Trend-driven information cascades on random networks

TL;DR

This paper introduces a generalized threshold model incorporating global trend-influenced nodes, revealing how such nodes can both accelerate and inhibit cascades, with implications for understanding collective behavior.

Contribution

It proposes a new model with global trend followers, analyzing their dual role in facilitating or hindering information cascades in networks.

Findings

Global nodes accelerate cascades after a trend emerges.

Global nodes can reduce the likelihood of trend emergence.

A moderate proportion of trend followers maximizes cascade size.

Abstract

Threshold models of global cascades have been extensively used to model real-world collective behavior, such as the contagious spread of fads and the adoption of new technologies. A common property of those cascade models is that a vanishingly small seed fraction can spread to a finite fraction of an infinitely large network through local infections. In social and economic networks, however, individuals' behavior is often influenced not only by what their direct neighbors are doing, but also by what the majority of people are doing as a trend. A trend affects individuals' behavior while individuals' behavior creates a trend. To analyze such a complex interplay between local- and global-scale phenomena, I generalize the standard threshold model by introducing a new type of node, called \textit{global nodes} (or \textit{trend followers}), whose activation probability depends on a global-scale trend; specifically the percentage of activated nodes in the population. The model shows that global nodes play a role as accelerating cascades once a trend emerges while reducing the probability of a trend emerging. Global nodes thus either facilitate or inhibit cascades, suggesting that a moderate share of trend followers may maximize the average size of cascades.