Despite a long his tory of generative practices in music, cre a- tion of large

scale form has tended to remain under direct control of the composer. In the field of musical metacre a- tion, such interaction is generally perceived to be desirable, due to the complexity of ge n eratin g form, and the necessary aesthetic decisions involved in its creation. However, the requirements for dynamic musical generation in games, i n- stallations, as well as in performance, point to a need for greater autonomy of creative systems in generating larg e

scale structure. This position paper surveys the complexity of musical form and existing approaches to its generation, and posits potential methods for more computationally cre a- tive procedures to address this open problem