On the statistical properties of viral misinformation in online social media

TL;DR

This paper analyzes the statistical behavior of highly viral misinformation in social media, revealing it follows a Poisson process with exponential interarrival times, and uses Bayesian methods to quantify uncertainty and predict future viral posts.

Contribution

It applies Extreme Value Theory and Bayesian analysis to characterize and predict the occurrence of viral misinformation in social media.

Findings

Viral misinformation events follow a homogeneous Poisson process.

Interarrival times between viral posts are exponentially distributed.

Bayesian methods quantify uncertainty in the rate of viral posts.

Abstract

The massive diffusion of online social media allows for the rapid and uncontrolled spreading of conspiracy theories, hoaxes, unsubstantiated claims, and false news. Such an impressive amount of misinformation can influence policy preferences and encourage behaviors strongly divergent from recommended practices. In this paper, we study the statistical properties of viral misinformation in online social media. By means of methods belonging to Extreme Value Theory, we show that the number of extremely viral posts over time follows a homogeneous Poisson process, and that the interarrival times between such posts are independent and identically distributed, following an exponential distribution. Moreover, we characterize the uncertainty around the rate parameter of the Poisson process through Bayesian methods. Finally, we are able to derive the predictive posterior probability distribution of the number of posts exceeding a certain threshold of shares over a finite interval of time.