Intervention Taxonomy
A classification of the interventions Communitas supports — what each does, when it's appropriate, and what makes it safe.
Communitas organizes interventions into five categories based on what they address. Each intervention is designed to be minimal, specific, and grounded in research on what actually improves community health.
Every intervention in Communitas must be opt-in, transparent, logged, and human-approved. No intervention reaches a member without their consent. No intervention executes without a steward’s review. Every intervention is recorded with enough context for audit and review. These are not guidelines — they are system constraints.
Connection interventions
Section titled “Connection interventions”Connection interventions create new relationships between members who would benefit from knowing each other but haven’t connected organically.
Opt-in introductions
Section titled “Opt-in introductions”Suggest a mutual introduction between two members in different clusters who share interests, complementary skills, or overlapping work. Both parties must consent before the introduction happens. Research on bridges in social networks shows that even a single new cross-cluster tie can improve information flow and reduce fragmentation.
Bridge invitations
Section titled “Bridge invitations”Invite members to participate in a conversation, project, or event outside their usual cluster. Unlike introductions (which connect two specific people), bridge invitations connect a person to a group. Appropriate when a cluster is becoming isolated or when a member’s expertise is relevant to a discussion happening elsewhere.
Mentor matching
Section titled “Mentor matching”Pair an experienced member with a newcomer or someone transitioning into a new role. Matching considers shared interests, communication style compatibility, and availability. Effective for reducing newcomer drop-off and distributing institutional knowledge beyond a small core.
Onboarding interventions
Section titled “Onboarding interventions”Onboarding interventions help new members form their first connections and find their footing.
Personalized first steps
Section titled “Personalized first steps”Generate a short, specific onboarding path for a newcomer based on their stated interests and the community’s current structure: where to introduce yourself, what to read first, who to talk to. Generic “welcome” messages are less effective than paths tailored to what the person actually cares about.
Newcomer paths
Section titled “Newcomer paths”Create structured entry points for different types of newcomers — contributors, learners, organizers — so that each person finds relevant activity quickly. Paths reduce the cognitive load of joining a complex community.
Buddy systems
Section titled “Buddy systems”Assign a specific existing member as a newcomer’s point of contact for their first few weeks. The buddy answers questions, makes introductions, and checks in. Research on newcomer retention shows that a single responsive human contact in the first week is one of the strongest predictors of long-term participation. See the newcomer onboarding experiment.
Facilitation interventions
Section titled “Facilitation interventions”Facilitation interventions improve the quality of conversations and make participation more rewarding.
Discussion prompts
Section titled “Discussion prompts”Suggest conversation starters or follow-up questions that draw out quieter members, connect a discussion to related threads, or shift a stalled conversation forward. Prompts are suggestions to stewards, not messages posted autonomously.
De-escalation suggestions
Section titled “De-escalation suggestions”When a conversation shows early signs of escalation — rising negativity, personal language, rapid-fire exchanges — suggest a de-escalation prompt to the moderator or participants. The goal is to intervene before conflict hardens, not to suppress disagreement. Research on conversational repair shows that early, light-touch interventions are more effective than heavy moderation after the fact. See the language facilitation experiment.
Recognition rituals
Section titled “Recognition rituals”Surface contributions that might otherwise go unnoticed: a member who consistently helps newcomers, a contributor whose work spans multiple clusters, someone who de-escalated a difficult conversation. Recognition distributed broadly counteracts participation concentration by making diverse contributions visible.
Governance interventions
Section titled “Governance interventions”Governance interventions help communities maintain and enforce their own norms.
Norms reminders
Section titled “Norms reminders”When a conversation drifts toward behavior that conflicts with community norms, surface the relevant norm — not as a warning, but as context. Effective when norms exist but aren’t top-of-mind, especially for newer members who haven’t internalized them yet.
Policy surfacing
Section titled “Policy surfacing”When a decision is being made that’s covered by existing community policy, surface that policy in context. Prevents communities from relitigating settled decisions and helps members understand the reasoning behind existing rules.
Conflict scaffolds
Section titled “Conflict scaffolds”For recurring interpersonal or structural conflicts, provide a structured process: acknowledge the disagreement, surface the relevant norms, suggest a resolution path, and document the outcome. Conflict scaffolds don’t resolve conflicts — they give communities a legitimate process for working through them. This reflects research on moderation as legitimacy.
Information interventions
Section titled “Information interventions”Information interventions ensure that relevant knowledge reaches the people who need it.
Thread surfacing
Section titled “Thread surfacing”When a discussion in one cluster is relevant to members in another cluster, suggest cross-posting or summarizing it for the affected group. Addresses information flow gaps without requiring members to monitor every channel.
Cross-cluster sharing
Section titled “Cross-cluster sharing”Periodically surface activity summaries across clusters so that members maintain awareness of what’s happening beyond their immediate group. Useful for preventing fragmentation in communities with many subcommunities.
Knowledge gardening
Section titled “Knowledge gardening”Identify valuable information buried in old threads, scattered across channels, or locked in individual expertise, and suggest organizing it into shared artifacts — documentation, FAQs, decision logs. The knowledge model supports this by tracking what information exists and where gaps remain.
Interventions are not features to ship — they are hypotheses to test. Each one should be evaluated against the health metrics it targets, using the experimental approach described in the experiment registry. What works in one community may not work in another. The taxonomy provides a starting vocabulary, not a fixed playbook.