Multidimensional Political Incivility on Twitter: Detection and Findings

Anonymous

Abstract: Hostile and uncivil political discourse online may negatively affect Democratic processes. In this work, we consider the task of {\it political incivility} detection. Unlike previous attempts, we utilize a multidimensional perspective of political incivility, differentiating between impoliteness--which is sometimes acceptable--and intolerance--which inherently violates the Democratic norms. We evaluate state-of-the-art classifiers on this task using MUPID, a large multidimensional political incivility dataset of 13K tweets, which we collected and annotated by means of crowd sourcing. Our results and analyses illustrate the challenges involves in this task. In particular, we observe that intolerance is often expressed using implicit language, that requires higher-level semantic understanding. In addition, we apply political incivility detection at large-scale, exploring the distribution of uncivil content across individual users and across U.S. states. Our findings align and extend existing theories of Political Science and Communication.

Paper Type: long

Research Area: Computational Social Science and Cultural Analytics

Contribution Types: NLP engineering experiment, Data resources, Data analysis

Languages Studied: English