Computer-Assisted Algorithmic Compositional

Approaches:
The Role of AI in Music Creative Practices

Mingmei Hao

A research paper submitted to the University of Dublin,

in partial fulfilment of the requirements for the degree of

Master of Science Interactive Digital Media

2020
Declaration

I have read and I understand the plagiarism provisions in the General Regulations of the University
Calendar for the current year, found at: http://www.tcd.ie/calendar

I have also completed the Online Tutorial on avoiding plagiarism ‘Ready, Steady, Write’, located at
http://tcdie.libguides.com/plagiarism/ready-steady-write

I declare that the work described in this research paper is, except where otherwise stated, entirely
my own work and has not been submitted as an exercise for a degree at this or any other university.

Signed: ___________________

20th May 2020

i
Permission to lend and/or copy

I agree that Trinity College Library may lend or copy this research Paper upon request.

Signed: ___________________

20th May 2020

ii
Acknowledgements

I would like to express my upmost thanks to my supervisor, Dr Natasa Paterson-Paulberg, for her
support and encouragement during the research.

iii
Abstract

Algorithm had rich precedents in music, and composers’ relationships with algorithms have been
witnessed since ancient times. With the advent of computers, music was forced to go beyond the
boundaries of the traditional musical languages. The mid-1950s was a transitional period between
traditional and innovative modes of thoughts in compositional practices. Music technology
stimulated the growth of music while altering musical forms. During the second half of the 20th
century, the positive results from the musical experiments have demonstrated the flexibility and
generality of computer-assisted compositional methods. This study offers a deep view of a wide
range of approaches for algorithmic compositions. In the modern landscape of algorithmic music,
the various approaches of algorithmic composition have opened new directions in music, including
the stochastic method, rule-structured technique and Artificial Intelligence (AI) -based systems. The
computer, as an assistant of the human composer, was designed to execute rules, learn rules and
create rules.

This study focuses on the musical evolution from the topics of musical patterns, algorithms, AI,
computational creativity to collaborative intelligence, by examining composer’s perspectives on
what compositional technique they approach, and how they approach it. A number of musical works
have demonstrated the power of human-AI interaction. Composers are seeking sophisticated,
advanced and intelligent approaches to extend the potential of computer-assisted music. The
implemented algorithm can represent composers’ creative processes, and algorithmic composition
is considered as a mode of creativity. From the current millennium, there is a growing trend for
composers to embrace state-of-art technologies and to infuse AI languages into music creative
practices to complement their musical minds.

iv
Table of Contents

List of Figures.............................................................................................................. vii

List of Abbreviations .................................................................................................. viii

Chapter 1. Introduction.................................................................................................. 1
1.1 Overview .............................................................................................................................. 1
1.2 Goals and Motivation ........................................................................................................... 2
1.3 Outline.................................................................................................................................. 3

Chapter 2. Background: The Evolution of AI ............................................................... 4

2.1. Introduction .......................................................................................................................... 4
2.2. Origins of Algorithmic Thinking ......................................................................................... 4
2.3. From Computers to Music Technology ............................................................................... 6
2.4. The Modern Landscape of Algorithmic Music and AI ........................................................ 9
2.4.1. The Stochastic Method........................................................................................ 9
2.4.2. The Rule-Structured Technique ........................................................................ 10
2.4.3. AI-based systems .............................................................................................. 11
2.5. Conclusion ......................................................................................................................... 14

Chapter 3. Methodology .............................................................................................. 15

3.1. Introduction ........................................................................................................................ 15
3.2. Development and Approach of the Questionnaire ............................................................. 15
3.3. Participants ......................................................................................................................... 16
3.4. Method ............................................................................................................................... 16
3.5. Data Collection .................................................................................................................. 17
3.6. Conclusion ......................................................................................................................... 18

Chapter 4. Findings and Reflection ............................................................................. 19

4.1. Introduction ........................................................................................................................ 19
4.2. Algorithmic Practice in Contemporary Music ................................................................... 19
4.2.1. The Use of Music Software and Programming Languages .............................. 19
4.2.2. Indeterminacy, Randomization and Recombination ......................................... 22
4.3. Up to my f-holes ................................................................................................................ 24
4.4. Turing Test: Text to Music #1 ........................................................................................... 28

v
 4.5. Future Prospects ................................................................................................................. 31
4.6. Reflection ........................................................................................................................... 33
4.7. Conclusion ......................................................................................................................... 34

Chapter 5. Conclusion ................................................................................................. 35

5.1. Summary ............................................................................................................................ 35
5.2. Contributions...................................................................................................................... 36
5.3. Limitations and Further Research ...................................................................................... 37

Appendices .................................................................................................................. 38
Appendix i. Sample Information Table..................................................................................... 38
Appendix ii. Participant Information Sheet ............................................................................... 38
Appendix iii. Sample of Questions ............................................................................................. 42

References ................................................................................................................... 43

vi
List of Figures

Figure 1: Score of Up to my f-holes, page one (Ring 2008) .............................................................. 25

Figure 2: Score of Up to my f-holes, page two (Ring 2008) .............................................................. 26
Figure 3: Score of Up to my f-holes, page three (Ring 2008) ............................................................ 27
Figure 4: A screenshot of Turing Test system demo (Adams 2019) ................................................. 28
Figure 5: A screenshot of minimilate object demo (Adams 2019) .................................................... 29
Figure 6: An image of William Dowdall’s performance in Contemporary Music Centre’s Salon
2016 (CMC 2016) .............................................................................................................................. 30

vii
List of Abbreviations

AI Artificial Intelligence
AIVA Artificial Intelligence Virtual Artist
EMI Experiments in Music Intelligence
MAGE Music Algorithm Generation Engine
MIDI Musical Instrument Digital Interface

viii
 Introduction

Chapter 1. Introduction

1.1 Overview

Algorithmic composition is an automation of the process of composing music using algorithms

(Maurer 1999). With the evolution of computational techniques, new methods of music production
were extensively explored and music was produced experimentally through the aid of digital tools.
An algorithmic approach to music goes back many centuries. From early examples throughout the
history, basic algorithmic solutions were employed in the algorithmic-like processes. Dannenberg
hints that the concepts of algorithms existed long before the advent of digital computing and
programming languages (Dannenberg 2000, p.1). W. A. Mozart's Musical Dice Game (1792) is one
of the earliest examples of algorithmic compositions. Trace back to the late eighteen century, the
game was implemented in a re-combinatorial process, as such the method involves the combination
of pre-composed phrases. The procedures of Mozart's music-generating techniques indicates the
significance of early algorithmic composition. In the late 20th century, with the extensive use of
computers, algorithms were applied on composing programs and compositionally-related tools. The
concept of computer-assisted composition refers to the music production process when composers
employ a series of rules. These rules are known as algorithms, with the term being used among
composers, engineers and scientists (Cope 2000, p.3). In the field of computer music, the notion of
algorithm becomes conceptual. Algorithm, as a distinctive, musical grammar, is used for the
production of musical structures. Algorithms in music are utilized in various ways, such as sound
synthesis, sampling and composing. With the development of theoretical and applied technologies,
many composers and programmers explored the greatest possibility from music to mathematics, and
far beyond. Music, math and patterns are closely linked to explain the rationale and logic behind the
practices.

