Analyzing Country-Level Heterogeneity in Public Attitudes Toward Migration Topics on Bluesky
Date:
Link: MPIDR
Abstract: Migration is a globally significant political issue, with discourse varying across countries. This study examines the regionally varying patterns of migration discourse on Bluesky, an emerging platform that saw an increase in number of users following Elon Musk’s acquisition of Twitter (now X) and its subsequent policy changes. We explore how Bluesky users discuss migration issues in the United States, the United Kingdom, and Germany, focusing on key topics, sentiment, and variations in word usage. To achieve this, we identify country-specific themes, sentiment dynamics, and construct dialectograms using embeddings trained on posts from each country to explore nuanced linguistic differences. During this process, since Bluesky lacks geolocation metadata, we employ an large language model (LLM) to annotate posts based on the country they reference. Our findings reveal distinct themes shaping Bluesky users’ migration discourse in the United States, the United Kingdom, and Germany. In all countries, negative sentiment posts outnumber positive sentiment posts approximately five times, and the predominant themes of negative sentiment posts were more partisan compared to those of positive sentiment posts. A bilateral comparison using dialectograms highlights words co-occur distinctively with words related to recently relevant subtopics in each corpus. This study provides new insights into regional variations in migration discourse on Bluesky and introduces an innovative approach for analyzing social data sources that lack traditional metadata.
Slides: PDF