Reconstructing signed relations from interaction data

Abstract

Positive and negative relations play an essential role in human behaviour and shape our communities. Polarization in the internal structure of these communities strongly affects their functioning. Imagine a community where two groups of individuals have positive relations within but negative relations across the groups. While such a polarized setting is stable according to structural balance theory, it may hinder constructive dialogue, foster division, and impede collective problem-solving.

To understand if a community is polarized in its relational structure, it is imperative to possess data on the signs of the relations. Unfortunately, such data is rare and commonly gathered through surveys. Conversely, interaction data is more abundant, for instance, in the form of proximity or communication data. We show how the underlying signed relations in communities can be extracted from interaction data.

Employing a statistical network approach, we infer signed relations using interaction data for five social systems and reconstruct their signed networks. Subsequently, we leverage the reconstructed signed networks to study group cohesion and structural balance. Finally, we establish a connection between these concepts and the emergence of polarization in social systems.

Relevant publications

G. Andres, G. Casiraghi, G. Vaccario and F. Schweitzer. Reconstructing signed relations from interaction data