There is little doubt that texture, surface, extension, value, and color are in one sense basic categories of physical phenomena, constituting in essence a fundamental stratum of experience and the sensible world. Modern psychology and epistemology, however, cohere in saying that this stratum is not the sense world of everyday, the primary datum of the functioning societal member. Neither is the abstract world of the physicist's concepts an everyday world. What is given in our day-to-day life is neither very abstract, nor very concrete, but a sort of functionally-bound middle world of norms and stereotypes.

It takes the disciplined seeing characteristic of foundational training in the arts to drive this bourgeois view of physical reality back to elemental sense data. Just as it requires the supremely abstract language of mathematics and the rarified "experience" of modern scientific experimentation to drive that same view offcenter in the direction of abstraction. The first exercises of a program of design education begin to perform the former role.

This study describes a system that has been developed to address a number of these issues. Based on research findings from the field of Artificial Intelligence which expounds on the need for multiple techniques to represent any complex area of knowledge, we have selected a particular approach that focuses on multiple techniques for design representation. We review this approach in depth by considering its many facets necessary when implementing a knowledge-based system. We then partially test the viability of this approach through a small case study, implementing a knowledge-based system for designing stairs. While this effort only deals with a small part of the total design process, it does explore a number of significant issues facing the development of computer-based design assistants, and suggests several techniques for addressing these concerns.

A building while being designed can be visualized and criticized from its solid model, and the model can take a variety of forms depending on its part): computer-based, drawn in orthographic or perspective projection, constructed of cardboard or wood, or described narratively by means of text, programmatic data, performance model or animation. However, practicing architecture is the process of recording and communicating the decision making process and the contractual obligations that result. In actual practice, in contrast to the designer directed ideal, more participants are brought in sooner at the beginning of a project and with more publicity, which in turn means keeping more, not fewer, records. As the profession evolves, records of the string of design decisions will become more automated, more carefully structured and more retrievable. More buildings will be "tracked" and exposed to review in this way because public environmental sensitivity will improve. The communication between a single designer and his own thoughts will become less and less important.

Two approaches to the integration of artificial intelligence and knowledge-based systems into architectural design practice are currently dominant. One attempts to create systems which can on their own produce designs, the other provides intelligent support for those doing design. It was, in part, the recognition of limitations in the ability of traditional CAD systems and building modelers to reflect what designers actually do that led to explorations into the idea of intelligent assistants. Development of such assistants was aided by research into the act and process of design through protocol and other studies. Although some work is currently being done in the development of artificial intelligence and knowledge based applications in architecture, and work continues to be done on the study of design methodologies, the bulk of available information in each of these areas remains in the realm of design disciplines related to but outside of architecture and do not reflect the explicit role of architectural design in the embodiment and expression of culture.

The relationship of intelligence to culture has resulted in some skepticism regarding the ultimate capacity of neural nets and symbolically programmed computers in general. Significant work has been done questioning the rational tradition in computer development for its failure to address phenomena which are not easily subject to scientific analysis. Further skepticism regarding the role of artificial intelligence and knowledge-based or expert systems in architectural design has been emerging recently. Such criticism tends to focus on two issues: the nature of drawing as an activity which involves both the generation and interpretation of graphic artifacts, and the nature of the human designer as an active agent in the design process.

Less than three years after the introduction of its first computer workstation, the Barcelona office is fully computerized, from carrying out even basic design directly with computer technology, to developing inhouse software and maintaining an internet node via modem. This rapid adoption of the technology, although a relatively smooth one, was not free from strange side-effects. Because of the continuing involvement of a large part of the existing staff, the transition to computer aided design required the appearance of hybrid methodologies which are neither the traditional ones, nor what one might expect to find in the newly established CAD practice.