In this paper, we present a system to support design work which is one of the human creative activities. We have applied the system to the domain of automobile design. Conceptual design work of automobile has two phases. In the first phase, each designer builds a new design concept. In the second phase, designers decide a target design concept among their concepts. The system first analyzes existing automobile data statistically and visualize the result in a metric space. In conventional statistical analysis, we give some interpretation to the result. But in our system, the statistical method is used for the different purpose. The system shows its user the space to trigger his/her concept building. The user builds a new design concept by reconfiguring the space. The concept is saved in the database, and when designer group decide their target concept, the system analyzes the database and assists their decision making. Through the experiments, we have got a clue to consider the phenomenon of concept formation process. Based on it, we propose a model of concept formation process. We have confirmed that the system is useful for conceptual design work. The designer of an automobile company said that it is interesting and is effective to promote conceptual design work.

A distributed and networking card-handling tool named KJ-Editor that simulates arranging index-cards on a desk as working in collaboration is described. Card-handling is one of the most useful methods for information representeation and idea-generation. In KJ-Editor, hundreds of cards can be generated on any place in a display and a sentence can be written on each of them. A generated card can be picked and moved by a mouse. Cards may be grouped by enclosing them with a curve. Relationships of cards and groups can also be marked by special lines. The chart of cards edited on KJ-Editor can be output by a printer and stored in a disk. When a user of the cooperative work makes some operations on the chart in KJ-Editor, the other collaborators can see the operations on thier own displays. That is so-called WYSIWIS (What You See Is What I See) facilities are implemented in KJ-Editor. An experiment that four collaborators made a specification of a middle-scale software- "LIFT" problem, that is well known as a common problem for requirements analysis-using KJ-Editor was conducted. The collaborators meet at a room and are provided with separate networked computers. They can make a face-to-face communication. According to our observation and analysis on this experiment, some features of cooperative work activity using KJ-Editor are identified : (1) a computer-supported card-handling tool is a useful resource for the group in mediating their cooperative work ; (2) pointing a card or an element of the chart by a mouse has an effect for concentrating the discussion, and (3) WYSIWIS facilities sometimes become obstacles for personal viewing of the card-arrangement and cause collaborators to be uncomfortable.

Knowledge Formulation is an essential task to design database schema, to develop knowledge bases, to author hypermedia bases, etc. Its goal is to find a proper information structure to formulate gathered pieces of information. The usual procedures for the knowledge formulation are the grouping and linking of cards. We've developed a system named CONSIST (CONStruction of Information STructure) to support this knowledge formulation task considering above features. This paper first describes our view to the support of this task and the basic functions of CONSIST. Next, it describes the evaluation of results of the experimental application of CONSIST to a sample problem. Finally, we discuss the necessary functions for more effective support for the knowledge formulation, which are now implemented on our successive system CONSIST-II.

GrIPS is a groupware that aims to enhance people's creative abilities by assisting them in the creation and organization of ideas. The prototype of GrIPS works on three workstations connected by LAN and supports creative meetings, such as planning meetings. The methodological basis of GrIPS is diagram drawing, which consists of two stages, the idea divergence stage and the idea convergence stage. In the first stage, the participants of the meeting are encouraged to collect ideas in a brain-storming manner. In the second stage, they create a set of cards, each of which describes a collected idea, and draw a diagram that illustrates the groupings of the cards and the connections between them. The effective presentaion to the others could be done with the final diagram. GrIPS never substitutes the computation for human creativity, but it helps human activity, providing three new functions : associative keyword retrieval, automatic diagram drawing and individual-basis sound mixing. The associative keyword retrieval lists up lots of relative words, which can be picked up as elements of ideas in the first stage. The automatic diagram drawing converts a complex diagram into an easy-to-understand diagram that has the same structure as the original in the second stage. The redrawing is done by heuristics. The individual-basis sound mixing simulates the acoustic effects of a meeting room, by mixing the voice of each participant in appropriate proportion. It aims to compensate for the loss of communicating reality due to the distributed working environment.

