Winner does not take all: contrasting centrality in adversarial networks

TL;DR

This paper introduces the concept of low-key leaders in adversarial networks, highlighting their prevalence and proposing a new model to generate such networks, with evidence from diverse real-world data.

Contribution

It defines low-key leaders based on contrasting centrality measures and presents a novel random graph model to simulate their emergence in adversarial networks.

Findings

Low-key leaders are common in real-world adversarial networks.

A new random graph model effectively generates networks with low-key leaders.

Empirical data from animal, political, and financial networks support the hypothesis.

Abstract

In adversarial networks, edges correspond to negative interactions such as competition or dominance. We introduce a new type of node called a low-key leader in adversarial networks, distinguished by contrasting the centrality measures of CON score and PageRank. We present a novel hypothesis that low-key leaders are ubiquitous in adversarial networks and provide evidence by considering data from real-world networks, including dominance networks in 172 animal populations, trading networks between G20 nations, and Bitcoin trust networks. We introduce a random graph model that generates directed graphs with low-key leaders.