Formalizing Conceptual Design Around Decisions and Their Impact in the Design Space

Bill Wood
Center for Design Research
Stanford University

Conceptual design, the earliest phase of the design process, has been identified as a key component in the final outcome of the design process. Formalizing conceptual design poses many problems: Ambiguity and uncertainty pervade the design evaluation model, increasingly so in team-based design settings. The design process must progress from abstract concepts to a concrete description of the behavior or form of candidate designs. The space of these candidates is quite large and may not be well known to the designers.

Decision making is the primary activity of the design team throughout the design process but is more far-reaching that just decisions about artifact form or behavior. Decisions pervade the design process: what are the important aspects of a design, how can they be formalized for comparing among design alternatives, what information might be useful, is it worth developing this information (in the form of prototyping, analysis, focus groups, etc.). Decision-based design must embrace all of these varied perspectives on design decision making.

My work in this area concerns the application of decision theory in early stages of design. The specific approach relies on a statistical model of the design space, a means of controlling the horizon of issues explored within this design space, and the application of formal decision methods to suggest the best path for the design process. The major issues of each of these components are discussed:

Modeling the design space: The options for modeling the design space are manifold. The most prevalent methods rely on parameter vectors which represent the form of a design. These vectors are augmented with design evaluation features calculated from the description. Problems with this method derive from its ability to capture ambiguity in the design evaluation and the ability to accurately model mathematically all relevant features. Concurrent engineering and life-cycle design introduce features which are difficult to model, in the language of optimization they are non-operational. Design experience (e.g. feedback from manufacturing, input from component catalogs, intermediate results from analysis, etc.) must be composed into a description of the design space which makes virtually any description or evaluation operational. Care must be taken not to introduce unsubstantiated bias into the design space approximation (e.g. assuming quadratic relationships among design parameters).

Adding abstraction to performance as decision foci: One of the basic decision paths in design is that from abstract concept (e.g. motor) to concrete instances (e.g. DC, permanent magnet, rare earth, brushless, frameless motor and controller). Just as the description of a design proceeds from the abstract to the concrete, the evaluation of a design runs from the abstract to the specific. The two are coupled: as a design becomes more concrete, the evaluation function can become more specific.

Application of decision theory to conceptual design: Given an appropriate model for the design space and some generic decision foci, decision theory can be applied to design and the design process. Information value theory can be used to understand the impact of uncertainty in the design state due to abstraction as well as ambiguity (i.e. absence of an aspect of the design from the evaluation model).

The overall process of design is shaped by these three components. Designers begin with abstract, high level objectives and abstract candidate designs. Information value form the design space raises issues to the the design team that could bear consideration (i.e. inclusion in the design evaluation). The design evaluation shapes the progress from abstract to concrete design candidates. Design team focus is constantly directed toward developing information which impacts design decisions. A least commitment design strategy evolves from this combination of methods, a strategy that is normative for conceptual design.