Re: Testing Discourse for Debian - Moderation concepts

[Neil McGovern <neil@halon.org.uk>]

I am going to try and split this out into two replies, so those
following along can see the different issues. The irony of the
difficulty on doing this within email may or may not be lost for others.

On Sun, Apr 12, 2020 at 02:43:31PM -0700, Ihor Antonov wrote:
> > You have to trust the moderators, 
> 
> So far I am not convinced that I can trust you to moderate. 
> 
> > and you have to have some mechanism to
> > evaluate that trust and to discuss it and possibly revoke it if something
> > goes horribly awry. 
> 
> Prevention should always be the first step. Something WILL go wrong but you are
> too blinded by the immediate sugar candy in front of you.
> 

I just want to state, I won't debate any issues around freedom of
speech. I believe that these do not apply in this context - especially
with Debian being a private entity.

Now, I do believe you have a comment on moderation, and how this is
done. This requires me to explain two concepts in Discourse - trust
levels and flags.

Firstly, trust levels. These are the levels of "trust" that the platform
has in any particular user. Instead of explaining it here, please have a
read of the following:
https://blog.discourse.org/2018/06/understanding-discourse-trust-levels/
The short version is that the more a particular account interacts with
the community in a positive way, the more trust the system has about
them, and the more privileges they are afforded to assist in
moderation.

Secondly, flags. Discourse has the opinion that moderation cannot be
proactive with a small group of users - this doesn't scale. Instead, it
encourages community members to flag posts. If a post receives
sufficient flags, it is then automatically hidden. Users may chose to
"unhide" the post for themseleves if they wish to view it.

These are then sent to the moderating team to agree, disagree or ignore
the flag. This will unhide the post, or keep it hidden and offer an
opportunity for the moderator to suggest the original author edits their
post in light of the number of flags they got. If an author does so, the
post automatically unhides.

All these actions are logged, and affects the trust levels above. In
fact, every time an admin performs any action on a user, this is
logged.

I hope this explains how I believe that moderation is more powerful on
Discourse, but also more practical, transparent and accountable.

Neil