The task of a design support system is conventionally conceived as one of providing the designer with solutions to specific parts of a design problem. In this paper we argue that this approach is fundamentally flawed. We identify two main modalities of support: the production of necessary consequences and the production of possible consequences of the current design description, and discuss the problem of devising an architecture capable of providing such support using the Edinburgh Designer System (EDS) as an example. We describe the difficulties inherent in integrating the derivation of possible consequences into the architecture of EDS and argue that while in principle such difficulties can be overcome, in practice the goal of providing globally consistent solutions to particular parts of the design problem is unattainable. We propose a different approach in which the design support system explores the consequences of various design decisions. The results of this exploration are represented as counterfactual conditionals which we believe more closely approximates the information required by a designer.
[1]
Tim Smithers,et al.
Design is intelligent behaviour, but what's the formalism?
,
1990,
Artif. Intell. Eng. Des. Anal. Manuf..
[2]
Oskar Dressler.
Problem Solving with the NM-ATMS
,
1990,
ECAI.
[3]
Johan de Kleer,et al.
Problem Solving with the ATMS
,
1986,
Artif. Intell..
[4]
Brian Logan,et al.
Being economical with the truth: assumption-based context management in the Edinburgh Designer System
,
1991
.
[5]
A. Newell.
Unified Theories of Cognition
,
1990
.
[6]
Raymond Reiter,et al.
A Logic for Default Reasoning
,
1987,
Artif. Intell..
[7]
Tim Smithers,et al.
Creativity and design as exploration
,
1993
.
[8]
J. P. Tsang.
A combined generative and patching approach to automate design by assembly
,
1991
.
[9]
Brian Logan,et al.
Design as intelligent behaviour: An AI in design research programme
,
1990,
Artif. Intell. Eng..