Social media represents one of the newest forums for aggression, one where hostility is experienced personally, and often contains sexist and racist content. To stem some of the pervasive negativity, Twitter enacted a widely publicized policy to reduce abusive messaging. Here we examine the effect of this change on sexist and racist messages through sentiment analysis of tweets. We analyze (over 3.8 million) messages containing the slur “bi!ch;” and additional messages with racist slurs or geo-located tags. Adopting a quasi-experimental approach comparing tweets sent before and after this policy shift, we find evidence that the policy had a significant impact on user behavior. Tweets after the policy were more positive than before the change. In particular, the policy decreased the negative sentiment of messages containing sexist or racist content. The findings highlight the viciousness of negative messaging on Twitter as well as the positive influence of institutional policy.
Presented in Session 9. Marriage, Family, Households, & Unions; Gender, Race, & Ethnicity