摘要
生成式人工智能的快速发展正在推动社会科学研究范式的革新。以社会科学理论与经验知识为基础,将“大模型+智能体”生成的数据作为有意义的事实数据进行研究的第五范式——智能社会研究范式正在兴起。在技术原理方面,这一新兴范式基于生成式人工智能的两大核心技术突破:一是大模型所代表的从“模式识别”到“模拟世界”的“机器学习革命”;二是大模型驱动的智能体所代表的从解决“专项任务”到求解“通用问题”的“通用人工智能萌芽”。在知识逻辑方面,该范式深度融合并赋能定量、定性、社会仿真与大数据这四大传统范式,以智能生成的数据克服传统数据的碎片化、偏差性与获取壁垒,并通过理论引导的交互机制拓展了知识发现可利用的事实空间。在应用路径方面,通过构建代表各类人类行为体与社会构成的智能社会模型,可以精确模拟全类社会系统,从而揭示复杂社会现象的涌现规律;科学发现的通用执行则赋能从文献综述、假设生成、数据采集与分析、理论检验与发现的“科学环”全流程,显著提升了社会科学研究的效能。
The rapid development of generative artificial intelligence is driving a paradigm shift in social sciences research.A new fifth paradigm,Intelligent Social Science(ISS)paradigm,is emerging,which is based on social science theories and empirical knowledge and treats the generated data from“large models+AI agents”as meaningful factual data for research.Technologically,this new paradigm is based on two major breakthroughs in generative AI:the“machine learning revolution”represented by large models,which moves from“pattern recognition”to“world simulation”;and the“germ of general artificial intelligence”driven by AI Agents,where LLMs powered intelligent agents evolve from solving“specific tasks”to tackling“general problems”.Logically,this paradigm deeply integrates and empowers the four traditional paradigms:quantitative,qualitative,social simulation,and big data.It overcomes the fragmentation,bias,and access barriers of traditional data through AI-generated data and expands the factual space for knowledge discovery through theory-guided interactive mechanisms.In application,precise simulation of universal social systems can be achieved by constructing intelligent social models representing various human actors and social structures,revealing the emergence laws of complex social phenomena;and the general implementation of scientific discovery empowers the entire“scientific cycle”process,from literature review,hypothesis generation,data collection and analysis to theory test and discovery,significantly enhancing the efficiency of social science research.
作者
王震宇
杨曦
张润泽
Wang Zhenyu;Yang Xi;Zhang Runze
出处
《智能社会研究》
2025年第1期152-176,243-244,共27页
关键词
智能社会科学研究
范式革新
生成式人工智能
大模型
智能体
intelligent social sciences
paradigm shift
generative artificial intelligence
large models
AI agents
