@Article{Gero1990,
author = {John S. Gero},
title = {Design Prototypes: A Knowledge Representation Schema for Design.},
journal = {AI Magazine},
year = {1990},
volume = {11},
number = {4},
pages = {26--36}
}
Design takes the goals defined by the problem and turns them into design descriptions, which describe how to build the required design artifact [p28].
Design, ultimately, has a goal, is constrained by the world, and involves making decisions, learning, and exploring [p28].
The naive model of design turns functionality into a design. Computer aided design turns structure into design. Bobrow (1984) creates a model that turns the goal of the designer into a set of behaviours that achieve the goal.[p28]
One can go on further from this to specify a model where the functionality translates to a set of behaviours, which are then translated to structures. Gero calls this synthesis [p29].
As a process, design encompasses these features: "... formulation, synthesis, analysis, evaluation, reformulation, and production of design description"[p29].
Design involves two main research issues: the frameworks to represent knowledge in the design process and appropriate transformation processes. The first is the only one addressed in the paper. [p29]
Designers schematize their knowledge about how to design. The three main classes of schemas are archetypes, prototypes, and stereotypes. Archetypes are paradigmatic cases of their class, essentially the types that exemplify the definition. Design may reuse but will generally not reproduce archetypes. Stereotypes are copies that are unchanged. Prototypes are the first that others are modelled on. [p30].
There seems to be no one schema that can satisfy all the requirements of design. [p30]
Design prototypes are schemas that bring together information from all relevant schemas into one general design schema for a specific design situation. They use prototype theory (Osherman and Smith 1991) and scripts (Schank and Abelson 1975).[p31]
Prototype theory determines the membership of a concept by its relation to the exemplar of its type. [p31].
The structure of design prototypes can be broken down into five basic types, apart from knowledge of the design prototype itself: [p32]
1) Relational knowledge - the dependencies between the various elements in the design (variables, functions, etc).
2) Qualitative knowledge - an adjunct to relational knowledge that reflects what occurs when elements are changed (including what changes can be made, such as ranges for variables).
3) Computational knowledge - the mathematical relations between the various elements of the design.
4) Constraints - what the design is constrained by, such as the expected behaviour when one is considering functional elements.
5) Context knowledge - what must come from or is determined by things external to the design itself.
There are three major types of design [p34 - 35]:
Routine: retrieve a design prototype, generate instances, and prune to the one that best fits the purpose.
Innovative: retrieval of design prototypes with alteration of the knowledge contained.
Creative: the introduction of new variables into a design prototype or, in rare cases, the creation of an entirely new prototype.