Echoing within and between : quantifying echo chamber behaviours on Reddit

This thesis examines whether echo chambers exist on Reddit, the self-titled 'front page of the internet'. As a social media platform Reddit is widely popular in Western countries but both relatively unknown and understudied by academics. This research will show that the structural features of the platform; in particular its organisation into distinct topic-based communities, called subreddits; makes it an ideal space to empirically test the existence of echo chambers on social media. In this research we consider how the network of subreddits naturally self-organises into topical 'public sphericules' of like-minded communities. Whereas most previous research has treated echo chambers as binary, here we treat echo chamberness as a spectrum from greatest chamberness to least. To achieve this, a key methodological advantage of the research is the use of multiple simple measures of chamberness applied to the 1,000 most active subreddits during our period of study, January 2019. By creating a background distribution of 'expected' levels of chamberness, we can understand the relative chamberness of any single subreddit. This enables us to consider at what point the natural process of interest homophily crosses from useful public sphericules into problematic 'meta echo chambers'. We focus our research on the set of political subreddits, and in particular two key case study subreddits: The_Donald, a group of supporters of United States President Donald Trump, which self-characterises as an echo chamber by explicitly banning dissenting views; and changemyview a place to go to have your questionable opinions changed, which seeks to be an 'anti-echo chamber'. Through this research we find that, contrary to our expectations, The_Donald is not actually the echo chamber it set itself up to be, and that the most active participants of changemyview show a notable bias towards alt right contrarian discourse, which we would not expect of a broad-minded anti-echo chamber. The research also more broadly considers whether social media echo chambers are exclusively the concern of political content, as the majority of existing literature suggests. Through the examples of the sports and porn communities, which are both internally densely connected and largely isolated from the wider subreddit network, we illustrate how content which is not socio-political can still fall victim to the concerning behaviours which motivate our fears of echo chambers. This thesis will show that the concerns which motivate the study of social media echo chambers -- increasing polarisation, fragmentation, and loss of common ground between opposing groups -- affect many of our views, not only the political.