博论题目：理解人类理性行为：司马贺跨学科建模方法研究
答辩人：汤杉杉
指导老师：刘劲杨
答辩时间：2024年5月25日
目 录
绪论
0.1 选题背景
0.2 司马贺生平
0.3 文献综述与研究现状
0.3.1 司马贺著述简介
0.3.2 国内外研究现状
0.4 研究方法与主要思路
0.4.1 研究方法
0.4.2 主要思路
0.4.3 拟完成的创新
第1章 司马贺建模研究的工作基础
1.1 研究取向溯源
1.1.1 科学实证取向
1.1.2 学科交叉取向
1.2 人类理性行为的科学研究
1.2.1 人类理性的哲学思考
1.2.2 理性的行为主义界定
1.2.3 行为分析与科学建模
1.2.4 行为建模与数学化理念
1.3 建模的世界观与方法论设定
1.3.1 从自然界到人工界
1.3.2 作为复杂系统的世界实在
1.3.3 复杂性的层级
1.4 司马贺行为建模的工作框架
第2章 个体理性行为的建模
2.1 “经济人”的超凡模型解释
2.1.1 “经济人”及假设
2.1.2 超凡模型的建立
2.1.3 批判完全理性
2.2 “管理人”的有限理性建模
2.2.1 理性的计算修正
2.2.2 行为模型与直觉模型
2.2.3 转向“管理人”
2.3 “适应人”的进化模型
2.3.1 理性行为的动态调整
2.3.2 进化模型与理性过程
2.3.3 “适应人”的生成
2.4 个体理性模型的变迁
第3章 组织管理行为的建模
3.1 古典组织的机器模型
3.1.1 集体理性的必要性
3.1.2 机器模型及问题
3.1.3 传统管理原则的弊端
3.2 组织内行为的动机模型
3.2.1 官僚模型及其分析
3.2.2 理解组织行为的关键
3.2.3 生产与参与的动机模型
3.3 组织与管理的决策模型
3.3.1 自主决策行为的模型
3.3.2 管理的实证研究纲领
3.3.3 探索管理科学
3.4 组织行为模型的整合
第4章 认知及智能行为的建模
4.1 认知的信息加工模型
4.1.1 心理活动的层次划分
4.1.2 认知的三种基本类型
4.1.3 信息加工系统的结构
4.1.4 信息加工模型的架构
4.2 物质符号系统及其假设
4.2.1 模拟认知的必要条件
4.2.2 符号系统的智能表现
4.2.3 物质符号系统的构建
4.2.4 物质符号系统假设的推论
4.3 模拟智能的计算机模型
4.3.1 逻辑理论家LT
4.3.2 通用问题解决者GPS
4.3.3 发现系统BACON
4.4 开辟符号主义路径
第5章 从行为建模到“人工科学”
5.1 人工物的创造与设计
5.1.1 构建设计的理论体系
5.1.2 设计教学的课程细目
5.1.3 社会设计的计划与愿景
5.2 人类行为科学的新模型
5.2.1 选择进路与控制进路
5.2.2 基于人与环境关联的建构
5.2.3 行为科学新定位
5.3 人工科学的跨学科研究原则
5.3.1 人工与自然的关联
5.3.2 整合不同研究视角
5.3.3 研究手段的多元协同
5.4 人工科学的当代拓展
第6章 司马贺建模方法评析
6.1 人类理性行为的系统建构
6.1.1 理性行为的层级及其关联
6.1.2 理性行为的复杂性特质
6.1.3 理性行为建模的价值与意义
6.2 行为建模方法的跨学科开拓
6.2.1 建模的方法论优势
6.2.2 建模激发科学创新
6.2.3 建模推动跨学科实践
6.3 行为建模的反思与展望
6.3.1 理性行为科学路径的局限
6.3.2 建模方法的困境及应对
6.3.3 行为建模的元模型
结 语：穿越迷宫的解谜
参考文献
附录
司马贺生平活动年表
司马贺主要著作提要
致谢
摘要
面对大科学时代重大问题的挑战，不同学科的交叉与融合已是研究范式的主流趋势。当代科学巨匠司马贺（亦译为“西蒙”，Herbert Alexander Simon，1916-2001）一生屡次攀登多座学科的高峰，是经济学、管理学、认知心理学及人工智能等诸多领域的先驱和重要开创者，曾获诺贝尔经济学奖、图灵奖、心理学界三个最高奖项等荣誉。如何展开跨学科研究？如何通过跨学科研究创立新学科？如何通过新学科的拓展以实现新领域的开辟？面对复杂环境，新的人类行为科学该如何建构？这些问题在司马贺贯彻一生的建模研究中得到创造性地探索和回答。
人类理性的性质是司马贺一生关注的重点，他将理性定义为一种“行为方式”，用“通过模型来理解人类行为”概括毕生的研究工作，其中建模方法被视为其思想的精髓。本论文研究的目的是通过分析与比较司马贺针对不同人类理性行为的建模过程与方法，形成关于人类理性行为的系统理解，并揭示“人工科学”这一开创性设想，整合司马贺在不同学科、不同领域、不同层级的建模工作，构建出跨学科人类理性行为的系统框架及其方法路径，以期推进当代跨学科方法论研究。
