This thesis uses computational text analysis to advance the sociological understanding of how culture, as reflected in public discourse, comes to change. Drawing upon theoretical foundations of analytical sociology and cultural sociology, I view public discourse as a collective phenomenon shaped by an interplay of individual-level dynamics and interdependencies between different societal spheres that contribute to public discussions. Computational text analysis enables analyzing the interdependencies and individual-level dynamics underlying shifts in public discourse and advances the sociological explanation of cultural change. Using the empirical setting of the Swedish immigration discourse, this work explicates mechanisms and conditions shaping changes in public discourse in the wake of unexpected events and political crises. This thesis contributes to existing work with new research designs for analyzing individual-level and macro-level cultural changes by further integrating computational text analysis into sociological research.

Essay I theorizes and tests the role of different individual-level mechanisms in generating aggregate-level shifts in social media discourses following disruptive events. I analyze how changes in (1) the way individuals talk about immigration (within-individual change) and (2) the composition of individuals who participate in online discussions (compositional change) contribute to the aggregate-level shifts in the Swedish online immigration debate following a series of Jihadist terrorist attacks in the 2010s in Sweden and in Europe. I quantify how shifts in the salience of culture, security, and economy in immigration discourse before and after attacks are jointly shaped by the two mechanisms. This study underscores the necessity of individual-level text data in distinguishing between mechanisms that can produce similar shifts in aggregate measures of discourse.

Essay II systematically analyzes the effects of terrorist attacks on public opinion towards im-migration across Europe and evaluates the moderating role of national-level media attention preceding the attacks. I examine this idea empirically by analyzing the impact of 38 jihadist terrorist attacks on public opinion on immigration in 19 European countries in 2013-19, using a multi-site natural experiment approach. In this paper, I combine survey data from the ESS with a multilingual corpus of media coverage from the largest European national media outlets. Contrary to the expectations, even under conditions of heightened media attention on immigration, I find no significant effects of terrorism. I demonstrate the importance of combining digital trace textual data from social media with traditional data sources for sociological inquiry.

Essay III studies long-term cultural change and compares the similarity between cultural meanings in the national newspapers and on social media over more than a decade. I compare changes in collective meanings of immigration — the macro-level dynamics within different spheres of public discourse — in the Swedish national newspapers and at the largest Swedish online forum, Flashback. I test whether the European ‘refugee crisis’ brought the two spheres of public discourse closer together or drove them further apart. This paper offers a novel methodological approach to drawing a plausible and comprehensive comparison between discourses generated by vastly different groups of actors using seeded topic modeling. The results support the convergence story, meaning networks in the newspapers and online became more aligned during 2015–16 European ‘refugee crisis.’

Essay IV examines the relationship between everyday interests and political preferences within Sweden’s multi-party context, thus analyzing a broader social context surrounding politically engaged Swedes who discuss politics and immigration online. Using data from Swedish Twitter from 2010-2020 and utilizing mixed-membership clustering to discover coherent groups of everyday interests in unstructured social media data, we analyze how non-partisan everyday interests align with specific political parties and the left-right political scale. The results demonstrate that the fragmentation of everyday interests in a multi-party political system aligns with political ideologies rather than partisan identities. Despite the observed polarization, certain interests transcend political divides, offering opportunities for dialogue across ideological lines.