Everyone designs who devises courses of action aimed at changing existing situations into preferred ones.
Herbert Simon (1969)
The nature of problems in general, and design problems in particular, has challenged philosophers, computer scientists, and designers alike. Simon’s definition of design as a problem-solving behavior aimed “at changing existing situations into preferred ones” is a common starting point (Simon 1969; Jackman et al. 2007). Everyone does it. It is a cognitive behavior. The goal is an improvement. Some “existing situations” are problems; but are all problems design problems? If I need my hat off the shelf in the closet and I simply stand up and reach for it, I may be changing an existing situation to a preferred one, but have I done design? The solution to the problem is immediately apparent—obvious. It simply needs to be executed. Only if I am unable to reach the hat do I need to think about it. Even then, having encountered this difficulty before, I may know I can stand on a chair and reach, or use a stick to work the hat off the shelf. I may know of several alternative solutions requiring a step or two to execute (fetching the chair or finding a stick), but the solution is routine. Only someone contemplating this problem for the first time actually needs to devise a new, or creative, response. In each case the existing situation is the same, as is the preferred one, indicating that response to problems depends on personal experience and that the response will not always be unique or novel.
Simon’s “course of action” draws attention to the process of retrieving the hat, but the desired result (hat on head) is also a physical configuration of the world that addresses the problem of a wet, cold head. Thinking to put on a hat involves knowledge that the hat is an appropriate solution, one among several. Both process and knowledge are involved in design (Kalay et al. 1990).
First, the taking in of scattered particulars under one Idea, so that everyone understands what is being talked about…. Second, the separation of the Idea into parts, by dividing it at the joints, as nature directs, not breaking any limb in half as a bad carver might.
Plato
In his 1964 Notes on the Synthesis of Form, Christopher Alexander uses this quote from Phaedrus, in which Plato talks about “the taking in of scattered particulars under one Idea, so that everyone understands what is being talked about,” acknowledging the difficulty of establishing the scope or boundaries of many problems (Alexander 1964). Alexander uses the problem of producing hot water for tea to illustrate the range of solutions, from kettle to household to municipal hot water possibilities. Locating, bounding, and defining the Idea that identifies the problem remains a challenge.
Having established an Idea, Plato’s second step is “the separation of the Idea into parts, by dividing it at the joints, as nature directs, not breaking any limb in half as a bad carver might” (Alexander 1964). The notion of problems being divisible into parts is crucial to the theory of classical problem solving proposed by Newell and Simon (1958) because it is part of divide-and-conquer strategies which attempt to reduce complex problems to solvable routine parts whose individual solutions can be aggregated into a complete solution.
The “taking in of scattered particulars under one Idea” and “separation of the Idea into parts” also suggests that all characteristics of the problem can be established before the solution process begins. This view sees both the problem and the solution plan as objective facts in the world, distinct and separate from us, the designers. Unfortunately, many problems are impossible to separate from their context; they are unavailable for laboratory study and their “joints” are difficult to define in advance, producing what Rittel and Webber (1973) describe as wicked problems. This observation that problems depend on social context has also seen significant exploration in the computer science field of human–computer interaction (HCI), where the theory of situated action has gained significant currency (Suchman 1987).
Researchers have also found that the perceived difficulty of a problem is not an intrinsic quality of the problem; it tends to depend on the level of expertise of the designer (Dorst 2004). What a neophyte might find a challenging design problem, an experienced designer may see as routine, or even obvious. The lack of fixity to problems makes it difficult to develop software that is universally appropriate—two different designers may need very different kinds of tools, or a single designer may need different tools at different times.
If the action required to transform the existing situation to the preferred one is recognized at the same time as the problem, the solution can be said to be obvious or known. This may happen because the exact same problem, or a very similar problem, has been encountered in the past and its solution is remembered.
More challenging than problems with obvious solutions are problems that might be described as nearly obvious or routine—those with well-known procedural solutions. “Routine design can be defined as that class of design activity when all the variables which define the structure and behavior are known in advance, as are all the processes needed to find appropriate values for those variables” (Gero 1994). Many engineering problems, such as the selection of a beam to carry a particular load, are well understood and characterized by mathematical formulas that can be solved in a straightforward fashion, given knowledge of appropriate material properties.
Where no obvious solution is available, where analogy to similar problems fails, and where no analytic formulation exists, designers deploy a process or set of interrelated strategies intended to help identify solutions. While descriptions of design process vary, one of the first steps will probably be to analyze the problem and identify its salient characteristics. In architecture this may take the form of a survey of required spaces, identifying the uses and required size of spaces, adjacency requirements, client’s style preferences, etc. The process of problem definition, dividing the problem “at the joints,” begins to establish the problem solution: If you need a house, an office building isn’t the solution.
