Language Evolution for Evading Social Media Regulation via LLM-based Multi-agent Simulation

  • 2024-05-05 10:02:54
  • Jinyu Cai, Jialong Li, Mingyue Zhang, Munan Li, Chen-Shu Wang, Kenji Tei
  • 0


Social media platforms such as Twitter, Reddit, and Sina Weibo play a crucialrole in global communication but often encounter strict regulations ingeopolitically sensitive regions. This situation has prompted users toingeniously modify their way of communicating, frequently resorting to codedlanguage in these regulated social media environments. This shift incommunication is not merely a strategy to counteract regulation, but a vividmanifestation of language evolution, demonstrating how language naturallyevolves under societal and technological pressures. Studying the evolution oflanguage in regulated social media contexts is of significant importance forensuring freedom of speech, optimizing content moderation, and advancinglinguistic research. This paper proposes a multi-agent simulation frameworkusing Large Language Models (LLMs) to explore the evolution of user language inregulated social media environments. The framework employs LLM-driven agents:supervisory agent who enforce dialogue supervision and participant agents whoevolve their language strategies while engaging in conversation, simulating theevolution of communication styles under strict regulations aimed at evadingsocial media regulation. The study evaluates the framework's effectivenessthrough a range of scenarios from abstract scenarios to real-world situations.Key findings indicate that LLMs are capable of simulating nuanced languagedynamics and interactions in constrained settings, showing improvement in bothevading supervision and information accuracy as evolution progresses.Furthermore, it was found that LLM agents adopt different strategies fordifferent scenarios.


Quick Read (beta)

loading the full paper ...