大模型驱动的社交智能体谣言易感性与干预策略研究

殷勇杰; 袁靖炜; 宫庆媛; 陈阳

doi:10.11959/j.issn.2096-6652.202545

您当前的位置：

首页 >

文章列表页 >

大模型驱动的社交智能体谣言易感性与干预策略研究

专题：社会计算与韧性社会 | 更新时间：2026-03-18

- 大模型驱动的社交智能体谣言易感性与干预策略研究
- Rumor susceptibility and intervention strategies of large language model-driven social agents
- 智能科学与技术学报 2025年7卷第4期页码：517-532
- 作者机构：
  
  1.复旦大学计算与智能创新学院，上海 200438
  2.复旦大学智能复杂体系基础理论与关键技术实验室，上海 200438
- 作者简介：
  
  [ "殷勇杰（2002- ），男，复旦大学计算与智能创新学院硕士生，主要研究方向为在线社交网络智能体行为建模与分析。" ]
  [ "袁靖炜（2004- ），男，复旦大学计算与智能创新学院在读，主要研究方向为多智能体系统。" ]
  [ "宫庆媛（1991- ），女，博士，复旦大学智能复杂体系基础理论与关键技术实验室青年副研究员，主要研究方向为在线社交网络用户行为大数据。" ]
  [ "陈阳（1981- ），男，博士，复旦大学计算与智能创新学院教授、博士生导师，上海市智能信息处理重点实验室副主任，主要研究方向为社会计算、计算机网络、大规模用户行为数据挖掘等。" ]
- 基金信息：
  
  国家自然科学基金项目(62102094)
- DOI：10.11959/j.issn.2096-6652.202545
  中图分类号： TP39
- 收稿：2025-09-30，
  
  修回：2025-12-15，
  
  录用：2025-12-21，
  
  纸质出版：2025-12-15
- 稿件说明：
移动端阅览
殷勇杰,袁靖炜,宫庆媛等.大模型驱动的社交智能体谣言易感性与干预策略研究[J].智能科学与技术学报,2025,07(04):517-532.

YIN Yongjie,YUAN Jingwei,GONG Qingyuan,et al.Rumor susceptibility and intervention strategies of large language model-driven social agents[J].Chinese Journal of Intelligent Science and Technology,2025,07(04):517-532.
殷勇杰,袁靖炜,宫庆媛等.大模型驱动的社交智能体谣言易感性与干预策略研究[J].智能科学与技术学报,2025,07(04):517-532. DOI： 10.11959/j.issn.2096-6652.202545.

YIN Yongjie,YUAN Jingwei,GONG Qingyuan,et al.Rumor susceptibility and intervention strategies of large language model-driven social agents[J].Chinese Journal of Intelligent Science and Technology,2025,07(04):517-532. DOI： 10.11959/j.issn.2096-6652.202545.

摘要

近年来，社交平台上涌现出众多大语言模型（large language model，LLM）驱动的社交智能体，其深度参与到用户的在线社交活动中。然而，社交平台的内容并不都是经过验证的真实信息，智能体的信息传播能力可能加剧谣言传播。衡量智能体对谣言的易感性、降低其对谣言的采信程度是亟待解决的问题。为了应对该风险，系统地探究了智能体在社交场景下对谣言的易感性与观点演变轨迹。研究结果表明，智能体不仅对未知谣言高度易感，其观点也会随着长期接触谣言而逐渐强化。此外，智能体在参与社交活动时，具有明显的采信谣言倾向。为缓解智能体对谣言的易感性，提出了“自提示”干预策略。该策略可以使智能体对谣言的采信率从75.91%显著降至13.50%，并有效促使持有中立观点的智能体转向反谣言立场。所提策略不仅揭示了LLM驱动的智能体的谣言易感机制，也为增强其抗谣言能力提供了可行路径，保障了LLM智能体在社交平台上安全部署，降低智能体带来的谣言传播风险。