Roads points out that 1968 marks the beginning of modern research into artificial intelligence and
music (Roads 1980, p.15). Forte argues that, the questions in music analysis would gradually
become questions in the field of AI (Forte 1967). By using state-of-art algorithms, computers
behave intelligently and creatively. As many higher level concepts emerged, such as interactive
sound, artificial creativity and musical intelligence, they are far more complex than the classic
algorithms and mathematical frameworks. However, since the onset of the digital transformation
and revolution, some major critics had arisen. The emergence of AI has been a challenge in the
music industry. Some believe the virtues of technology are over praised, and that AI is killing

1
 Introduction
creativity. It was argued that at the terminus of technology, people will eventually go back to pen
and paper. In terms of audience reception in algorithmic culture, it has been postulated that the
initial purpose of composing music using computers is not aesthetically pleasing to its audience.

Although many concerns have not been addressed yet, the volume of AI-related discussions are
growing. As Cope argues, algorithmic procedures already exist in the creative process, the computer
is just a tool (Cope 2000, p.2). Caianiello suggests that the computer is an extension of creativity,
and the difference of human and machine intelligence should be obscured (Caianiello 1984, p.30).
As Roads points out, in every situation where the idea of musical intelligence is proposed, new
possibilities and further directions become clear (Roads 1980, p.23). The discussion on the
technological change and the role of computer music continues to expand, and provides a different
understanding of human musical cognition and machine intelligence.

1.2 Goals and Motivation

The process of composing algorithmic composition involves the use of formal procedures. This
study attempts to interpret what are the “formal procedures” and how do “algorithms or rules” work
within a creative setting. The purpose of the study is to excavate algorithmic techniques from pre-
computer age to post-digital age, and to explore how different algorithmic approaches are shaped by
social discourse. This paper addresses the questions on how algorithms and technologies subvert the
conventional means of how music is produced. The study then recognises the need to gather
subjective reviews and critical reflections from experienced individuals, in order to receive a variety
of responses of different musical approaches. In order to respond to the need, the study aims to gain
insights into composers’ point of views of computer-assisted composition and how much of the
piece is effected by this. With the method of qualitative research, the aim can be achieved through
contributions from expert composers. With the development of artificial intelligence, many
composers implemented AI techniques to their music creative processes. The motivation of the
research is to provide a broad understanding on algorithmic compositional practices and the
application of AI techniques to computer music. The ultimate goals of the study is to understand the
role of AI in contemporary music as well as the influence of AI techniques to modern human
composers. Other aspects of AI techniques are the possible results of human-AI interaction and
collaborative intelligence. The intersection of music, computational technology and AI is becoming
a compelling area to explore. The study aims to combine multidisciplinary research (mainly focus

2
 Introduction
on art, computer science, mathematics and music) and comprehensive investigations of music
professions to provide a broad understanding on the relationship of algorithms, artificial intelligence
and human.

1.3 Outline

This paper contains five chapters including an introduction, the background, methodology, findings
and reflection, and a conclusion. Chapter 1 is the introduction to the paper, creates an overview of
algorithmic composition from the pre-computer age to the digital age, and introduces research aims
as well as the structure of the paper. Chapter 2 provides a historical viewpoint on the development
of algorithmic practices in music, drawing from a selection of literature and sources. Further,
chapter 3 examines the process of conducting the research and the implementation of the
questionnaire. Chapter 4 presents the analysis and interpretation of the findings, accompanied by
musical examples. Chapter 5 unites previously discussed findings and algorithmic compositional
approaches, concludes the paper, and provides an outlook toward future research.

3
 Background: The Evolution of AI

Chapter 2. Background: The Evolution of AI

2.1. Introduction

Drawing from a selection of literature and sources, this chapter offers a historical viewpoint on the
development of algorithmic practice in music. The chapter is organised into five sections. The first
section gives a brief introduction of the chapter, and the second section discusses the origin of
algorithmically derived composition. The third section covers the topics from the background of
computers to music technology, and the fourth section shows a modern landscape of algorithmic
music, as well as the current trends and future possibilities of AI in music practices. The last section
concludes the history of algorithmic composition in a nutshell.

2.2. Origins of Algorithmic Thinking

Algorithm, can be found across cultures in ancient era as early as 1200 BC, in Babylonian times.
The term algorithm is derived from the older forms of Greek word “arithmos” and Arabic word
“algorism”, which literally means a series of numbers (Cope 2000, p.2). It originates from a
mathematical book called Rules of Restoration and Reduction in the ninth century. The word’s
original meaning is linked to an algebraic system. In the late twelfth century, the rules of Hindu-
Arabic numerals and the variants of algorism were translated into Latin, and the word began to
spread into Europe. The form and meaning of the word algorithm evolved over millennia of time.
The modern meaning of algorithm wasn't introduced until nineteenth century. The ancient definition
of algorithm and its subsequent derivations represent a remarkable modernity in terms of the
mathematical logic (Boyer 1951).

In the field of music, some aspects of music have long been involved with algorithmic segments.
Algorithm had rich precedents in music, and musicians’ relationships with algorithms have been
witnessed in the history of music. The origin of algorithmic thinking in music dated back to the
beginning of the fifth century BC. Around 500 BC, Greek philosopher and mathematician
Pythagoras found music has mathematical foundations, and the mathematical characters exist in
basic musical concords (Riedweg 2008, p.2). He convinced that music is convertible into numbers
by any forms, based on the discovery that the intervals in music originate with numbers. An octave
(1:1), perfect fifth (3:2) and fourth (4:3) all have integer ratios (Crocker 1963, p.197). Pythagoras
revealed the nature of music, without inventing sophisticated mathematics. While the “simplest
mathematics” in musical intervals later expressed as arithmetic ratios. This numerical interpretation
could be demonstrated on monochord, an ancient single-string instrument invented by Pythagoras.

4
 Background: The Evolution of AI
The instrument consists of a string, a moveable bridge and a sound box. It explains how the natural
interval relationship is linked with the length of the string vibrating and the pitch produced
(Riedweg 2008, p.27). Monochord has become an early evidence of the evolving awareness on
Pythagorean musicology. The ancient scientific method of measuring intervals and pitches was
developing and evolving over the centuries. Monochord had been revived and used as an illustrating
tool throughout the Middle Ages. In the nineteenth century, several modern divisions of the
monochord had been proposed based on the old mechanism. As a progenitor of mathematics and
music, Pythagoras’s legendary discovery of the numerical relationship in music had a major impact
on the modern shape of algorithmic composed music.

