Affect, the Master Pulling the Strings: Emotional states and their influence on cognitive strategies and social polarization: A semi-automated analysis of online textual data

This thesis explores the impact of emotional states on cognitive strategies, behavior, and groupbased polarization in discussions related to the COVID-19 pandemic. To analyze shifts in emotional states and corresponding cognitive strategies I will focus on discussions around the disruptive period of Covid-19. The reason for this is that disruptive events are moments in which emotions are stirred up and habituated cognitive strategies shaken. The pandemic increased the potential for polarization and mistrust in government, democracy, and science. Therefore, it’s crucial to investigate if the methods used in this thesis can detect these potentially dangerous dynamics through textual data. Using comments from two prominent Austrian newspapers comments sections, the study aims to determine whether emotional shifts during the pandemic affected the level of group-based polarization over a period of time spanning from February 2020 to May 2021. The paper discusses the impact of COVID-19 on political consequences, emotional states most present during the pandemic, and the connection between cognition, affect, and polarization. The study utilizes the dimensional theory of affective intelligence as its main theory and employs a dictionary-based approach for semi-automated text analysis. Building on the given assumptions, it aims to infer via different combinations of the extracted measures whether a discussion during a given timestamp is characterized by higher or lower polarization. However, the results do not support any of the proposed hypotheses, and the analysis does not yield any significant results on how emotional states change during the pandemic or how they affect group-based polarization, given the methods and data used. This thesis concludes by describing the theoretical and methodological limitations encountered during the study and suggests alternative approaches, including utilizing BERT or conducting an analysis of semantic breadth to analyze emotions expressed in text.