面对人类理性行为这个大问题，司马贺所开展的一系列研究可追溯到其基本的研究取向与工作思路。“理性”是哲学史上一个长久性的重要概念。但不同于理性思维的哲学反思，司马贺受科学实证精神与学科交叉取向的影响。基于复杂系统思考的世界观与方法预设，司马贺从行为分析的视角对理性加以重新界定，主张运用数学化的理念将对人类理性的理解转化为对理性行为的科学建模。司马贺关于人类理性行为的建模方法研究分别从个体理性行为、组织管理行为、认知与智能行为三个层级展开，形成系统研究框架。
理性行为通常对应于“理性个体”，因此什么是理性个体的行为就成为司马贺的研究进点。人类理性行为建模的起点是对基于完全理性的“经济人”假设的思考和批判。在司马贺看来，经济人假设所主张的理性行为未考虑真实的受限情况，故称之为“超凡模型”。为寻找能够替代的新模型，司马贺先对理性进行了计算简化与修正，构建出符合人类理性有限性的“近似”理性模型，转向“管理人”。后将理性行为视为最终实现理性的动态过程，将理性适应与进化之间的相似性相结合，提出进化模型以表明个体是有意识、有目的地进行自我调整以适应环境的“适应人”，由此实现个体理性行为从超凡模型到行为模型与直觉模型，再到进化模型的变迁。
社会活动的理性行为还表现为个体之上的组织行为。在已有理论中，关于组织中人之行为的命题各有侧重。为此，司马贺经梳理阐释了组织行为的背景和诱发条件，对应地构建出组织管理行为的机械模型、动机模型和决策模型。为实现管理与决策的科学化、理性化，司马贺强调一套科学的、成熟的组织行为理论应考虑到三类组织模型的整合。在组织管理行为建模的过程中，针对组织可能面临的整体决策和不确定性，既需考虑成员的工具性，又要关注动机、士气、利益冲突等因素之于组织生产与活动参与的影响，并将决策制定过程视为理解组织现象与理性行为的关键，以消解传统管理原则的弊端、突破单一组织理论的局限。
随着研究的深入，司马贺发现，要真正理解人类理性行为，除了展开个体、组织的分析，还需深入到心智层面，即认知与智能行为的建模。作为认知心理学、人工智能的开创者之一，司马贺认为对人类认知中识别、问题解决和学习过程的分析可以解释理性行为的奥秘，由此构建了认知的信息加工模型。司马贺将由输入装置、输出装置、记忆装置和控制四部分构成的信息加工系统与包括输入、输出、存储、复制、建立符号结构和条件性迁移共六项功能的物质符号系统进行类比，运用计算机程序和算法语言等，先后构建了一系列计算机模型，如逻辑理论家、通用问题解决者和发现系统等，形成一套系统的认知与智能行为模拟体系。
为实现各层级行为建模的循环和综合，司马贺提出“人工科学”的学科设想。通过理论体系的构建、课程细目的设置以及社会设计的规划，司马贺回应了人工物的创造与设计问题。此外，司马贺主张把具有适应能力的有机体、复杂性特征突出的外部环境以及人与环境之间的关系统一置于一个由子模型和组合关系构成的综合模型中，由此形成以“行为的”和“功能的”为基本定位的人类行为科学新模型。司马贺把对人类理性行为的见解与建模方法的成果凝练于人工科学的蓝图中，在社会科学领域发起一场科学综合的运动。
依据全文研究，司马贺凭借跨学科建模方法的运用，实现了有关人类理性行为认识的复杂性转向。理性行为的建模可类比为“穿越迷宫的解谜”：理性行为不是一种基于完全信息的确定性计算，而是必须面对不确定性，在有限理性原则的指导下，为达到问题解决的目的而进行的启发式解题过程，呈现为由方案的搜索、决策的制定与执行等环节构成的行为系统，具有有限性、目标导向性、动态生成性与开放性的特征，能够灵活自主地适应并应对不确定的复杂环境。司马贺的建模方法不仅丰富了研究工具、提供了实证操作手段，而且超越以往的简单性方法，既加深了对人工界的了解，拓宽科学研究的视域和问题域，又有力地推动了知识的综合创新与学科的交叉借鉴，同时为未来发现和创造全新知识、实现学科的深度融合提供了方法论上的示范，深刻影响了当代跨学科方法的发展。
关键词：司马贺；理性行为；建模；方法论；复杂性
Abstract
Facing great challenges of serious problems in the big science era, the interdisciplinary paradigm has taken the lead in different fileds. Herbert Alexander Simon (1916-2001) is a scientific giant who has made outstanding contributions to multiple disciplines. He is an important pioneer in many scientific fields such as economics, management, cognitive psychology and artificial intelligence. Simon also receives numerous honors like the Nobel Prize in Economic Sciences, the Turing Award, the Award for Outstanding Lifetime