The Logic of Architecture: Design, Computation, and Cognition by William J. Mitchell

Abstract

TECHNOLOGY AND CULTURE Book Reviews 625 The Logic ofArchitecture: Design, Computation, and Cognition. By William J. Mitchell. Cambridge, Mass.: MIT Press, 1990. Pp. xi + 292; illustra­ tions, notes, bibliography, index. $35.00 (cloth); $19.95 (paper). This book by a well-known researcher in computer-aided architec­ tural design formulates some basic principles of architectural theory using a first-order predicate logic as its vehicle. The aim is to provide foundations for an axiomatic treatment of architecture. William Mitchell hopes that his unconventional quest will help workers in artificial intelligence produce items capable of addressing and aiding the complex visual thinking of architects. In this pursuit he attempts to address philosophical, psychological, historical, and artistic dimen­ sions of the art of architecture. Mitchell’s strategy is to separate the activity of architectural design into three realms—a design world, a critical language, and a construc­ tion world. The design world can exist on paper, in a computer data base and display system, or in any medium capable of being manip­ ulated into any of many (perhaps infinite) distinct states by the designer. The critical language allows statements that may correspond to states in the design and construction worlds. States in the design world are tested against programs, which reside in the critical language and state preconditions for solutions. The construction world is normally the “real” world, where the design is implemented physically. According to Mitchell, “A design problem exists when you can say what you want in the critical language, but cannot immediately see how to produce a state in the design world that depicts what you want” (p. 64). Under this paradigm, architects need speech as much as pragmatic design work. It idealizes the power of the critical language to assert needs. In fact, design problems seldom are expressed very clearly. Designers discover many needs and wants through the design process and interaction with the client. An architectural composition is a state of the design world, but most states of an open design world are noise. A useful formal deductive system must establish procedures to narrow the range of states to a manageable subset that may make sense. Mitchell suggests as an abstract solution the provision of a limited vocabulary of formal elements and the establishment of a typological syntax governing the permissible arrangements of these elements. Thus languages of architectural form can be defined or established. To illustrate this procedure Mitchell uses examples from classical architecture. Classical orders have been highly codified down to fine details since the Periclean age. Their broader aspects are therefore easy to axiomatize if their subtleties, the careful equipoise of different highly interrelated proportions or the fine detailing of curvatures, are ignored. Mitchell’s axiomatization avoids these, and its rigidity ex­ cludes many lively and eccentric examples by Michelangelo or Bor­ 626 Book Reviews TECHNOLOGY AND CULTURE romini. If even the relatively simple grammar of classical orders requires a syntax capable of dealing with ambiguity in order to obtain satisfactory results, under open-ended conditions the process requires a suppleness that a first-order logic may not be able to provide. Mitchell’s discussion of languages of architectural form with wellordered grammars centers on the most obvious and easy example, Palladian villas. Palladio’s reliance on relatively simple forms and rigorous application of geometric rules, at least in his published designs, make him seem deceptively regular. If 18th-century English aristocrats could imitate successfully the Palladian syntax in their own country houses, so should computers. Mitchell’s attempt at axiomatic development of typical plans demonstrates unwittingly the futility of his approach. His rules multiply as they extend into finer details to the point that for porticos he must provide one rule for each of the dozen or so villas he considers. Even so, the tiresomely many rules given are insufficient to generate such classic examples as the Villa Rotunda at Vicenza, the Villa Thiene at Quinto Vicentino, or the unexecuted Villa Mozenico on the Brenta canal without additional syntax. Obvi­ ously Palladio did not know, beforehand, what a Palladian villa was. Mitchell’s attempt at an axiomatic mechanization of architecture has a distinguished ancestry. Perhaps the most elaborate rule-driven design system, not...