Abstract

In recent years

numerous social agents empowered by large language model (LLM) have emerged on social media

which play a significant role in online social interactions. However

since not all messages shared on social media are verified as genuine

the involvement of these agents could amplify rumor propagation. Therefore

it becomes increasingly important to measure agents' susceptibility to rumors and reduce their acceptance of rumors. To tackle this issue

agents' susceptibility to rumors and how their opinion evolve on social media were systematically examined. The findings demonstrate that agents are highly susceptible to unknown rumors

tend to reinforce their beliefs over time through prolonged exposure. Furthermore

agents show a strong tendency to believe rumors during social interactions. To reduce agents' susceptibility to rumors

a “self-prompting” intervention strategy was proposed

which significantly reduced rumor acceptance rate among agents from 75.91% to 13.50% and effectively motivated agents with a neutral stance to take on anti-rumor positions. This research not only deepens our understanding of the mechanisms behind the rumor susceptibility of LLM-driven agents

but also provides an effective pathway to improve their anti-rumor capabilities

thereby offering support for the safe deployment of agents on social media and reduction of rumor propagation.

关键词

Keywords

references

FERRARA E, VAROL O, DAVIS C, et al. The rise of social bots[J]. Communications of the ACM, 2016, 59(7): 96-104.

汤家伟, 刘育杉, 高敏, 等. Cerberus: 基于深度学习的跨网站社交机器人检测系统[J]. 智能科学与技术学报, 2024, 6(4): 482-494.

TANG J W, LIU Y S, GAO M, et al. Cerberus: cross-site social bot detection system based on deep learning[J]. Chinese Journal of Intelligent Science and Technology, 2024, 6(4): 482-494.

HIMELEIN-WACHOWIAK M, GIORGI S, DEVOTO A, et al. Bots and misinformation spread on social media: implications for COVID-19[J]. Journal of Medical Internet Research, 2021, 23(5).

许灵毓, 钟义信, 陈志成. 社交机器人对社会舆论的影响因素研究[J]. 智能系统学报, 2024, 19(1): 122-131.

XU L Y, ZHONG Y X, CHEN Z C. Research on the influence factors of social robots on social opinions[J]. CAAI Transactions on Intelligent Systems, 2024, 19(1): 122-131.

倪清桦, 鲁越, 林飞, 等. 平行音乐: 大模型时代的人机混合音乐创演[J]. 智能科学与技术学报, 2024, 6(2): 150-163.

NI Q H, LU Y, LIN F, et al. Parallel music: human-machine hybrid music creation and performance in the era of large models[J]. Chinese Journal of Intelligent Science and Technology, 2024, 6(2): 150-163.

黄峻, 林飞, 杨静, 等. 生成式AI的大模型提示工程: 方法、现状与展望[J]. 智能科学与技术学报, 2024, 6(2): 115-133.

HUANG J, LIN F, YANG J, et al. From prompt engineering to generative artificial intelligence for large models: the state of the art and perspective[J]. Chinese Journal of Intelligent Science and Technology, 2024, 6(2): 115-133.

田永林, 王兴霞, 王雨桐, 等. RAG-PHI: 检索增强生成驱动的平行人与平行智能[J]. 智能科学与技术学报, 2024, 6(1): 41-51.

TIAN Y L, WANG X X, WANG Y T, et al. RAG-PHI: searching for parallel people and parallel intelligence driven by enhanced generation[J]. Chinese Journal of Intelligent Science and Technology, 2024, 6(1): 41-51.

GU S K, YIN Y J, GONG Q Y, et al. A large-scale dataset of interactions between weibo users and platform-empowered LLM agent[C]//Proceedings of the 34th ACM International Conference on Information and Knowledge Management. New York: ACM Press, 2025: 6392-6396.

GAO C, LAN X C, LU Z H, et al. S3: social-network simulation system with large language model-empowered agents[EB]. 2023.

GAO Y, ZHANG M M, LYSYAKOV M. Does social bot help socialize? evidence from a microblogging platform[J]. Information Systems Research, 2025.

WANG B, HE W Y, ZENG S L, et al. Unveiling privacy risks in LLM agent memory[C]//Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Stroudsburg: ACL, 2025: 25241-25260.

BANG Y J, CHEN D L, LEE N, et al. Measuring political bias in large language models: what is said and how it is said[C]//Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Stroudsburg: ACL, 2024: 11142-11159.

DANRY V, PATARANUTAPORN P, GROH M, et al. Deceptive explanations by large language models lead people to change their beliefs about misinformation more often than honest explanations[C]//Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems. New York: ACM Press, 2025: 1-31.

PENG M, CHEN N, TANG J H, et al. How does misinformation affect large language model behaviors and preferences?[C]//Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Stroudsburg: ACL, 2025: 13711-13748.