Contributions to Psychology and so on. How can interdisciplinary research be effectively conducted? How can new disciplines be forged through interdisciplinary research? How can these new disciplines be expanded to facilitate exploration into novel domains and the resolution of complex issues? These urgent methodological questions find their answers in the modeling approach which is used throughout Simon’s whole life.
Simon focuses on the issue of “the nature of human rationality”. He defines “rationality” as “a way of behavior” and uses “understanding of human behavior by models” to summarize his all works. The method of “modeling” can be regarded as the essence of his thought. The purpose of this study is to form a multi-level understanding of human rational behavior and its systematic understanding by analyzing and comparing Simon’s modeling processes and methods, to reveal the pioneering concept of “the sciences of the artificial” and integrate Simon’s modeling approaches in different disciplines, fields and levels, to construct a systematic framework and methodological path for human rational behavior in order to thereby promote the research of contemporary interdisciplinary methodology.
Facing the big problem of human rational behavior, Simon’s research can be traced back to his fundamental research orientation and general working ideas. Rationality, as a philosophical concept, has a long history. However, different from philosophical thinking of rationality, Simon is influenced by the spirit of scientific empiricism and interdisciplinary orientation. Based on the worldview and methodological assumptions of complexity and systems thinking, he redefines rationality from the perspective of behavioral analysis and advocates for the use of mathematical concepts to transform the understanding of human rationality into scientific modeling of rational behavior. The modeling of human rational behavior can form a systematic research framework, which is carried out from three levels: individual rational behavior, organizational management behavior and cognition and intelligent behavior.