In the eighteenth century, Musikalisches Würfelspiel (Musical Dice Game) was an attempt of an
algorithmically derived composition in the Classical Period. However, the notion of algorithm was
not yet defined at that time. Since the dawn of music, musicians have been employing methods
which are often considered as algorithmic in some sense. Algorithm was used as a broad term and
Mozart's Musical Dice Game is one of the most often cited examples (Dawin 2010, p.2). The rule of
the game is to use a dice to produce musically randomly in a re-combinatorial process, whereby the
pre-composed phrases are combined at random. The process is based on a Markov (or Markoff)
chain model, which is a mathematical system experiences transitions within two states (Dawin 2010,
p.4). The result of composing and re-composing tends to be an infinity as long as the dice remains
rolling. The probability of transitioning is a stochastic process. The principle of Markov chain is
memorylessness, which means the transition probability depends solely on the current state
(Srivastava 2014). The memory-less property enables the model to be analysed mathematically. The
Markov chain is named after mathematician Andrey Markov and developed in the nineteenth
century. Today the framework and its applications are widely used in algorithmic composition and
machine learning intelligence, while in Mozart’s era the blueprint of algorithmic composition
practice remained unveiled over centuries. The method implemented in the musical dice game was
a precedent on later explorations, although the specific terms were not defined at that time.

Besides Pythagoras and Mozart, Johann Sebastian Bach’s approach to canonic composition is
another precedent on algorithmic procedures for music generation. Bach embodies the genres of
canon and fugue. Canon is a contrapuntal technique, it refers to the rule of continuous imitation
among a composition (Collins 1995, p.93). The early definition of canon indicates the genre of the
puzzle canon. In a composition the initial melody is devised first, then it transposed to the

5
 Background: The Evolution of AI
remaining parts of the piece. The time intervals in-between must be guessed enigmatically. In the
puzzle canon, the entire composition is constructed in this way. Many of Bach’s canons were
composed in this style, they are considered as parts of a long tradition of contrapuntal approaches in
the Baroque era (Collins 1999, p.27). During the performance, musicians interpret the original score
and solve the musical riddles in canonic variations. The composition The Musical Offering (1747) is
one of Bach’s remarkable contributions in a fugal form. Another notable work is The Art of Fugue
(1740), an unfinished piece written in the eighteenth century. Bach coded his name (B-flat, A, C,
and H) and embedded in the music as a hidden motif (Simoni 2003). The Art of Fugue represents
the apotheosis of musical procedures on the imitative counterpoint. During the course of a
composition, the transposition and modulation processes are indispensable parts of algorithmic
practices. For Bach and the other early eighteenth-century classical music composers, the rules
applied to the canonic melodies are indeed the algorithm.

In the early 20th century, modern music reaches a turning point. According to Ashby, “music
composition in analogy to science, has evolved to a process of problem solving” (Ashby 2001,
p.585). The idea of scientific models and compositional techniques have offered solutions to the
musical problems. Arnold Schoenberg was one of the advocates. In 1921, Schoenberg composed
the first piece using twelve pitch classes. Later in 1923, he refined the method and devised a twelve-
tone system, a compositional technique of serializing musical elements, such as pitches, rhythms,
and dynamics. The regular use of notes, tones and series became influential on atonal music and
serial music (Kostka 2016, p.198). Schoenberg’s serialism is a method to integrate musical
parameters into a particular series. It is a significant modern experiment of composing and it is
connected with the basics of algorithmic composition. The twelve-tone method was exclusively
used for many decades. One of Schoenberg’s students Scott Bradley, an American composer,
applied the twelve-tone technique for scoring animations and films. During the 1950s, Bradley
practiced the technique in the musical scores of the classical Hollywood cartoon, Tom and Jerry
(1940-1958) (Goldmark 2007, p.70). One example was in a scene from “Puttin' on the Dog” (1953),
Bradley used twelve-tone scale for a disguised dog. The twelve-tone method was a prelude of
aesthetic and intellectual transitions to modern music. The introduction of twelve-tone serialism has
brought the musical composition into the discourse of intellectual history (Ashby 2001, p.586).

2.3. From Computers to Music Technology

In the history of music, the connection between music and mathematics predates musical computer
science (Collins 2018, p.1). By the mid-20th century, with the development of technology and the

6
 Background: The Evolution of AI
advent of computers, music was forced to go beyond the boundaries of the traditional musical
languages (Milstein 1992, p.62). Around the time of World War II, the decade was a transitional
period between traditional and innovative modes in music. In 1948, “musique concrete”, literally
means concrete music, was developed by French composer Pierre Schaeffer (Manning 2004, p.19).
Musique concréte was an early form of electronic music. In the early 1950s, with the establishment
of music studios in Cologne, the concept of “elektronische musik”, literally means electronic music,
was created by German composer Herbert Eimert. Despite the debatable relationship between
musique concréte and elektronische musik (Manning 2004, p.68), the concrete and synthetic
techniques were combined by many avant-garde composers. They mixed instruments, voices,
electronic sounds, and non-musical elements into their compositions (Maurer 1999).

The avant-garde, a term originated from French, referring to a group of vanguards or the practice of
experimental treatments in terms of social reforms. The praxis of avant-garde challenged the
traditional forms of literature, science as well as musical arts. The avant-gardists in music refer to
tradition-breakers who practice new compositional strategies. During the post-World War II period,
many radical compositions were produced, and music was heavily shaped by a variety of rules
(Jakubowska 2011). Among a group of pioneers, American composer Henry Cowell and his pupil
John Cage were profoundly influential in the mid-20th century. Henry Cowell is one of the earliest
composers intended to explicitly explain the musical materials (Cowell and Godwin 1969, p.xvi).
As a prolific composer, he composed over seven hundred musical works as well as many academic
publications (Sitsky 2002, p.115). By the late-1910s, Henry Cowell wrote his first avant-garde
piano piece, The Tides of Manaunaun (1917). Around that time, he became acquainted with many
experimental musicians and modernists, including Arnold Schoenberg, who devised the twelve-tone
method (Sitsky 2002, p.112). Cowell’s thoughts on radical compositional approaches were
explained in his later published book New Musical Resources in 1930. He created a repertoire of his
unconventional innovations and established a relation between the “theory of musical relativity”
and his music (Cowell and Nicholls 1996, p.xi). By the end of 1931, Cowell designed the first
electronic drum machine, “Rhythmicon”, with scientist Leon Theremin (Sitsky 2002, p.114). He
composed several original compositions for the instrument, including Ryhthmicana (1931).
Ryhthmicana was the first musical work in record that exploiting the fundamentals of the modern
electronic technology (Smith 1973, p.145). Cowell was also known for his early efforts on the
approach of indeterminacy, and the idea was further developed by John Cage. John Cage was
another avant-garde composer of post-war avant-garde. His music was especially influenced by
Henry Cowell’s avant-garde piano repertoire. 4’33’’ (1952) is an often cited example of avant-garde

