In this paper, we propose a fundamental concept for design of multi-agent systems. Using a computer simulation of organization theory, we propose the limits of agents' capacity, and the extent to which their limited capacity may overcome. In order to simplify the problem, we have used as an experimental environment, a "block world" similar to the one used in the planning field of artificial intelligence.

The main conclusion that has been derived from this simulation is that, in order to overcome the limited capacity of the agent, it becomes necessary to form a cooperation system. Furthermore, if trade-off can be established between the different capacity of the agents, then it may be possible to determine an optimal agent composition.

Functions and properties of a user interface(UI) in decision support systems(DSS) are discussed and its implementation on our prototype DSS, actDSS, will be described.

First, functions which a general UI should have are examined based upon ‛the eight golden rules' proposed by Shneiderman. Next, a functional model for UI in DSS will be proposed, where a UI in DSS has three type of support, communication support, control support and user support. Especially we focus our attention on the control support and clarify the most important properties of UI in DSS, ‛undo', ‛macro' and ‛mode-less' properties, with the automaton like formulation of them Finally, their implementation on our prototype DSS, actDSS, especially the mechanism of realization of these properties will be described.

This paper describes a newly developed system for quality contorol of the factories producing multi-item products such as foods and seasonings. The old way of quality control in multi-item production factories, which we call the < 1st stage >, is to analyze a set of products by expert people. The 1st stage systems include much inefficiency, and so productivity of quality control is limited.

To deverop and introduce new system, the review and reconstruction of the old system were carried out. The new system, which we call the < 2nd stage >, developed here decomposes the products into analytical components which are predetermined by product specifications. In this case the information system developed plays important roles to decompose and integrate analytical information of the products.

Using relational database and LAN, the 2nd stage system not only realized 200% effective transactions of routine works as compared with the 1st stage one, but made it possible to analize quality control system from various view points. At the same time the system established the quality control for raw materials without increasing man power.

It is possible to understand that the concepts and methodologies of this system is new, and the system is applicable to other multi-item businesses such as phermaceutical and fine chemical ones.

In this paper, first, we construct a mathematical model of social system, which is generated by Total System and the Selecting and Forming Environment Principle such as autonomous units select and form thier own environments.

Second, we show the mechanism and process of the self-transformation between Hierarchical System and Decentralized Autonomous System.

And last, many examples from management fields are shown. Moreover we discuss the possibility of the planning and realization of artificial management system or business organization by means of introducing Information system.