META‐DESIGN CONSIDERATIONS IN DEVELOPING MODEL MANAGEMENT SYSTEMS*
探讨了开发模型管理系统时的认知因素，包括人类局限、信息存储与检索、问题解决策略，并引入基于图的框架来研究自动建模中的模型评估标准、多准则综合及评估时机。
ABSTRACT This paper examines cognitive considerations in developing model management systems (MMSs). First, two approaches to MMS design are reviewed briefly: one based on database theory and one based on knowledge‐representation techniques. Then three major cognitive issues—human limitations, information storage and retrieval, and problem‐solving strategies—and their implications for MMS design are discussed. Evidence indicates that automatic modeling, which generates more complicated models by integrating existing models automatically, is a critical function of model management systems. In order to discuss issues pertinent to automatic modeling, a graph‐based framework for integrating models is introduced. The framework captures some aspects of the processes by which human beings develop models as route selections on a network of all possible alternatives. Based on this framework, three issues are investigated: (1) What are proper criteria for evaluating a model formulated by an MMS? (2) If more than one criterion is chosen for evaluation, how can evaluations on each of the criteria be combined to get an overall evaluation of the model? (3) When should a model be evaluated? Finally, examples are presented to illustrate various modeling strategies.