Problems and designs are usually understood in terms of the set of characteristics that are sought (cozy, modern, energy efficient, etc.) and measures (cost, sizes, number of bedrooms, number of users, annual energy budget, etc.) that must be achieved. It sometimes helps to think of each of these characteristics and measures as separate distinct dimensions, perhaps with some known values (e.g. maximum cost), but with many unknowns. Collectively, the characteristics and measures define a multidimensional design space or state space within which the ultimate solution must be sought. It is the set of all unique design solutions that can be described by the selected characteristics. A design proposal can be assessed along each of the axes of this space (count the bedrooms, estimate the costs, simulate the heat loss, etc.). The values establish a state in that space (a point or a region consisting of closely related points). This is an important idea that appears repeatedly in this book and much related literature.
One of the first modern descriptions of design space and state space search came from Herbert Simon, who won the 1979 Nobel Prize in economics. In his 1969 book, The Sciences of the Artificial, he presented his ideas about how to study the man-made, artificial world and its production. A powerful concept, design space remains one of the backbone ideas of design computing.
Implicit in this idea is another—that design is search. This idea is consistent with observations of both paper-based and digital design activity (Woodbury and Burrow 2006; Jackman et al. 2007). It underpins many projects that aim to support or augment human design skills, and matches the description of design as iteration. The biggest problem, as we will see later, is the enormous number of candidate states.
State space search strategies rely on measurable aspects of the design, but not everything you value about an environment can be measured. There may be affective qualities such as “relaxed” or “corporate” which are important, but not directly represented in the design, nor computable from what is represented. That is, no matter how detailed, state space representations are always incomplete approximations of the overall project.
Closely associated with the idea of a design space is the idea of an objective function, also called a utility or fitness function, discussed in more depth in Chapter 12. For each design this function combines both the measures along the design space dimensions and a weighting based on the importance or impact of each measure to produce a value that represents an assessment of the design. This value provides a means by which designs may be compared.
If you think of the face of the earth as a two-dimensional design space of locations, the elevation of the surface at that location might be thought of as the objective function. The problem of finding a high or low spot is a simple optimization problem.
During problem definition or exploration, the characteristics that the design needs to satisfy might be expressed as constraints (“five bedrooms”) or performance goals (“net-zero energy”). As part of an objective function these can be thought of as weights applied to measures along the axes. However, while we may think of the separate axes of the design space as distinct, it is often a mistake to think of them as orthogonal, or independent. For instance, the volume of heated space in a building impacts the energy consumption, so the number of bedrooms affects the ability to get to net-zero energy. One of the challenges and wonders of design is the ways in which multiple design goals can sometimes be met by a single solution, as when windows provide both views out and daylight in. Of course, it is also common for constraints and goals to conflict, as happens with “minimize cost” and “maximize space,” precisely because they are interrelated. Because nominally different constraints or goals actually interact and depend on the value system of those making assessments, discovering the core challenges of a design problem can be quite difficult, as Rittel and Webber discussed. Design problems with too many criteria may be overconstrained (impossible to solve), while those with too few will be underconstrained (with too many solutions).
There is another view of design problems that sees them co-evolving with their solutions (Dorst 2004). This view acknowledges the human difficulties of fully defining the problem up front; sometimes it is helpful to redefine the problem. The particular solution and the problem definition emerge together from a process of puzzle making (Archea 1987).
The view of design as puzzle making or co-evolution is very different from the divide, conquer, and aggregate models descended from the Platonic view. It acknowledges the active role of the designer in formulating the problem. While Newell and Simon’s work was intended to lead to artificial intelligence, puzzle making leaves human cognition at the center of the process. It also suggests that wicked problems may arise from a mis-fit of problem and solution, requiring renegotiation of either or both.
Identifying the true problem motivating a design project is not always easy. One, possibly apocryphal, story involves a growing company contemplating construction of a new building. They are said to have solved their space needs by switching to a “hot desk” model of space assignment after their architect pointed out that the number of workers traveling on any given day was about the same as the number meant to be housed in the new building. By assigning open desks to employees on an as-needed basis, they saved the cost of a new building.
