After some time we developed another type of bot: the moderator. This one you cannot see, and if you behave you will barely know it is there. Building it has been the hard part — it needs a feel for the conversation so it does not clamp down when it should not.
The idea
The idea is a service watching over messages. If you are new, it will watch carefully. It also checks in every so often — so do not get too comfortable. The reason it does not stream every message is cost. The backend already has its own protection that covers all messages; the moderator is the more context-aware layer on top.
Reports also go through the moderator. No moderation system is perfect, so a human is still kept in the loop.
Reporting a user
If you report someone and the report looks legitimate, the moderator will take a look and see whether the messages pass a threshold. If they do, it decides on an action: a warning, a mute, or removal.
SFW mode
We have added a safe-for-work mode. The main lobby is always safe for work. This part is still a work in progress — the aim is for the moderator to tell the difference between someone just being rude and something that really needs action. It also applies to private messages (again, work in progress), and for now that mostly relies on the report feature.
Impact so far
We have run some tests, and the basics hold up. Only time will tell for the rest.
Comments
No comments yet. Be the first to share a thought.