A Quantative Study of Social Media Echo Chambers

Abstract: The changing online environment - where the breadth of the information we are exposed to is algorithmically narrowed - has raised concerns about the creation of "echo chambers"; in which individuals are exposed mainly to information already in alignment with their preconceived ideas and opinions. This thesis explores the role of Twitter as a social media and as an information network, and investigates if exposure to and participation in political discussions resembles echo chambers. The findings by analyzing the Twitter friendship network shows that users on Twitter tend to prefer to follow like-minded individuals to some extent, but not in the domain of what constitutes an echo chamber. However, analyzing all communication events associated to a set of influential politicians, newspapers, journalists and bloggers during a ten-day period in connection to the British general election in June 2017, reveals that users are more likely to engage with ideologically similar peers, than with users with different political beliefs from themselves. Data is collected using both the Twitter rest and streaming API, and is analyzed using methods from graph theory and social network analysis. One part of the project involves collection and analysis of several million tweets, in which case the cluster computing platform Apache Spark is used. The other part is concerned with finding the degrees of separation between accounts in the Twitter network, in which case the API is queried step-by-step.