Add logging and data collection for network rewiring and homophily actions

[lamurian]

User Story

As a developer, I want to log network rewiring and homophily actions at both the agent and model levels so that I can analyze how frequently agents adapt their social connections and how these dynamics affect collective stress and resilience patterns.

Description

The current model does not track how often agents perform network rewiring (_rewire_to_similar_agent()) or network homophily adaptation (_adapt_network()). These mechanisms play a crucial role in social structure evolution and should be logged systematically for both individual agents and the overall model. The logs should integrate seamlessly into the existing Mesa DataCollector in src/python/model.py, allowing access to both per-agent and aggregate metrics.

Proposed Flow

• Update src/python/agent.py:

  • In the Person class:
    • Add counters for rewiring and homophily actions:

      self.rewire_count = 0
      self.homophily_count = 0

    • Increment rewire_count inside _rewire_to_similar_agent()
    • Increment homophily_count inside _adapt_network()
  • Optionally reset or aggregate counts per simulation step depending on analysis needs

• Update src/python/model.py:

  • Extend Mesa DataCollector configuration to include:

    agent_reporters={
        "RewireCount": "rewire_count",
        "HomophilyCount": "homophily_count",
        ...
    },
    model_reporters={
        "TotalRewires": lambda m: sum(a.rewire_count for a in m.schedule.agents),
        "TotalHomophily": lambda m: sum(a.homophily_count for a in m.schedule.agents),
        "MeanRewires": lambda m: np.mean([a.rewire_count for a in m.schedule.agents]),
        "MeanHomophily": lambda m: np.mean([a.homophily_count for a in m.schedule.agents]),
        ...
    }

  • Ensure the data collector collects this information every simulation step

• Testing:

  • Verify that rewire_count and homophily_count increase when corresponding methods are called
  • Confirm data collector captures these values at both agent and model levels
  • Check consistency of aggregation in model reporters (sum/mean values)

Reference

• src/python/agent.py → _rewire_to_similar_agent() and _adapt_network() methods in Person class
• src/python/model.py → Mesa DataCollector configuration

Notes:

• Keep logging lightweight to avoid simulation slowdowns.
• Consider resetting per-day counters if daily aggregation is desired.
• Ensure compatibility with existing data collection pipeline and output formats.
• This enhancement enables quantitative analysis of network evolution in response to affect and stress dynamics.

[lamurian]

Current Network Rewiring Implementation

1. Homophily-Based Connection Preferences

The model implements homophily through similarity calculations when agents consider rewiring their connections:

def _adapt_network(self):
    """Adapt network connections based on stress patterns and social support effectiveness."""
    cfg = get_config()
    
    # Check if agent should consider network adaptation
    stress_breach_count = getattr(self, 'stress_breach_count', 0)
    adaptation_threshold = config.get('network', 'adaptation_threshold')  # Default: 3
    
    if stress_breach_count < adaptation_threshold:
        return
    
    # Calculate adaptation metrics
    rewire_prob = config.get('network', 'rewire_probability')  # Default: 0.01
    homophily_strength = config.get('network', 'homophily_strength')  # Default: 0.7
    
    # Check each neighbor for potential rewiring
    for neighbor in current_neighbors:
        if self._rng.random() < rewire_prob:
            # Calculate similarity with neighbor (for homophily)
            affect_similarity = 1.0 - abs(self.affect - neighbor.affect)
            resilience_similarity = 1.0 - abs(self.resilience - neighbor.resilience)
            overall_similarity = (affect_similarity + resilience_similarity) / 2.0
            
            # Calculate support effectiveness
            support_effectiveness = self._get_support_effectiveness(neighbor)
            
            # Decide whether to keep or rewire connection
            keep_connection_prob = (overall_similarity * homophily_strength +
                                 support_effectiveness * (1.0 - homophily_strength))
            
            if self._rng.random() > keep_connection_prob:
                # Rewire: find a new potential connection
                self._rewire_to_similar_agent(current_neighbors)