Rational behavior usually corresponds to rational individual, so Simon’s research firstly focuses on what is the behavior of rational individual. The starting point of the modeling of human rational behavior is the consideration and critique of the hypothesis of “Homo Economicus”. In Simon’s view, due to the neglect of the limitations of rationality in real behavior, this idealized approach could be abstracted as “The Olympian Model”. In the quest for alternative models, Simon simplifies and modifies the calculation of rationality, then he constructs the model of bounded rationality. This transition marks a shift from the “Homo Economicus” to the “Homo Administrativus”. After that, contemplating the similarity between rational adaptation and evolution, Simon regards rational behavior as a dynamic process of ultimately achieving rationality and proposes an evolutionary model. It advocates for individuals as “Homo Adaptivus” who adjust themselves consciously and purposefully to adapt to the environment, so as to realize the change of individual rational behavior from “The Olympian Model” to “The Behavioral Model” and “The Intuitive Model”, and then to “The Evolutionary Model”.
Rational behavior in social activities is also manifested as organizational behavior above the the level of individual. Among the existing theories, the propositions about the behavior of people in the organization have different emphases. Therefore, after understanding the background and triggering conditions of organizational behavior, Simon correspondingly constructs the mechanical model, motivation model and decision-making model of organizational management behavior. To achieve the scientification and rationalization of management and decision-making, Simon emphasizes that a scientific and mature theory of organizational behavior should consider the integration of three types of organizational models. In the process of modeling organizational management behavior, when facing the overall decision-making and uncertainty of the organization, it is necessary to consider not only the instrumentality of the members of the organization, but also the influence of motivation, morale, conflict of interest and other factors on the organization’s production and participation activities. Simon considers the decision-making process as the key to understanding organizational phenomena and rational behavior, so as to eliminate the drawbacks of traditional management principles and break through the limitations of a single organization theory.
As research progressed, Simon finds that to truly understand human rational behavior, it is necessary to delve not only into individual and organizational analyses but also to delve into the realm of cognition and intelligent behavior. As a pioneer in cognitive psychology, Simon believes that the analysis of recognition, problem solving and learning process in human cognition can explain the mystery of rational behavior. From this perspective, he constructs the model of cognitive information processing. Simon analogizes the information processing system composed of input, output, memory, and control components, to the “Physical Symbol System”, which includes six functions: input, output, store, copy, build symbol structure and conditional transfer. Then subsequently invents a series of computer models by using computer programs and languages, such as Logic Theorist, General Problem Solver and BACON to establish a system for simulating cognitive and intelligent behavior.
To achieve the cyclic and comprehensive modeling of behaviors, Simon proposes the disciplinary concept of “the sciences of the artificial”. Through the construction of theoretical frameworks, the design of curriculum details, and the planning of social design, Simon addresses the issue of the creation and design of artifacts. In addition, Simon advocates integrating adaptable organisms, environments with prominent complexity features, and the relationships between those into a comprehensive model composed of sub-models and combination relations. The new model of human behavior science is based on “behavioral” and “functional” and it condenses the insights of human rational behavior and the results of modeling methods into the blueprint of the sciences of the artificial, and initiates a movement of scientific synthesis in the field of social sciences.
Based on this study, Simon achieves a complex shift in understanding human rational behavior through the application of interdisciplinary modeling. The modeling of rational behavior can be analogous to “puzzle-solving through a maze”: it is not a deterministic calculation based on complete information but rather a heuristic problem-solving process guided by principles of bounded rationality when facing uncertainty. It is presented as a behavioral system composed searching for solutions, making decisions, and executing actions and has the characteristics of limitations, goal orientation, dynamic generation, and openness, capable of flexibly adapting to and coping with uncertain and complex environments. Simon’s modeling approach not only enriches the research tools and provides empirical operational means but also transcends the previous simplistic methods. It deepens the understanding of the artificial world, broadens the horizon and problem domain of scientific research and effectively promotes the cross-disciplinary of difference knowledge. Furthermore, it provides a methodological demonstration for creating new theories and promoting the deep integration of disciplines in the future, which has a profound impact on the development of contemporary interdisciplinary methods.
Key words: Herbert A. Simon； Rational Behavior； Modeling； Methodology；Complexity
站内搜索
