类ChatGPT大模型赋能数字政府的合比例性控制  

ChatGPT-Like Big Model Enabling Proportionality Control in Digital Government

作　　者：付鉴宇 Fu Jianyu(School of Administrative Law,Northwest University of Political Science and Law,Xi'an 710063,China;Collaborative Innovation Center for Building the Rule of Law,Northwest University of Political Science and Law,Xi'an 710063,China)

机构地区：[1]西北政法大学行政法学院,西安710063 [2]西北政法大学法治陕西建设协同创新中心,西安710063

出　　处：《科技与法律(中英文)》2025年第2期63-74,共12页Science Technology and Law(Chinese-English Version)

基　　金：陕西省社会科学基金年度项目“行政法治视角下陕西数字经济的软法治理研究”(2023E046);陕西省教育厅重点项目“法治陕西建设背景下数字经济软法的功能定位及法治完善研究”(23JY077);国家社会科学基金西部项目“国家网络安全审查法律制度研究”(22XFX021)。

摘　　要：ChatGPT展示出生成式人工智能新高度,有望更实质深层地赋能数字政府进阶建设,但也必须预警其潜在风险“。限制公权”与“保障私权”的平衡统一是行政法治的核心命题,类ChatGPT大模型赋能数字政府则容易产生多种表征的公权力“控制性危机”和私权利“限制性危机”。比例原则是裁量治理的首要原则,三项子原则能够充分支撑类ChatGPT大模型赋能数字政府的民主行政、科学行政和依法行政三维目标,与行政法治具有紧密的内在关联,递归实现“权力—权利”关系的均衡配置。在具体的实践路径上,可依据生成式人工智能的独特技术机理,遵循“训练与输入”及“生成与输出”两阶分析框架,通过嵌入事前论证、清单管理和动态规制等系列制度机制,分别落实适当性原则、最小损害原则与均衡性原则,以实现对类ChatGPT大模型赋能数字政府的合比例性控制。ChatGPT demonstrates a new level of generative artificial intelligence,which is expected to empower the ad-vanced construction of digital government in a more substantive and deeper way,but it is also necessary to warn of its potential risks.The balance and unity of"limiting public power"and"safeguarding private rights"is the core proposi-tion of the rule of law in administration,and ChatGPT-like big model empowered digital government is prone to pro-duce multiple manifestations of public power"controlling crisis"and private rights"restrictive crisis".The principle of proportionality is the first principle of discretionary governance,the three sub-principles can fully support the Chat-GPT big model empowered digital government of democratic administration,scientific administration and administra-tion in accordance with the law of the three-dimensional objectives,and the rule of law and the administrative rule of law has a close intrinsic relationship,recursively realize the"power-right"relationship balanced configuration In the specific practical path,based on the unique technical mechanism of generative artificial intelligence,a two-stage ana-lytical framework:"pre-training and input stage"and"generation and output stage",has been set up.By embedding a series of institutional mechanisms such as prior justification,list management and dynamic regulation,the principles of appropriateness,minimum damage and balance are implemented respectively to realize the proportional control of the ChatGPT-like big model-enabled digital government.

关 键 词：行政法治 类ChatGPT大模型 数字政府 比例原则 

分 类 号：D912[政治法律—法学]