FISM (Flexible Interpretive Structural Modeling) is a computer-aided system which effectively and flexibly supports human thinking process and human idea generating process. The process of thinking and idea generating by FISM is called FISM Session. One of characteristics of FISM is that it supports idea conceiving and idea arranging in human thinking process mutually and supplementary. Another characteristic is that FISM has some strategies based on mathematical background to support arrangement of idea. In this paper, details of functional specifications of FISM are described. FISM system is composed of three subsystems FISM/WS, FISM/CdI and FISM/PC. FISM/WS and FISM/CdI are implemented on UNIX workstation. FISM/WS supports all steps of FISM Session, for instance, the embedding step and the partitioning step, etc. FISM/CdI is a card-imaged interface tool. It supports the element extraction step and the result display step in FISM Session. FISM/PC is the personal computer version of FISM/WS, and it has basically the same function as FISM/WS. Effectiveness and use of FISM are shown by using the example in business field.

Even we human cannot solve the frame problem. Nevertheless, in daily life we can act as if there is no such problem, because we use some heuristics to bypass it. In this paper, we propose that the concept of causality is the heuristics. We also claim that temporal ordering in causal relation is not essential. Any partial ordering suffices but once the order is fixed, it should be followed even in reverse order reasoning such as inferring the past from future. We formalize causal reasoning and apply it to YSP.

Qualitative reasoning deals with quantities qualitatively, and it can estimate the value of quantities that change continuously during dynamic behaviour. Many applications of qualitative reasoning, for example to diagnoses and designs, have been tried. However, there is one problem. The use of qualitative values causes ambiguities in reasoning. We use inequalities to solve this problem. In this approach, we can deal with quantities both qualitatively and quantitatively. If a quantity is known only qualitatively, its value is described as an inequality. However, if a quantity is known quantitatively, its value is described as an equality without changing its value to a qualitative one. Therefore, we can reduce ambiguities caused by changing a quantitative value to a qualitative one. To use nonlinear inequalities, we developed a nonlinear inequalities constraint solver, CONSORT (Constraint Solver for Nonlinear Inequalities). CONSORT combines three constraint solvers, which use Interval method, Simplex method and Groebner base method. By transferring results amongst the three constraint solving methods, CONSORT can solve a broader range of inequality problems than is possible using on individual method. Moreover, CONSORT can analyze positive and negative feedbacks by using the interval method constraint solver. CONSORT was used as a constraint solver of Desq (Design Support System using Qualitative Reasoning) which supports to determine the values of design parameters within electronic circuits.

To generate good text, many kinds of decisions should be made. Many researchers have been engaged in searching for both an architecture that would determine a proper order for these decisions and heuristics that would make an appropriate decision locally at each decision point. However, even if such an architecture or heuristics were found, there are still certain kinds of problems that are difficult to consider during the generation process. Those problems include, for example, structural ambiguity and sentence complexity. Some of them can be easily detected and solved by introducing a revision process after generation. In this paper, we argue the importance of text revision with respect to natural language generation, and adopt a computational model of text revision. We also discuss its implementation issues and introduce dependency directed backtracking in order to realize efficient revisions. Our model can be implemented easily because the revision process allows the system to make a decision at each point without any anticipation of the future decisions. Finally, we evaluate our method on an experimental Japanese text generation system.

This paper and succeeding paper [安部93] propose a method to identify Dynamical Systems qualitatively according to qualitative descriptions about a behavior of observed system. System Identification problem is well known especially in control theory. Several "Quantitative" methods have been reported to solve this problem. However, these methods can be applied to a small part of Dynamical Systems. Proposed method can be applied to a variety of Dynamical Systems for which "Quantitative" method can not be applied. This is realized by a kind of quantizing of Euclid space into Qualitative Value space. Proposed method performs the following two steps. (1) Identification of Qualitative States. (2) Identification of Qualitative Differencial Equations. This paper is focussed on the step (1). Qualitative State is a set of Qualitative Values and Qualitative Derivatives which characterize observed Dynamical System. Qualitative States must satisfy several constraints, namely, causal structure of the system. This paper propose a method to identify such Qualitative States according to temporal sequences of Qualitative Values. Assuming continuous behavior of systems, several mathematical relations between Qualitative Values and Qualitative Derivatives are also presented to derive this method.

This paper and another paper [安部93] propose a method to identify Dynamical Systems qualitatively according to qualitative descriptions about a behavior of observed system. Proposed method performs the following two steps. (1) Identification of Qualitative States. (2) Identification of Qualitative Differencial Equations. This paper is focussed on the step (2). Qualitative States must satisfy several constraints. Such constraints describe a temporal causal structure of observed system and called "Qualitative Differential Equations". This paper proposes a method to identify Qualitative Differential Equations which describe observed Dynamical System. This method becomes possible by identifying relations between Qualitative Values and Qualitative Derivatives according to Qualitative States derived by the step (1).