This may all sound esoteric and obscure, but it was mis-alignment between problem definitions and proposed building designs that helped motivate the US General Services Administration (GSA) to promote BIM technology for design, beginning in 2003 (Hagen 2009). The GSA specifies and justifies buildings based on spatial needs, both in terms of areas and circulation (many federal buildings are courthouses with intricate public, judicial, and defendant circulation systems), but proposals and plans submitted by architects were difficult to compare against the specifications. The first-phase BIM deliverable requirement for GSA projects beginning in 2007 was for a Spatial Program BIM, effectively requiring those aspects of a design to be interpretable by computer in a standardized way.
Generation and refinement of solutions within the design space is the subject of Chapter 12, but two points are important to note here. The first is that the way in which the problem is defined or described—Simon’s “existing conditions” and their “preferred” alternatives—goes a long way to determining the shape of the solution by establishing the vocabulary. This observation reinforces the importance of careful problem specification, especially if automated tools are to be used.
The second and related challenge of solutions to problems is “the stopping problem.” To be useful, an automated process should be able to recognize when a solution has been reached. Each iteration of the design is different, but not necessarily better, and there is no way to identify an objectively “best.” In the absence of a good stopping rule the design process may continue until some external event (budget or calendar) draws it to a close. Defining problems in such a way that a solution can be recognized when it emerges remains a grand challenge.
Design is a problem-solving behavior. Distinctions can be made between obvious, routine, and creative design problems, the latter requiring action of sufficient complexity that planning is necessary. Design expertise influences the perceived difficulty of the design problem. While it is generally asserted that problems can be divided into their natural parts during analysis or during the solution process, research into wicked and situated problems supports the notion of puzzle making or co-evolution of problem definitions and solutions. Characteristics by which the problem is defined become characteristics by which a solution is measured and constraints defined. These can be taken to define a design state space that is useful for conceptualizing the problem–solution pair. This space provides a domain over which a utility or objective function might be defined, providing a framework within which to assess solution quality and perform design iterations until a suitable design is recognized or a stopping rule satisfied.
Alexander, Christopher. 1964. Notes on the synthesis of form. Cambridge, MA: Harvard University Press.
Dorst, Kees. 2004. The problem of design problems, in Proceedings of design thinking research symposium 6. http://research.it.uts.edu.au/creative/design/index.htm.
Kalay, Yehuda, L. Swerdloff, and B. Majkowski. 1990. Process and knowledge in design computation. Journal of Architectural Education 43: 47–53.
Rittel, Horst and Melvin M. Webber. 1973. Dilemmas in a general theory of planning. Policy Sciences 4: 155–169.
Simon, Herbert. 1969. The sciences of the artificial. Cambridge, MA: MIT Press.
Suchman, Lucy. 1987. Plans and situated actions: The problem of human–machine communication. Cambridge: Cambridge University Press.
Alexander, Christopher. 1964. Notes on the synthesis of form. Cambridge, MA: Harvard University Press.
Archea, John. 1987. Puzzle-making: What architects do when no one is looking, in Computability of Design. Edited by Y. Kalay, 37–52. New York, NY: John Wiley.
Dorst, Kees. 2004. The problem of design problems, in Proceedings of design thinking research symposium 6. http://research.it.uts.edu.au/creative/design/index.htm.
Gero, John. 1994. Introduction: Creativity and cognition, in Artificial intelligence and creativity: An interdisciplinary approach. Edited by T. Dartnall, 259–268. Dordrecht: Kluwer.
Hagen, Stephen, Peggy Ho, and Charles Matta. 2009. BIM: The GSA story. Journal of Building Information Modeling (Spring): 28–29
Jackman, J., S. Ryan, S. Olafsson, and V. Dark. 2007. Meta-problem spaces and problem structure. In Learning to solve complex, scientific problems. Edited by D. H. Jonassen. Mahwa, NJ: Lawrence Erlbaum.
Kalay, Yehuda, L. Swerdloff, and B. Majkowski. 1990. Process and knowledge in design computation. Journal of Architectural Education 43: 47–53.
Newell, Alan and H. A. Simon. 1958. Elements of a theory of human problem solving. Psychological Review 65: 151–166.
Rittel, Horst and Melvin M. Webber. 1973. Dilemmas in a general theory of planning. Policy Sciences 4: 155–169.
Simon, Herbert. 1969. The sciences of the artificial. Cambridge, MA: MIT Press.
Suchman, Lucy. 1987. Plans and situated actions: The problem of human–machine communication. Cambridge: Cambridge University Press.
Woodbury, Robert and Andrew L. Burrow. 2006. Whither design space? AIE EDAM: Artificial Intelligence for Engineering Design, Analysis, and Manufacturing 20: 63–82.