Two weeks ago, I posted an idea about how we might introduce custom sorting of comments.
The general idea was that users could anonymously and individually rate other users as 'poor' 'neutral' or 'good' in regards to their comments. As a result the ratings would customize the comment sort for a user based on his/her ratings.
After some more discussion, it became clear that the most useful signal for influencing the comment sort was if a user's comments were considered to be 'poor'; It is much easier to draw a distinction between a 'neutral' or 'poor' commenter than it is between a 'neutral' and 'good' commenter', as a result, the signal is more reliable.
The point was also raised that rating users could become taxing, and this could make the effect less useful.
For these reason's our thinking on the issue has evolved, and so has the possible implementation.
Here is the current scheme:
In addition to being able to 'ignore' or 'mute' a user, you can 'hush' them. 'Hushing' a user increases the rate at which that user's comments fall to the bottom of a thread. Thus, all things being equal (time and votes), comments from users that you have hushed will be found at the bottom of comment threads.
Hushing a user will only effect your comment sort, not the sort for other users.
It is possible that the amount that a particular user is hushed could serve as a signal for some sort of global bias. However, generally speaking, that is not an approach that we prefer to take. Instead of applying community signals to top-down moderation, we prefer to allow individual moderation to lead to global effects. For example, if many users ignore a specific user, that user's posts are less likely to be shared and seen by the rest of the community without need for top-down moderation. Hushing keeps with this approach. If a significant number of users hush a specific user, that user's comments are less likely to be voted-up, or responded to, as a result, the effect of hushing can alter the comment sort across the entire community.
Another reason why this implementation seems preferable to the one originally suggested, is that it is functionally similar to 'ignore' and 'mute', and would be both easier to implement and to explain.
Your thoughts are much appreciated.