ElecTwit: A Framework for Studying Persuasion in Multi-Agent Social Systems

TL;DR

ElecTwit is a simulation framework for studying persuasion in multi-agent social media environments, revealing diverse techniques and model behaviors that impact social dynamics and influence during elections.

Contribution

The paper introduces ElecTwit, a realistic simulation environment for analyzing persuasion techniques in multi-agent social systems, surpassing prior game-based models.

Findings

Wide range of persuasion techniques used across models

Model architecture influences persuasion effectiveness

Emergence of phenomena like 'kernel of truth' messages

Abstract

This paper introduces ElecTwit, a simulation framework designed to study persuasion within multi-agent systems, specifically emulating the interactions on social media platforms during a political election. By grounding our experiments in a realistic environment, we aimed to overcome the limitations of game-based simulations often used in prior research. We observed the comprehensive use of 25 specific persuasion techniques across most tested LLMs, encompassing a wider range than previously reported. The variations in technique usage and overall persuasion output between models highlight how different model architectures and training can impact the dynamics in realistic social simulations. Additionally, we observed unique phenomena such as "kernel of truth" messages and spontaneous developments with an "ink" obsession, where agents collectively demanded written proof. Our study provides a foundation for evaluating persuasive LLM agents in real-world contexts, ensuring alignment and preventing dangerous outcomes.