Adapting Bot Detection Models for Romania’s Disinformation Ecosystem

Abstract

The proliferation of social media bots and fake accounts has significantly disrupted information ecosystems, posing substantial challenges in detecting and mitigating disinformation. While machine learning and deep learning models have shown varying levels of success on platforms like Twitter and Facebook, they often fail to account for region-specific nuances critical for effective bot detection. Facebook and Twitter have been widely used in disinformation research due to their large user bases and historically open API access, facilitating large-scale data collection. This study addresses these gaps by proposing a hybrid detection framework tailored to Romania's disinformation landscape. Each society is shaped by its unique historical, cultural, linguistic, and geopolitical factors, influencing how disinformation spreads and resonates with different audiences. The proposed approach emphasizes well-established narratives that significantly influence vulnerable populations, including young adults with limited capacity for fact-checking and older adults with low levels of digital literacy. This research will begin by reviewing existing literature on bot detection methodologies and narrative analysis, identifying their strengths, limitations, and applicability to regional contexts. By integrating conventional detection methodologies with a refined analysis of niche and high-risk narratives, this research investigates how disinformation campaigns gain momentum and escalate, providing a deeper understanding of their dynamics and impact. The results will reveal patterns and strategies employed in the propagation of disinformation, contributing to the development of more targeted and effective detection systems. This method also facilitates the early identification of emerging disinformation clusters, offering timely and proactive intervention opportunities. This paper is part of a broader PhD research program centered on analysing narratives and narrative strategies in disinformation, with all findings supporting this overarching goal. The findings emphasize the importance of regional customization in bot detection frameworks, particularly for countries like Romania, where disinformation leverages historical, cultural, and socio-political triggers. These insights will strengthen the resilience of information ecosystems and hold significant value for cybersecurity professionals, social media platforms, and policymakers dedicated to combating manipulation and fostering a more secure digital environment.