学生论文
论文查询结果 |
返回搜索 |
|
论文编号: | 15397 | |
作者编号: | 2120233658 | |
上传时间: | 2025/6/10 1:13:14 | |
中文题目: | AI赋能组织学习与组织绩效提升 ——基于改进NK模型的仿真实验 | |
英文题目: | AI-Empowered Organizational Learning and Organizational Performance Enhancement: A Simulation Experiment Based on an Improved NK Model | |
指导老师: | 林润辉 | |
中文关键字: | 人工智能;集体学习;多样性;网络结构;概率化干预;NK 模型 | |
英文关键字: | Probabilistic AI intervention; Skill diversity; NK model; Friction coefficient; Network Structure; Multi-Agent Simulation; Organizational Learning | |
中文摘要: | 本研究探究了技能多样性、网络结构、任务复杂度、摩擦系数与人工智能介入之间的交互关系对组织学习的影响。现有研究虽分别考察了这些因素的单独作用,但对其交互效应与AI调节作用的系统性理解不足。针对此理论缺口,研究提出“五维交互模型”,旨在揭示AI如何在不同环境中调节多样性、网络结构与任务复杂度的协同效应。 研究整合了集体智能涌现理论、网络知识传递机制、NK模型复杂度理论和信息摩擦理论,构建了多维交互理论框架。基于NK模型开发的多智能体计算实验平台,通过两组系统化实验设计:概率化AI介入实验和确定性AI比例配置实验,检验了多种参数组合对组织学习绩效与系统稳健性的影响。实验变量包括多样性水平(不同收益函数区分)、网络类型(ER/WS/完全图/线性)、任务复杂度(K=0/3/7)、摩擦系数设置(包括高摩擦、默认摩擦和无摩擦)以及AI介入方式(概率化干预频率0.0-0.3或确定性AI比例0%~30%)。 研究发现,在概率化AI介入方面,最优干预频率随任务复杂度非线性变化,呈倒U型关系,中复杂度任务适合中频干预(0.1),高复杂度环境则以低频干预(0.05)为佳。AI介入对多样性效应产生分化影响,在高复杂度环境强化高多样性团队,而在低复杂度环境则更多提升低多样性团队。网络结构与干预频率存在匹配关系,稀疏网络适合高频干预,密集网络适合低频干预。在确定性AI比例配置方面,研究揭示AI比例呈非线性正向影响,5%~10%区间效益最高,与复杂度形成“缓冲效应”,与多样性呈“替代关系”,与网络结构展现“补充效应”。多维分析表明不同网络结构下AI优先改善的维度各异,如线性网络中学习速率提升最明显。尤为重要的是,摩擦系数实验颠覆了“零摩擦即最优”假设,高摩擦环境展现出反直觉优势,其最高收益与最少方案并存,支持“有益摩擦”理论。 本研究的理论贡献在于构建了AI赋能组织学习的多维交互框架,提出并验证了“AI多维增强器”理论和“概率化AI介入”的独特价值。实践层面,研究为组织提供了基于情境的AI配置策略和摩擦管理方法,为优化人机混合系统提供了理论依据与实践指导。 | |
英文摘要: | This study explores how skill diversity, network structure, task complexity, friction coefficients, and AI interventions jointly shape organizational learning. Addressing the limited understanding of their interactions, we propose a Five-dimensional Interaction Model to decode AI's role in moderating these factors. Integrating collective intelligence theory, network dynamics, NK model complexity, and information friction, we developed a computational platform for multi-agent experiments. Two designs were tested: probabilistic AI intervention (frequency: 0.0–0.3) and deterministic AI allocation (proportion: 0%–30%), with variables including diversity (heterogeneous payoff functions), networks (ER/WS/complete/linear), task complexity (K=0/3/7), friction settings, and AI modes. Key findings reveal that:(1)Probabilistic AI: Optimal intervention frequency exhibits an inverted U-shaped relationship with task complexity-0.1 for K=3 vs. 0.05 for K=7. AI amplifies high-diversity teams in complex tasks but benefits low-diversity groups in simple contexts. Sparse networks pair with high-frequency interventions, dense networks with low-frequency.(2)Deterministic AI: 5%–10% AI proportion yields peak performance, showing buffering effects with complexity, substitution with diversity, and complementarity with networks.(3)Friction paradox: High friction unexpectedly outperforms frictionless conditions, supporting "beneficial friction" via constrained exploration.(4)Optimal configuration: Medium diversity, small-world networks, and low-frequency AI (0.05–0.1) balance performance and robustness. The study advances a multidimensional framework for AI-enhanced organizational learning, proposing the "AI Multi-dimensional Enhancer" theory. Practically, it offers context-driven AI deployment strategies and friction management principles for human-AI collaboration. | |
查看全文: | 预览 下载(下载需要进行登录) |