
Describe conduct, not personalities
“Be nice” is not a complete moderation rule. It can be interpreted differently by different people and may be used to suppress legitimate criticism. Describe the behaviour you expect instead: challenge an idea without targeting a person, do not publish someone else’s private information, avoid repeated unwanted contact and keep promotional material within the stated limits.
Use examples to show where the boundary lies. A member saying that an event was poorly organised is not the same as repeatedly insulting the organiser. A firm disagreement is not automatically harassment. Context matters, including repetition, targeting and the practical effect on someone’s ability to participate. Rules should help moderators explain these distinctions rather than simply label a person as difficult.
Publish a reporting route and explain its limits. Members should know whether a report goes to volunteer moderators, staff or the platform. Do not promise anonymity or confidentiality that the tools cannot support. A report may need limited sharing with an authorised reviewer, and some platform actions may make its source easier to infer.
Use proportionate, explainable decisions
Moderation options may include clarification, a request to edit, content removal, a temporary restriction or exclusion. These are not always steps in a compulsory ladder. A credible threat or exposed personal information can require immediate protective action. An unclear first-time formatting mistake usually calls for a different response.
Assess the content, context and applicable rule before selecting an action. When it is safe and appropriate, tell the affected member what happened, which rule applied and what they can do next. Avoid public arguments about an individual’s motives. A concise explanation can be firm without becoming punitive theatre.
This separates a valid topic from an unsafe method of presenting it. It also gives the member a route back into the discussion rather than treating all criticism as an attack. Document enough context that another moderator can understand the decision without retaining unnecessary copies of the exposed information.
Separate incidents from ordinary disagreement
An incident needs an owner and a clear escalation path. Identify who can change access, contact platform support and make organisational decisions. If content suggests an immediate danger, use appropriate local emergency services and the platform’s relevant reporting tools; this website is not an emergency service. Do not ask an untrained moderator to investigate a threat alone.
Limit access to incident records. Record the reason for an action, the minimum relevant evidence, the responsible reviewer and any follow-up requirement. Decide how long records are needed and when they should be reviewed for deletion. Avoid circulating harmful material across team chats merely to keep everyone informed.
Automated filters can help manage volume, but they need human review and context. A filter may miss coded abuse or wrongly flag a quoted term in an educational discussion. Monitor errors and provide a practical way to challenge a decision. Do not describe automation as neutral simply because it applies the same matching rule repeatedly.
The Santa Clara Principles offer a useful reference on transparency and accountability in content moderation. The Trust & Safety Professional Association also publishes professional resources. These references inform practice; they do not replace platform rules, local legal obligations or specialist advice for serious cases.
Allow review and protect the team
An appeal process should state where to write, what information is useful and what outcomes are possible. Where staffing permits, have someone other than the original decision-maker review the case. If the same person must review it, be honest about that limitation. A fair process does not guarantee that an appeal succeeds, but it should allow relevant context and errors to be considered.
When a decision changes, correct the practical consequences and explain the correction to those who need to know. Do not publish sensitive details to demonstrate transparency. Aggregate explanations of recurring decisions can clarify policy without turning individual members into case studies.
- Consistency
- Comparable behaviour is assessed using the same principles, with relevant context recorded.
- Accountability
- Someone owns the decision, can explain it and can correct an error.
- Sustainability
- The work has coverage, boundaries and support rather than relying on permanent personal availability.
Review difficult cases with care for the moderators involved. Rotate exposure where possible, allow breaks from distressing material and give people a route to ask for support. If the workload exceeds capacity, narrow the service or pause growth rather than normalising exhaustion. Strong boundaries protect the people maintaining the space as well as the people participating in it.
Reference: the Santa Clara PrinciplesTransparency and accountability in content moderation.Reference: Trust & Safety Professional AssociationProfessional resources for trust and safety work.