Environmental Conflict, Citizen Perceptions, and Digital Narrative Dynamics: An NLP Study of Lithium Mining Protests in Serbia

Abstract

The global transition toward a green economy has created unprecedented demand for critical minerals, fundamentally reshaping geopolitical and ecological landscapes. While lithium is essential for this "materials revolution," the burdens of green extractivism are often disproportionately outsourced to semi-peripheral regions, igniting fierce conflicts over environmental justice and national sovereignty. This study examines the digital narrative construction and affective mobilization surrounding the anti-lithium mining protests in Serbia between 2020 and 2025. Using Natural Language Processing (NLP) assisted by Large Language Models (LLMs) to analyze 2,609 social media posts across Facebook, X (Twitter), and LinkedIn, this research quantifies the digital "voice" of the public under a competitive authoritarian regime. The findings reveal that while social media primarily serves to organize and coordinate the protests (81.4%), the underlying grievances blend environmental justice (14.2%), resource nationalism (2.4%), and critiques of green colonialism (2.0%). Together, these distinct themes forge a unified and powerful narrative of "Environmental Nationalism." Crucially, the longitudinal sentiment analysis identifies a statistically significant "Anger Gap", a 7 to 10 percentage point surge in high-arousal anger across platforms during active protest phases. This emotional tipping point, transforming latent ecological fear into mobilized political anger, serves as the psychological glue that binds individual anxieties into a collective resistance identity. Ultimately, this study demonstrates how localized environmental grievances are digitally weaponized to challenge both state-led industrial policy and the asymmetric power dynamics of the global green transition.