Dig a little deeper and it gets weirder : social media and manipulation

The research presented in this thesis provides a detailed discussion and analysis of the phenomenon of content ranking manipulation on social media sites, in which the ranking of content submitted to social media sites is artificially manipulated in order to increase its prominence, at the expense of other content submitted to the site. The thesis documents the various types of manipulation that were identified on Reddit, a popular social media site; in addition, the impact of this manipulation is discussed in the context of different types of communities that exist inside social media sites. A framework for discussing manipulation and its impacts upon social media site users is proposed, and is discussed in the context of existing social media sites. Over the last decade, social media sites have rapidly risen to prominence as one of the most popular types of site on the World Wide Web. Sites such as Facebook, Twitter and Reddit act as a hub in which users may access and read content‚ÄövÑvp - for example, links to articles, photographs, videos and more - while at the same time submitting new content to the site. Social media sites are largely un-curated: users do not have to be approved by the administrators before they are able to add content to the site. In order to provide some means of ensuring that submitted content is high-quality, social media sites typically use some form of quantitative measurement to determine how prominent each piece of submitted content should be on the site. This is typically implemented using a voting mechanism, in which users are allowed to vote on individual pieces of content, and the votes are aggregated to determine the ranking. On social media sites, users who are submitting content to the social media site also have the ability to vote on content submitted by others. Due to the desire of submitters to have their content be prominently displayed, various techniques may be employed in order to ensure that their content receives more votes than other's content. These techniques constitute the types of manipulation identified in this research and presented in this thesis. The existing literature features several examples of studies into manipulation; however, these studies focus on in-depth analysis of specific manipulation techniques, such as Douceur's (2002) discussion of the Sybil attack. By contrast, this research takes a significantly broader approach to the discussion of manipulation by first identifying a definition of what manipulation is, based on interviews with community administrators and moderators, and then applying social research techniques to determine what kinds of manipulation exist. As a result, this thesis identifies new kinds of manipulation and enables further, more focused research on manipulation. The research takes the form of a three-phase study, in which administrators, moderators and users of a large social media site participated. Methods based on a grounded approach to qualitative data analysis were extensively employed in this research, and were used in the analysis of the data collected in all three phases. In the first phase of the research, administrators and moderators of Reddit were interviewed, and analysis revealed a number of high-level categories of manipulation seen by these users. The second phase of the research employed an innovative data collection tool that operated inside the web browser of all participants in the study, which gathered additional information on end-user perspectives of manipulation; this served to both confirms the completeness of the categories of manipulation identified in the first study, and to provide more detail for each category of manipulation using user perspectives of manipulation. The third phase of the research involved interviewing participants from the second phase, and gathered data regarding the impact that manipulation had on their experience of the social media site. Analysis of the collected data was performed using a social research approach, based on grounded-theory based methods to derive a systematic analysis of the data gathered in all three phases of the research. The resulting analysis is then discussed, and findings are derived. The research has identified a broad variety of different forms of social media site manipulations, each of which is broken down into multiple sub-categories and discussed. Each of these categories has a variety of different forms, which are discussed in this thesis, and examples presented. The research found that users generally share the same negative views of manipulation as administrators and moderators do, but note that some kinds of manipulation actually has a positive impact on their experience of the site. The research relates manipulation to earlier work done on the subject of relevance (Saracevic 1975, 2007), and establishes that attempts to manipulate the ranking of a social media site are attempts to modify the systemic relevance of content for all users by modifying other forms of relevance (such as cognitive, affective, and topical) for smaller groups of users. The contributions of this research are an identification of categories of manipulation on social media sites, a discussion of the design, implementation and results of a novel approach to data collection in an online study, and a model of manipulation and its impact upon users of different kinds of online communities. These contributions extend and broaden the scope for future research into social media site manipulation as a wide topic, which has previously been limited to individual forms of manipulation.