7
 Background: The Evolution of AI
music. It is a silent piece consists of three movements, and it was premiered in 1952, with the title
refering to the length of the performance, which is four minutes and thirty-three seconds. 4’33’’ is
the epitome of Cage’s rejection of tradition (Sitsky 2002, p.xiii). Cage challenged conventional
music by applying an inventive, unprecedented and contentious method. He believed that silence is
an organized sound, and his musical work embraced silence as a musical gesture (Sitsky 2002, p.6).
In 1951, Cage composed a solo piano piece, Music of Changes (1951) using I Ching charts. I Ching
is ancient Chinese text used for divination. Cage used it as a tool and approached composing music
by similarly using chance elements. Music of Changes is the first instrumental work that involves
chance and mechanical procedure, and it is a remarkable piece using indeterminate compositional
technique (Sitsky 2002, p.94). As a leading figure of the progressive avant-garde movement, John
Cage redefined the limit of music understanding in a cultural context. However, the avant-garde
music was deprecated as aggressively unattractive by some critics. Indeed, the mid-1950s was a
transitional period between the old and new music forms. Composers invented various “calculating
tools” to organise sound, from rule-based instrumental piece (The Tides of Manaunaun), the
musical imitation on the silence of nature (4’33’’) to the change-controlled music (Music of
Changes), creating radical compositional techniques and challenging musical tradition (Sitsky 2002,
p.95). Furthermore, the development of technology has opened a new field of composing. The
avant-gardists’ musical ideas and approaches are now place within a broader context of musical
modernism (Cowell and Nicholls 1996, p.113).

As Manning asserts, the development of music in any period is subject to the technological progress
which it parallels (Manning 2004, p.8). From both a technical and musical point of view, music
technology stimulates the growth of music while altering musical forms. A major breakthrough of
computer music occurred in the 1950s. Lejaren Hiller is an American composer and professor. He is
a leading figure of experimental music and has created a number of pioneering works during the
20th century. In 1952, as ILLIAC I, the first series of supercomputers ILLIAC became operational,
Hiller decided to compose a musical score with the assistance of a machine. In 1956, he composed a
piece “Illiac Suite” in collaboration with Leonard Issacson using the computer ILLIAC I. Illiac
Suite is the earliest composition generated with the aid of a computer (Higgins and Kahn 2012,
p.150). In 1958, Hiller founded Experimental Music Studios, the first canonical electronic studio in
America. In the late 1960s, HPSCHD was a subsequent work he created with John Cage. HPSCHD
consisted of harpsichord solo and electronic sounds, it was a pioneering and ambitious piece built
using computers. HPSCHD’s premiere performance, occured on May 1969, and brought computer-
assisted composition into the public sphere for the first time (Manning 2004, p.201). Another

8
 Background: The Evolution of AI
computer program developed in the 1960s was MUSICOMP, one of the first computer systems for
automated composition. It was created by Lejaren Hiller and Robert Baker, and it took many years
to programme (Hiller and Baker 1964, p.62). Computer Cantata (1963) was a composition by
MUSICOMP, it was comprised of electronic sounds, natural sounds and human voice (Higgins and
Kahn 2012, p.149). The original purpose of creating Computer Cantata was to test the experimental
logic and the potential combination of concrete and synthetic sounds. Hiller’s compositions were
dependent on digital methods, many of them were created in rule-based computational systems.
During the following decade, Hiller, Cage and other composers continued to explore various
techniques and styles, and many musical works were created. The compositions were based on a set
of rules and procedures, following electronic and mechanical approaches. The digital revolution in
music during the 20th century was considerably inspired by the invention of the computer. It has
brought musicians and technology together into a field of computer music composition. During the
second half of the 20th century, the positive results from the musical experiments demonstrated the
flexibility and generality of computer-assisted compositional methods, and provided a possibility
for the further development on a more sophisticated, advanced musical logic: algorithmic
composition (Hiller and Baker 1964, p.62).

2.4. The Modern Landscape of Algorithmic Music and AI

2.4.1. The Stochastic Method

From the early efforts on formalism to algorithmic music, composers imposed maximal complexity
in terms of the development of more advanced machine models. According to Schwanauer and
Levitt, computer-assisted algorithmic composition is based on three approaches: random generation
of note attributes, concise and powerful grammars, and intelligent mathematical models
(Schwanauer and Levitt 1993, p385). The stochastics method, the rule-structured technique and the
artificial intelligence system are current approaches to algorithmically generated music.

By the late-20th century, the well-defined compositional models were established with programming
generative algorithms. Iannis Xenakis was a post-World War II algorithmic composer, he is known
for his stochastic compositions during the 1960s. In his ground-breaking book Formalized Music
(1963), he asserts stochastics is the foundation for composing (Manning 2004, p.87). Stochastics is
a term derived from mathematics. Xenakis furthered the basics of stochastics and applied this to his
composing process. The computer was used as a tool to deduce a piece from data’s structure using
probability weightings. The creative decisions were determined by chance and the probabilistic
factors. The concept of randomness has existed since early times, it can be found in Mozart's

9
 Background: The Evolution of AI
Musical Dice Game and John Cage’s chance-oriented music whereby his I Ching charts also
implied the idea of random variables. By the late 1960s, Xenakis employed the stochastic method in
both instrumental and electronic compositions. The stochastic approach was considered as one of
the applications in algorithmic compositional environments. In terms of the relationship between
human and machine, Xenakis believes the output of the machine is a set of computational materials
which aids in composing music. In contrast, Lejaren Hiller, a composer who approaches a rule-
based technique to his compositions, instead integrates the human contribution and computer
generation process together (Maurer 1999). However Xenakis continued to explore the possibility
of mathematical models as an integral part of the composing process itself. In 1966, Xenakis
founded EMAMu, an institute of interdisciplinary research for music and other forms of arts
(Manning 2004, p.87).

2.4.2. The Rule-Structured Technique

Another approach to algorithmic composition is a rule-structured technique. Early efforts on the use
of rules include Bach’s canon and fugue, Arnold Schoenberg’s twelve-tone system and serialism.
Lejaren Hiller is an example of modern composers who developing complex rule systems. As
opposed to the stochastic method, the composing process in a rule-based system is not chance-
oriented (Nierhaus 2009, p.239). The data is constructed in a formal, pre-decided procedure, and the
musical grammar guides the structure of the composition. The specific musical patterns are unified
as database or subroutines with the composer programming the rule into a computer system, such as
Hiller’s MUSICOMP. Another example is CHORAL, a system created by Kemal Ebcioglu.
CHORAL used BSL (Backtracking Specification Language), a logical programming language to
harmonize chorales of Bach’s style (Schwanauer and Levitt 1993, p385). During the compositional
process, Ebcioglu developed a rule-based method and divided the chorale into manageable units.

