摘要
生成式人工智能(Generative Artificial Intelligence)技术的崛起,为社会科学研究方法的扩展带来了契机。基于变换器(Transformer)架构的大语言模型由于经过大量预训练能够更有效地理解和处理复杂的上下文信息,吸引了众多社会科学家的关注。社会科学家对生成式人工智能应用的探索,一方面集中于其在数据分析、充当实验中介和协助写作等研究过程中的效果;另一方面关注生成式人工智能所具有的模拟人类对象的能力。论文以偏见研究为具体案例,从“生成式人工智能作为研究工具”和“生成式人工智能作为研究对象”两个方面回顾和探讨了社会科学如何调整自身的研究实践以面对大语言模型应用的趋势,并总结生成式人工智能给社会科学研究方法带来的助益。
The emergence of Generative Artificial Intelligence(GAI)technology presents an opportunity for expanding social science research methods.Transformer-based large language models(LLMs),distinguished by their ability to understand and process complex,contextual information,have attracted significant interest from social scientists.Applications of GAI in social sciences focus on two aspects:its effectiveness in research processes such as data analysis,experimental mediation,and writing assistance,and its ability to simulate human subjects.Taking social bias as an exemplary research area,this paper reviews and discusses how social science researchers adopt novel methodological practices to address the trend of GAI applications from two perspectives:using GAI as an instrumental agent and as research subject.The paper also summarizes both the benefits and shortcomings of using GAI in social science research,then discusses its potential in revolutionizing knowledge production.
作者
卢云峰
黄星尧
吴语菲
周旅军
LU Yunfeng;HUANG Xingyao;WU Yufei;ZHOU Lyujun
出处
《贵州民族大学学报(哲学社会科学版)》
2025年第3期70-91,共22页
Journal of Guizhou Minzu University(Philosophy and Social Sciences)
基金
北京大学武汉人工智能研究院开放课题“数字社会的核心特征与治理模式研究:以武汉光谷为例”阶段性成果。
