Dissertation

Migration, culture, and inequalities in algorithmically-mediated societies

Abstract

The widespread use of online platforms has transformed how people interact, express opinions, and access information. As a byproduct of these digital interactions, users continuously generate large-scale digital trace data, opening up unprecedented opportunities for social science research. In this thesis, digital trace data from Facebook, Twitter (now X), TikTok, and Wikipedia 
are used to address key questions ranging from cross-country cultural similarity to large-scale human migration, patterns of gender and social inequalities, and online behavior on algorithmically mediated platforms. Across six studies, it demonstrates how digital trace data can complement or overcome the limitations of traditional data sources, which are often slow, costly, or unavailable. The first two chapters link culture and migration by introducing a novel measure of cross-national cultural similarity based on Facebook users’ interests and showing that this measure improves predictions of international migration flows beyond standard economic and geographic factors. The third chapter proposes a novel use of Wikipedia page view data to detect information-seeking behavior during crises, showing that readership patterns can serve as a near real-time proxy for forced migration, as demonstrated during the 2022 Ukrainian refugee crisis. Addressing global inequality, the fourth chapter uses Facebook data to examine gender balance among users interested in Science, Technology, Engineering, and Mathematics (STEM) fields worldwide and provides a detailed analysis of gender disparities across age and education groups in Brazil. The 
fifth chapter investigates social vulnerability by analyzing Twitter data to characterize missing children in Guatemala, offering insights that complement scarce and delayed official statistics. Finally, the sixth chapter evaluates user watching behavior on short-form video platforms by analyzing donation-based TikTok data, revealing the limited predictability of watching behavior and 
highlighting methodological challenges arising from increasing platform data access restrictions. Collectively, these studies demonstrate how digital trace data can advance the measurement and understanding of complex social phenomena, particularly in regions of the Global South and in crisis contexts where conventional data are sparse or delayed. By integrating computational methods with theoretical perspectives from the social sciences, this thesis advances the field of computational social science and highlights the societal value of using digital trace data for research.