ADR-017: Cohort-Based Community Layers

Date: 2025-12-29 Status: Accepted Deciders: Development Team Related: ADR-003 (Multi-Tenant RLS), ADR-011 (Reputation Decay)

Context

Human communities naturally organize in layers based on interaction frequency and trust depth. Anthropological research (Robin Dunbar, etc.) shows:

• Inner circles: 5-7 people (high frequency, deep trust)
• Active community: ~50 people (regular interactions, medium trust)
• Extended network: ~150 people (Dunbar's Number, baseline trust)

Current Karmyq implementation treats all community members equally, missing opportunity to model natural social structures.

Decision

Implement cohort-based community layers reflecting natural human social organization.

Layer Structure

Inner Circle (5-7 people)

• High frequency interactions (weekly or more)
• Deep trust relationships
• Core decision-making group
• Regular mutual assistance
• Quick response to urgent requests

Active Community (~15-50 people)

• Regular interactions (monthly)
• Medium trust relationships
• Engaged in community governance
• Consistent participation in exchanges
• Shared norms and culture

Extended Network (~50-150 people)

• Occasional interactions (quarterly or less)
• Baseline trust relationships
• Broader community participation
• Resource sharing network
• Discovery layer for new connections

Layer Assignment

Self-Organizing: Layers determined by interaction patterns, not manual assignment Algorithm:

// Interaction frequency score
const innerCircleThreshold = 4; // interactions per month
const activeCommunityThreshold = 1; // interactions per month

function calculateLayer(userId, communityId) {
  const last6MonthsInteractions = getInteractions(userId, communityId, 6);
  const interactionsPerMonth = last6MonthsInteractions.length / 6;

  if (interactionsPerMonth >= innerCircleThreshold) {
    return 'inner_circle';
  } else if (interactionsPerMonth >= activeCommunityThreshold) {
    return 'active_community';
  } else {
    return 'extended_network';
  }
}

Layer Benefits

Inner Circle Members:

• Priority notification for urgent requests
• Access to "inner circle only" exchanges
• Stronger voice in governance decisions
• Cross-community introduction privileges

Active Community Members:

• Standard notification priority
• Access to most exchanges
• Full governance participation
• Community event organizing rights

Extended Network Members:

• Lower notification priority (opt-in for urgent)
• Access to public exchanges
• Observation of governance (limited voting)
• Path to active participation

Dynamic Movement

Members naturally move between layers based on participation:

• Increased interaction → move to more engaged layer
• Decreased interaction → move to less engaged layer
• Transparent calculation visible to member
• No stigma attached to layer (it's descriptive, not prescriptive)

Consequences

Positive

• Models Reality: Reflects how humans naturally organize
• Reduces Noise: Notifications targeted by interaction frequency
• Encourages Depth: Inner circles build deeper trust
• Scales Gracefully: Extended network can grow without overwhelming actives
• Measurable Engagement: Clear metrics for community health

Negative

• Exclusion Feelings: Extended network members may feel "second-class"
• Gaming Potential: People may artificially boost interaction frequency
• Calculation Complexity: Algorithm must balance multiple factors
• Cultural Sensitivity: Some communities may reject hierarchy

Alternatives Considered

Alternative 1: Flat Community Structure

• Why rejected: Doesn't model human social reality; overwhelms engaged members with noise

Alternative 2: Manual Role Assignment

• Why rejected: Creates explicit hierarchy; subject to politics and favoritism

Alternative 3: Fixed Member Tiers

• Why rejected: Rigid; doesn't adapt to changing participation patterns

Implementation Notes

Phase 1: Read-Only Layer Display (v9.0)

• Calculate layers based on interaction history
• Display in member profiles
• No functional differences (just informational)

Phase 2: Layer Calculation Deployed (Sprint 16 — soft version)

• calculate_community_layer() PostgreSQL function added (migration 026)
• Layer field returned on community members API response
• Layer chip visible to admins in the community members view (informational only)
• No notification priority or governance gating — layers are descriptive this sprint
• Future governance: inner circle as natural moderators, active community as norm proposers

Phase 3: Cross-Layer Dynamics (v11.0+)

• Mentorship programs (inner → extended)
• Layer transition rituals
• Inter-layer trust building activities

Database Schema

-- No new tables needed! Just computed values
CREATE OR REPLACE FUNCTION calculate_community_layer(
    p_user_id UUID,
    p_community_id UUID
) RETURNS VARCHAR(50) AS $$
DECLARE
    v_interactions_per_month DECIMAL;
BEGIN
    SELECT COUNT(*) / 6.0 INTO v_interactions_per_month
    FROM (
        -- Requests created
        SELECT created_at FROM requests.help_requests
        WHERE requester_id = p_user_id
          AND created_at > NOW() - INTERVAL '6 months'

        UNION ALL

        -- Offers made
        SELECT created_at FROM requests.help_offers
        WHERE offerer_id = p_user_id
          AND community_id = p_community_id
          AND created_at > NOW() - INTERVAL '6 months'

        UNION ALL

        -- Messages sent
        SELECT created_at FROM messaging.messages
        WHERE sender_id = p_user_id
          AND created_at > NOW() - INTERVAL '6 months'
    ) interactions;

    IF v_interactions_per_month >= 4 THEN
        RETURN 'inner_circle';
    ELSIF v_interactions_per_month >= 1 THEN
        RETURN 'active_community';
    ELSE
        RETURN 'extended_network';
    END IF;
END;
$$ LANGUAGE plpgsql;

API Changes

// GET /communities/:id/members response
{
  "success": true,
  "data": {
    "members": [
      {
        "id": "user-uuid",
        "name": "Alice",
        "layer": "inner_circle",
        "interactionsPerMonth": 5.2,
        "layerSince": "2025-01-15T00:00:00Z"
      }
    ],
    "layerCounts": {
      "inner_circle": 7,
      "active_community": 42,
      "extended_network": 98
    }
  }
}

References

• Dunbar, R. "Grooming, Gossip, and the Evolution of Language" (1996)
• Dunbar's Number research (cognitive limit of ~150 stable relationships)
• Social Brain Hypothesis
• Anthropological studies of hunter-gatherer band sizes