3 Lessons of Genetic Algorithms for Computational Innovation

For some time, I have been struck by the connection between the mechanics of innovation and genetic algorithms (GAs)-search procedures based on the mechanics of natural selection and genetics. In this short paper, I explore those connections by invoking the fundamental metaphor of innovation as an explanation for GA power of effect. Thereafter, I reverse the argument, by setting out to construct competent GAs-GAs that solve hard problems quickly, reliably, and accurately-through a combination of effective (1) design methodology, (2) design theory, and (3) design. While, we won’t have the opportunity to review the technical lessons in detail, the abstract does examine three crucial qualitative issues: (1) the key race between selection and the innovation operators, (2) the idea of a control map that helps us understand the genetic algorithm’s sweet spot, and (3) the primary hurdle or impediment to competent GA design, a hurdle that has been overcome by three different algorithms that obey the same principle: the need to identify important substructures before deciding among them. The implications of these lessons in practical GA design and the construction of a computational theory of innovation are briefly explored.