Digital Discourse Transformation: Analysis of Neutral Discussion Escalation into SARA Debates on Indonesian Social Media Platforms

Abstract

The digital era has fundamentally transformed public communication in Indonesia, presenting critical challenges through the escalation of neutral discussions into debates based on Ethnicity, Religion, Race, and Intergroup Relations (SARA) on social media platforms. This research aims to identify linguistic and interactional patterns that serve as markers of discourse transformation, analyze the role of platform algorithms in accelerating escalation, compare escalation characteristics across X, Threads, Instagram, and TikTok platforms, and develop predictive models for early detection of identity-based discourse escalation. The study employs a mixed-methods design with Digital Critical Discourse Analysis framework, integrating Social Network Analysis and controlled digital experiments. Data collection involved 1,247 discussion threads across four platforms and 28 in-depth interviews using stratified purposive sampling and maximum variation sampling. Analysis utilized statistical testing, machine learning pipeline with BERT-based models, and thematic analysis with inter-rater reliability ≥0.80. Results revealed four distinct transformation phases characterized by decreasing lexical diversity (TTR 0.67 to 0.29), increasing negative sentiment (0.12 to -0.73), and network fragmentation (density 0.34 to 0.12). The developed Transformative Discourse Model achieved 89.7% accuracy in predicting escalation events with 4-14 hours early detection capability. Platform-specific analysis showed TikTok as fastest escalation (14.2 hours) and Threads as slowest (31.8 hours). The research contributes Indonesian Digital Discourse Corpus, cross-platform comparative framework, and evidence-based intervention protocols, supporting digital literacy strengthening and radicalism prevention in Indonesian cyberspace.