CEU Electronic Theses and Dissertations, 2025
| Author | Chen, Yijing |
|---|---|
| Title | Addressing Bias and Oversimplification in Measurements of Political Polarization |
| Summary | Advances in digital technologies have fundamentally reshaped how individuals participate in political communication. Today, the production, circulation, and (mis)interpretation of political information unfold within a sociotechnical system that is increasingly interpersonally networked, algorithmically curated, and infused with AI-generated content. Thus, challenges such as political polarization are now embedded in a more complex media environment, which calls for renewed examinations of classic theories, the continuous adaptation of methods to new datasets, and critical reflections on how key concepts like political polarization are defined and measured. Motivated by this growing complexity, my dissertation seeks to operationalize--and critically reflect on the current operationalizations of--classic theories (i.e., intermedia agenda setting) and social constructs (i.e., political polarization) in today's information ecosystem. Specifically, it examines the stability of a classic theory–intermedia agenda setting (IAS)–in a fragmented media environment, and addresses three key limitations of existing polarization research: (1) the conflation of attitudinal and behavioral measures, (2) the reliance on linear, uni-dimensional metrics in modeling political beliefs, and (3) the measurement bias introduced by focusing exclusively on online engagement data that is easily retrievable via public APIs. The dissertation consists of one conceptual chapter and three empirical chapters. Chapter 1 establishes the theoretical framework, delineating the key developments in political communication in the contemporary media environment and articulating how polarization is conceptualized and operationalized from attitudinal and behavioral perspectives. Chapter 2 assesses the stability of IAS theory by analyzing agenda alignment across different types of news media during the 2016 and 2020 U.S. presidential elections, revealing the increasing divergence and the shifting IAS roles of different media. Chapters 3 and 4 turn to political polarization and focus on the attitudinal and behavioral perspectives, respectively. Chapter 3 introduces a novel application of Response-Item Networks (ResIN) to measure polarization using attitude data, modeling belief systems as statistical networks of interconnected beliefs. This approach captures ideological polarization as a structural transformation in belief systems and provides nuanced, multi-level metrics. Chapter 4 reflects on the current behavioral measurements of polarization that rely on political engagement documented by public platform APIs (i.e., visible engagement). Using a combined dataset of survey responses and user-donated YouTube traces from Hungary, the analysis reveals how a sole focus on visible engagement can distort our understanding of the broader ideological landscape. Overall, this dissertation bridges network science, computational social science, and political communication through conceptual clarification, methodological innovation, and empirical reflection. The main contributions include: (1) clarifying the conceptual distinction between attitudinal and behavioral polarization, enabling more consistent interpretation across studies; (2) expanding belief network analysis (BNA) through the use of ResIN, offering new tools for visualizing ideological divides and quantifying belief system polarization with greater granularity; (3) introducing a novel framework to address measurement bias in behavioral polarization studies by integrating digital traces and survey responses; and (4) advocating for a more globalized approach to polarization research through analyzing data from the U.S., Hungary, and other European countries. |
| Supervisor | Omodei, Elisa; Iniguez Gerardo |
| Department | Network Science PhD |
| Full text | https://www.etd.ceu.edu/2025/chen_yijing.pdf |
Visit the CEU Library.
© 2007-2025, Central European University