SU Z, ZHANG J, QU X Y, et al. CONFLICTBANK: a benchmark for evaluating knowledge conflicts in large language models[C]//Proceedings of the 38th International Conference on Neural Information Processing Systems. 2024: 103242-103268.

HASHER L, GOLDSTEIN D, TOPPINO T. Frequency and the conference of referential validity[J]. Journal of Verbal Learning and Verbal Behavior, 1977, 16(1): 107-112.

MCGUIRE W J. Resistance to persuasion conferred by active and passive prior refutation of the same and alternative counterarguments[J]. The Journal of Abnormal and Social Psychology, 1961, 63(2): 326-332.

STANOVICH K E, WEST R F. Individual differences in reasoning: implications for the rationality debate?[J]. Behavioral and Brain Sciences, 2000, 23(5): 645-665.

LIU Y H, LIU Y X, ZHANG X Q, et al. The truth becomes clearer through debate! multi-agent systems with large language models unmask fake news[C]//Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York: ACM Press, 2025: 504-514.

LI X Y, ZHANG Y X, MALTHOUSE E C. Large language model agent for fake news detection[EB]. 2024.

PENNYCOOK G, RAND D G. Cognitive reflection and the 2016 U.S. presidential election[J]. Personality and Social Psychology Bulletin, 2019, 45(2): 224-239.

BROWN T, MANN B, RYDER N, et al. Language models are few-shot learners[J]. Advances in Neural Information Processing Systems, 2020, 33: 1877-1901.

WEI J, WANG X Z, SCHUURMANS D, et al. Chain-of-thought prompting elicits reasoning in large language models[C]//Proceedings of the 36th International Conference on Neural Information Processing Systems (NeurIPS). Red Hook: Curran Associates, 2022: 24824-24837.

LI N, GAO C, LI M Y, et al. EconAgent: large language model-empowered agents for simulating macroeconomic activities[C]//Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Stroudsburg: ACL, 2024: 15523-15536.

AHER G V, ARRIAGA R I, KALAI A T. Using large language models to simulate multiple humans and replicate human subject studies[C]//Proceedings of the International Conference on Machine Learning. New York: PMLR, 2023: 337-371.

PARK J S, O'BRIEN J, CAI C J, et al. Generative agents: interactive simulacra of human behavior[C]//Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology. New York: ACM Press, 2023: 1-22.

SHEN Y F, ZHAO Z H, XUE X, et al. A framework for analyzing abnormal emergence in service ecosystems through LLM-based agent intention mining[C]//Proceedings of the 2025 IEEE International Conference on Web Services (ICWS). Piscataway: IEEE Press, 2025: 484-490.

WANG L, MA C, FENG X Y, et al. A survey on large language model based autonomous agents[J]. Frontiers of Computer Science, 2024, 18(6): 186345.

PIAO J H, LU Z H, GAO C, et al. Social bots meet large language model: political bias and social learning inspired mitigation strategies[C]//Proceedings of the ACM on Web Conference 2025. New York: ACM Press, 2025: 5202-5211.

WANG Z H, CAI S F, CHEN G Z, et al. Describe, explain, plan and select: interactive planning with LLMs enables open-world multi-task agents[C]//Proceedings of 37th Conference on Neural Information Processing Systems (NeurIPS). Red Hook: Curran Associates, 2023, 36: 34153-34189.

FARQUHAR S, KOSSEN J, KUHN L, et al. Detecting hallucinations in large language models using semantic entropy[J]. Nature, 2024, 630(8017): 625-630.

KIM S, YUN S, LEE H, et al. Propile: probing privacy leakage in large language models[C]//Proceedings of the 37th Conference on Neural Information Processing Systems (NeurIPS). Red Hook: Curran Associates, 2023: 20750-20762.

李亚玲, 蔡京京, 柏洁明. 生成式大模型引发的隐私风险及治理路径[J]. 智能科学与技术学报, 2024, 6(3): 394-401.

LI Y L, CAI J J, BAI J M. Privacy risks induced by generative large language models and governance paths[J]. Chinese Journal of Intelligent Science and Technology, 2024, 6(3): 394-401.

CHEN C Y, SHU K. Combating misinformation in the age of LLMs: opportunities and challenges[J]. AI Magazine, 2024, 45(3): 354-368.

