The parallel evolution of online and media immigration discourses
Author(s): Anastasiia Menshikova, Miriam Hurtado Bodell, Måns Magnusson, Marc Keuschnigg
Wednesday 14 | 15:00-15:20
Room: TP53
Session: Media and communication
Different ways to interpret reality exist in different segments of society, leading to collective outcomes such as polarization of opinions and behaviors or the erosion of social cohesion. Our work identifies similarities and differences in interpretations of immigration in two entirely different but equally important segments of public discourse: nationwide traditional media and online social media. We focus on the case of the Swedish immigration discourse in 2006 - 2020 to gain insights into the parallel evolution of immigration interpretations among professional journalists and users of an anonymous online forum.
We analyze 15M posts from Sweden’s largest online discussion forum and 16M newspaper articles from Sweden’s four largest daily newspapers. With a novel implementation of a joint seeded topic model, we detect various interpretations in a combined online and newspaper corpus, tackling methodological challenges such as differences in vocabulary between professional journalists and social media users and ensuring comparability and equivalence of interpretations in the two corpora.
We develop a collection of interpretations of immigration and conduct a systematic comparison by analyzing the salience of interpretations in the traditional media and at an online forum as multivariate time-series data. We compare how the immigration discourse in both spheres reacts to major events. In addition, we identify points of change in both domains and compare the timing and the nature of these changes for both corpora. Overall, our results demonstrate that the image of immigration that professional journalists create and the ways social media users interpret immigration fundamentally differ, with the former’s framing mirroring their explanation of the flow of news events and the latter’s framing mainly mirroring complaints about the influx of newcomers and the failing of public policy.
We demonstrate an application of computational text analysis for developing macro-level measurements of cultural dynamics arising in highly institutionalized (newspapers) and highly interactive (social media platforms) environments. Our findings explicate the macro-level conditions that situate voting behavior, residential segregation, or labor market inequalities in Sweden.