For You, Not for Truth How TikTok Manipulates Perception and Belief through Engagement-Driven Algorithms

Abstract

This thesis investigates the TikTok platform in the context of its economic, social and potential geopolitical impact in the United States. By analyzing its algorithmic design, this study will determine whether TikTok shapes public perception on certain key issues by amplifying misinformation, thereby creating potential national security vulnerabilities. Moreover, this thesis examines the national and international context within which TikTok operates, focusing not only on regulatory requirements but important issues like data privacy, potential motivations of key stakeholders and potential structural remedies currently under discussion. Combining qualitative policy analysis with quantitative machine learning methods, the study evaluates how TikTok’s engagement-maximizing infrastructure introduces risks at the intersection of technology, geopolitics, and digital governance. The qualitative component draws on legislative debates, corporate disclosures, stakeholder incentives, and comparative case studies—including India’s TikTok ban, the Huawei precedent, and the GDPR framework—to contextualize TikTok’s regulatory and national security challenges. In parallel, a quantitative analysis of 19,084 TikTok video transcripts employs a zero-shot classification pipeline and Random Forest ensemble model to identify predictors of misinformation and emotional amplification. The results show that emotionally charged, high-engagement content—particularly false claims—disproportionately benefits from algorithmic promotion. Taken together, the findings suggest that TikTok’s recommendation system is not politically neutral but is structured to prioritize virality over veracity. The thesis concludes with a set of policy recommendations that address algorithmic transparency, platform accountability, and foreign influence safeguards without undermining democratic values or digital rights.