Friday, October 19, 2018 — 1:30 p.m. ~ 128 Sirrine
Trumping Hate on Twitter: Measuring the prevalence of online hate speech, with an application to the 2016 U.S. election”
Abstract: Despite a growing body of research devoted to defining and detecting online hate speech and extremist rhetoric, the existing scientific literature lacks a systematic framework for assessing how the content and popularity of these harmful messages change over time. We offer a new approach to measuring the real-time prevalence of online hate, using both context-specific data and data produced by a large random sample of users; employing multiple methods of text classification; and measuring not only the volume, but also the proportion, and number of unique users producing it. Here we apply our framework to test the widely-held proposition that Donald Trump’s divisive 2016 campaign and election has popularized online hate speech and white nationalist rhetoric in the American Twittersphere. Highlighting the need for such a systematic approach—contrary to the conventional wisdom—our analysis of over one billion tweets demonstrates that online hate did not become more popular on Twitter either over the course of the campaign or in the aftermath of Trump’s election.