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- More content accessible: Many people in the core of the community have unpublished rough notes on topics, or interesting ideas that aren’t suitable for e.g. academic publication.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).
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
Feature request: integrate the content from the EA fora into LessWrong in a similar way as alignmentforum.org
Risks&dangers: I think there is non-negligible chance the LW karma system is damaging the discussion and the community on LW in some subtle but important way.
Implementing the same system here makes the risks correlated.
I do not believe anyone among the development team or moderators really understands how such things influence people on the S1 level - it seems somewhat similar to likes on facebook, and it's clear likes on Facebook are able to mess up with peoples motivation in important ways. So the general impression is people are playing with something possibly powerful, likely without deep understanding, and possibly with a bad model of what the largest impacts are (focus of ordering of content, vs. subtle impacts on motivation)
In situations with such uncertainty, I would prefer the risks to be less correlated
edit: another feature request: allow to add co-authors of posts. a lot of texts are created by multiple people, ant it would be nice if all the normal functionality worked
Are there any plans to evaluate the current karma system? Both the OP and multiple comments expressed worries about the announced scoring system, and in the present day we regularly see people complain about voting behaviour. It would be worth knowing if the concerns from a year ago turn out to have been correct.
Related to this, I have a feature request. Would it be possible to break down scores in a more transparent way, for example by number of upvotes and downvotes? The current system gives very little insight to authors about how much people like their posts and comments. The lesson to learn from getting both many upvotes and many downvotes is very different from the lesson to learn if nobody bothered to read and vote on your content.
Are there particular instances of complaints related to voting behavior that you can recall?
I remember seeing a couple of cases over the last ~8 months where users were concerned about low-information downvotes (people downvoting without explaining what they didn't like). I don't remember seeing any instances of concern around other aspects of the current system (for example, complaints about high-karma users dominating the perception of posts by strong-voting too frequently). However, I could easily be forgetting or missing comments along those lines.
Currently, you can see the number of votes a post or comment has received by hovering over its karma count. This does let you distinguish between "many upvotes and many downvotes" and "no votes". Adding a count of upvotes and downvotes would provide more information about the distribution of strong votes (e.g. one strong upvote vs. several weak downvotes, or vice-versa). I can see how that could be useful, and I'll bring it up with the Forum's tech team to hear their thoughts. Thank you for the suggestion!
It's not an instance of complain, but take it as a datapoint: I've switched off the karma display on all comments and my experience improved. The karma system tends to mess up with my S1 processing.
It seems plausible karma is causing harm in some hard to perceive ways. (One specific way is by people updating on karma pattern mistaking them for some voice of the community / ea movement / ... )
As humans, we are quite sensitive to signs of social approval and disapproval, and we have some 'elephant in the brain' motivation to seek social approval. This can sometimes mess up with epistemics.
The karma represents something like sentiment of people voting on a particular comment, weighted in a particular way. For me, this often did not seemed to be a signal adding any new information - when following the forum closely, usually I would have been able to predict what will get downvoted or upvoted.
What seemed problematic to me was 1. a number of times when I felt hesitation to write something because part of my S1 predicted it will get downvoted. Also I did not wanted to be primed by karma when reading other's comments.
On a community level, overall I think the quality of the karma signal is roughly comparable to facebook likes. If people are making important decisions, evaluating projects, assigning prices... based on it, it seems plausible it's actively harmful.
This seems like a strong plan, and I'm glad you've thought things through thoroughly. I'll just outline points of agreement, and slight differences.
I certainly agree with the approach of building EA Forum 2 from LessWrong 2. After all, the current EA Forum version was built from LessWrong 1 for similar reasons. We had a designer sketch the restyled site, and this was quite a positive experience, so I'd recommend doing the same with the successor. Basically, the EA Forum turned out quite a bit more beautiful than the LessWrong, and the same should be possible again. I think there are some easy wins to be had here, like making the LW2's front-page text a bit darker, but I also think it's possible to go beyond that and make things really pretty all-around.
I agree with keeping LW2's new Karma system, and method of ordering posts, and I think that this is a major perk of the codebase. I'm also happy to see that you seem to have the downsides well-covered.
One small diff is when you say "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." Personally, I think that forum administrator... (read more)
I've not yet read it myself, but I'm curious if anyone involved in this project has read "Building Successful Online Communities: Evidence-Based Social Design" (https://mitpress.mit.edu/books/building-successful-online-communities). Seems quite relevant.
I'm concerned with the plans to make voting/karma more significant; I would prefer to make them less significant than the status quo rather than more. Voting allows everyone's biases to influence discussion in bad ways. For example, people's votes tend to favor:
I think the proposed karma system, particularly when combined with the highly rated posts being listed higher, is a quite bad idea. In general, if you are trying to ensure quality of posts and comments while spreading the forum out more broadly there are hard tradeoffs with different strengths and weaknesses. Indeed, I might prefer some type of karma weighting system to overly strict moderation but even then the weights proposed here don't seem justifiable.
