Experimentally Reducing Partisan Incivility on Twitter

Опубликовано 26 января 2018, 22:47

Cross-partisan incivility is a pressing concern in American politics, especially online. When partisans communicate incivilly, they are less likely to learn from each other and more likely to distrust each other. This paper analyses the way that people learn about norms of political behavior online. I conduct an experiment that tests how different forms of moral suasion---appeals to morality intended to influence behavior---affect how Democrats and Republicans learn about norms of partisan speech. Using bots that shared the political identity of the subjects, I sent messages that appealed to the moral principle theorized to be most convincing to either liberals ("care") or to conservatives ("authority"). Using a sample of subjects who had been frequently incivil in political discussions on Twitter during the 2016 US presidential election, I found that both forms of moral suasion were equally effective at dissuading both Democrats and Republicans from online incivility. These effects were significantly moderated by the anonymity of the subjects, especially among Republicans: subjects who elected to have an anonymous profile were much less likely to change their behavior. On some subsamples, the reduction in incivility persisted for up to a month after treatment.

See more at microsoft.com/en-us/research/v...