From a Hashtag to a Movement:Modeling Conversation Around #MeToo onTwitter

Abstract

In this paper I evaluate the effectiveness of three different modelframeworks from epidemiology, SIR, SISR, and SIRS, at capturing thedynamics and the spread of social movements online. For this project I lookspecifically at data from #MeToo and the corresponding movement. Thedata suggests that the hashtag movement can be divided into groups ofindividuals in the U.S. and abroad. I take into account these between groupdynamics. I estimate model transition probability parameters from theTwitter data on #MeToo and compare them with optimal parametersfound by minimizing mean squared error of the model output with respectto key metrics from the data. I run simulations of the models over anunderlying social network that is representative of Twitter and compareresults quantitatively and qualitatively to the data. SISR appears to be themost effective single model at capturing the dynamics of both the hashtagand the movement. Extensions are made to consider the effect of exogenousreinjection of infection into the network and network community structureon the model output.