Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Simulation of Language Evolution under Regulated Social Media Platforms: A Synergistic Approach of Large Language Models and Genetic Algorithms

Jinyu Cai, Yusei Ishimizu, Mingyue Zhang, Munan Li, Jialong Li, Kenji Tei

TL;DR

This paper presents an LLM-driven multi-agent framework to simulate language evolution under regulated social platforms, introducing constraint and expression language strategies and using a genetic-algorithm–driven evolution operated by LLMs. A supervisory agent models platform moderation to enforce rules, enabling adversarial co-evolution between users and regulators. The framework is evaluated in an abstract password game and a realistic illicit-pet-trade scenario, with a 40-participant human study and ablation experiments confirming GA's benefit for long-term adaptation. Results show that increasing dialogue rounds improves unobstructed conversation length and information transmission, though the models exhibit variability and limitations in strategy diversity and ecological realism. The work highlights practical implications for designing moderation-aware simulations and points to future work on fine-tuning for social-media contexts and incorporating broader, multi-user dynamics.

Abstract

Social media platforms frequently impose restrictive policies to moderate user content, prompting the emergence of creative evasion language strategies. This paper presents a multi-agent framework based on Large Language Models (LLMs) to simulate the iterative evolution of language strategies under regulatory constraints. In this framework, participant agents, as social media users, continuously evolve their language expression, while supervisory agents emulate platform-level regulation by assessing policy violations. To achieve a more faithful simulation, we employ a dual design of language strategies (constraint and expression) to differentiate conflicting goals and utilize an LLM-driven GA (Genetic Algorithm) for the selection, mutation, and crossover of language strategies. The framework is evaluated using two distinct scenarios: an abstract password game and a realistic simulated illegal pet trade scenario. Experimental results demonstrate that as the number of dialogue rounds increases, both the number of uninterrupted dialogue turns and the accuracy of information transmission improve significantly. Furthermore, a user study with 40 participants validates the real-world relevance of the generated dialogues and strategies. Moreover, ablation studies validate the importance of the GA, emphasizing its contribution to long-term adaptability and improved overall results.

Simulation of Language Evolution under Regulated Social Media Platforms: A Synergistic Approach of Large Language Models and Genetic Algorithms

TL;DR

This paper presents an LLM-driven multi-agent framework to simulate language evolution under regulated social platforms, introducing constraint and expression language strategies and using a genetic-algorithm–driven evolution operated by LLMs. A supervisory agent models platform moderation to enforce rules, enabling adversarial co-evolution between users and regulators. The framework is evaluated in an abstract password game and a realistic illicit-pet-trade scenario, with a 40-participant human study and ablation experiments confirming GA's benefit for long-term adaptation. Results show that increasing dialogue rounds improves unobstructed conversation length and information transmission, though the models exhibit variability and limitations in strategy diversity and ecological realism. The work highlights practical implications for designing moderation-aware simulations and points to future work on fine-tuning for social-media contexts and incorporating broader, multi-user dynamics.

Abstract

Social media platforms frequently impose restrictive policies to moderate user content, prompting the emergence of creative evasion language strategies. This paper presents a multi-agent framework based on Large Language Models (LLMs) to simulate the iterative evolution of language strategies under regulatory constraints. In this framework, participant agents, as social media users, continuously evolve their language expression, while supervisory agents emulate platform-level regulation by assessing policy violations. To achieve a more faithful simulation, we employ a dual design of language strategies (constraint and expression) to differentiate conflicting goals and utilize an LLM-driven GA (Genetic Algorithm) for the selection, mutation, and crossover of language strategies. The framework is evaluated using two distinct scenarios: an abstract password game and a realistic simulated illegal pet trade scenario. Experimental results demonstrate that as the number of dialogue rounds increases, both the number of uninterrupted dialogue turns and the accuracy of information transmission improve significantly. Furthermore, a user study with 40 participants validates the real-world relevance of the generated dialogues and strategies. Moreover, ablation studies validate the importance of the GA, emphasizing its contribution to long-term adaptability and improved overall results.

Paper Structure

This paper contains 32 sections, 2 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Background and Related Work
  3. Large Language Models
  4. Slang Detection and Identification
  5. LLMs in Evolutionary Game Theory and Social Simulation
  6. Framework Design
  7. Overview
  8. Participant Agents
  9. Memory Module
  10. Reflection Module
  11. Selection
  12. Crossover
  13. Mutation
  14. Planning Module
  15. Dialogue Module
  16. ...and 17 more sections

Figures (4)

  • Figure 1: Framework Overview. The framework consists of two participant agents and a single supervisory agent. The iterative process follows steps 2 through 5, where participant agents continuously refine their language strategies while the supervisory agent identifies regulation violations.
  • Figure 2: Average Continuous Dialogue Turns and Information Transmission Accuracy Across Dialogue Rounds
  • Figure 3: Box plots of user study scores across different metrics in two scenarios. The red x symbol denotes the mean value.
  • Figure 4: Performance with/without GA