农业图书情报学报 ›› 2026, Vol. 38 ›› Issue (10): 15-24.doi: 10.13998/j.cnki.issn1002-1248.26-0634

• 人工智能应用场景与可操作治理专题 •    下一篇

自主智能体嵌入下智能制造企业数据资产生命周期韧性提升机制研究

黄磊1,2, 胡锦1(), 秦胜君1,2   

  1. 1. 广西科技大学 经济与管理学院,柳州 545006
    2. 广西科技大学 智能决策系统分析与技术评估实验室,柳州 545006
  • 收稿日期:2026-09-11 出版日期:2026-10-05 发布日期:2026-11-10
  • 通讯作者: 胡锦
  • 作者简介:

    黄磊(1984- ),男,副教授,广西科技大学经济与管理学院,研究方向为人工智能治理、大数据治理、技术评估

    秦胜君(1984- ),女,副教授,广西科技大学经济与管理学院,研究方向为智能决策、大数据分析、复杂系统

  • 基金资助:
    广西教育厅教学改革课题“AIGC赋能经管类课程教学质量多维提升创新研究与实践”(2025JGA260); 广西科技大学博士基金项目“广西人工智能基础研究知识扩散动力机制研究”(24S08)

Optimization Mechanism of Data Asset Lifecycle Resilience for Smart Manufacturing Enterprises in the Context of Autonomous Agent Embeddedness

HUANG Lei1,2, HU Jin1(), QIN Shengjun1,2   

  1. 1. School of Economics and Management, Guangxi University of Science and Technology, Liuzhou 545006
    2. Laboratory for Intelligent Decision-Support Systems Analysis and Technology Assessment, Guangxi University of Science and Technology, Liuzhou 545006
  • Received:2026-09-11 Online:2026-10-05 Published:2026-11-10
  • Contact: HU Jin

摘要:

[目的/意义] 自主智能体深度嵌入智能制造企业数据资产生成与流转全过程,在提升价值创造效率的同时引发数据主权、决策权归属与责任追溯等治理难题,增值与风险矛盾凸显,亟须一体化的韧性治理框架。 [方法/过程] 借鉴ISO 55013:2024《资产管理:数据资产管理指南》数据全生命周期管理标准,试将自主智能体嵌入下数据资产生命周期划分为规划、采集、存储、加工、流通、应用、处置7个阶段,综合运用文献研究与逻辑推演方法构建韧性治理理论框架。研究拟提出价值增益与治理摩擦并存的双重效应框架,并将其划分为净增值、侵蚀、中性三维价值状态类型;界定自主性边界模糊、决策权争夺、责任归属困境3类治理摩擦及其沿生命周期的阶段分布;构建人机系统互嵌模式象限,设计被动监督、主动审核、协同决策3层人在回路(HITL)机制,使之与效率、决策、合规三维韧性分层匹配,并推导出4个理论命题。 [结果/结论] 研究表明,HITL机制的分层匹配与动态适配是提升智能制造企业数据资产生命周期韧性的有效组织机制,旨在为数据治理与AI治理的融合补充具有可操作性的理论分析框架。

关键词: 人工智能治理, 智能制造, 数据资产, 生命周期, 韧性机制

Abstract:

[Purpose/Significance] Autonomous agents are evolving from analytical tools into quasi-operational industry stakeholders. In addition, autonomous agents are becoming deeply embedded in the entire process of data asset generation and circulation within smart manufacturing enterprises. While enhancing the efficiency of data collection, processing, and application, autonomous agents also give rise to governance challenges. For instance, autonomous agents cause a blurring of data sovereignty, contested attribution of decision rights, and difficulties in accountability tracing. However, there is a governance dilemma of balancing data value generation and governance risk. The precondition and evaluation mechanism of data asset lifecycle resilience demands a comprehensive theoretical framework with the AI governance principle of human-in-the-loop (HITL). Smart manufacturing provides a large number of application scenarios characterized by intensive human-machine collaboration and complex intelligent decision-making chains. Hence, this study intends to counter the dilemma by addressing the following research question: What mechanism needs to be constructed to ease the conflict between data asset lifecycle resilience and AI governance of smart manufacturing in the context of autonomous agent embeddedness. [Method/Process] Based on ISO 55013:2024, Asset management—Guidance on the management of data assets, and following the stage delineations of DAMA-DMBOK2 and the Pilot Program for Whole-Process Management of Data Assets, this study aims to construct a hybrid conceptual framework for the data asset lifecycle with human-machine interaction. The hybrid conceptual framework divides the data asset lifecycle in the context of autonomous agent embeddedness into seven stages: planning, collection, storage, processing, circulation, application, and disposal. By applying a systematic literature review, this study constructs a theoretical model of smart manufacturing data asset lifecycle resilience governance in the context of autonomous agent embeddedness. The theoretical model integrates basic principles of socio-technical systems theory, principal-agent theory, dynamic capabilities theory, and contingency theory. Hence, this study intends to develop a theoretical base for further empirical study in the domain of data governance and AI governance of smart manufacturing. [Results/Conclusions] First, value gains and governance conflicts coexist at every stage of the smart manufacturing data asset lifecycle. Data assets transition dynamically among the three types of value states, including net gain, erosion, and neutrality. Second, three types of conflict persist throughout the entire smart manufacturing data asset lifecycle, and propagate along pathways of data quality deterioration, decision delay, and increasing compliance risks. Third, the HITL intervention framework and the multi-phase structure of smart manufacturing data asset lifecycle resilience are structurally isomorphous. The degree of HITL intervention decreases as the level of automation of data asset generation increases. Meanwhile, the irreplaceability of HITL intervention at key points in the lifecycle increases when the application of autonomous agents increases. The design of the HITL intervention mechanism dynamically depends on critical factors including task risks, the level of automation, and governance maturity. Consequently, this study identified that the governance performance of smart manufacturing data asset lifecycle resilience can be optimized by applying the HITL intervention with features of dynamic adaptation and stratified matching. The study of the HITL intervention framework in the data asset lifecycle resilience governance of smart manufacturing provides a theoretical basis for further academic discussion.

Key words: artificial intelligence governance, smart manufacturing, data asset, lifecycle, resilience mechanism

中图分类号:  F272.7,TP18

引用本文

黄磊, 胡锦, 秦胜君. 自主智能体嵌入下智能制造企业数据资产生命周期韧性提升机制研究[J]. 农业图书情报学报, 2026, 38(10): 15-24.

HUANG Lei, HU Jin, QIN Shengjun. Optimization Mechanism of Data Asset Lifecycle Resilience for Smart Manufacturing Enterprises in the Context of Autonomous Agent Embeddedness[J]. Journal of library and information science in agriculture, 2026, 38(10): 15-24.