By formalizing and geometrizing the music, the different methods (stochastics and rule-structured)
applied to the mentioned examples have broadened the conception of computer-assisted algorithmic
composition (Xenakis 1963, p.ix). Inspired by the algorithmic methodology and rule structures,
some algorithmic composers started to explore further possibilities of musical intelligence from a
different perspective. They aimed to combine AI methods with musical applications. Applying AI
to algorithmic composition is a unique method in music production and during the mid-20th century,
electronic and computer music composers were seeking for a human-like response from the
computer systems (Maurer 1999). The development of the intelligent, responsive system often
associated with the applications of artificial intelligence in music production process. In 1957,
Hiller and Issacson’s composed Illiac Suite, the first composition using AI method. In 1960,
10
 Background: The Evolution of AI
Russian researcher Zaripov published the first paper about AI-assisted music (Zaripov 1960, p.479).
In the late 1960s and early 1970s, many programs were developed using formal grammars
augmented by heuristic techniques. At the beginning of the decade, the emergence of AI-related
music theory was significantly notable. The theoretical foundations of artificial intelligence
languages were established by that time period. In 1967, American musicologist Allen Forte
concluded that musical analytical questions would become questions of AI (Schwanauer and Levitt
1993, p.3). The year 1968 marks the formal intersection of music and artificial intelligence. Shortly
after Forte, two important papers were published that year. The first scholarly paper “Pattern in
Music”, was written by Herbert Simon and Richard Sumner (Roads 1980, p.15). They suggested
that music composition is the process of pattern processing. Simon and Sumner’s study draws a
direct parallel of pattern induction and sequence extrapolation of the intelligence test (Rowe 2001,
p.169). The second paper published in 1968 was Terry Winograd’s “Linguistics and the Computer
Analysis of Tonal Harmony”. Winograd’s theory was based on the model of systemic grammar. He
implemented the method into a harmony-analysis system (Roads 1980, p.16). In both models, they
established a paradigm for proposing musical formalism. Computers are designed to identify and
learn musical patterns, they are well-equipped to be involved in the musical discourse (Rowe 2001,
p.169).

2.4.3. AI-based systems

The fundamental of the stochastic and ruled-based techniques is similar to the third method,
artificial intelligence. However in AI systems, computers have a greater capacity to create their own
musical formulas (Maurer 1999). The intelligent systems are often associated with the process of
machine learning. For both researchers and composers, the learning computer opened a new
possibility in music creative process. David Cope is one of the practitioners of applying AI
techniques to music. He is a leading figure on the advancement of artificially composed music.
Begin in the early 1980s, Cope has been developing his expert system EMI (Experiments in Music
Intelligence). Cope defined the function of an expert system is to incorporate artificial intelligent
techniques to emulate a human expert (Cope 1987, p.30). EMI is a collection of computer programs
for the simulation of musical styles (Cope 1999, p.79). It composes music by analysing the musical
patterns through database and re-composing phrases in that style. Based on a pattern-matching
process, EMI is programmed to re-assemble each component to a new piece (Cope 1999, p.79).
Cope introduced the concept of recombinancy, and he believes it exists within every creative
process (Cope 2018). He implemented the recombinancy technique using the association network.
The association network is a database where the musical information is analysed and recombined,
and a new composition is formed automatically using the given subroutines. In the following

11
 Background: The Evolution of AI
decade, Cope explained EMI’s methodology in his 1991 published book Computers and Musical
Style (1991) and its sequel Experiments in Musical Intelligence (1996). From 1981 to 2003, EMI’s
career produced a considerable amount of classical music compositions within two decades. Cope
programmed augmented transition networks to produce music in the style of Bach, Chopin,
Beethoven, Brahms, Stravinsky, Gershwin, Mahler, and many more. In the context of artificial
intelligence, EMI is based on the framework of a musical Turing test (Nierhaus 2009, p.122). Some
listeners couldn't distinguish EMI generated music from the authentic human-produced classical
music. For example, EMI’s simulated style of Bach is virtually identical with Bach’s original works
(Nierhaus 2009, p.124). In 2003, Cope began developing a new system “Emily Howell”. As a result,
the early 21st century marks the end of EMI’s era. “From Darkness, Light” was Emily’s first album
released in 2009, including three preludes and fugues. Unlike EMI, Emily’s music doesn't remain
particular styles, it has the ability to generate new styles. Emily creates a computer model of
musical creativity, and its output becomes unpredictable. The composing process integrates pattern-
matching and recombinance in Cope’s association network. From the late 20th century to the early
21st century, Cope have produced many pioneering compositions based on computer-assisted
simulation of specific musical styles. He also explored new artificial intelligence languages to
integrate with human minds (Cope 2018). After EMI and Emily, Cope further explored aspects of
machine learning and algorithmic intelligence. He acknowledges the algorithmic nature of his
musical works, and he believes that computer will continue to play a significant role in music
production during the next millennium (Cope 1999, p.82).

Another AI method is genetic programming, a technique that the computer evolves to solve its own
problem (Keats 2006, p.72). It seeks to imitate nature’s evolution to generate programs artificially.
Genetic programming is based on the principles of heuristic technique. It was developed by
computer scientist John Koza and he patented the invention in 1988. Genetic programming is a
paradigmatic algorithm that can apply to specific domains such as music generation. This technique
has been applied to musical tasks in the areas of composing as well as improvising (Alpern 1995,
p.12). GenJam is a model developed in a framework of genetic programming for bebop
improvisations. Bebop is a modern style of jazz emerged in the 1940s, it often features fast tempos.
GenJam was created by John Biles and he has been developing the program since 1993 (Biles 1994,
p.131). The system operates based on learning and breeding, which includes the process of
initialization, selection, reproduction, crossover, and others. An initial jazz melody is mapped in the
population as the parent program, then the new melody is sent to the fitness function as the children
output of the initial individual (Biles 1994, p.131). Biles explored beyond the relationship between

12
 Background: The Evolution of AI
music generation and machine learning. Machine learning algorithms are used as creative
compositional tools, it is a ground-breaking approach from executing rules, learning rules to
creating rules (Fiebrink and Caramiaux 2018, p.3). Genetic programming transcends the machine
learning capabilities, it is a combination of genetic algorithms and artificial intelligence (Keats 2006,
p.70).

From the musical intelligence systems to the genetic programming, the basics of two branches of AI
methods are similar. The AI compositional techniques have been successfully applied in the field of
music as music can be composed using digital algorithms in an iterative process. In the 1980s,
Cope’s EMI model successfully structured classical music. Cope’s model is very similar to the
structure of Markov chain (Nierhaus 2009, p.124). In the late 1980s, John Koza foresaw an
inevitable growth of integrating music technology and intelligent systems. He could be viewed as a
prophet of the underlying power of AI, ultimately believing that human beings will live in the age
of creative machines. The different approaches (stochastics method, rule-structured technique and
AI system) to algorithmic composition are distinctive in terms of the algorithmic means and output,
however in addition to the three approaches mentioned above, there is trend of a hybrid approach to
algorithmic compositions for automated music. “Music Algorithm Generation Engine (MAGE)” is
an experimentation on such method. MAGE is a system that combines several AI methods, includes
stochastic techniques and genetic algorithms to generate music piece (Crawford and Fox 2016,
p.215). Besides the compositional approaches based on score, musical patterns and genres, some
systems focus on sound synthesis and MIDI tools to generate musical materials (Maurer 1999). An
example from recent years includes AIVA, the first virtual electronic composer created in 2016
(AIVA 2019). AIVA composes music for film soundtracks, games, commercials and many more
fields. Another example is MuseNet, a system developed by OpenAI in 2019. By analysing a MIDI
database, MuseNet is capable of composing multi-instrumental music in different styles. In April
2020, OpenAI introduced an advanced system Jukebox, which is designed to generate music cross
genres in various modes (OpenAI 2020).

