Investigating the Impact of Media Bias on News Readers

Samar Haider

Abstract: How news media present facts surrounding an event can have a significant impact on readers' perception of them. News sources on opposing sides of the political spectrum may spin their coverage in very different ways, leading to a divergence in the worldviews of their readership and increasing polarization in society. While the phenomenon of media bias has been extensively covered in various social science disciplines, including political science, communications, economics, and sociology, the precise measurement of its impact on readers' opinions, and which factors influence them the most, has received less attention. In this project, we investigate the influence of this bias in news articles on readers' perceptions of real world events. We deconstruct this bias into two components: selection and framing. Selection bias relates to which subset of topics or facts are presented to the reader, while framing bias relates to how those topics or facts are presented. To isolate the two components and study their impacts separately, we propose a novel pipeline that utilizes LLMs to extract facts from collections of news stories on a particular event and then generate new articles with precisely controlled amounts of selection and framing bias. This nuanced approach allows for a more detailed understanding of how bias in media presentations can shape public perception, ultimately contributing to the broader discourse on media influence and bias.