The Case for Concise Computational Creativity Systems

Abstract

In formalizing the notion of creativity – defining that concept so that it can be dealt with in the context of computing – researchers have sought to develop the most concise possible definition that captures the essence of creativity. But in building computational creativity systems, these same researchers are content to develop extremely elaborate and complex systems that implement a cornucopia of different ideas and approaches, or perhaps implement single ideas that are themselves very complex. Using examples from the literary arts, I argue that it is helpful to write very small programs that model specific techniques in different domains of creativity. These programs can then be used to identify how those specific techniques contribute to creativity, serving as the foundation for further work. By aiming for brevity and developing small, elegant systems, researchers can explore new aesthetic territory and better relate their results to the particular ways in which their systems operate. Keywords: constraint, brevity, modeling, domain-specific techniques, story generation, poetry generation