annajordanous / A-Fitness-Function-for-Creativity-in-Jazz-Improvisation-and-Beyond

The code featured in this academic paper from 2010 - here is the abstract of the paper Can a computer evolve creative entities based on how creative they are? Taking the domain of jazz improvisation, this ongoing work investigates how creativity can be evolved and evaluated by a computational system. The aim is for the system to work with minimal human assistance, as autonomously as possible. The system employs a genetic algorithm to evolve musical parameters for algorithmic jazz music improvisation. For each set of parameters, several improvisations are generated. The fitness function of the genetic algorithm implements a set of criteria for creativity proposed by Graeme Ritchie. The evolution of the improvisation parameters is directed by the creativity demonstrated in the generated improvisations. From preliminary findings, whilst Ritchie's criteria does guide the system towards producing more acceptably pleasing and typical jazz music, the criteria (in their current form) rely too heavily on human intervention to be practically useful for computational evaluation of creativity. In pursuing more autonomous creativity assessment, however, this system is a promising testbed for examining alternative theories about how creativity could be evaluated computationally.