The Impact of Social Attractiveness on Casual Group Formation: Power-Law Group Sizes and Suppressed Percolation

Basic Information

• Paper ID: 2507.12585
• Title: The Impact of Social Attractiveness on Casual Group Formation: Power-Law Group Sizes and Suppressed Percolation
• Authors: Matheus S. Mariano, José F. Fontanari (Instituto de Física de São Carlos, Universidade de São Paulo)
• Classification: physics.soc-ph, cond-mat.stat-mech, nlin.AO
• Publication Date: October 13, 2025 (arXiv v2)
• Paper Link: https://arxiv.org/abs/2507.12585v2

Abstract

This study revisits an interaction model driven by social attractiveness, an agent-based framework in which point-like agents randomly move within a two-dimensional arena and exhibit heterogeneous social attractiveness, causing them to remain near high-attractiveness "celebrity" companions. The research reveals that, compared to a null model, the attractiveness-driven model exhibits average degree growing linearly with system size at fixed density, leading to tighter group formation and suppressed percolation transitions. Crucially, the model exhibits a power-law distribution P(n) ∝ n^(-2.5), with an exponent independent of density, highlighting the critical role of individual attractiveness in shaping social aggregation in physical space.

Research Background and Motivation

1. Problem to Address: Understanding the dynamical mechanisms of casual group formation, particularly the impact of heterogeneous social attractiveness on group size distributions and network structure. Traditional random models oversimplify individual behavior and lack mechanisms for heterogeneous social attractiveness.
2. Problem Significance: Face-to-face interactions form the foundation of human social life, from cocktail parties to street pedestrian flows. The formation and dissolution of casual groups are ubiquitous. Understanding these dynamic processes is crucial for social science.
3. Limitations of Existing Approaches:
   • Pioneering random models from the 1960s provided foundational insights but oversimplified individual behavior
   • Lack of explicit mechanisms for heterogeneous social attractiveness
   • Early models typically described environments without high-impact individuals dominating
4. Research Motivation: Bridging classical sociological insights with modern computational methods by introducing a "celebrity effect" to more realistically model casual group formation processes.

Core Contributions

1. Recharacterized the attractiveness-driven interaction model: Provided comprehensive quantitative analysis of the model's equilibrium properties, with particular focus on structural characteristics of spatial contact networks.
2. Discovered key scaling behavior: Demonstrated that at fixed density, average degree grows linearly with system linear size L, contrasting sharply with the null model.
3. Identified suppression of percolation transitions: Tight clustering around celebrities prevents formation of a single large group, eliminating the characteristic percolation transition found in null models and random geometric graphs.
4. Established power-law group size distribution: Found robust power-law distribution P(n) ∝ n^(-2.5), with exponent β = 2.5 independent of density.
5. Provided in-depth analysis of computationally intensive simulations: Overcame challenges of long equilibration times, providing the most comprehensive quantitative characterization of this model to date.

Detailed Methodology

Task Definition

Study face-to-face interactions of N point-like agents within a two-dimensional arena, where agents possess heterogeneous social attractiveness. The goal is to analyze the size distribution and network structural properties of resulting casual groups.

Model Architecture

Basic Setup:

• N agents moving within a square arena of side length L (periodic boundary conditions)
• Density ρ = N/L²
• Interaction distance d = 1, movement step size v = 1

Agent Attributes:

• Attractiveness aᵢ ∈ 0,1: Quantifies an agent's ability to attract others to remain stationary
• Activation probability rᵢ ∈ 0,1: Probability that an inactive agent becomes active
• State: Active (can interact and move) or inactive

Dynamics Rules (asynchronous updates, δt = 1/N):

1. Inactive agents: Attempt activation with probability rᵢ
2. Active agents:
   • Movement decision probability: pᵢ(t) = 1 - max{aⱼ | j ∈ Nᵢ(t)}
   • If deciding to move: randomly select direction, move distance v
   • After moving, become inactive with probability 1-rᵢ if isolated

Group Definition: Groups are defined as connected components in the contact graph, where edges exist between active agents at distance ≤ d.

Technical Innovations

1. Heterogeneous social attractiveness mechanism: Introduces parameter aᵢ to simulate the "celebrity effect" in reality, where high-attractiveness agents (aᵢ ≈ 1) cause others to remain longer.
2. Comparative analysis framework: Contrasts with null model (aᵢ = 0, agents continuously moving) to clearly demonstrate the impact of attractiveness mechanism.
3. Multi-scale analysis: Systematically investigates effects of system size L and density ρ on network structure and group distribution.

