Published in "Editor's Notes", in Computer Music Journal, MIT Press, v. 16, #3 Fall 1992.
With regard to the the idea of a "meaningful taxonomy" of interactive computer-based methods of musical creation, it should be borne in mind that terms such as "intelligent instrument", "algorithmic composition", and "interactive composition program" came into being at a point in computer music history when the primary reason for such terminology was the simple need to communicate to others what unprecedented things we were trying to do, in the absence of any established terminology describing creative automation in music.
Differences implied by such terms were often highly influenced by now long-past technological limits such as realtime computer throughput: An "instrument" was something which could generate sonic response interactively in realtime. "Algorithmic composition" often resulted in printed note-lists or non-realtime sound computation, precluding interactivity. "Intelligent instruments" were distinguished from other computer-based performance instruments by their greater use of encoded decision-making logic relative to stored pre-specified note/event data.
Now that the available technology is much more powerful and widespread, and the field of creative possibilities is so much more densely populated, such factors as degree of realtime interactivity, allocation of subtasks to person or machine, and path of information flow are more often determined by human design than technical limit. Terms which were adequately descriptive when first coined are now insufficient for differentiation within the much larger range of approaches now actively being pursued.
Rather than a taxonomy in the Aristotelian sense (finite categories with defined boundaries, usually hierarchical in structure), I would suggest we explore a representation modeling musical creative processes as a multi-dimensional space, in which methods and systems can be localized as positions and movements along various continua encompassing all characteristics of importance. Vocabulary and language inevitably imply conceptual models (Sapir-Whorf). I would suggest starting a new nomenclature with a conceptual model first, and then allowing new vocabulary to result from it. Such a representation of creativity automation would best be evolved through dialogue among all involved in our field and others, so as to encompass not only the full range of approaches now actively pursued with computers in our art and other arts, but all possibilities, past and future - insofar as is possible.
Such axes as are suggested below, as possible examples for such a multidimensional model, involve many non-quantifiable variables. Therefore positioning of individual creative methods and systems along the various axes of such a matrix would be relative rather than absolute, and would also best be represented as vectors of relative length (ranges along each continuum) rather than as point locations on the axes.
Axes of such a multi-dimensional representation of interactive musical generation might include such continua as the following: degree of human participation (completely passive listening versus total creative responsibility); amount of physical coordination, practice, and/or prior musical knowledge required for human interaction; number of variables manipulable in realtime; number of variables manipulable by the user (not the coder) prior to reatime output ("configurability"); amount of time typically needed to learn to use a system; balance of task allocation (human versus computer) in decision-making in the compositional realms of pitch, timbre, articulation, macrostructure, et cetera, and/or in labor-intensive tasks such as notation or transcription; extensiveness of software-encoded representation of musical knowledge (materials, structures, procedures); predictability and repeatability (versus randomness from the user's viewpoint) of musical result from a specific repeatable human interaction; inherent potential for variety (output as variations of a recognizable piece or style, versus source not being recognizable by listeners); ratio of user's non-realtime preparation time to realtime musical output; degree of parallelization of human participation; degree of parallelization of automated processing; number of discrete non-intercommunicative stages of sequential processing (e.g. composer-performer-listener or coder-user-listener versus integrated develpment and use by single person or small community); degree of multi-directionality of information flow; degree of parallelization of information flow; openness versus finiteness of form, duration and logic system; et cetera.
The function of such a multi-dimensional representation, or even of a community-wide dialogue attempting to model one, would exceed that of a standard taxonomy. It would allow easy perception of areas of the creative matrix which have not yet been explored, or which have proved musically productive or counterproductive historically, facilitating investigation as to why. It would help identify preference ranges for such factors as the number of independent variables with which a typical player/user comfortably interacts in realtime. Ideally, a sufficiently general representation involving musical task allocation, process structure, and information flow, would encompass the range of creational possibilities we have inherited from previous eras and alternative cultures, as well as undiscovered potentials, and allow greater understanding of all music making processes.
- Laurie Spiegel
New York City, NY