Successful conversations usually happen as a result of selection
circumstances that make it more likely that interesting people participate.
Early LessWrong was interesting because of the posts, then there was a phase
when many were still learning, and so were motivated to participate,
to tutor one another, and to post more.
But most don’t want to stay in school forever, so activity faded, and the
steady stream of new readers has different characteristics.
It’s possible to maintain a high quality blog roll, or an edited stream of
posts.
But with comments, the problem is that there are too many of them, and
bad comments start bad conversations that should be prevented rather
than stopped, thus pre-moderation, which slows things down.
Controlling their quality individually would require a lot of moderators,
who must themselves be assessed for quality of their moderation decisions,
which is not always revealed by the moderators’ own posts.
It would also require the absence of drama around moderation decisions,
which might be even harder.
Unfortunately, many of these natural steps have bad side effects or are hard
to manage, so should be avoided when possible.
I expect the problem can be solved either by clever algorithms that
predict quality of votes, or by focusing more on moderating people
(both as voters and as commenters), instead of moderating comments.
On Stack Exchange, there is a threshold for commenting (not just asking or
answering), a threshold for voting, and a separate place (“meta” forum)
for discussing moderation decisions.
Here’s my guess at a feature set sufficient for maintaining good
conversations when the participants didn’t happen to be selected for generating
good content by other circumstances:
All votes are tagged by the voters, it’s possible to roll back the effect
of all votes by any user.
There are three tiers of users: moderators, full members,
and regular users.
The number of moderators is a significant fraction of the number of
full members, so there probably should be a few admins who are outside
this system.
Full members can reply to comments without pre-moderation, while regular
users can only leave top-level comments and require pre-moderation.
There must be a norm against regular users posting top-level comments
to reply to another comment.
This is the goal of the whole system, to enable good conversations between
full members, while allowing new users to signal quality of their
contributions without interfering with the ongoing conversations.
Full members and moderators are selected and demoted based on voting
by moderators (both upvoting and downvoting, kept separate).
The voting is an ongoing process (like for comments, posts) and weighs
recent votes more (so that changes in behavior can be addressed).
The moderators vote on users, not just on their comments or posts.
Each user has two separate ratings, one that can make them a full member,
and the other that can make them a moderator, provided they are a
full member.
Moderators see who votes how, both on users and comments, and can use
these observations to decide who to vote for/against being a moderator.
By default, when a user becomes a full member, they also become
a moderator, but can then be demoted to just a full member if other
moderators don’t like how they vote.
All votes by demoted moderators and the effects of those votes, including
on membership status of other users, are automatically retracted.
A separate meta forum for moderators, and a norm against discussing
changes in membership status etc. on the main site.
This seems hopelessly overcomplicated, but the existence of Stack Exchange
is encouraging.
Successful conversations usually happen as a result of selection circumstances that make it more likely that interesting people participate. Early LessWrong was interesting because of the posts, then there was a phase when many were still learning, and so were motivated to participate, to tutor one another, and to post more. But most don’t want to stay in school forever, so activity faded, and the steady stream of new readers has different characteristics.
It’s possible to maintain a high quality blog roll, or an edited stream of posts. But with comments, the problem is that there are too many of them, and bad comments start bad conversations that should be prevented rather than stopped, thus pre-moderation, which slows things down. Controlling their quality individually would require a lot of moderators, who must themselves be assessed for quality of their moderation decisions, which is not always revealed by the moderators’ own posts. It would also require the absence of drama around moderation decisions, which might be even harder. Unfortunately, many of these natural steps have bad side effects or are hard to manage, so should be avoided when possible. I expect the problem can be solved either by clever algorithms that predict quality of votes, or by focusing more on moderating people (both as voters and as commenters), instead of moderating comments.
On Stack Exchange, there is a threshold for commenting (not just asking or answering), a threshold for voting, and a separate place (“meta” forum) for discussing moderation decisions. Here’s my guess at a feature set sufficient for maintaining good conversations when the participants didn’t happen to be selected for generating good content by other circumstances:
All votes are tagged by the voters, it’s possible to roll back the effect of all votes by any user.
There are three tiers of users: moderators, full members, and regular users. The number of moderators is a significant fraction of the number of full members, so there probably should be a few admins who are outside this system.
Full members can reply to comments without pre-moderation, while regular users can only leave top-level comments and require pre-moderation. There must be a norm against regular users posting top-level comments to reply to another comment. This is the goal of the whole system, to enable good conversations between full members, while allowing new users to signal quality of their contributions without interfering with the ongoing conversations.
Full members and moderators are selected and demoted based on voting by moderators (both upvoting and downvoting, kept separate). The voting is an ongoing process (like for comments, posts) and weighs recent votes more (so that changes in behavior can be addressed). The moderators vote on users, not just on their comments or posts. Each user has two separate ratings, one that can make them a full member, and the other that can make them a moderator, provided they are a full member.
Moderators see who votes how, both on users and comments, and can use these observations to decide who to vote for/against being a moderator. By default, when a user becomes a full member, they also become a moderator, but can then be demoted to just a full member if other moderators don’t like how they vote. All votes by demoted moderators and the effects of those votes, including on membership status of other users, are automatically retracted.
A separate meta forum for moderators, and a norm against discussing changes in membership status etc. on the main site.
This seems hopelessly overcomplicated, but the existence of Stack Exchange is encouraging.