Modelling the transmission dynamics of online social contagion

TL;DR

This study models the spread of online social contagion through social media, estimating transmission parameters and demonstrating the model's predictive ability across different datasets.

Contribution

It introduces a mathematical infectious disease model to quantify online social contagion dynamics and estimates key parameters like R0 and generation time.

Findings

Median R0 ranged from 1.9 to 2.5

Generation times between 1.0 and 2.0 days

Model showed reasonable predictive power with R^2 0.52-0.70

Abstract

During 2014-15, there were several outbreaks of nominated-based online social contagion. These infections, which were transmitted from one individual to another via posts on social media, included games such as 'neknomination', 'ice bucket challenge', 'no make up selfies', and Facebook users re-posting their first profile pictures. Fitting a mathematical model of infectious disease transmission to outbreaks of these four games in the United Kingdom, I estimated the basic reproduction number, $R_0$, and generation time of each infection. Median estimates for $R_0$ ranged from 1.9-2.5 across the four outbreaks, and the estimated generation times were between 1.0 and 2.0 days. Tests using out-of-sample data from Australia suggested that the model had reasonable predictive power, with $R^2$ values between 0.52-0.70 across the four Australian datasets. Further, the relatively low basic reproduction numbers for the infections suggests that only 48-60% of index cases in nomination-based games may subsequently generate major outbreaks.