An `AI in Design' View of Design
Problem Description

Design Type

Characterize the design problem along the abstractness dimension.

Discussion. Many design systems generate a full design starting from a partially specified design belonging to one of the above enumerated types [Brown 1995]. Most of the design systems so far have been successful in the areas of parametric and attribute (configuration) design, where the problems are specified more precisely. There are also systems which were developed to transform designs from one type to another (e.g. functional type => embodiment type). The more abstract (i.e., less concrete) the initial stage of the design, the more powerful the heuristics and the design modelling tools need to be [Freeman & Newell 1972] [Goel 1997].

Characterize the design problem along the routineness dimensions.

Discussion. This attribute is determined by the type, amount, and organization of the relevant knowledge that the system possesses. Routine design systems rely on experiential knowledge [Brown 1995]. That problem solving process is based on retrieving and directly using well-organized knowledge that is suited to the task. Non-routine design reasoning is always based on combining knowledge in new ways [Dyer et al. 1986] [Hodges et al. 1992] [Williams 1992] [Goel 1997], and probably involves planning. The majority of design systems still tend to be routine design systems. Non-routine design assumes a potential to construct new design solutions using various reasoning processes [Cagan & Agogino 1991] [Murthy & Addanki 1987] [Zhao & Maher 1992].

Up : Next

/Papers/IJDC/sect3.1.html Fri May 23 17:44:09 EDT 1997