Genetic Algorithms as a Computational Theory of Conceptual Design

Abstract

The essentially inductive processes of conceptual design have received scant at- tention in those portions of the design literature concerned with effective com- putation or mathematical rigor. This paper draws a connection between the discriminative and recombinative processes of conceptual design and genetic al- gorithms (GAs)-search procedures based on the mechanics of natural genetics and natural selection. Recent empirical and theoretical results with a type of GA called a messy genetic algorithm support the conjecture that GAs can solve all problems no harder than the functions of bounded deception in time that grows no more quickly than a polynomial function of the number of decision variables. These results suggest that inductive designers-far from wasting their efforts when they bet on some combination of past designs-are engaging in a computa- tionally effective means of solving very difficult, even misleading, design problems. Although more theoretical and computational work is needed, the paper shows one path to a more rigorous theory of conceptual design, a path that should help put design on the same mathematical foundations as analysis without detracting from the joy, or the necessity, of human invention.