Since the 1950s, composers have employed single or combined methods to compose music with the
assistance of different intelligent systems. However, as Forte hints, the interaction between music,
mathematics and machines has been controversial (Forte 1967, p.32). Music technology has been a
threat to earlier traditions. It is not easy to generalize the relationship between a computer’s
capabilities and human creativity. Indeed, how AI methods are compatible with human input is a

13
 Background: The Evolution of AI
question which has been of interest for a long time. Some musicians have given reluctant answers
on the role of the computer and AI-oriented music, many finding it difficult to balance computer
assistance and human input. Since the mid-20th century, from over five decades of numerous
examples and experiments, composers’ attempts weren’t proved as great successes artistically and
technologically, and the criticism persists. However the experiments are important in terms of the
development of computational techniques. In a larger context, collaborative intelligence becomes a
possibility between composer and computer. As Cope asserts, AI is not replacing the human, it is
adhere to human (Cope 2018). In a modern landscape of algorithmic music, the various techniques
and methods of algorithmic composition have opened new directions in music. Stochastic
processing, rule-based systems and AI techniques are pioneering methods that expand the potential
of computer-assisted music.

2.5. Conclusion

This chapter draws from the historical view on the development of algorithmic composition, from
the early times, the information age to recent periods. To conclude, the use of random processes in
music is significantly notable before the advent of computers. Algorithms and the human have had a
long and deep relationship in the history of music. This association has been existing across cultures
and across eras. From the 1950s to 1980s, composers attempted to encode musical knowledge and
patterns into the algorithm itself. Algorithm is becoming an indispensable part in the composing
process, it is a process of recombining musical DNA. The intelligent systems evolved from
executing rules, learning rules to creating rules. An AI system can be based on a database of various
compositional strategies. The stochastics method, rule-structured technique and AI system are three
major approaches. The revolutionary approach to AI assisted music adds extra dimensions to
algorithmic compositions.

During the recent decades, despite the controversy and critics, numerous computer systems and
compositions demonstrated the power of human-AI interaction, with resultant musical works
proving a positive response to collaborative intelligence. As Cope asserts, musical intelligence is an
extension of the human’s mind. From the current millennium, there is a growing trend for
composers to embrace state-of-art technologies and to infuse AI languages into musical applications
to achieve new results, and they are seeking for novel, sophisticated and intelligent responses from
machines.

14
 Methodology

Chapter 3. Methodology

3.1. Introduction

This chapter presents the process of conducting the research and the implementation of the
questionnaire. The chapter is organised into five sections. The first section creates an overview of
the chapter, and the second section discusses the development of the proposed research method. The
third section covers the topic of the recruitment of participants and sampling strategies, and the
fourth section examines the collection and management of qualitative data. The last sections
concludes the rationale of the applied research method and the measuring instrument.

3.2. Development and Approach of the Questionnaire

The paper aims to gain insights into composers’ perspectives of computer-assisted compositions,
and how much the musical pieces are impacted by technology. Due to the nature of the research, a
questionnaire with qualitative dimension is considered appropriated. With a qualitative research
methodology, it enables the researcher to gather data and professional judgments in a coherent way
(Kouritzin et al., 2009, p.176). The approach of the inquiry is established on the immediate need for
qualitative studies of expert composers. Since the selection of the research instrument is determined
by the primary purpose and scope of the research, the discussion of sampling is essential. This paper
aims at a specific population, and there is a strong focus on small-scaled sample groups.
Considering the traditional distinction between quantitative and qualitative method is the sampling
types: either probability or purposive samples in the survey, the research recognises the usefulness
of purposive samples (Vogt et al., 2012, p.220).

At a practical level, the survey research is particularly appropriate to the investigation of various
approaches of compositions and how music is produced in a contemporary context. In order to
create the basic parameters of the proposed research tool, in the practice of the research, it is
important to plan an effective method and follow three key procedures “concepts, reliability and
validity”. To increase the reliability of the research, the researcher needs to employ a valid method
to define the concept (Vogt et al., 2012, p.319).

15
 Methodology
3.3. Participants

In order to gather idiosyncratic knowledge from sophisticated composers, a qualitative research

instrument is implemented. It allows the researcher to gather viewpoints through a small group of
musical professionals. By orchestrating the research framework, the survey investigates in-depth
case studies of the chosen participants, which interprets the true value of the research (Nuhn et al.,
2002). Seven composers were recommended by college staff as the possible participants, since the
availability of participants is limited in the experienced individuals. They were contacted via email
with an attached research outline and a brief introduction of the researcher. Three of them expressed
their interests and accepted the initial invitation. They were: Hugh O'Neill, Judith Ring and
Sebastian Adams (See Appendix i sample information table for details). Due to the non-sensitive
topics and non-vulnerable populations involved in the subject matter, participants did not find
themselves uncomfortable of the research topic and no ethical concern arose. Moreover, the
research does not intend to focus on a deep analysis on gender differences. Instead, the recruitment
focus on the suitability of participants based on musical knowledge, therefore the sampling is not
dichotomized into gender groups. The small sample reflects a relatively healthy gender balance: two
male and one female. In terms of the age considerations, all participants are mature and they are
aged between 29 and 48. Since the participants come from different backgrounds and age groups,
the diversity of sample profiles increases the representativeness of the qualitative data.

3.4. Method

The research was implemented by the means of questionnaire. Participants were recruited through
email communications, and they were not required to present physically as a group. Each
participant was communicated individually via email. Their contributions were based on the format
of the questionnaire. The questionnaire was designed in a form which allows in-depth explorations
of participants’ inputs, with the expectation to maximize the sampling results. By employing
qualitative research strategies, all questions were designed open-ended for a detailed evaluation of
the answers (See Appendix iii for details). The advantages of the open-ended questions include the
possibility to investigate the answers that the respondents give spontaneously (Denscombe 2008,
p.360). Open-ended questions tend to produce an added diversity of responses compared with the
close-ended form. In terms of the length of the answers, respondents are more likely to type long
and detailed answers for the non-paper based questionnaire. Considering of various approaches on
questionnaire structures, the research therefore is best suited to conduct non-paper based open-
ended questions. The questionnaire consisted of ten questions. All questions were based on the
topics of different compositional approaches in the context of contemporary music practice, which

16
 Methodology
identified in the background chapter. Participants were informed that the questionnaire would take
approximately thirty minutes to complete, and it may take longer depends on the answers. In terms
of the format of the answers, any forms of inputs were encouraged, including written essay, musical
score, recording, video material, sample of composition, and other musical examples. Participant
information sheet, informed consent form and other written documents were provided electronically
(See Appendix ii for details), as part of the practicing of social distancing guidance under COVID-
19 circumstances.

