结构化论辩理论在医疗决策中的应用研究  

The Application of Structured Argumentation Theory in Medical Decision-Making

作　　者：崔建英[1] 安宇辉 Cui Jianying;An Yuhui

机构地区：[1]中山大学逻辑与认知研究所、哲学系,广东广州510275

出　　处：《学术研究》2025年第4期42-48,177,共8页Academic Research

基　　金：国家社会科学基金重大项目“语用逻辑的深度拓展与应用研究”(19ZDA042);广东省哲学社会科学规划项目“基于语用论辩论证理论的同行评议研究”(GD24CZX02)的阶段性成果。

摘　　要：神经网络被广泛应用于决策支持系统,但其无法对决定(从输入到输出)给出明确解释,由此被称作人工智能的“黑盒”效应。结构化论辩理论通过论证间的支持与击败关系解释输入—输出(从前提到结论),有助于合乎伦理规范和可解释医疗决策模型的设计。我们基于结构化论辩框架P-ASPIC+构建医疗决策模型MDM,将其应用于一个医疗决策案例的刻画:首先,将患者的就诊信息量化并存储到知识库;其次,构造推理框架,借助求解器TOAST计算在给定语义下的可接受论证集;最后,通过解释生成器生成文本解释辅助主体医疗决策。结构化论辩框架建构决策模型的方法,既可为决策者提供合适的选择,也可清晰阐明这些选择背后的支持理由,提高了决策的透明度和可解释性,有助于提升医疗决策的质量和信任度。Neural networks are widely used in Decision Support Systems(DSSs),however,they cannot provide clear explanations for decisions(from input to output),which is referred to as the“black-box”effect in Artificial Intelligence.Structured Argumentation Theory explains input-output relations(from premises to conclusions)through support and defeat relations between arguments,facilitating the design of ethical and explainable Medical Decision Models(MDM).This study constructs an MDM based on the Structured Argumentation Theory P-ASPIC+and applies it to a medical decision-making case.First,patient consultation data is quantified and stored in a knowledge base.Next,an inference framework is constructed,and a solver,such as TOAST,is used to compute an acceptable set of arguments under the given semantics.Finally,an explanation generator produces textual justifications to assist the decision-making process.The method of constructing decision models using Structured Argumentation Theory not only provides decision-makers with suitable choices but also clearly elucidates the supporting reasons behind these choices,thereby enhancing the transparency and explainability of the decisions.This enhances the transparency and explainability of decisions,ultimately contributing to more reliable and trustworthy medical decision-making.

关 键 词：标准可能性逻辑 医疗决策模型 结构化论辩框架 

分 类 号：B81-05[哲学宗教—逻辑学]