Large Language Model-driven Multi-Agent Simulation for News Diffusion Under Different Network Structures

Xinyi Li,Yu Xu,Yongfeng Zhang,Edward C. Malthouse

2024-10-17

Abstract:The proliferation of fake news in the digital age has raised critical concerns, particularly regarding its impact on societal trust and democratic processes. Diverging from conventional agent-based simulation approaches, this work introduces an innovative approach by employing a large language model (LLM)-driven multi-agent simulation to replicate complex interactions within information ecosystems. We investigate key factors that facilitate news propagation, such as agent personalities and network structures, while also evaluating strategies to combat misinformation. Through simulations across varying network structures, we demonstrate the potential of LLM-based agents in modeling the dynamics of misinformation spread, validating the influence of agent traits on the diffusion process. Our findings emphasize the advantages of LLM-based simulations over traditional techniques, as they uncover underlying causes of information spread -- such as agents promoting discussions -- beyond the predefined rules typically employed in existing agent-based models. Additionally, we evaluate three countermeasure strategies, discovering that brute-force blocking influential agents in the network or announcing news accuracy can effectively mitigate misinformation. However, their effectiveness is influenced by the network structure, highlighting the importance of considering network structure in the development of future misinformation countermeasures.

Social and Information Networks,Artificial Intelligence,Multiagent Systems

What problem does this paper attempt to address?

The paper attempts to address the issue of how fake news spreads on social media under different network structures and its influencing factors, and explores effective countermeasures. Specifically, the paper investigates the following aspects by introducing a large language model (LLM)-driven multi-agent simulation approach:

1. **Key Factors in the Spread of Fake News**: The paper explores the impact of agent personality traits (such as the Big Five personality traits) and network structure on the spread of fake news. By simulating the information dissemination process under different network structures, the authors verify the role of these factors in news diffusion.

2. **Effective Strategies to Counter Fake News**: The paper evaluates the effectiveness of three intervention strategies, including:
- **Encouraging Comment Sharing on News**: It is hypothesized that by requiring agents to add comments when sharing news, they can be prompted to think more carefully about the content they are sharing, thereby reducing impulsive sharing behavior.
- **Announcing the Accuracy of Shared News**: Simulating a fact-checking algorithm that confirms the authenticity of news after it has been disseminated to a certain proportion and informs agents who have received the news of its inaccuracy.
- **Blocking Certain Influencers**: Identifying and blocking key influencers in the network to curb the spread of fake news.

3. **Impact of Network Structure on the Effectiveness of Intervention Strategies**: The paper emphasizes the importance of network structure in developing future countermeasures against fake news. Different network structures (such as random networks, scale-free networks, and high-betweenness networks) have varying impacts on the effectiveness of the aforementioned intervention strategies.

Through these studies, the paper aims to provide new perspectives and methods for understanding the mechanisms of fake news dissemination and developing effective countermeasures.