This paper describes how a computational system for designing can learn useful, reusable, generalized search strategy rules from its own experience of designing. It can then apply this experience to transform the design process from search based~knowledge lean! to knowledge based~knowledge rich!. The domain of application is the design of spatial layouts for architectural design. The processes of designing and learning are tightly coupled.
Abstract.In this paper, we propose and begin to formalise an approach to machine learning in design called situated learning with the purpose of providing a foundation to developing better design tools in an agent-based framework. Situated learning theory postulates that the situations that an expert is exposed to forms the developmental conditions of expertise. We extend and adapt that theory for computer-aided design with the primary objective of learning the use of existing knowledge, rather than simply the knowledge itself. The idea behind situated learning is to learn situations and associate them with some knowledge with the intention of using the knowledge in similar situations. MotivationMachine learning for design has largely been concerned with acquiring design knowledge from examples. An implicit assumption behind such work has been that the knowledge is always applicable. However, the situation within which the knowledge is learned plays a role in its future applicability. Thus, in addition to learning design knowledge it is important also to learn the situation so that a basis exists for the reuse of that knowledge. One way to achieve this is to represent situations and contexts explicitly. However, current machine learning design tools do not represent situations and contexts and therefore are situation and context free. We propose a situated learning paradigm applicable in designing. In this approach, we borrow and adapt a theory from educational instruction called "situated learning" which states that situations which one experiences as a novice form the developmental conditions of an expert (Ingold 1995, Brown et al. 1989. Our primary concern is with the use of knowledge rather than the knowledge itself. Our first aim is therefore to make a design tool learn situations which existed when some knowledge was learnt, with the intention that when similar situations exist in the future it could apply the knowledge. A secondary goal of this approach is to emerge high level structure-behaviour relationships that serve as a means of improving on existing knowledge, and consequently improve the performance of the design tool especially in the conceptual design stages. In this paper we discuss the foundations of such a design tool based on agents and begin to provide a formalisation of the key concepts of situated learning in an agent-based framework.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.