Characterizing Gender Differences in Misogynistic and Antisocial Microblog Posts

Abstract

This chapter presents an observational study into the genders of authors posting abusive misogynistic insults and hate speech on Twitter. We first characterize the different uses of potentially abusive and misogynistic expletives in Twitter using a novel diversity-based sampling strategy and use Amazon’s Mechanical Turk (MTurk) crowdsourcing platform to construct a labeled dataset of abusive, misogynistic insults. This misogyny dataset and datasets from prior work on harassing and hate speech in Twitter provide datasets for evaluating classification algorithms that can automatically identify this antisocial content. Leveraging 1.8 billion tweets from Twitter’s 1% public sample stream between 1 January 2015 to 31 December 2015, results show this antisocial content is relatively rare on Twitter, accounting for less than 1% of English tweets. After applying a gender classifier to this data, results demonstrate the population of English-tweeting Twitter users in 2015 who posted abusive misogynistic content were more likely to be classified as female when compared against a random Twitter sample from the same timeframe. For harassing and hateful speech, however, we find authors are more likely to be classified as male when compared against a random Twitter sample. Though this work does not cover threats of sexual violence, these results are consistent with prior research into misogynistic and hateful insults both on- and offline and have consequences for interventions designed to improve the quality of content in online spaces.