The questionnaire was distributed by email, and each participant received identical questions.
Considering of the busy schedules of participants, they were not required to respond the questions
elaborately, however they were provided the opportunity to answer tailoring subsequent questions.
After a week’s time, the questionnaire results were emailed to the researcher as agreed. During the
research conducting process, the researcher followed four basic procedures as mentioned by
Rogelberg. First, no harm should come to a participant during the research. Second, the participants
are aware of the potential consequences and the consent form should sent to them. Third, the
participants understand they voluntarily agree to take part in the research. Fourth, the measures of
accessing data should make sure the data is stored confidentially (Rogelberg 2004, p.180).

3.5. Data Collection

The answers from the questionnaire were received in the format of written email, and the data was
already transcribed. Most of the data format was textual based, and the answers were organised into
a docx document. Besides written responses, other forms of data, including musical examples and
related information were assembled in a folder. All data collected from the questionnaire was stored
in a Google Drive folder, and it was used for solo purpose of research analysis. Nobody outside the
research will have access to the data, and the electronic data stored online will not be used for future
studies without permission of participants.

Furthermore, considering the common limitations of qualitative method, the results from the
questionnaire were carefully evaluated to make sure they were representative and generalizable. As
Rogelberg asserts, comparing to a quantitative research, the qualitative research does not gather
enough data from a selected population (Rogelberg 2004, p.175). Therefore the researcher intended
to improve the adequacy of the sample by applying various strategies. The research aimed to focus
on the richness and complexities of data in order to present a deep analysis of the findings.

17
 Methodology
3.6. Conclusion

This chapter presents the methodology behind the questionnaire design. The research applied a
rigorous method to the practicing of data collection, selection, and evaluation. Survey research
provided a legitimate approach into sampling strategies and data collection methods. The chapter
has identified the need for an effective and efficient research instrument. The choice of qualitative
research established a solid research base for the investigation of computer-assisted compositions,
and the method guides the discussion of findings in the following chapter.

18
 Findings and Reflection

Chapter 4. Findings and Reflection

4.1. Introduction

This chapter presents the results and creates focus topics derived from the questionnaire. The
chapter contains seven sections. The first section gives a brief overview of the chapter, and the
second section examines various compositional approaches from three participants in a
contemporary context. The third and the fourth section analyses two musical examples. The first
piece is Up to my f-holes (2008), a cello and tape composition by Judith Ring. The second example
is Sebastian Adams’s Turing Test: Text to Music #1 (2016), and this section discusses the
development of a generative notation system and how it could produce generative music. The fifth
section examines different compositional approaches of participants and future prospects of the
human-AI interaction. The sixth section provides a critical reflection of the findings. This section
addresses major concerns as well as some further thoughts from modern composers’ point of view.
The last section summarizes the chapter. It brings findings, critical reflections and future prospects
of computer-assisted compositions together.

4.2. Algorithmic Practice in Contemporary Music

4.2.1. The Use of Music Software and Programming Languages

The results from the questionnaire emphasizes the lens of unique interpretations on composers’
compositional approaches. The applied qualitative method reflects a degree of flexibility. With
open-ended questions, participants were free to respond. Only a few responses were left “yes” or
“no”, and most responses were rich and explanatory. Results of the survey largely represent a clear
need of software or hardware environments for three composers. A music system typically
comprises of buttons, keys and sensors which allows the composer to manipulate.

All participants expressed they use music production software in the process of composing, and
they installed the software either in their own laptops or in music studios. Max/MSP, Finale, and
Sibelius are most used programs. Two of three participants are using at least one of them. All
participants use more than two software to create and manage the audio content. They use a wide
range of musical applications from commercial to free, including Logic Pro X, Cubase, SPEAR,
FFmpeg, FMOD, Note Performer, Wwise, SuperCollider, Premiere Pro, Ableton, Reaper, Pro Tools,
Csound, Opusmodus and Audacity. One of the participants O'Neill explained his choices on
different software. O'Neill is a composer and improviser who is specialised in music theory and

19
 Findings and Reflection
performance. He mainly uses Logic Pro X. He uses it as a way to organise ideas and hear the
possibilities between found sound and musical texture. In Logic Pro X he tends not to use many
effects except as a way to tidy recorded sound, compression, and equalization etc. O'Neill also uses
MaxMSP to generate compositional material from recorded samples and as an input for instruments
to achieve a particular texture or abstraction of the instrumental sound. Occasionally the composer
uses Csound for the precision it offers. As for notation processes, O'Neill uses Finale, he found it to
be the most precise program in terms of unusual notation. “You can be very exact in Finale, more so
than in Sibelius”, he indicated. O'Neill tries out new tools and interactive music systems quite
regularly. Recently, he has started using Opusmodus which offers a more complete and intuitive
way to compose algorithmically. The overall responses demonstrated the variousness of composers’
choices on musical tools. Participants use digital tools directly or indirectly in composition
processes. These tools include digital audio workstation, music notation software, composing
software, audio editing software and synthesis software.

Besides music software, the findings also suggested that some participants use certain programming
languages. One participant indicated she does not use programming language. Two participants
expressed they use programming languages. Adams is a composer who is experienced in writing
notated instrumental and experimental music. He responded in the questionnaire that the coding
approaches he uses are basic JavaScript and Max MSP. He also uses the Bach externals in Max to
generate notation systems. The other participant O'Neill, who uses programming languages
expressed that Lua is used to write scripts for live instrumental processing. It serves as a front end
for the control of Supercollider. Like JavaScript, Lua is a multi-paradigm programming language.
Findings suggested for the participants who use programming languages, they often compose music
with functional programming and have showed great interests in problem solving and object-
oriented programming. In music software programming environments, two of three participants
employ high-level coding directly in their music. All participants demonstrated a good
understanding of programming concepts.

On the topic of how they utilize computer systems with or without embedding object-oriented
programming, participants expressed that they use digital tools in certain aspects of the
compositional process, depending on the individual program and musical genres. Ring explained
her compositional approaches in Finale, Note Performer and Logic Pro X. She stated that Finale is
used to hone the compositional material and sometimes writes directly onto the score in the

