” effect.

We want to en­sure that the Fo­rum re­mains a place for more ad­vanced dis­cus­sion. There­fore at least ini­tially, we will not link to the Fo­rum promi­nently. Once the stan­dards of dis­cus­sion are solid, we will grad­u­ally be­gin to link to the Fo­rum more promi­nently: karma and mod­er­a­tion should en­sure that the best con­tent is high­lighted, and lower-qual­ity com­ments lose promi­nence. We will mon­i­tor how this changes dis­cus­sion on the Fo­rum. How­ever, we think that ex­pos­ing new­com­ers to high qual­ity con­tent, al­low­ing them to con­tribute, and giv­ing them feed­back on their con­tri­bu­tions could be an ex­cel­lent way for them to im­prove their un­der­stand­ing, and be­come more in­volved in the com­mu­nity. We think that the benefits of this way of on­board­ing new com­mu­nity mem­bers will likely out­weigh the costs.

Availa­bil­ity of older con­tent: Ex­ist­ing Fo­rum posts and com­ments will con­tinue to be available. It will still be pos­si­ble to com­ment on old posts.

Bro­ken links: We will en­sure that links to old Fo­rum posts and com­ments redi­rect to the ap­pro­pri­ate places on the up­dated site.

Se­quence se­lec­tion and mod­er­a­tion: CEA will have two staff as mod­er­a­tors of the Fo­rum, which will give us more con­trol over which posts and se­quences are pro­moted (e.g. on the front page). Although CEA has a view on which causes we should pri­ori­tize, we rec­og­nize that the EA Fo­rum is a com­mu­nity space that should re­flect the com­mu­nity. We will mod­er­ate based on agreed stan­dards, which will be fo­cused on en­sur­ing good dis­cus­sion rather than any par­tic­u­lar con­clu­sions. We will pub­lish our mod­er­a­tion and cu­ra­tion plans for feed­back, and we en­courage you to hold us to those pub­lic stan­dards.

What comes next

We are already test­ing a closed beta among CEA staff, to give CEA’s de­vel­op­ment team time to iron out bugs and fa­mil­iarize them­selves with the code­base, and the mod­er­a­tion team to ex­per­i­ment with the new mod­er­a­tion fea­tures.

In mid-Au­gust, we hope to have a semi-pub­lic beta of top Fo­rum con­trib­u­tors, which we will grad­u­ally open to more users. Around this time, we aim to share our de­tailed mod­er­a­tion plans in or­der to seek feed­back.

If all goes well, we hope to lock the cur­rent Fo­rum (i.e. pre­vent new posts), and switch to the new Fo­rum by late Au­gust/​early Septem­ber. Some time af­ter that, all old con­tent and karma will be trans­ferred to the new do­main.

Longer-term, be­sides the above-men­tioned sub­fo­rums and lo­cal groups sec­tions, we are still con­sid­er­ing a Sin­gle Sign On sys­tem with Effec­tiveAltru­ism.org (cur­rently used by EA Funds and the Giv­ing What We Can pledge), more de­sign changes, and work­ing with the LessWrong team to im­ple­ment new fea­tures that both code­bases can benefit from, or (if ap­pro­pri­ate) build­ing new fea­tures spe­cific to the EA Fo­rum.

Im­por­tantly, we want Fo­rum users to have a say in how this de­vel­ops. We’d wel­come poin­t­ers to con­sid­er­a­tions we’ve missed, or feed­back on any part of our plans.


Marek, Max, Sam, Ju­lia & JP