Mapping to musical actions in the FILTER system

In this paper we discuss aspects of our work in develop-ing performance systems that are geared towards human-machine co-performance with a particular emphasis on improvisation. We present one particular system, FILTER, which was created in the context of a larger project related to artificial intelligence and performance, and has been tested in the context of our electro-acoustic performance trio. We discuss how this timbrally rich and highly non-idiomatic musical context has challenged the design of the system, with particular emphasis on the mapping of machine listening parameters to higher-level behaviors of the system in such a way that spontaneity and creativity are encouraged while maintaining a sense of novel dialogue.