LIN S, HILTON J, EVANS O. TruthfulQA: measuring how models mimic human falsehoods[C]//Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Stroudsburg: ACL, 2022: 3214-3252.

PEREZ E, RINGER S, LUKOSIUTE K, et al. Discovering language model behaviors with model-written evaluations[C]//Proceedings of the Findings of the Association for Computational Linguistics (ACL) 2023. Stroudsburg: ACL, 2023: 13387-13434.

GE Y B, KIRTANE N, PENG H, et al. LLMs are vulnerable to malicious prompts disguised as scientific language[EB]. 2025.

ZHANG H N, DIAO S Z, LIN Y, et al. R-tuning: instructing large language models to say 'I don't know'[C]//Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). Stroudsburg: ACL, 2024: 7113-7139.

LI M, CHEN L C, CHEN J H, et al. Selective reflection-tuning: student-selected data recycling for LLM instruction-tuning[C]//Proceedings of the Findings of the Association for Computational Linguistics (ACL) 2024. Stroudsburg: ACL, 2024: 16189-16211.

CHEN Y, HU J Y, XIAO Y, et al. Understanding the user behavior of foursquare: a data-driven study on a global scale[J]. IEEE Transactions on Computational Social Systems, 2020, 7(4): 1019-1032.

BREHM J W. A theory of psychological reactance[M]. New York: Academic Press, 1966.

ZHANG X N, LIN J Y, SUN L B, et al. Electionsim: massive population election simulation powered by large language model driven agents[EB]. 2024.

YANG Z Y, ZHANG Z B, ZHENG Z R, et al. Oasis: open agent social interaction simulations with one million agents[EB]. 2024.

DAWES J. Do data characteristics change according to the number of scale points used? An experiment using 5-point, 7-point and 10-point scales[J]. International Journal of Market Research, 2008, 50(1): 61-104.

WANG P, ANGARITA R, RENNA I. Is this the era of misinformation yet: combining social bots and fake news to deceive the masses[C]//Proceedings of the Web Conference 2018-WWW'18. New York: ACM Press, 2018: 1557-1561.

SHAO C C, CIAMPAGLIA G L, VAROL O, et al. The spread of low-credibility content by social bots[J]. Nature Communications, 2018, 9: 4787.

Team Q. Qwen2 technical report[EB]. 2024.

DUBEY A, JAUHRI A, PANDEY A, et al. The llama 3 herd of models[EB]. 2024.

GLM T, ZENG A H, XU B, et al. ChatGLM: a family of large language models from GLM-130B to GLM-4 all tools[EB]. 2024.

ORLANDO G M, LA GATTA V, RUSSO D, et al. Can generative agent-based modeling replicate the friendship paradox in social media simulations?[C]//Proceedings of the 17th ACM Web Science Conference 2025. New York: ACM Press, 2025: 510-515.

WU J Y, GUO J F, HOOI B. Fake news in sheep's clothing: robust fake news detection against LLM-empowered style attacks[C]//Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York: ACM Press, 2024: 3367-3378.

袁唯淋, 赵卫伟, 胡振震, 等. 智能情报融合综述: 对抗视角下的开源情报融合分析[J]. 智能科学与技术学报, 2024, 6(3): 284-300.

YUAN W L, ZHAO W W, HU Z Z, et al. Research on intelligence fusion: a holistic analysis of open-source intelligence fusion from the perspective of confrontation[J]. Chinese Journal of Intelligent Science and Technology, 2024, 6(3): 284-300.

陈君海, 项凤涛, 黎拓新, 等. 融合证据分析的贝叶斯神经网络虚假信息检测方法[J]. 智能科学与技术学报, 2025, 7(3): 316-328.

CHEN J H, XIANG F T, LI T X, et al. Evidence-aware Bayesian neural networks for fake news detection[J]. Chinese Journal of Intelligent Science and Technology, 2025, 7(3): 316-328.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

开源社区中项目热度阶段预测方法研究——以大语言模型项目为例

RadiFlow：AI Agents重构放射科工作流程

融合知识图谱和大语言模型的罗氏沼虾养殖问答系统的设计与实现

基于ACP和LLM犯罪学研究进展：综述与展望

Cerberus：基于深度学习的跨站社交网络机器人检测系统