| 751 | 0 | 68 |
| 下载次数 | 被引频次 | 阅读次数 |
随着人工智能技术的普及，自动化决策系统深入行政领域，催生的行政决策自动化实践虽能显著提升行政效能、优化资源分配并改善决策质量，但也可能导致行政决策偏离法治原则，甚至违背权力制约的行政理念。行政决策自动化运行流程可被系统划分为数据样本构建、决策模型生成、人工复核检验及决策结果应用四个关键阶段，运行全过程环环相扣。行政决策自动化算法偏差主要存在制度调试的优化迟滞、训练数据偏差的质效症候、技术失灵的叠加效应、系统桎梏的算法黑箱等症结，对社会结构、伦理秩序等造成深层的冲击。为了解决算法偏差在可解释性、可追溯性、社会价值等方面造成的负面影响，需提升法治兼容性、筑牢公信力基石、保障系统合法性。
Abstract：With the popularization of artificial intelligence technology, automated decision-making systems have deeply penetrated the administrative field. While these practices can significantly enhance administrative efficiency, optimize resource allocation, and improve the quality of decision-making, they may also lead to deviations from the rule of law in administrative decision-making, even violating the administrative principle of power restraint. The operational process of automation of administrative decision-making can be systematically divided into four key stages: data sample construction, decision model generation, manual review and verification, and application of decision results, with each stage interlinked in a cyclical manner. Key issues in automation of administrative decision-making algorithms include delayed optimization in system debugging, quality and efficiency symptoms arising from training data bias, the compounded effects of technical failures, and algorithmic black boxes constrained by system limitations, which exert profound impacts on social structures and ethical order. To address the negative effects of algorithmic bias in terms of explainability, traceability, and social value, it is essential to enhance the compatibility with the rule of law, strengthen the foundation of public credibility, and ensure the legitimacy of the system.
[1]Monarcha-Matlak A.Automated Decision-making in Public Administration[J].Procedia Computer Science,2021,(14).
[2]刘祺，曾昭腾.数字政府建设中的公私合作治理机制建构[J].中国行政管理，2025,(2).
[3]Roehl U B U,Hansen M B.Automated,Administrative Decision-making and Good Governance:Synergies,Trade-offs,and Limits[J].Public Administration Review,2024,(6).
[4]Araujo T,Helberger N,Kruikemeier S,et al.In AI We Trust?Perceptions about Automated Decision-making by Artificial Intelligence[J].AI & society,2020,(3).
[5]Citron D K,Pasquale F.The Scored Society:Due Process for Automated Predictions[J].Wash.L.Rev.,2014,(1).
[6]Viechnicki Peter,William D Eggers.How Much Time and Money Can AI Save Government?Cognitive Technologies Could Free up Hundreds of Millions of Public Sector Worker Hours[M].New York:Deloitte University Press,2017.
[7]韩春晖，刘雅琦.算法行政中法律适用的范式革新与风险因应[J].江苏行政学院学报，2025,(4).
[8]朱峥.自动化行政中的程序裁量及其规范构造[J].比较法研究，2025,(3).
[9]杜雨钊，唐力.人工智能辅助司法裁判的正当性边界与制度约束[J].新疆大学学报(哲学社会科学版),2025,(3).
[10]李晓睿，郭婕.生成式人工智能嵌入数字政府建设的价值、风险与应对[J].科技智囊，2025,(7).
[11]赵鹏，张硕.论自动化行政决定的说明义务[J].山西大学学报(哲学社会科学版),2024,(2).
[12]黄甄铭.大数据视阈下政府决策机制模型建构[J].行政与法，2020,(8).
[13]Parycek P,Schmid V,Novak A S.Artificial Intelligence (AI) and Automation in Administrative Procedures:Potentials,Limitations,and Framework Conditions[J].Journal of the Knowledge Economy,2024,(2).
[14]江必新，王鑫.数字行政行为算法歧视的法律规制[J].学术论坛，2024,(6).
[15]张涛.自动化行政对行政程序的挑战及其制度因应[J].华中科技大学学报(社会科学版),2022,(5).
[16]匡亚林.社会救助自动化决策的运行机理、算法偏差及治理路径[J].社会保障评论，2025,(2).
[17]Allars M.Automated Decision-Making and Review of Administrative Decisions[J].Ga.L.Rev.,2023,(3).
[18]Buoso E.Fully Automated Administrative Acts in the German Legal System[J].European Review of Digital Administration & Law-Erdal,2020,(1-2).
[19]查云飞.人工智能时代全自动具体行政行为研究[J].比较法研究，2018,(5).
[20]沈颖尹.数智社会算法治理的法律规范研究——基于发展与治理平衡视角[J].中国特色社会主义研究，2025,(3).
[21]张凌寒.算法自动化决策与行政正当程序制度的冲突与调和[J].社会科学文摘，2021,(2).
[22]邢露元，沈心怡，王嘉怡.生成式人工智能训练数据风险的规制路径研究[J].网络安全与数据治理，2024,(1).
[23]Washington A L.How to Argue with an Algorithm:Lessons from the COMPAS-ProPublica Debate[J].Colo.Tech.LJ,2018,(1).
[24]Michael K.In this Special Section:Algorithmic bias—Australia's Robodebt and Its Human Rights Aftermath[J].IEEE Transactions on Technology and Society,2024,(3).
[25]吴进进，符阳.算法决策：人工智能驱动的公共决策及其风险[J].开放时代，2021,(5).
[26]Hassija V,Chamola V,Mahapatra A,et al.Interpreting Black-box Models:A Review on Explainable Artificial Intelligence[J].Cognitive Computation,2024,(1).
[27]匡蕾.生成式人工智能的数据安全风险及回应型治理策略[J].网络安全技术与应用，2025,(2).
[28]田野.生成式人工智能辅助政府自动化决策：角色定位及发展路径[J].价格理论与实践，2024,(9).
[29]Lucas Jr H C.Social Security Administration's Progress in Modernizing Its Computer Operations[J].Information & Management,1985,(5).
[30]Burrell J.How the Machine “Thinks”:Understanding Opacity in Machine Learning Algorithms[J].Big Data & Society,2016,(1).
[31]邹开亮，卢筱璇.算法自动化行政处罚决策中相对人程序权利的保障困境与出路[J].长春大学学报，2025,(3).
[32]翟岩，李小波.可知的不可知：算法透明的内涵澄清与“模块化”制度体系构建[J].河南社会科学，2025,(6).
[33]李景豹.智能革命与法理革新：大模型技术对基础法理的挑战及其制度因应[J].理论导刊，2025,(6).
[34]陆卫明，张文，王明.全球人工智能治理从“共识”到“共治”:动力、挑战与路径[J].西安交通大学学报(社会科学版),2025,(2).
基本信息:
DOI：10.13975/j.cnki.gdxz.2025.06.002
中图分类号:TP18;D035;D912.1
引用信息:
[1]匡亚林,矫杨.行政决策自动化的算法偏差及治理路径[J].公共治理研究,2025,37(06):24-37.DOI:10.13975/j.cnki.gdxz.2025.06.002.