摘要
经典的决策分析将偏好与效用视为决策人认知黑箱的激励响应。效用函数的集结经过了从实验归纳、公理化构造到近期发展的两者相结合的对话方法的过程。与这一主流方向不同,本文企图在探讨认知系统内部结构和规律的基础上,达到对偏好这一外部认知表现的深层机制的认识,从而实现对偏好形成过程的模拟。本文首先引进有关角色系统和规范的概念,提出了描述规范系统变化动机的双内学习模型,以公理形式给出若干心理学的发现。分析了规范系统的结构及其变化规律。在此基础上,本文发展了一种描述单人决策问题的行为/决策模型。取消'决策者存在一致偏好结构'的传统假设,而代之以'决策者对给定问题存在规范层次系统'的要求。这使得模型及其求解过程与大量的心理学发现相一致。克服了经典价值论中的有关悖论。此外,本文以充足的理由,首次引进和谐性原则作为一种决策规则。使经典的经济人模型渗入了人类非理性的成份。文章的最后,我们构造了一种求解模型的初步算法。重点考虑了同层规范变量的相关性和不同层次规范M—E关系的构造这两个重要问题,并给出了一个简单的演算示例。
In classical decision theory, preference (or utility) is regarded as response to the stimulus input to the cognitive system of a decision maker. Initially, the psychophysical function (utitlity function) was induced from experimental results. Then by deduction, the utility function was derived from a given set of hypotheses which were approximately congruent with the prevalent humanbeing's behavior. Now the elicitation method has developed into a hybird one also called the interactive decision method. On the contrary to the common run, this paper attempts to disclose the inner structure and principles of the cognitive system, understand the behavior mechanism of preference and finally finish simulating the elicitation progress of decision makers. At first, this paper introduces several concepts in relation to actor system, proposes a bi-inner-model learning frame, work that is used to explain the process of cognitive systems. Then, using this explaining framework, some psychological characteristies of perception are tested, the configuration of norm hierarchy along with its reconstruction are analyzed, and a general principle followed by norm change—the lowest variability prineiple—is proposed. Based on the above concepts, we develop a kind of general decision/behavior modet, which is appropriate to describe all decision situations of single decision maker. In stead of the assumption that the decision maker has a consistent preference structure, in this paper, we assume the existence of multi-relation hierarchy system, which shows the requirements of norms, and analytically exhibits the internal mechanism of preference or general utility. Therefore, the decision behavior and the model indicated is even more consistent with the psychological findings, the decision process incorporates the analysis of the evolution caused by learning motivity, some paradoxes on classical preference and utility theory have been overcome. Additionally, with enough reasons, this paper originally intuoduces harmonic relation as a decision criterion so that the classical 'economic roan'decision model is penetrated with irrationality of scientific nature. At last, we focus attention on procedure for solving generalized multicriterion decision. Two important aspects, which are dependence of attributes on the same level, and construction of means-end relation among attributes on the different levels, are discussed in detail. An elementary algorithm and an illustrative example are presented.
出处
《系统工程学报》
CSCD
1991年第2期79-91,共13页
Journal of Systems Engineering
