Computational Framing Analysis for Polarized Topics Online

Framing, which elevates pieces of information in communication to make them more noticeable, affects people’s perceptions and choices. Given that framing influences behavior, it is a cardinal consideration for society, especially on polarized topics online, such as COVID-19 and climate change. Although the analysis of frames is crucial, frames are exceptionally challenging to conceptualize. Additionally, annotated framing data is only scarcely available. Due to that, I aim to leverage the breakthroughs in natural language processing to advance computational framing research on two fronts. First, I developed framing detection algorithms for three distinct exploratory levels, i.e., frame labels, frame dimensions, and frame structure. My work shows trade-offs between the validation of established and the exploration of novel frames using a multi-perspective approach. Second, I studied the relations between content and users concerning the prevalence of the frames employed in online systems. A substantial interplay between user behavior and the framing of content in information systems is revealed in a research direction yet to be explored. At its core, my research integrates social science research with computational approaches, broadening the field and revealing several new research directions. Besides fostering an increased understanding of framing, I developed novel methodologies for framing analysis and released their artifacts for public use, e.g., as open-source tools. My findings can inform the design of future information systems to balance the users’ online behavior regarding the framing diversity of the content.

Link to PDF: Thesis
Link to Thesis Description on the University page: TUGonline
Link to Presentation Slides: Slides