Comprehensive analysis of machine learning classification metrics, algorithmic models, and the nuances of the neutral class

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The rapid proliferation of social media networks has fundamentally transformed the global information ecosystem,creating a critical tension between the preservation of freedom of speech and the mitigation of systemic risks topublic safety and national security. This paper establishes a novel theoretical mathematical model designed tosimulate and optimize the legal regulation of information flows within digital platforms. By treating legal andstructural regulation as an explicit control variable, the proposed framework quantifies its cascading effects oncore societal performance indicators. The system dynamics incorporate a multi-layered matrix of contextual riskfactors, including real-time disinformation indices, historical user behavioral patterns, spatial-temporal constraints,and geopolitical event proximity. Through simulation-based optimization, the model demonstrates how optimalregulatory thresholds can be algorithmically derived to prevent the dual failure modes of authoritarian overcensorship and unmitigated information disorder. While the framework provides a robust conceptual architecturefor algorithmic policymaking, it is presently bounded by its theoretical nature; validation has been executed strictlythrough synthetic scenarios and simulated environments rather than empirical production data. Consequently,this study lays the formal mathematical groundwork for adaptive, risk-aware digital governance tools, outliningcritical pathways for future empirical validation across diverse socio-cultural and international legal frameworks.