人工智能风险体系与模块化评价指标构建实证研究  

Construction and Empirical Study of Modular Index System for Measuring Risk Level of Artificial Intelligence

作　　者：苏文成[1,2] 洪舒悦 卢章平[1,2] 刘桂锋[1,2] 潘颖[1,2] Su Wencheng;Hong Shuyue;Lu Zhangping;Liu Guifeng;Pan Ying(Institute of Science and Technology Information,Jiangsu University,Zhenjiang 212013;Jiangsu University Library,Zhenjiang 212013)

机构地区：[1]江苏大学科技信息研究所,镇江212013 [2]江苏大学图书馆,镇江212013

出　　处：《情报杂志》2025年第1期136-145,154,共11页Journal of Intelligence

基　　金：国家社会科学基金青年项目“具身认知视阈下图书馆空间重塑机理与路径研究”(编号:22CTQ015)研究成果。

摘　　要：[研究目的]在人工智能蓬勃发展的时代,开展人工智能风险体系构建及其对应的模块化评价指标研究,以全面反映AI发展所带来的潜在风险,为不同领域的研究者提供决策依据。[研究方法]首先,基于扎根理论和专家调查法,系统地抽取和设计模块化人工智能风险评价指标;其次,采用层次分析法,对各指标权重进行测算,形成量化后的得分表,并对风险等级进行划分;最后,借助问卷调查法和专家调查法,对该体系进行应用实证研究,以验证其科学性及可行性。[研究结果/结论]人工智能风险体系包含由9个一级指标和41个二级指标组成的模块化评价指标。该体系涵盖技术、伦理、法律和社会等多个维度。基于此体系可以针对某一人工智能事件或功能,使用主观评价的方法对每一个风险子项进行打分,并最终得出总分。模块化指标可以灵活拆分,以量化测度不同领域的AI风险,兼顾全维度与具象维度的AI风险测度问题。[Research purpose]In the era of vigorous development of artificial intelligence,research on the construction of artificial intelligence risk system and its corresponding modular evaluation index is carried out to fully reflect the potential risks brought by the development of AI and provide decision-making basis for researchers in different fields.[Research method]Firstly,based on grounded theory and expert investigation,the risk assessment indicators of modular artificial intelligence are systematically extracted and designed.Secondly,analytic hierarchy process(AHP)is used to calculate the weight of each index,form a quantified score table,and divide the risk level;Finally,utilizing questionnaire survey and expert survey,this paper makes an empirical study on the application of this system to verify its scientificity and feasibility.[Research result/conclusion]The risk system of artificial intelligence includes 9 first-level indicators and 41 second-level indicators.The system covers many dimensions such as technology,ethics,law and society.Based on this system,each risk sub-item can be scored by subjective evaluation method for an artificial intelligence event or function,and finally the total score can be obtained.The existence of modular indicators can be flexibly split to quantitatively measure the AI risk in different fields,taking into account the AI risk measurement problem in both full and concrete dimensions.

关 键 词：人工智能 风险体系 风险评估 模块化评价指标 扎根理论 层次分析法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]