Data Driven Perspectives on Societal Issues: Addressing Challenges in the Digital Age

Abstract

This dissertation examines three interconnected societal challenges emerging from digital transformation and globalization: the spread of misinformation on social media, the regulation of digital platforms, and the effectiveness of economic policy responses to global crises. Using large-scale behavioral datasets combined with robust empirical methods including econometric identification strategies, natural language processing, and computational techniques, the research provides evidence-based insights into contemporary societal issues. <br><br> The first part analyzes misinformation dynamics, focusing extensively on Community Notes, X's crowdsourced fact-checking initiative. The study of 15,000+ annotated posts reveals that misleading posts receive 35.85% fewer retweets than accurate posts. Further analysis shows that misinformation perceived as believable receives 217.09% more retweets, while posts perceived as harmful receive 41.32% fewer retweets, suggesting that easily believable yet non-harmful misinformation spreads most effectively. A survey of 1,800+ participants demonstrates that community-based fact-checking is perceived as significantly more trustworthy than simple misinformation flags across the political spectrum, with explanatory context—rather than source—driving this trust boost. An analysis of 91,452 flagged posts reveals that AI-generated misinformation differs systematically from human-generated content: it originates from more influential accounts, is more visually engaging and positively toned, and significantly outperforms non-AI content in virality despite being rated as slightly less harmful or believable. <br><br> The second part addresses digital platform regulation by analyzing 156+ million moderation decisions submitted to the EU's Digital Services Act Transparency Database. The findings reveal significant variation in moderation volumes across platforms, with TikTok accounting for 64% of submissions despite having fewer EU users than competitors. Notably, 60.67% of moderation decisions were fully automated, though automation levels varied substantially by platform, with TikTok relying primarily on automation while X and YouTube relied on manual review. These inconsistent practices suggest varied interpretations of regulatory obligations and raise concerns about enforcement consistency. <br><br> The third part evaluates economic policy effectiveness by examining temporary fuel tax reductions implemented in France, Germany, and Italy in 2022 using a staggered Difference-in-Differences design. The analysis finds very high pass-through rates, indicating that tax cuts were effectively transmitted to consumers, with gasoline showing higher pass-through than diesel. However, the study cautions that such tax cuts, while effective short-term measures, raise concerns about distributional equity and long-term climate goals. <br><br> By integrating computational techniques with econometric rigor and grounding analysis in real-world behavioral data, this dissertation demonstrates how Big Data can illuminate contemporary societal risks and evaluate policy interventions, providing insights for policymakers, platform designers, and researchers addressing the interconnected challenges of digital transformation and global crises.