|
| 文章编号：1672-3104(2025)04-0143-10 |
| |
|
预训练语言模型引入数字政府建设的风险应对——基于行政关系的视角 |
| |
曹冬英 |
| |
|
（云南师范大学历史与行政学院，云南昆明，650500） |
| |
| 摘 要： 从行政主客体关系的角度来看，将预训练语言模型引入数字政府建设能够推动行政主体与相对人之间的沟通升级、促进行政行为内外流程的多维完善、提高行政决策的理性水平。在正视预训练语言模型带来的正面效能的同时，需要清醒地认识到它可能引发的行政相对人的数据权利被侵扰、行政伦理基本原则面临挑战、行政责任主体模糊不明等风险。为了应对风险，行政主体应当以权利保障为核心落实相关法律法规、以价值理性为取向构建伦理保障机制、以权责清晰为基点优化责任结构关系。 |
| |
| 关键词： 预训练语言模型；数字政府；行政关系；ChatGPT；数字治理 |
| |
|
| |
Risk management of introducing pre-trained language models into digital government construction: From the perspective of administrative relationships |
| |
CAO Dongying |
| |
（College of History and Administration, Yunnan Normal University, Kunming 650500, China） |
| |
| Abstract: From the perspective of the relationship between administrative subjects and objects, introducing pre-trained language models into the construction of digital government can promote the communication upgrading between administrative subjects and their counterparts, facilitate the multi-dimensional improvement of internal and external processes of administrative actions, and enhance the rational level of administrative decision-making. While facing up to the positive effects brought by pre-trained language models, it is necessary to clearly recognize the risks that they may cause, such as the infringement of the data rights of administrative counterparts, the challenge to the basic principles of administrative ethics, and the ambiguity of the administrative responsibility subjects. To deal with risks, administrative entities should implement relevant laws and regulations with rights protection as the core, construct ethical guarantee mechanisms with value rationality as the orientation, and optimize the responsibility structure relationship with clear rights and responsibilities as the starting point. |
| |
Key words: pre-trained language model; digital government; administrative relations; ChatGPT; digital governance |
| |
|
