Freedom and Necessity in Computer Aided Composition: A Thinking Framework and its Application

Johannes Kretz Zentrum fr innovative Musiktechnologie (ZiMT) der Universitt fr Musik und darstellende Kunst Wien, Vienna, Austria Email: zimt@mdw.ac.at
Abstract: This paper presents some of the authors experiences with computer aided composition (CAC): the modeling of physical movements is used to obtain plausible musical gestures in interaction with constraint programming (rule based expert systems) in order to achieve precisely structured, consistent musical material with strong inner logic and syntax in pitch material. The "Constraints Engine" by Michael Laurson implemented in OpenMusic (IRCAM) or PWGL (Sibelius Academy) can be used to set up an interactive framework for composition, which offers a balance of freedom (allowing chance operations and arbitrary decisions of the composer) and necessity (through strict rules as well as through criteria for optimization). Computer Aided Composition is moving far beyond being algorithmic or mechanical. This paper proposes an approach based on evolutionary epistemology (by the Austrian biologist and philosopher Rupert Riedl). The aim is a holistic synthesis of artistic freedom and coherent structures similar to the grown order of nature. Keywords: Computer Aided Composition, CAC, physical modeling, music, composition, artificial intelligence, expert systems, evolutionary epistemology. "Things derive their being and nature by mutual dependence and are nothing in themselves." (Nagajuna, second century Buddhist philosopher) "Nothing, in the development of this world, can be understood in isolated manner." (Rupert Riedl, 1999)

1. Introduction
One of the particular challenges of composing is (and always was) finding the balance between simplicity and complexity, between freedom and necessity, and which is even more difficult the creation of a piece of art, that can be perceived at various levels of attention, for example in that way, that the most obvious level (the surface) can be understood and enjoyed CHAOTIC MODELING & SIMULATION - Papers from CHAOS 2010 International Conference http://www.cmsim.net/sitebuildercontent/sitebuilderfiles/conf_papers.html

Kretz, Johannes easily without special pre-knowledge (even intuitively), but other, deeper, less obvious layers of the work provide more subtle contents satisfying the more sophisticated interests of the connoisseurs. Besides it turns out to be particularly difficult to formalize a sophisticated process of creation, where various structural layers of material from the surface to the inner structures have to interact permanently, and where various parameters (pitch, time) in various layers of complexity (melodies, chords, rhythmic cells, formal parts) influence each other. In this paper we present a philosophical framework, allowing to represent musical thinking in an organic and meaningful way, giving the artist tools, where freedom, random choices and arbitrariness have its place equally to well structured organization of elements. Finally we will also give examples of the frameworks practical application with the help of computers.

2. From Aristotle to Evolutionary Epistemology

2.1. Humanities versus Natural Sciences In music, like in many other fields of human activity, the question of causality is essential. A music, which gives us the impression of absence of causes and effects is usually perceived as unsatisfying by listener. Therefore it might be worth looking in detail into a model of understanding causality in general. Since the age of enlightenment we are confronted with two opposing thinking paradigms, fundamentally separating the scientific field of humanities from the natural sciences. The approach of humanities (fig.1) is mainly deductive, where truths are derived from abstract, fundamental certainties, which cannot be put into question. Based on these truths a hierarchical view on the world is unfolded by progressive specialization through deduction, providing a universe of nested classifications. On the other hand natural sciences (fig.2) build their view of the world empirically from the most elementary observations of physics, also establishing a hierarchical view of the world in nested classifications by progressive spezilalization, just starting from the other end. This leads in the case of humanities to a world, where everything has a final sense but lacks objectivity, since the fundamental thruths are given on the basis of religion, culture, ideology or coincidence, being not always the most objective guides through life. In the case of natural sciences on the other hand, we obtain a maximum degree of objectivity, but questions about the higher sense and goals can only be answered very vaguely. Since the last decades of the 20th century it becomes more and more obvious that a more sophisticated multi-causal framework of thinking would be very useful, which possibly could help to overcome the splitting of our world into humanities and natural sciences.

