Bias, inequality, and polarization in modern digital information systems
MetadataShow full item record
Digital technology has the potential to "democratize information" – making ideas, opinions, and knowledge accessible anywhere, anytime, and to everyone. But is this potential truly realized or will it ever be realized? Do systems enabled by digital technology exhibit or even enhance information bias, skewness, and polarization? How can we overcome them? In this dissertation, I investigate these questions in two major but distinct digital information systems: open collaboration systems (i.e., Wikipedia) and mass media broadcast networks (i.e., broadcast television in the United States). Open collaboration platforms have fundamentally changed the way knowledge is produced, disseminated, and consumed. Wikipedia is arguably one of the most successful examples of such platforms, serving millions of information seekers daily. Despite many benefits provided by the decentralization of knowledge production on Wikipedia, does the open nature and lack of broad oversight and coordination leave the question of information poverty and skewness to the mercy of the system’s natural dynamics? And if so, what can be done to address this? In Chapter 1, I examined this question using both causal inference from a natural experiment and empirically informed diffusion simulations. Another important and pervasive information system is that of televised mass media. Whereas Wikipedia is relatively open and does not have strong information gatekeeping, televised mass media has various forms of information gatekeeping, particularly through media ownership, government regulation and journalistic practice. But how does this gatekeeping affect skewness and polarization in the real-world information that is conveyed to the public? To investigate these questions, I study televised news information systems in the United States with a massive scale unstructured text data and various state-of-the-art text mining techniques in Chapter 2 and Chapter 3 of this dissertation. The text transcripts include the complete televised content from more than 800 television channels across all 210 designated media markets in the United States over a 5-year period between 2013 and 2018. Chapter 2 of this dissertation examines how media ownership impact political slant and information diversity in the news using massive-scale text transcripts. I found that when large owners act coherently, they can skew information to emphasize views, perspectives and framing that they advocate. This is important because previous studies have shown that broadcast media can have a dramatic impact on political and social outcomes and undeniably shapes the national dialogue surrounding important issues. In Chapter 3 of this dissertation, I study the skewed coverage of gun violence incidents in local televised news. I found that some types of gun violence, such as suicide, accidents, domestic violence and sex crimes are systematically covered less relative to other types such as assault weapon incidents, are systematically covered more. Importantly, areas of high vs. low gun ownership received different exposure to different incident types through local news coverage, further dividing an already divided population. I conducted ting a nationally representative survey found that the general public’s view on different type of gun violence is skewed in a manner that is consistent with "the warped mirror" that our media conveys.