2. Similarity-Based Rewiring

When agents decide to rewire connections, they seek out similar agents:

def _rewire_to_similar_agent(self, exclude_agents):
    """Find and connect to a new agent with similar stress characteristics."""
    # Get all agents except current neighbors and self
    all_agents = [agent for agent in self.model.agents if agent != self]
    available_agents = [agent for agent in all_agents if agent not in exclude_agents]
    
    if not available_agents:
        return
    
    # Find agents with similar affect/resilience levels
    similarities = []
    for agent in available_agents:
        affect_similarity = 1.0 - abs(self.affect - agent.affect)
        resilience_similarity = 1.0 - abs(self.resilience - agent.resilience)
        overall_similarity = (affect_similarity + resilience_similarity) / 2.0
        similarities.append((agent, overall_similarity))
    
    # Sort by similarity (highest first)
    similarities.sort(key=lambda x: x[1], reverse=True)
    
    # Try to connect to most similar agent
    for similar_agent, _ in similarities[:3]:  # Try top 3 most similar
        # Connection logic would go here
        pass

Network Rewiring Implementation

1. Stress-Triggered Adaptation

Network adaptation is triggered when agents experience repeated stress:

• Trigger Condition: stress_breach_count >= adaptation_threshold (default: 3)
• Rewiring Probability: Each existing connection has a chance to be rewired (default: 0.01)
• Balance Factor: Agents balance between homophily and support effectiveness

2. Decision Framework

For each neighbor, agents make a rewiring decision based on:

# Weighted decision between similarity and support effectiveness
keep_connection_prob = (overall_similarity * homophily_strength + 
                       support_effectiveness * (1.0 - homophily_strength))

Where:

• homophily_strength = 0.7 (70% weight on similarity, 30% on support effectiveness)
• overall_similarity = average of affect and resilience similarity
• support_effectiveness = neighbor's ability to provide support

3. Support Effectiveness Calculation

def _get_support_effectiveness(self, neighbor):
    """Get the effectiveness of a neighbor as a support provider."""
    # Based on neighbor's current state
    neighbor_fitness = (neighbor.resilience + (1.0 + neighbor.affect) / 2.0) / 2.0
    return min(1.0, neighbor_fitness + 0.2)  # Add some base effectiveness

Configuration Parameters

The network adaptation system is controlled by these key parameters:

# Network adaptation parameters
NETWORK_ADAPTATION_THRESHOLD=3      # Stress events needed to trigger adaptation
NETWORK_HOMOPHILY_STRENGTH=0.7      # Weight given to similarity vs support (0-1)
NETWORK_REWIRE_PROBABILITY=0.01     # Probability of considering rewiring each connection

# Network topology (Watts-Strogatz)
NETWORK_WATTS_K=4                   # Mean degree (neighbors per agent)
NETWORK_WATTS_P=0.1                 # Rewiring probability for small-world effect

Key Behavioral Mechanisms

1. Homophily-Driven Clustering

• Agents prefer connections with others who have similar:
  • Affect levels (emotional states)
  • Resilience levels (coping capacity)
  • Stress experiences (through similarity in stress responses)

2. Adaptive Network Evolution

• Stress Response: High stress triggers network adaptation
• Quality-Based: Ineffective connections are more likely to be rewired
• Similarity Seeking: Agents actively seek similar others when rewiring

3. Emergent Network Patterns

• Emotional Clusters: Similar emotional states form connected groups
• Resilience Communities: High-resilience agents tend to connect with each other
• Stress Propagation: Stress can spread through both similar and dissimilar connections

Integration with Other Systems

1. Stress System Integration

• Trigger: stress_breach_count increments with each stressful event
• Reset: Counter resets after network adaptation occurs
• Threshold: Adaptation only occurs after multiple stress events

2. Social Support Integration

• Effectiveness Tracking: Agents evaluate neighbor support quality
• Decision Weighting: Support effectiveness influences rewiring decisions
• Resource Costs: Network adaptation requires psychological resources

This implementation creates realistic social network dynamics where agents naturally form homophilous connections while occasionally adapting their networks based on stress experiences and support effectiveness.