AgentFarm

Social Behavior Analysis

Core Analysis Categories

1. Social Network Analysis

  • Network Structure: Analyzes the connections between agents based on their interactions
  • Connection Metrics: Calculates metrics like:
    • Network density (how interconnected agents are)
    • Average outgoing/incoming connections per agent type
    • Unique interaction pairs
  • Agent Type Relationships: Examines how different agent types interact with each other

2. Resource Sharing Patterns

  • Sharing Metrics: Tracks:
    • Total resources shared between agents
    • Number of sharing actions
    • Average resources per sharing action
  • Altruistic Sharing: Identifies instances where agents share resources without immediate benefit
  • Sharing Matrix: Analyzes which agent types share with which other types
  • Temporal Distribution: Examines how resource sharing evolves over time

3. Cooperation vs. Competition

  • Action Classification: Categorizes agent actions as cooperative or competitive
  • Cooperation-Competition Ratio: Calculates the balance between cooperative and competitive behaviors
  • Agent Type Analysis: Compares how different agent types balance cooperation and competition
  • Temporal Trends: Tracks how cooperation/competition patterns change throughout the simulation

4. Spatial Clustering

  • Cluster Detection: Identifies groups of agents that cluster together spatially
  • Cluster Composition: Analyzes the agent type makeup of each cluster
  • Diversity Index: Calculates how diverse each cluster is (using Shannon entropy)
  • Isolation Metrics: Tracks which agents tend to remain isolated vs. clustered
  • Agent Type Clustering: Examines which agent types tend to form or join clusters

5. Reproduction Social Patterns

  • Social Context: Analyzes whether reproduction occurs in:
    • Isolation
    • Homogeneous groups (same agent type)
    • Heterogeneous groups (mixed agent types)
  • Agent Type Reproduction: Compares reproduction rates across different agent types
  • Social Influence: Examines how social context affects reproduction success

Analysis Capabilities

Cross-Simulation Analysis

  • Aggregates social behavior metrics across multiple simulations
  • Identifies consistent patterns that emerge regardless of initial conditions
  • Calculates average metrics and their variance across simulations

Pattern Extraction

  • Identifies emergent social patterns like:
    • Cooperation networks
    • Resource sharing communities
    • Competitive hierarchies
    • Spatial segregation or integration

Insight Generation

  • Automatically extracts key insights about social dynamics
  • Identifies which agent types are most social, cooperative, or competitive
  • Recognizes unusual or unexpected social behaviors

Visualization

  1. Social Network Visualizations:
    • Agent connection graphs
    • Network density charts
    • Connection metrics by agent type
  2. Resource Sharing Visualizations:
    • Sharing patterns by agent type
    • Sharing matrix heatmaps
    • Temporal distribution of sharing
  3. Cooperation/Competition Visualizations:
    • Cooperation vs. competition pie charts
    • Ratio analysis by agent type
    • Temporal trends in cooperation/competition
  4. Spatial Clustering Visualizations:
    • Cluster composition pie charts
    • Diversity vs. cluster size scatter plots
    • Clustering ratios by agent type
  5. Reproduction Pattern Visualizations:
    • Social context of reproduction
    • Reproduction events by agent type
    • Social context breakdown by agent type

Reporting

The system generates comprehensive reports that include:

  • Key findings and insights
  • Detailed metrics across all social dimensions
  • Agent type-specific insights
  • Emergent patterns
  • Recommendations for further investigation

Applications

This social behavior analysis framework allows you to:

  1. Understand Emergent Social Structures: See how complex social networks and behaviors emerge from simple agent rules

  2. Compare Agent Types: Analyze how different agent types (system, independent, control) interact socially and which develop advantages

  3. Identify Successful Strategies: Determine which social behaviors correlate with survival and reproduction success

  4. Track Social Evolution: Observe how social structures and behaviors evolve over the course of simulations

  5. Detect Unexpected Patterns: Identify surprising or emergent social behaviors that weren’t explicitly programmed

The analysis script provides a command-line interface to run this analysis on your simulation data, generating visualizations and a comprehensive report that summarizes all social behavior patterns observed in your simulations.