摘要
文章研究了基于大语言模型(LLMs)的推荐系统人工智能代理模拟模型对社交网络的影响。针对社交媒体中推荐系统对用户行为和社会观点的潜在影响,提出了一种新的模拟框架,通过构建多样化的模拟代理来模拟真实用户群体。模拟代理利用一组预定义并分配的属性与描述特征进行编码,相关特征以7分李克特量表形式呈现,以此模拟具有多样化个性特征和动态偏好的用户。实验设计了多样性、平衡性和相符性3种推荐系统设置,用于评估它们对用户参与度和群体极化的影响。结果表明,不同的推荐系统设置对用户行为存在显著影响:其中,相符性设置虽能提升用户参与度,但可能加剧群体极化;而多样性和平衡性设置则可使用户反应更趋均衡。文章为理解与优化推荐系统在社交网络中的应用提供了新视角,也为后续研究提供了理论基础与实践指导。
This paper studies the impact of Artificial Intelligence agent simulation model of recommendation system based on Large Language Models(LLMs)on social networks.To address the potential impact of the recommendation system on user behavior and social opinions in social media,it proposes a new simulation framework to simulate the real user group by constructing a variety of simulation agents.The simulation agent is encoded with a set of predefined and assigned attributes and description features,and the related features are presented in the form of a 7-point Likert scale to simulate users with diverse personality characteristics and dynamic preferences.Three recommendation system settings,diversity,balance and consistency,are designed to evaluate their impact on user engagement and group polarization.The results show that different recommendation system settings have a significant impact on user behavior:among them,the consistency setting can improve user engagement but may aggravate group polarization;the diversity and balance settings can make the user response more balanced.This paper provides a new perspective for understanding and optimizing the application of recommendation systems in social networks,and also provides theoretical basis and practical guidance for subsequent research.
作者
王杉
安琪
WANG Shan;AN Qi(Dianchi College,Kunming 650228,China;Kunming City University,Kunming 650106,China)
出处
《现代信息科技》
2025年第21期52-58,共7页
Modern Information Technology
