Design Tasks as Ill-defined Complex Problems

Pirita Seitamaa-Hakkarainen

Adapted from ”Weaving Design Process as a Dual space Search”, Seitamaa-Hakkarainen, Pirita, Department of Home Economics and Craft Science Research Reports 6. University of Helsinki, 2000.

Designing is generally considered to be a form of complex problem solving (Simon, 1969; Goel, 1995). In the most design studies, carried out within the framework of cognitive science, the analysis of design problem solving focuses on finding common features that can be generalized across different domains of design (Goel, 1995). Designing can be defined as a cognitive process intended to produce a solution to a design task.

The design task typically requires creation of an object that has some utility, some practical usefulness as well as a purely decorative or aesthetic purposes, in the most cases, it should have both. The effectiveness of the problem solving depends on how one approaches the problem situation (i.e., design task situation). A designer makes plans or prototypes for an end product, and in that process of design includes intentions and plans. These are then carried out in varying phases into a suggestion to be produced. Thus, designing is a conscious attempt to a solve design task by using relevant design methods and knowledge.

In general, aesthetic and functional dimensions of a design product in the professional context are assessed in terms of client and user satisfaction. However, any design task requires a very complicated process of searching for a workable (i.e., aesthetic as well as functional) solution that can be reached in a practical and effective way.

Designing involves various elements that must be considered and related to each other, within the constraints in order to create a functional and aesthetic solution. To conclude, it is important to analyze what kind of characteristics are included in design tasks, and what kinds of knowledge are required to solve them (Goel, 1995). In order to study design process one needs to analyze design tasks, their structure, and the context in which these occur.

Nature of Complex Problems

Problems which require a process of structuring and restructuring, in which solutions emerge only gradually through a process of defining external and internal constraints, and, moreover, which rely on an extensive body of domain knowledge are called complex problems (Adelson & Soloway, 1988); design tasks prototypical examples (Simon, 1969; Akin, 1986; Goel, 1995; Goldschmidt, 1997).

The designer is rarely in a position to identify all possible solutions to the task at hand. It is therefore most important to establish the choices that satisfy the requirements of solution (Akin, 1986; Goel & Pirolli, 1992). However, despite the ill-defined and complex character of the problem which designers face, they still usually manage to find solutions.

Defining the problem has several aspects. One must, initially, recognize that the problem exists in order to begin to solve it, and the way one defines the problem affects attempts to solve it. In this context, the concept of task environment (Newell & Simon, 1972) is essential.

The analysis of task environment means that the task has certain characteristics which make it more or less problematic for the problem solver. A task usually begins in a certain state with certain conditions, or certain pieces of information, which are present at the onset of work on the problem. Also, in the task environment it is desirable that the initial state be transformed to the goal state.

The designer, however, has at her disposal certain ways to change the given state to the goal state of the problem. According to Simon (1969) the initial state in problem solving refers to the situation at the outset of a problem, including any existing external constraints (such as time limits). The goal state refers to the desired outcome of the problem, and operators refers to the actions permitted or applied in order to solve a problem (Simon, 1969).

Different types of problems can be defined according to how well the initial state and goal state are specified (Goel, 1995). One useful approach to distinguish types of problems is to characterize them according to the type of goal. Problems can be found requiring only a single solution, and all the given information points toward that solution. Problems with one solution are called a well-defined problems or sometimes convergent and closed-ended problems. Ill-defined problems (labeled sometimes as divergent problems), on the other hand, do not have a single optimal solution, but rather the problem can lead to several different, equally correct solutions. Based on the openness or closedness of the definitions of the task environment, design problems are generally considered as ill-defined or ill-structured problems as distinguished from well-defined or well-structured problems (Simon, 1969, Akin, 1986).

Simon (1973) described ill-defined problems as those that are more complex, have less specific criteria for knowing when the problem is solved, and do not supply all the information required for solution. The distinction between ill- and well-defined problems is based on the amount of information guiding the search of solution given in the task environment.

Convention has been that in an ill-defined problem the information given in the task environment is either unknown or incoherent. Mayer (1983) presented the view that problems can be divided into at least four types, depending on how the initial state and goal state are determined. For example 1) the problem state and goal state are both well-defined, 2) the problem state is well-defined and goal state is ill-defined, 3) problem state is ill-defined and the goal state is well-defined, 4) problem state and goal state are both ill-defined. Thus, in each design problem it should be possible to determine how much information is specified at the onset of the tasks.

In the beginning of the design process, the initial situation is not usually completely specified; the goal state is never fully specified in advance nor are all requirements explicit. Thus, the path from initial state to goal situation is usually completely unknown (Simon, 1973).

The designer must generate and represent a great deal of additional information during the problem solving, in order to find plausible solutions. The designer does not work directly with the task environment; rather he or she forms a representation of the problem, based on perception and knowledge. The openness of the task implies a need for redefining and restructuring the task assignment several times.

Task redefinition involves the designer subjectively identifying and defining the task. Newell and Simon (1972) called this representation problem space. They defined problem space as a set of knowledge states and operators that move the problem solver from one knowledge state to another knowledge state.

Problem space also contains evaluation functions. However, when the initial state is usually vague and the goal state either unknown or ambiguous and no obvious solutions are available, a subject acquires more information during problem solving. Hence it becomes possible to change and modify the initial problem space.