Freedom and Necessity in Computer Aided Composition

Figure 1. Viewpoint of Humanities

Figure 2. Viewpoint of Natural Sciences

2.2. The four causes of Aristotle In this context the differentiated thinking model of the four causalities by Aristotle described around 350 B.C. in his book on Metaphysics [1] offers an approach, which we also find in the core of Evolutionary Epistemology [2] [3] [4] [5] [6] [7]. Aristotle mentions four causes, which interact in any development of complex entities: causa efficiens (energy) causa finalis (targets) causa materialis (bricks, elements) causa formalis (construction plan)

Causa efficiens represents effects of energy and matches mainly with the causality model of physics (fig 2.). Causa finalis on the other hand is the final aim or purpose of human activity and corresponds to the causality model of humanities (fig.1). Additionally there is another pair of antagonistic causes: Causa materialis describes the effects of the building material on the built objects, while causa formalis describes the effects of the construction plan on the built sub-entities. This framework can be applied very well on composing music (see fig.3). Here causa efficiens is creative energy, the desire for experimentation, even the creative use of mistakes and errors (in the language of biology: mutations). On the other side causa finalis provides the aesthetic targets, the purpose of the music, (which can even be extra-musical like in dance music, opera, sacred music etc.). Nevertheless, composing would be extremely difficult, if it had to happen only between the causalities of the two causes, as would biological evolution be extremely slow in its success, if only mutation and selection would steer it. The chances are just to small that one can write a large work of music just by try and error. Therefore both in evolution theory [2] and in music it is important to introduce a second pair of causes: The creation of materials, bricks, elements in music: motives, patterns, chords, rhythmical cells etc. makes the creative process more efficient. Of course it poses some constraints for the possible results, but nevertheless the

Kretz, Johannes increased efficiency and speed of creation outweighs the loss of possibilities by far, both, in biology and in composition. The hierarchical nested interaction of various layers of construction entities ranging from the level of materials (perspective of causa materialis) to the level of a construction plan (causa formalis) allows the creation of extended works with a reasonable amount of effort.

Figure 3. interaction of Aristotles causes

In addition both, in traditional composition and in computer aided composition as well as in biological evolution [4], another aspect of causality has to be highlighted: If n elements (musical entities like notes, motives, chords etc.) are interacting, and each entity is interacting with each other, the number of possible interactions increases enormously for growing n. It is represented by the triangular number (handshake problem):
T n 1 2 3 ... (n 1) n n2 n 2

This means that interaction is getting too complex, when the number of entities is exceeding a certain number. (Compare fig.4). A reasonable solution for this would be differentiating between internal and external interactions of entities forming groups of entities, interacting with other groups of entities, on various levels of hierarchy (see fig. 5 to 7).

Figure 4. increasing complexity

Figure 5. internal / external interaction

Figure 6. network of networks

Figure 7. network of networks of networks

Freedom and Necessity in Computer Aided Composition

3. Musical application
Musical gestures representing physical movement have always played a certain role in music history. Also in contemporary computer aided composition the simulation of physical movements can serve as a main source of convincing plausible gestural data. Since we are used to understand this physical behavior of objects in the real world, the simulation of these movements with the means of music makes a lot of sense for our perception, too. These gestures can follow the known laws of physic (gravity, aerodynamics, etc. like in fig.8, where two bouncing rubber balls are simulated) or even add some creativity like in fig.9, where the two simulated rubber balls are in love, attracting each other, so that their movements turn into a more interesting dance, where each ball tries to meet the other, while their inertia mostly prevents an actual meeting.

Figure 8. Physical modeling

Figure 9. Creative modeling

When applying these gestures to musical material, it is evident that a direct translation of the position parameters into the domain of chromatic pitch would produce flat, rather unsatisfying musical results. While the curves can be useful to shape (or influence) the surface of music, other tools are needed to control the inner, structural side of music in interaction with these curves. (See fig. 10 and 11.) One of the most promising methods for this is constraint programming. The user creates in interaction with the computer an expert system, where desired results are obtained by defining a search space (the basic material for possible solutions), a set of strict rules, allowing or forbidding certain characteristics of the solution, also (if needed) heuristic rules, which favor certain criteria of optimization. This approach originating from programming languages like PROLOG [8] was already implemented in IRCAMs Patchwork environment [9]. The same engine was then ported to OpenMusic [10] by rjan Sandred, and is finally a core part of todays most advanced CAC environment PWGL [11] by Michael Laurson. Note, that this approach of combining elements from a search space

Kretz, Johannes according to strict and heuristic rules corresponds very well with the interaction of causa materialis (the effects of the characteristics of the search space and the strict rules) and causa formalis (the steering towards certain desired global qualities with the help of heuristic rules). The following examples are aimed to illustrate this approach. In the case of second horizon for piano and orchestra by Johannes Kretz the search space consisted of a set of allowed chords with all their inversions, octave permutations and transpositions. Additional constraints were applied: The range of the piano should not be exceeded. Two adjacent chords should have not more the three common notes, but at least one. The shape of the rubber ball movement was implemented as heuristic rule, so that the engine always chose a solution as close as possible to the ideal shape with the help of a system of bonus points. But the other (strict) rules had absolute priority. Fig.12 and 13 show possible results. It can be clearly seen, that the original shape is not represented perfectly, and that some of the inner structural necessities cause compromises of the shape. But exactly this compromise between outer and inner conditions of creation gives interesting, organic musical results. Compare [12] and [13] for details.

