Modeling Asymmetric Relationships from Symmetric Networks

TL;DR

This paper introduces a probabilistic method to infer unobserved asymmetric relationships from symmetric networks, enabling better understanding of underlying directed interactions.

Contribution

It presents a novel regression-based framework for reconstructing asymmetric networks from symmetric data, with improved inference and predictive capabilities.

Findings

Successfully applied to bilateral investment treaty network

Extracts politically relevant information inaccessible to other methods

Demonstrates superior predictive performance

Abstract

Many relationships requiring mutual agreement between pairs of actors produce observable networks that are symmetric and undirected. Nevertheless the unobserved, asymmetric network is often of primary scientific interest. We propose a method that probabilistically reconstructs the unobserved, asymmetric network from the observed, symmetric graph using a regression-based framework that allows for inference on predictors of actors' decisions. We apply this model to the bilateral investment treaty network. Our approach extracts politically relevant information about the network structure that is inaccessible to alternative approaches and has superior predictive performance.