Experimental Setup

Simulation Parameters

• System sizes: L = 25, 50, 100; Densities ρ = 0.5, 1, 2
• Number of agents: N = ρL², up to 5000 agents
• Time range: Simulations run to t = 10⁵ to ensure equilibrium

Evaluation Metrics

1. Average degree k: Average number of connections in contact graph
2. Maximum group fraction s: Proportion of active agents in largest group
3. Group size distribution P(n): Probability distribution of groups of size n
4. Pair correlation function g(r): Distribution of agent pairs at distance r

Comparison Methods

• Null model: All agents with aᵢ = 0, similar to random geometric graphs (RGG)
• Random geometric graph theory predictions: As theoretical baseline

Implementation Details

• 10³ to 10⁵ independent simulations per parameter combination
• Equilibrium detection: Exponential function fitting to determine equilibrium values
• Characteristic relaxation time τₖ ∝ L²

Experimental Results

Main Findings

1. Scaling Behavior of Average Degree:

• Null model: k depends only on density ρ, independent of L
• Attractiveness-driven model: k∞ ∝ L (at fixed density), indicating groups become tighter as system grows

2. Suppression of Percolation Transition:

• Null model: Percolation transition occurs at critical density ρc ≈ 1.44
• Attractiveness-driven model: s∞ grows smoothly with density, no percolation transition

3. Group Size Distribution:

• Null model: Exponential decay at low density, bimodal distribution at high density
• Attractiveness-driven model: Robust power-law distribution P(n) ∝ n^(-2.5)

Ablation Studies

Pair correlation function g(r) analysis verified spatial clustering effects:

• Null model: g(r) = 2πr/L² (uniform distribution)
• Attractiveness-driven model: Significant peak at r = d = 1

Key Experimental Findings

1. Long Relaxation Times: Attractiveness-driven model requires τₖ ∝ L² to reach equilibrium, far exceeding null model.

2. Tight Clustering Effect: Dense groups form around celebrities, with average degree k∞ growing linearly with L.

3. Density-Independent Power-Law Exponent: β = 2.5 remains constant across all tested densities.

Related Work

Classical Foundations:

• Coleman & James (1961): Equilibrium size distribution of freely-forming groups
• White (1962): Opportunity model of casual group systems

Modern Developments:

• Starnini et al. (2013, 2016): Original attractiveness-driven interaction model
• Random geometric graph theory: Gilbert (1961), Penrose (2003)

Advantages of This Work:

• First comprehensive analysis of spatial network properties of this model
• Discovery of percolation transition suppression as new phenomenon
• Establishment of robustness of power-law group size distribution

Conclusions and Discussion

Main Conclusions

1. Heterogeneous social attractiveness fundamentally alters statistical properties of casual groups
2. Celebrity effect leads to suppression of percolation transitions, a new phenomenon in spatial interaction models
3. Power-law group size distribution P(n) ∝ n^(-2.5) is a robust characteristic of this model

Limitations

1. High computational complexity: Long relaxation times limit study of large systems
2. Simplified attractiveness distribution: Uses uniform distribution; reality may be more complex
3. Two-dimensional constraint: Real social interactions may involve higher dimensions

Future Directions

1. More realistic attractiveness distributions: Such as wealth distributions
2. Introduction of repulsive hard cores: Model human comfort distance
3. Higher-order interactions: Consider three-body and higher interactions

In-Depth Evaluation

Strengths

1. Strong methodological innovation: First systematic analysis of spatial network properties of attractiveness-driven model
2. Rigorous experimental design: Large-scale simulations, systematic multi-parameter study
3. Significant theoretical implications: Discovery of new physical phenomena like percolation transition suppression
4. Clear writing: Well-structured, in-depth analysis

Weaknesses

1. Computational constraints: System size limited by L² scaling of relaxation time
2. Power-law fitting challenges: Statistical difficulties in asymptotic regime, requires large samples
3. Model simplification: Real social interactions more complex than model

Impact

1. Academic contribution: Provides new theoretical framework for sociophysics
2. Practical value: Applicable to crowd dynamics, urban planning, and related fields
3. Reproducibility: Detailed parameter settings and methodology descriptions

Applicable Scenarios

1. Social science research: Understanding mechanisms of casual group formation
2. Complex network analysis: Percolation phenomena in spatial networks
3. Crowd dynamics modeling: Predicting crowd aggregation in public spaces

References

1. Coleman J S and James J (1961) The Equilibrium Size Distribution of Freely-forming Groups
2. Starnini M et al. (2013) Modeling Human Dynamics of Face-to-Face Interaction Networks
3. Dall J and Christensen M (2002) Random geometric graphs
4. Newman M E J (2018) Networks

This paper, through introducing a heterogeneous social attractiveness mechanism, deeply reveals the impact of the "celebrity effect" on casual group formation, discovering important phenomena such as percolation transition suppression and power-law group distribution. It provides a new theoretical framework for understanding group dynamics in real social contexts.