Agent-based support for interactive search in conceptual software engineering design

While recent attempts to search a conceptual software engineering design search space with multi-objective evolutionary algorithms have yielded promising results, the practical application of such search-based techniques remains to be addressed. This paper reports initial findings of the application of software agents in support of an interactive, user-centered conceptual software design scenario. The supporting role of a number of single responsibility agents is described and results for a case study indicate that the application of such agents to search-based design scenarios provides efficient, high performance and effective support. The notion of interactive, joint human-computer activity appears to map well to conceptual software design scenarios: focus on superior design concepts and thence to useful and interesting designs provides a natural and effective way of narrowing the population based search. In addition, agents and the human designer appear to interact as cooperative "team players", jointly influencing the evolutionary algorithm based search. Nevertheless, challenges remain, including expanding the scale of implementation of underlying technologies to support distributed, collaborative design.