Content area
Abstract
This dissertation builds toward a theory of the evolution of modularity, by investigating specific techniques for promoting the evolution of different kinds of modularity, and using them to discuss a larger framework. It examines environmental and biological-process-inspired methods for promoting the evolution of modularity, both within different levels of abstraction and in terms of hierarchical organization, and makes a case for the specific mechanisms involved. Sequential-subtask environments are found to promote the evolution of both single-level and hierarchical modularity. Adding development to the evolutionary process also promotes the evolution of hierarchical modularity. When sequential-subtask environments and development are combined, the results are strongly hierarchically modular. Development appears to prevent single-level modularity from dropping over the course of evolution. Using number of species evolved as a proxy for the complexity of the evolutionary search space (fitness landscape) suggests that in sequential-subtask environments the search space is more complex compared to that in single-task or parallel-task ones, and in evolutionary developmental systems compared to evolutionary ones. A taxonomy of different methods for promoting modularity is described, and the results of my experiments are placed within that taxonomy.
