EXPLORING STYLE EMERGENCE IN ARCHITECTURAL DESIGNS JOHN S. GERO AND LAN DING Key Centre of Design Computing Department of Architectural and Design Science University of Sydney NSW 2006 Australia e-mail:{john,lan}@arch.usyd.edu.au Abstract. This paper presents an evolutionary approach to style emergence in architectural designs. Emergence is the process of making features explicit which were previously only implicit. Style is considered as a set of common characteristics of a group of designs. It is interpreted using a language model as an analogy and is represented at the genetic level. An evolutionary system based on genetic engineering is developed. It emerges style by locating the genetic structures which produce that style. Preliminary results are presented. 1. Introduction Style is an important aspect of architectural designs. Many studies have dealt with the concept or cognition of style but have lacked any process of deriving style (Simon, 1975). Emergence is the process of making features explicit which were previously only implicit. Emergence in design plays an important role in supporting designers’ creativity. Although some process models of emergence have been proposed and implemented (Gero et al, 1995), none has been proposed for emergence of style. In this work we consider style emergence using an evolutionary system based on genetic engineering. Evolutionary systems have been applied in design domains which involve design optimization and design learning (Bullock, 1995; Gero, Louis and Kundu, 1994). The genetic engineering approach (Gero, 1992) comes from the analogy with human intervention in biological evolution. It examines the performance of sub-populations and locates the genes or gene structures which produce the desired performance and then evolves them as new genes in some appropriate manner. Genetic engineering has two advantages when it is applied in design contexts. Firstly, it can set up the relationships between the design space and the design’s performance. Secondly, it can rapidly evolve the design space and generate desired design results.