Figure 10. physical modeling of a bouncing ball

Figure 11. representation of fig. 8 in music notation

Figure 12. interaction of the shape of fig.8 with harmonic rules

Figure 13. Effect of additional rhythmical rules

Freedom and Necessity in Computer Aided Composition Figures 14-20 show examples taken from the composition full scope fancy (2004) by Johannes Kretz. Here the search space was derived from a dodecaphonic row (fig. 14), which was used to generate a pool of chords (fig.15 and 16). The curves in fig. 17 and 18 should be approximated by the outer two voices of the final result, but since they were given only as heuristic rules, while the inner structure of the chords and melodies was defined by strict rules, we obtain both, an approximate physical behavior on the surface and a strict inner syntax in the details. (Compare fig.19 - 21).

Figure 14: dodecaphonic row

Figure 15. chord series derived from fig. 14.

Figure 16: extended chord series as search space

Figure 17. bass line: ballistic curve

Figure 18: treble line: ballistic curve

Figure 19. approximation through structured musical material

Kretz, Johannes

Figure 20. Beginning of fig. 19 in detail Figure 21 again shows the difference between the result of physical modeling alone and the result of the interaction between physical modeling and syntactical strict rules.

Figure 21. pure physical movement vs. structured approximation

5. Conclusions
The described method of computer aided composition differs significantly from more traditional approaches like algorithmic or stochastic composition. While a certain degree of randomness is allowed, when strict or heuristic rules dont suggest any difference between allowed decisions, the main focus here is the interaction of models in two levels: Firstly the interaction between strict rules and criteria of optimization has the potential to find solutions, that would be very hard to obtain by composing only with paper and pencil or other methods of computation. Secondly another iterative process between the computer and the composer controls the evolution of music: typically rules are added or modified after listening to the results of computation, until they match well with the aesthetics of the composer. In this way the results are approximating better and better to the desired results in an evolutionary process similar to the biological one.

Freedom and Necessity in Computer Aided Composition

References
[1] Aristotle: Methaphysics, book V 2, 350 B.C. [2] Rupert Riedl: Evolution und Erkenntnis, Munich 1982. [3] Rupert Riedl, (translated by R.P.S. Jefferies) Order in Living Organisms: A Systems Analysis of Evolution, orginally: Die Ordnung des Lebendigen. Munich 1975. [4] Rupert Riedl: Die Strategie der Genesis. Munich 1976. [5] Rupert Riedl: Biology of Knowledge, the evolutionary basis of reason, Chichester, John Wiley & Sons 1984. [6] Rupert Riedl: Zufall, Chaos, Sinn. Stuttgart 1999 [7] Rupert Riedl: Darwin, Zeus und Russels Huhn, Vienna 1994. [8] Balaban, M. / Murray, N.V.: Machine Toungues X: Prolog. In: CMJ Vol.9 No. 3 Fall 1985, Cambridge 1985. [9] Laurson, M. / Duthen, J.: Patchwork, a graphical language in Pre-Form. In: T.Wells and D. Butler (Ed.): Proc. of the 1989 ICMC. San Francisco 1989: p.172175 [10] Agon, C.: OpenMusic: Un langage visuel pour la composition musicale assist e par ordinateur, PhD Thesis, IRCAM -- Univ. Paris 6, 1998. [11] Laurson, M. and Kuuskankare, M.: PWGL: A Novel Visual Language based on Common Lisp, CLOS and OpenGL. In Proc. of ICMC02, p.142-145, Gothenburg, Sweden 2002. [12] Kretz, Johannes: Navigation of Structured Material in second horizon for Piano and Orchestra (2002). Publ in: Agon C. et al.(eds.): The OM Composers book 1. IRCAM, Paris 2006. p.107-127. [13] Kretz, Johannes: Continuous Gestures of Structured Material, Experiences in Computer Aided Composition, publ. in: Baboni-Schinlingi, J. (ed.): Prisma 01. Milano 2003. p.185-194.