The search process consists of progressive changes in the subject’s cognitive actions, which can be modeled as changes in knowledge states. Usually there are several actions or operators that can be applied. Some of them move the problem state closer to the solution, but some of them lead to dead ends.

Heuristic search strategies help the designer to focus attention on the part of the problem space most likely to contain a desired solution, and thus minimize the number of operations required to reach the solution (Akin 1986; Simon, 1973). On the basis of theoretical and practical background knowledge, heuristic methods guide thoughts and actions into promising search paths instead of those which are less productive.

Components of Complex Problems

Goel and Pirolli (1992) showed that there are important generalizations about design activity that apply across specific design fields. They conceptualized design-problem space by using several invariants, which they argued collectively constitute and structure a design-problem space, and are common to all typical design tasks.

The most important invariant has been labeled design constraints. According to Goel & Pirolli (1992; Goel, 1995) there are two kinds of constraints on design environments: ones that are negotiable and the others that are not (e.g., natural laws). The physical (internal) constraints broadly determine what kinds of design solutions are possible, but they are not immediate consequences of the design task itself.

Design constraints are not logical in nature, and are, therefore, more malleable. As a consequence, designers have a great tendency to change the problem situation, so that it conforms to their knowledge and experience. Goel and Pirolli (1992; Goel, 1995) outlined some unique properties of the design process; a personalized stop rule and a reversing direction of a transforming function, because in these the designer has freedom to change the initial situation and goal situation based on experience, possibly providing more effective solutions.

Chi argued (1997; see also Goel & Pirolli, 1992) that in the early investigations of the problem solving processes, which were based strongly on the information processing model, the tasks which were used were well-defined problems. Thus, the problem space was quite easy to derive and predict.

Research on those well-defined problems required that the problem space was first completely derived, i.e., all possible solution paths were described. After that, the path a particular subjects progresses along was identified, so that it was possible construct and test a runnable computational model (Chi, 1997). In an ill-defined design problem, on the other hand, it is impossible to determine the problem space. A designer must retrieve a great deal of technical, visual and procedural knowledge during the design process, as well as effectively frame problems, and produce appropriate solutions during design.

Therefore, in design problem-solving, solutions are never predictable. According to Goldschmidt (1997) these aspects turn designing into an indeterministic process which is very difficult to model.

Bromme and Tillema (1995) criticized much of the research of analyzing expertise is limited to either artificial or exemplary problems. According to Bromme and Tillema (1995; see also Bereiter & Scardamalia, 1993; Schön, 1983) professional knowledge develops as a product of professional action, and it emerges through work and performance in that profession, through the practical situations.

Professional knowledge progresses gradually in a process within a working context.. According to Bromme and Tillema (1995) this real-life situation is often neglected in expert research shaped by the information-processing approach, and expert research mainly clarifies the connection between complex knowledge structures and successful execution of complex tasks.

The traditional view of the research on cognition and expertise has recently been extended toward the study of physically and socially distributed cognition (Hutchins, 1991; Norman, 1993). Thus, the emphasis is placed more on the role and meaning of the social culture (i.e., division of the cognitive labor), social-culturally developed tools or physical artifacts that help one to take on more complicated tasks to solve.

Moreover, it has been shown that the multiple use of external representations (i.e., visualization and writing) reduces cognitive processing load and reasoning capacity (Norman, 1993; see also Fergusson, 1992;).


Adelson, B., & Soloway, E. (1988). A model of software design. In M.T.H. Chi,. R. Glaser. & M. Farr (Eds.), The nature of expertise (pp.185-208). Hillsdale, NJ: Lawrence Erlbaum.

Akin, Ö. (1986). Psychology of architectural design. London: Pion Limited.

Bereiter, C., & Scardamalia, M. (1993). Surpassing ourselves: An inquiry into the nature and implications of expertise. Chicago, IL: Open Court.

Bromme, R., & Tillema, H. (1995). Fusing experience and theory: the structure of professional knowledge. Learning and Instruction, 5, 261-267.

Chi, M. T. H. (1997). Quantifying qualitative analyses of verbal data: A practical guide. The Journal of the Learning Science, 6, 3, 271-315.

Ferguson, E. S. (1992). Engineering and the mind’s eye. Cambridge, MA: The MIT Press.

Goel, V. & Pirolli, P. (1992). The structure of design problem space. Cognitive Science, 16, 395-429.

Goel, V. (1995). Sketches of thought. Cambridge, MA: MIT Press.

Goldschmidt, G. (1997). Capturing indeterminism: Representation in the design problem space. Design Studies, 18, (4), 441-456.

Hutchins, E. (1991) The social organization of distributed cognition. In L. B. Resnick, J.M. Levine & S.D. Teasley (Eds.), Perspectives on socially shared cognition (pp. 283-307). Washington, DC: American Psychological Association.

Norman, D. A. (1993). Things that makes us smart. Defending human attributes in the age of the machine. New York: Addison-Wesley.

Mayer, R. E., (1983). Thinking, problem solving, cognition. NY: W.H. Freeman.

Newell, A. & Simon, H.A. (1972) Human problem solving. Englewood Cliffs, NJ: Prentice Hall.

Schön, D. A. (1983). The reflective practitioner: How professional think in action. New York: Basic Books.

Simon, H. A. (1969). The science of artificial (3rd ed. 1990). Cambridge, MA: The MIT Press.

Simon, H. A. (1973). The structure of ill-structured problems. Artificial Intelligence, 4, 181-204.