Echo Chambers without Conversation? Enriching Research on Polarization and Fragmentation on Twitter with the Analysis of Reciprocal Communication

Abstract

Echo chambers on social media are described as homophilic clusters that are characterized by a repeated confirmation of users' political opinions and a lack of confrontation with other opinions, a process that leads to the solidification and radicalization of beliefs. The empirical research literature on the existence and effects of echo chambers yields quite mixed results, with some studies finding evidence for the existence of echo chambers and others not. In this article, we argue that network analytic research about echo chambers on Twitter would benefit from an investigation of reciprocal communication. Current research finds evidence for echo chambers for political topics in retweet networks. However, such approaches may not adequately reflect the degree of fragmentation on Twitter because a retweet is a form of information diffusion that does not support the reciprocity necessary for political discussions. To capture reciprocal communication, we instead suggest to focus on replies. We then show that typical approaches to data collection based on hashtags or keywords capture only a small fraction of replies about any given topic. With the introduction of the conversation_ID by Twitter it is now possible to collect all replies to original tweets, resulting in much larger collections of replies. We illustrate an approach that focuses on reciprocal communication through replies with the construction of the #debate2020 dataset. Here, original tweets and replies are represented in a tree structure as threaded reciprocal communication. We argue that it is in threaded replies where we might find evidence for echo chambers in patterns of mutual affirmation or contestation and delegitimization of dissenting positions.