知识复杂性的测度方法综述与比较

A Review and Comparison of Measurement Methods of Knowledge Complexity

[研究目的]复杂知识是突破关键核心技术的基础支撑,也是企业或区域建立竞争优势的关键要素。该文旨在全面把握知识复杂性测度方法的研究现状与前沿进展,为知识复杂性的深入研究提供测度理论基础。[研究方法]通过梳理国内外知识复杂性研究的相关文献,详细介绍三种主要的知识复杂性测度方法的核心思想、理论基础和计算过程,并进一步对比分析不同方法的适用层次和优缺点,最后提出未来研究方向。[研究结论]反射法主要基于知识的空间分布异质性间接衡量知识复杂性,虽然原理简单且操作相对容易,但无法揭示知识复杂性的本质特征;元素依赖性指数利用知识组合频次衡量知识元素之间的相互依赖程度,进而反映知识复杂性,忽视了其他影响知识组合的因素;结构多样性指数通过知识组合网络拓扑结构的多样性反映知识复杂性,综合考虑多个网络结构指标,并经过大量数值模拟检验,但存在估计结果不稳定、仅适用于二值网络等问题。

[Research purpose]Complex knowledge is the basic support for breakthroughs in key core technologies,and it is also a key element for enterprises or regions to establish competitive advantages.This paper aims to fully grasp the research status and cutting-edge progress of knowledge complexity measurement methods,so as to provide a theoretical basis for the in-depth study of knowledge complexity.[Research method]By combing the relevant literature on knowledge complexity research at home and abroad,we introduce the core ideas,theoretical basis and calculation process of three main knowledge complexity measurement methods in detail,and further compare and analyze the applicable levels,advantages and disadvantages of different methods,and finally propose future research directions.[Research conclusion]The reflection method is mainly based on heterogeneity of spatial distribution of knowledge to indirectly measure knowledge complexity.Although the principle is simple and the operation is relatively easy,it cannot reveal the essential characteristics of knowledge complexity;the element dependence index mainly uses knowledge combination frequency to measure the degree of interdependence and further reflects knowledge complexity,but it ignores other components that affect knowledge combination;the structural diversity index reflects knowledge complexity through the diversity of topological structures of knowledge combination network,which comprehensively considers multiple network structure indicators,and has been tested by a large number of numerical simulations,but there are also problems such as unstable estimation results and only applicable to binary networks.