[Paper Review] Social Connection Induces Cultural Contraction: Evidence from Hyperbolic Embeddings of Social and Semantic Networks

Lingfei Wu, Linzhuo Li+1 more|arXiv (Cornell University)|Jul 26, 2018

Complex Network Analysis Techniques74 references3 citations

TL;DR

This paper proposes that dense social connections in academic physics networks induce cultural contraction—reducing topic diversity—by mapping social and semantic networks into hyperbolic (Poincaré disk) spaces to visualize hierarchical and diverse structures. Using hyperbolic embeddings, it shows that collaboration density contracts semantic space more than shared topics alone, creating convergent hubs or polarized clusters despite long-term research commitments.

ABSTRACT

Research has repeatedly demonstrated the influence of social connection and communication on convergence in cultural tastes, opinions and ideas. Here we review recent studies and consider the implications of social connection on cultural, epistemological and ideological contraction, then formalize these intuitions within the language of information theory. To systematically examine connectivity and cultural diversity, we introduce new methods of manifold learning to map both social networks and topic combinations into comparable, two-dimensional hyperbolic spaces or Poincaré disks, which represent both hierarchy and diversity within a system. On a Poincaré disk, radius from center traces the position of an actor in a social hierarchy or an idea in a cultural hierarchy. The angle of the disk required to inscribe connected actors or ideas captures their diversity. Using this method in the epistemic culture of 21st Century physics, we empirically demonstrate that denser university collaborations systematically contract the space of topics discussed and ideas investigated more than shared topics drive collaboration, despite the extreme commitments academic physicists make to research programs over the course of their careers. Dense connections unleash flows of communication that contract otherwise fragmented semantic spaces into convergent hubs or polarized clusters. We theorize the dynamic interplay between structural expansion and cultural contraction and explore how this introduces an essential tension between the enjoyment and protection of difference.

Motivation & Objective

• To investigate how social connectivity influences cultural and epistemic diversity in academic research.
• To formalize the tension between structural expansion (network growth) and cultural contraction (diversity loss) using information theory.
• To develop a method for visualizing and quantifying cultural and social hierarchies in semantic and network spaces.
• To test whether dense collaborations contract the space of discussed topics more than topic similarity alone.
• To explore the dynamic interplay between communication flows and the preservation or erosion of cultural diversity.

Proposed method

• Employ manifold learning to embed social networks and topic co-occurrence patterns into two-dimensional hyperbolic spaces (Poincaré disks).
• Use radial distance from the center of the Poincaré disk to represent hierarchical position—closer to center indicates higher centrality or prominence.
• Use angular spread around the disk to represent diversity—wider angles indicate greater diversity of connected actors or ideas.
• Map academic collaborations and topic combinations from 21st-century physics literature into the same hyperbolic space for direct comparison.
• Apply information-theoretic principles to formalize cultural contraction as a function of connectivity density.
• Use empirical data from physics research to validate the model and quantify the impact of collaboration density on semantic space contraction.

Experimental results

Research questions

• RQ1 To what extent do dense social collaborations contract the semantic space of research topics in physics?
• RQ2 How does the structural density of collaboration networks compare to topic similarity in driving cultural convergence?
• RQ3 What role does hyperbolic geometry play in capturing hierarchical and diverse structures in social and semantic networks?
• RQ4 How does communication flow through dense networks affect the diversity of ideas and research programs?
• RQ5 What is the dynamic tension between network expansion and cultural contraction in epistemic communities?

Key findings

• Dense university collaborations contract the semantic space of research topics more than shared topics alone, leading to convergence or polarization.
• Hyperbolic embeddings successfully capture both hierarchy and diversity, with radial position indicating centrality and angular spread indicating topic diversity.
• Communication flows in dense networks reduce the overall diversity of ideas, forming convergent hubs or polarized clusters.
• Despite long-term commitments to specific research programs, physicists in tightly connected networks discuss a narrower range of topics.
• The contraction of semantic space is driven primarily by connectivity density rather than topic overlap, indicating a structural effect of collaboration.
• The model reveals an essential tension between the enjoyment of cultural difference and its protection in epistemic communities.