Reducing the Use of Polarizing Language at Scale: A Twitter Experiment

• In Event: Putting Social Identities to Work for Democracy

Abstract

Citizens' perception of levels of polarization is in part shaped by the language with which political opponents refer to each other. In the context of Mexico, the government and opposition have developed a lengthy vocabulary to refer to each other in a derogatory manner, sometimes with words exceeding partisan tribalism and associated with class or race.
Social media like Twitter is a virtual space where content that uses this "polarizing language" is generated en masse. Even if it is only a subset of users that generates this content, if most of the political content includes the use of polarizing language this might increase the perception of polarization for the average citizen. Most importantly, it might set a new standard or social norm of what is considered acceptable political debate.
In the first part of the project, we design and implement an intervention to reduce the use of polarizing language by Twitter users. In partnership with a NGO in Mexico, we code a bot that reacts to Tweets using polarizing language. We probe and pit against each other various messages to persuade users to abandon the use of polarizing language, ranging from a reminder of the value of respect in a democracy to exposing the use of such politically loaded language to their close social network. After the delivery of the message we measure subsequent use of polarizing language by treated accounts as the main outcome.
This first part of the project will be followed up by further work exploring mechanisms of behavioral change as well as if the results have an impact on the perceptions of polarization from average citizens. We believe that the intervention has important consequences in the design of scalable policies to reduce polarization in developing and developed democracies.

Authors

• José Ramón Enríquez, Harvard University

• Julio S. Solis Arce, Harvard University