An investigation of the role of implicit and explicit constraints in indexing case-based architectural systems.

Abstract

This study is an attempt to employ computers as design-aids in the core of the design process, the synthesis phase. Most of the efforts done to provide design-aids are devoted to the analysis and evaluation phases. The share of the synthesis phase of this effort is very slim due to the difficulty of understanding and accommodating some aspects such as creativity. This study suggests that computers could assist in the synthesis phase of the design process if they had access to knowledge and experience similar to that designers rely upon. The study is based on the evidence that states that prior knowledge (experience) could play an important role in stimulating and inspiring the thinking process, and thus creativity. This study addresses an important problem that represents a major obstacle in building a case-based architectural system. A case-based system is an "intelligent" computerized environment built around the idea of adapting and transforming similar experiences in order to construct new solutions. Storage and retrieval--and consequently indexing schemes--are vital to these systems. An indexing scheme is a systematic guide to locating items contained in a collection. Although the idea seems simple, its applicability in architectural design is not because of the differences between the indexing schemes' requirement of the case based approach and the indexing schemes of buildings in the architectural domain. In order to avoid demanding a large set of indices, which would seem boring and impractical, indexing schemes of the case-based approach require a limited number of indices. Unfortunately, a building, as a complex artifact, can be indexed by large number of indices. An indexing scheme was proposed that allows similar experiences to be retrieved with a relatively small set of indices. The key idea in the solution is to record the implicit (problem-independent) constraints and use them with the explicit (problem-dependent) constraints in the retrieval process. The retrieved set is a result of two steps of alternative evaluation. A system was design to test the indexing scheme. It consists mainly of a case-base where knowledge about buildings, environments, and architects is stored; a control mechanism that handles the selection process; and a user-system interface facility that eases the communication between the two parties.