从“可解释”到“可接受”:一种作战决策背景下的人机智能融合新视角  

From Interpretability to Acceptability:A New Perspective of Human-machine Intelligent Fusion under the Background of Combat Decision Making

作　　者：吴冠霖 包卫东[2] 牛莅原 曹江 王吉 WU Guanlin;BAO Weidong;NIU Liyuan;CAO Jiang;WANG Ji(Institute of War Studies,Academy of Military Science of the PLA,Beijing 100091,China;College of Systems Engineering,National University of Defense Technology,Changsha Hunan 410003,China;Graduate School,Academy of Military Science of the PLA,Beijing 100091,China)

机构地区：[1]军事科学院战争研究院,北京100091 [2]国防科技大学系统工程学院,湖南长沙410003 [3]军事科学院研究生院,北京100091

出　　处：《指挥与控制学报》2023年第1期33-44,共12页Journal of Command and Control

基　　金：国家自然科学基金(62102445)资助。

摘　　要：如何在人类-人工智能混合环境中实现人机智能的有效融合,将人的意向性与机的形式化有机结合进而提升决策水平,是作战决策智能化发展最为迫切的问题之一.对当前人机智能融合主要解决思路“可解释性”方法的有关研究进行梳理,从基本概念发展、主要研究方向、关键基础技术3个方面总结了最新的研究进展,分析了作战决策背景下该原则的局限性以及实际运用中出现的问题.在此基础上提出“可接受性”概念并从“决策者需求”、“智能方法优化方向”出发明确其具体能力要求,以提供一种新的指导原则促进人工智能应用于实际作战决策.How to realize the effective fusion of human-intelligence in huma-intelligence hybrid environment and to improve the decision-making level with the organically combination of the intention of human and the formalization of human is one of the most pressing problems.The relevant research of“interpretability”methods of the present main solution thought of the human-machine intelligent fusion is sorted.From three aspects of the basic concept development,main research directions,key basic technologies,the latest development status is summarized,the limitations of the principle under the background of combat decision-making and the problems in its practical application are analyzed.On this basis,the concept of“acceptability”is proposed and its specific capability requirements are clarified from“decision-maker requirements”and“direction of intelligent method optimization”,which provides a new guideline of promoting human-AI decision application in practical combat decision-making.

关 键 词：可解释性 作战决策 人机智能融合 可接受性 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] E91[自动化与计算机技术—控制科学与工程]