What problem is being solved by giving up to 16 times maximum weight that would not be solved with gi... (read more)
Speculative feature request: anonymous commenting and private commenting
Sometimes people might want to comment anonymously because they want to say something that could hurt their reputation or relationships, or affect the response to the criticism in an undesirable way. For example,. OpenPhil staff criticising a CEA or 80K post would have awkward dynamics because OpenPhil funds these organizations partly. Having an option to comment anonymously (but let the default be with names) will allow more free speech.
Relatedly, some comments could be marked as &quo... (read more)
Thank you Marek and the whole CEA team for taking on this project! I love your initiative and what you outline seems like a very valuable and necessary step for the EA community. If things work out as you imagine, EA could be one of the first science-driven communities with a strong "community-reviewed" journal type offering (in this vein it may make sense to introduce different types of "publications" – idea, project report, scientific publication, etc. – with different standards for review and moderation). Very inspiring!
A question t... (read more)
The problem with down-voting is that it allows for views to be dismissed without any argument provided. It's kind of bizarre to give a detailed explanation why you think X is Y, only to see someone has down-voted this without explaining a tad bit why they disagree (or why they "don't find it useful"). I just can't reconcile that approach with the idea of rational deliberation.
One solution would be to demand that every down-vote comes with a reason, to which the original poster can reply.
If we upvote someone's comments then we trust them to be a better authority, so we should give them a greater weight in vote totals. So it seems straightforward that a weighted vote count is a better estimate of the quality of a comment.
The downside is that can create a feedback loop for a group of people with particular views. Having normal votes go from 1x to 3x over the course of so many thousands of karma seems like too small a change to make this happen. But the scaling of strong votes all the way up to 16x seems very excessive and risky to me.
Anothe... (read more)
Hey, first of all, thanks for what I'm sure what must have been a lot of work behind this. Many of these ideas seem very sensible.
Am I right in assuming that the scale for the upvotes was intended to be roughly-but-not-exactly logarithmic? And do downvotes scale the same way?
The post says that no user has more than 1,000 karma. But Peter_Hurford has more than 8,000 karma. I'm bringing this up not to quibble but rather because I'm wondering whether the threshold was meant to be set differently or perhaps users will lose some fraction of their karma during the switch.
Also, for link posts, I think it might be a good idea to require cross-posting the summary or an excerpt.
If you add a tag feature, can you make it so that authors can add tags to posts imported from EA Forum 1.0? I think it'd be great if someone interested in animal suffering could easily see all the EA Forum posts related to animal suffering.
And would you be willing to add a feature that allows you to tag individuals? (For this to work, you'd have to provide notifications in a more prominent way than the current 'Messages' system.)
I've always found the "karma" system utterly repulsive and deeply disturbing (across online forums in general) . It's a tool that can so easily catalyze bias and censorship, to the point that it becomes way more dangerous than it is useful. And the addition of older members having higher votes is extremely dangerous in preventing new members to question ideas from the dominant majority, hence leading to dogmatism.
The assumption that this will be prevented by already existing variety of views is not at all good enough guarantee: on the one hand, a... (read more)
DON'T BE A NECROMANCER!
LessWrong 2.0 resurrected all posts deleted by original posters which then had to be individually deleted, by users who may or may not be aware that this had happened. Please ensure this isn’t replicated with Effective Altruism Forum 2.0. If I can’t control my content I’ll move somewhere else that I can.
Would it be possible to introduce a coauthoring feature? Doing so would allow both authors to be notified of new comments. The karma could be split if there are concerns that people would free ride.
I'd like to echo a strong concern that karma based voting will lead to groupthink etc.
I'd feel substantially better if it was karma based upvotes only, and no karma based downvotes. Karma based downvotes allow community insiders to effectively kill posts.
Feature request: show reading times
It would be useful to show approximate reading times for posts, because readers can decide whether to commit to a long article or not. This saves valuable time of EA's, and improves the engagement with the post.
This is my first comment, but I'm currently writing and hoping to post some detailed reports later. Are there any restrictions on new users posting? I haven't tried posting yet, so I'm not aware if there are currently any restrictions in place but will there be new restrictions on new users in the future other than the karma system?
Would it be possible to have a 'sort by area' option? To see what people in the local community are writing, reading, commenting on, working on etc. May need location tags or location to be listed in profiles.
Leading from this, will the views each post gets be measured? Is that currently used to rank 'top' posts? Is the ratio of people reading with a hub profile to without a hub profile measured?
Having a quick suggestions/feedback form at the side of the main page or FAQ section may be useful. I've introduced them onto a few other online communities and they've been more popular and used than expected.
Would it be possible for you add a minimum font size requirement? Posts like this one are hard for me to read.