数字政府建设引入预训练语言模型的风险与克服  

The Risks and Overcoming of Introducing Pre-trained Language Models in Digital Government Construction

作　　者：曹冬英 Cao Dongying

机构地区：[1]云南师范大学历史与行政学院,昆明650500

出　　处：《中国高校社会科学》2025年第2期101-107,159,共8页Social Sciences in Chinese Higher Education Institutions

基　　金：国家社科基金后期资助项目“数字政府治理研究”(19FGLB032);福建省习近平新时代中国特色社会主义思想研究中心项目“从‘数字福建’到‘数字中国’的治理变迁研究”(FJ2022XZB026)阶段性成果。

摘　　要：数字政府建设过程中引入预训练语言模型(PTMLs)能够推动实现公共服务主客体之间的沟通升级、公共服务流程的多维完善、公共服务相关决策的有效优化,提升公共服务水平。引入预训练语言模型可能引发公共服务客体数据权利受损、公共服务伦理挑战变大以及责任主体模糊化等风险。未来,应以个体权利为核心加强公共服务客体保护,以增强价值理性为目标优化公共服务规制,以权责清晰为基点实现公共服务均等化。The introduction of pre trained language models(PTMLs)in the process of digital government construction can promote the upgrading of communication between the subject and object of public services,the multidimensional improvement of public service processes,and the effective optimization of public service related decisions,thereby promoting the overall improvement of the level of public services.The introduction of pre trained language models may lead to risks such as damage to the data rights of public service objects,increased ethical challenges in public services,and blurring of responsibility subjects.We should strengthen the protection of public service objects with individual rights as the core,optimize public service regulations with the goal of enhancing value rationality,and achieve equalization of public services based on clear rights and responsibilities.

关 键 词：预训练语言模型 数字政府建设 数字时代公共服务理论 

分 类 号：D630[政治法律—政治学]