Allen Bernholtz at Harvard Graduate School of Design

Computer drawing of different perspective views, source: Design Quarterly 66/67

Around 1965, Allen Bernholtz was professor of Architecture and Computer Technology at Harvard Graduate School of Design, working also in collaboration with Harvard Center for Computer Graphics and Spatial Analysis. His methodology of analyzing and synthesizing design problems was aligned with Christopher Alexander’s approach, with whom they were in close cooperation, while both at Harvard. Bernholtz saw that the potential of the use of computers in design lies in the revision of design process. The computer could assist the process as a decision making and information ordering aid. This would free the designer for the “higher levels of the process, which are completely human” Bernholtz says — referring to the creative and intuitive part of design. In the article “Some thoughts on computer role playing and design” published in Ekistics in 1968, he proposed the concept of game playing as a vehicle for investigating design decision-making processes. Games could provide a spatial and temporal reduction of design problems and, therefore, give insight into the interplay of different agents as well as the series of decision-making in the sequence of successive plays. According to Bernholtz, this investigation could be extended to man-machine interaction as a creative problem-solving process and lead eventually to “a computer-oriented language for architects” that would accept as input graphic representations.

Bernholtz, together with his colleage Edward Bierstone, adopted Christopher Alexander and Marvin Manheimer’s computer program HIDECS3 (1963) to decompose and recompose the problem of a house design as an instance. Initially, through constraining the design problem and analyzing the relationships between constraints, the architect should try to achieve a “mutual adaptability of form and requirement”, a “consistent aesthetic and functional clarity”. Requirements are introduced as negative or “misfit” factors, namely things that can “go wrong”. After establishing these requirements, the designer determines the ones that interact, defines the links that are connected in a graph. The graph is then mathematically interpreted by the computer to suggest the subdivision of the system into subsystems. To Bernholtz a problem clearly “stated is half-solved”. The subsequent synthetic process performed by the designer continues until a diagram of a form that satisfies the complete system of misfits and links is produced. The process does not yield a perfect final form but many forms, which possess unity and clarity of purpose. Also, it does not get shortened due to the aid of computer and still demands “unbiased and imaginative work”.

Diagrams showing the requirements decomposed by a specific computer program into sub-systems. source: Design Quarterly 66/67
Level by level re-composition into a hierarchy of subsets representing the architectural problem of a house design, source:Diagrams showing the requirements decomposed by a specific computer program into sub-systems. source: Design Quarterly 66/67
Diagrams showing the solution of a subsystem by the architect. source: Diagrams showing the requirements decomposed by a specific computer program into sub-systems. source: Design Quarterly 66/67
Formal picture of the final design solution of the house. source: Design Quarterly 66/67

Another similar project working on constrained design problems was in cooperation with the Center for Computer Graphics and Spatial Analysis. The computer program called RUMOR was generated random house plans and then it was evaluating them. The operator-designer was establishing a matrix of relationships, creating criteria for evaluation. “No effort has been made to generate only good plans” (Bernholtz, 1969).

RUMOR, The random generation and evaluation of plans, The two illustrations represent a joint plan composed of a living room, dining room, one bathroom, a TV room, a washroom and a sewing/laundry room. source: The Architecture Machine, N.Negroponte

Bernholtz approach is closely influence by Alexander’s ideas on decomposition and synthesis, although working on the smaller scale of a family house instead of environmental and urban situations. This implies associations of his work with MIT’s Architecture by Yourself, in terms that both determine spatial relation between units. Units are abstract quantities of a certain size that signify areas or rooms. The user (designer in the case of Bernholtz and potential inhabitants in the case of MIT) imposes requirement criteria and the computer becomes the reflection of their choices showing possibilities and limitations.


  • Nicholas Negroponte, The Architecture Machine, MIT Press, 1970
  • Allen Bernholtz — Edward Bierstone, Computer-Augmented Design, Design Quarterly issue 66/67, “Design and the Computer”
  • A.Bernholtz, S. Fosburg, A Generalized Program for Transforming Relationship Values into Plan Layout, Advanced Computer Graphics, 1971