[Paper Review] Structural Diversity and Homophily: A Study Across More than One Hundred Big Networks

Yuxiao Dong, R. A. Johnson+2 more|arXiv (Cornell University)|Feb 23, 2016

Complex Network Analysis Techniques40 references3 citations

TL;DR

This paper introduces structural diversity of common neighborhoods as a key factor influencing link formation in networks, proposing a Common Neighborhood Signature (CNS) to classify 120 large-scale networks into three distinct superfamilies. It reveals that higher structural diversity can either promote or inhibit link formation depending on the network context—positive in Facebook, negative in LinkedIn—challenging the universality of structural homophily and enabling improved link prediction and network modeling.

ABSTRACT

A widely recognized organizing principle of networks is structural homophily, which suggests that people with more common neighbors are more likely to connect with each other. However, what influence the diverse structures embedded in common neighbors have on link formation is much less well-understood. To explore this problem, we begin by characterizing the structural diversity of common neighborhoods. Using a collection of 120 large-scale networks, we demonstrate that the impact of the common neighborhood diversity on link existence can vary substantially across networks. We find that its positive effect on Facebook and negative effect on LinkedIn suggest different underlying networking needs in these networks. We also discover striking cases where diversity violates the principle of homophily---that is, where fewer mutual connections may lead to a higher tendency to link with each other. We then leverage structural diversity to develop a common neighborhood signature (CNS), which we apply to a large set of networks to uncover unique network superfamilies not discoverable by conventional methods. Our findings shed light on the pursuit to understand the ways in which network structures are organized and formed, pointing to potential advancement in designing graph generation models and recommender systems.

Motivation & Objective

• To investigate how the structural diversity of common neighborhoods—measured by the number of connected components—impacts link formation beyond traditional structural homophily.
• To identify whether and how varying configurations of common neighbors influence the likelihood of dyadic connections in real-world networks.
• To develop a novel network characterization method, the Common Neighborhood Signature (CNS), to uncover hidden network superfamilies not detectable via conventional metrics.
• To understand the divergent networking behaviors in platforms like Facebook and LinkedIn, where structural diversity has opposing effects on link formation.
• To inform the design of better graph generation models and recommender systems by revealing underlying organizational principles of network structure.

Proposed method

• Define structural diversity as the number of connected components in the subgraph formed by the common neighbors of a node pair.
• Measure link existence rates across 120 large-scale networks (real and random) while controlling for the number of common neighbors.
• Propose the Common Neighborhood Signature (CNS) as a network-level fingerprint based on the distribution of structural diversity across all dyads.
• Apply clustering techniques to CNS vectors to identify network superfamilies that group networks with similar structural diversity patterns.
• Compare the CNS-based classification with traditional network properties (e.g., degree distribution, clustering coefficient) to validate its distinctiveness.
• Use statistical modeling to analyze the conditional probability of link existence given structural diversity, revealing platform-specific trends.

Experimental results

Research questions

• RQ1 How does the structural diversity of common neighborhoods affect the probability of link formation, independent of the number of common neighbors?
• RQ2 Do networks like Facebook and LinkedIn exhibit opposing patterns in how structural diversity influences link formation, and if so, why?
• RQ3 Can structural diversity be used to define a new network signature (CNS) that reveals hidden network superfamilies not detectable by standard metrics?
• RQ4 To what extent do existing random graph models fail to reproduce the structural diversity patterns observed in real-world networks?
• RQ5 How does structural diversity challenge or extend the principle of structural homophily in network formation?

Key findings

• Structural diversity significantly influences link existence rates even when the number of common neighbors is held constant, demonstrating that not all common neighborhoods are equally conducive to link formation.
• In Facebook, higher structural diversity (more disconnected components) increases the likelihood of link formation, indicating that diverse, loosely connected common contexts facilitate online friendship formation.
• In LinkedIn, higher structural diversity negatively affects link formation, suggesting that tightly connected common neighborhoods (e.g., shared colleagues) are more conducive to professional connections.
• The CNS successfully classifies 120 networks into three distinct superfamilies, with Facebook and Friendster forming one superfamily and LinkedIn forming a separate one, highlighting fundamental differences in network organization.
• None of the standard random graph models studied (e.g., Erdős–Rényi, Watts-Strogatz, Barabási–Albert) can replicate the structural diversity patterns of the Facebook/Friendster superfamily, indicating a gap in current generative models.
• The study reveals that the principle of structural homophily—where more common neighbors increase link probability—can be violated when structural diversity is considered, as fewer mutual connections can sometimes lead to higher link formation rates.