基于LLMs的AI代理探索推荐系统对社交网络的影响

打开文本图片集

中图分类号：TP391.3；TP18 文献标识码：A 文章编号：2096-4706（2025）21-0052-07

Impact of Al Agent Exploration Recommendation System Based on LLMs on Social Networks

WANG Shan1， AN Qi² （1.Dianchi Colege，Kunming 650228，China;2.Kunming City University，Kunming 65o106，China）

Abstract：This paper studies the impact of Artificial Intelligence agent simulation model of recommendation system basedonLarge Language Models（LLMs）onsocial networks.Toaddressthepotential impactof therecommendationsystem onuserbehaviorandsocialopinions insocialmedia，itproposesanewsimulationframeworktosimulatetherealusergroup by constructingavarietyof simulationagents.Thesimulationagent is encoded withasetofpredefiedandasignedatributes and description features，andtherelatedfeaturesarepresentedintheformofa7-pointLikertscale tosimulateuserswith diverse personalitycharacterstcsanddynamic preferences.Treerecommendationsystemsetings，diversitybalanceandcosstency are designedtoevaluatetheirimpactonuser engagementandgrouppolarization.Theresultsshowthatdiferentrecommendation systemstingshaveasignifcantimpactonuserbehavior：amongthem，theconsistencysetingcanimproveuserengagement， but mayaggravate grouppolarization;thediversityand balancesettingscanmaketheuserresponse more balanced.This paper providesanewperspectiveforunderstandingandoptimizingtheapplicationofrecommendationsystemsinsocialnetworks，and also provides theoretical basis and practical guidance for subsequent research.

Keywords： recommendation system; LLMs; AI agent simulation; user engagement; group polarization

0 引言

互联网社交媒体（如微博、小红书）的蓬勃发展，已对社交互动、沟通模式与信息消费方式产生根本性重塑。（剩余9562字）