社交媒体空间公众大模型认知:主题、态度与传播  

Public Perception of Large Language Models on Social Media:Topics,Attitudes,and Communication

作　　者：孟天广 张静[1] 曹迥仪 MENG Tian-guang;ZHANG Jing;CAO Jiong-yi(School of Social Sciences,Tsinghua University,Beijing 100084,China;Laboratory of Computational Social Science and State Governance,Tsinghua University,Beijing 100084,China)

机构地区：[1]清华大学社会科学学院,北京100084 [2]清华大学计算社会科学与国家治理实验室,北京100084

出　　处：《苏州大学学报（哲学社会科学版）》2024年第5期181-190,共10页Journal of Soochow University(Philosophy & Social Science Edition)

基　　金：国家社会科学基金重大项目“基于大数据的智能化社会治理监测、评估与应对策略研究”(项目编号:18ZDA110);清华大学文科自主科研计划(项目编号:2022THZWJC09)的阶段性成果。

摘　　要：近年来,以ChatGPT为代表的大语言模型引发社会公众对人工智能的高度关注,进而在社交媒体空间逐步形成大模型认知。本文基于56万余条微博原创文本和529万余条转发微博数据,分析了社交媒体空间公众大模型认知,呈现了社会公众对大模型的关注主题、态度偏好及其传播机制。研究发现社交媒体空间中对大模型的主题关注集中在经济类议题,不同用户的主题关注偏好具有明显差异。社会公众对大模型技术呈现出明显的技术乐观倾向,技术乐观程度因用户身份和主题关注而异。大模型认知传播网络呈现出区隔化特征,财经类自媒体和机构媒体主导传播网络,进而影响普通用户构成的大众传播网络,体现出社交媒体中大模型认知的差异化传播路径。In recent years,large language models have attracted more and more social attention.Based on 560,000 original Weibo posts and 5.29 million Weibo reposts,this paper analyzes the public perception of large language models on social media,including hot topics,attitudes,and communication mechanisms.The study finds that economic issues attract the most attention and users with different identity tags have noticeable differences in thematic preferences.The public expresses optimistic sentiments towards large language models,with the degree of optimism varying by user identity tags and thematic preferences.Our results reveal the differentiated communication network of large language model discussions on social media.The communication network is highly segmented,with financial self-media accounts and institutional media accounts dominating the communication network.

关 键 词：大模型 社交媒体 认知传播 传播机制 社会治理 

分 类 号：G206[文化科学—传播学]