Eve library

Moderator Training

Guide

Revision 1

On this page

User guide 32.16.2.13. For human moderators working alongside Kuan Yin.

Kuan Yin does a lot of the heavy lifting, but you are the judgment, the warmth, and the accountability the system can't provide. This guide orients you to the role and — just as importantly — to looking after yourself in it.

Your role with the system#

  • Kuan Yin surfaces and scores; you decide the cases it routes to a human. Treat its outputs as informed suggestions, not verdicts.
  • You confirm irreversible actions. Bans, terminations, and permanent suspensions are never auto-applied — your review is the gate.
  • Lead with the compassionate frame. The default question is "what would help this person and this community grow?", not "what's the maximum penalty?".

Core workflows#

  1. Reviewing a flagged decision — read the rationale, the context snapshot, and the user's history/trust. Uphold, modify, or overturn, and always record why (it feeds transparency and appeals).
  2. Handling appeals — assume good faith; look for context the model missed.
  3. Crisis response — when crisis is detected, a person's safety comes first: follow the escalation path, share region-appropriate resources, and loop in the crisis team. You'll get specific crisis-intervention training before this is part of your queue.
  4. Coordinating a shield — during raids or targeted harassment, work with the defense tools and the affected creator.

Looking after yourself#

This work exposes you to harm, and that exposure is real and cumulative. The platform is built to protect you:

  • Exposure limits — a cap on harmful-content review per day.
  • Wellness time — protected breaks built into your shift.
  • Blur-first — severe content is blurred until you choose to view it.
  • Rotation & clean breaks — harmful review is interleaved with lighter work.
  • Screening & support — secondary-trauma screening, confidential counselling, and peer support are provided, not optional extras.

If you're struggling, say so. Stepping back is supported, and there's exit-support and long-term health monitoring if you move on from the role.

Onboarding path#

New moderators ramp gradually — lighter, lower-severity queues first, with resilience training — before handling the hardest cases. You're never thrown into the deep end.