20
 Findings and Reflection
program. She also uses the playback facility to hear back what she has done. “Playback engine is a
new way of working for me”, as she stated. In recent times, Ring purchased Note Performer, which
is an artificially intelligent instrumental sound library that can be loaded into Finale for playback.
She stated the quality of the sound set is good and her playback “sounds much more real”. Note
Performer is also compatible with Sibelius but Ring has not attempted to use it yet, she is happy to
work in the Finale set. Ring uses Logic Pro X to layer samples of the instruments that she recorded
with different musicians. “I would use Logic to blend different textures together to form a ‘tape
part’ or fixed media element for my composition”, she noted. Similarly, O'Neill expressed his
approach in Logic Pro X. When preparing mock-ups he uses Logic Pro X. Computer systems are
prepared with a lot of detail as is necessary when inventing new playing techniques or abstracting
extended techniques. O'Neill also uses MaxMSP more frequently when he is generating musical
material. He has created patches that work with various types of sounds, a patch that works with
percussive sounds, or one that works with single pitch sounds like (talking), and so on. He
processes found sounds and recorded sounds to generate musical information, pitch sequences or
rhythmic patterns. O'Neill also transforms sounds to be used directly in a composition. The last
participant Adams explained his experience of using Max and Sibelius. He uses Bach externals in
Max to create generative computer notation systems. By applying a complicated sets of adjustable
weighted-but-randomised parameters, the system can produce music that could be sight-read in
real-time. Adams often uses similar tools to generate streams of data from a concept. For example, a
series of pitches and rhythms etc. Occasionally, he uses computers to aid with calculating numbers
relating to structural or harmonic elements of a piece. In Sibelius, he often writes music straight into
the program, using its playback functionality as part of the compositional process. He considers
playback is a quick method for copying and adjusting already typed music. Adams concluded that
“generally, I use the computer anywhere as it will make a task quicker or simpler. I rarely write
pieces completely, or even mostly without the computer”. Similarly, O'Neill pointed out that “I
would say…in nearly every part of the compositional process I would use computers as they have a
few main advantages”. Computers simplify difficult tasks and they allow for the use of more
complex ideas in the compositional process. Three participants expressed their opinions on the role
of the computer in their compositional processes. To a large extent they interact with computer-
based composition systems. They compose music using their own familiar software and the
software is becoming an integral part of composers’ working environments.

21
 Findings and Reflection
4.2.2. Indeterminacy, Randomization and Recombination
When asked the question if they use specific instructions or algorithms to generate music, all
participants expressed that they use specific algorithms as compositional tools. Ring stated that in
Wwise, she can detail any algorithmic composition she intends to approach. Randomisation of pitch,
volume, start time, finish time, spatial placement, and percentage can all be controlled by herself.
Ring also writes melody generators in other programs and she uses them to generate bass lines in
pop music, or solo lines. She often makes algorithmic composition tools in M4L for Ableton. This
allows her to control randomness and to have a compositional system that she guides, rather than
composes. She writes a lot of generative tools, which feed into Ableton mostly. This applies in a
different way when using controllers. She explained, “for example, if I am controlling six
parameters with a controller through physical movement, but I want to control twelve musical
parameters, I will write a little process to include the six parameters and manipulate as sonically
appropriate”. It keeps the relationship of movement, data and sound, without needing a literal
connection. Ring also described her experience on experimental composition. Within her more
experimental music, she takes data from movement and voice data to influence other parameters.
The idea behind it is that there is a harmonious relationship between different elements, even if that
is sometimes a tenuous link. In O'Neill’s answer, he described his algorithmic approach in the
compositional process. Algorithmically he uses Max to extract information from recorded sound. “I
will allow whatever sets of information dictate the flow of the composition”, as he indicated. He
will often record a particular gesture performed on an instrument and use the results as musical
information. This is generally done as a pre-compositional process and will usually dictate the
parameters of the piece. O'Neill also mentioned he continues to discover different algorithmic
compositional approaches with new tools. He has recently started to use Opusmodus but cannot yet
include it in his list of software as he has yet to write a piece with it. Opusmodus provides an
innovative method of composing and analysing music. O'Neill’s initial exploration of Opusmodus
has proven that it is an integrated, self-contained software. The musical possibilities in using this
software are very exciting. Likewise, Adams explained the general algorithmic approaches which
involve the computer. He generally uses very simple algorithmic approaches, simply dealing with
repeating or augmenting series, transcription of one parameter into another (for example, turning
letters of a text into pitch values), and particularly by using weighted-randomness to control a
property in general without specifying the specific property. “I like to approach problems by
generalizing them and not by doing laborious work by hand, so I delegate as much as possible to the
computer”, he summarized. Results from this question reflected that all compositions from three
participants are partially or fully generated from computer systems. All participants’ experience of
algorithmic approaches ranged from applying basic algorithms to complex computational
22
 Findings and Reflection
techniques. The findings demonstrated that music activities are automated by software tools to
varying degrees.

On the next question when participants were asked which part of the composition is controlled by
the composer, they expressed their opinions in details about human elements in computer-aided
algorithmic compositions. Ring believes that form is the most important thing for her to control.
She creates a structure that she progresses through but often other processes control the shape
within sections that she can respond to. In her performance work, the system aids her in the
performance and composition. However in the studio based work, she sometimes is aiding the
computer. The other participant O'Neill argued that there have been a few occasions when
instructions have been given for the performer to improvise within certain parameters. This is
aleatoric techniques and he is combining technology with earlier styles of aleatoricism. Aside from
that he would control every part of the composition, and the entire score is controlled by the
composer. Adams gave the similar answer that he believes he is the creator of the music instead of
computers. The general response showed that, participants acknowledged the compositional process
represents a large degree of automation. Some parts of music are controlled by the computer while
some are controlled by the composer. Findings implied that in certain circumstances, computers
play the role of the improvisational or performance of the music.

When asked the question if they use any indeterminate approaches that akin to processes of Cage,
two of three participants answered they use indeterminate approaches much of the time, and one
participant does not use the method. One participant explained he creates indeterminate scores in
his improvisations based on random numbers on a modular synthesizer. The answers were
expanded in the next question when they were asked what kind of randomization or recombination
of musical elements they use. Ring expressed that, “this is different every time” and that she tries
experimenting with many elements that are chosen by the ear, rather than by a rules-based process.
It involves tweaking things all the time until she gets something she likes. It also involves taking
something designed for another purpose and seeing how it fits or does not. “This helps me
challenge my approach”, Ring added. In some cases, “something I write later on in a piece can end
up sounding better at the beginning, or somewhere else with the structure”. The other two
participant’s answers focused on the use of Max. O'Neill described the techniques he uses in detail.
He creates patches in Max to generate musical material. “I would sometimes use the traditional
elements such as retrograde inversion of that found in serialism.” He perceives a composition as an

23
 Findings and Reflection
eco-system of sorts, in which each element has a connection to all other elements and to the
composition as a whole. “Any composition is, for me, a recombination of initial elements used”, he
concluded. When discussing the randomization of information, O'Neill responded he would use
randomization in some of the techniques involving Max and in the generation of musical material.
However any use of randomization is usually done within strict parameters and with an idea as to
what the result might be. “I would tend to use this technique loosely, always picking and editing the
final result”. Similarly, in Adams’s answer, he described his approach in Max. He developed a
generative notation system using Bach externals in Max. In the computer system, all musical
parameters except duration and instrumentation of the piece can be randomized, always with
weights assigned to different values so that the result will be random in its specifics but pre-decided
in a more general way. Overall, the whole sample of composers believe the algorithmic patterns
exist in their compositions. Results suggested that composers encode musical patterns into the
algorithm. The composing process is a process of recombining, or randomizing the musical
elements. These contributions reiterate the concept in the background chapter that composers can
apply algorithmic techniques in rule-based compositions. In David Cope’s computer systems, the
composing process integrates pattern-matching and recombinance in the association network (Cope
2001, p.77). O'Neill’s approach on recombination is an interpretation of Cope’s method. Findings
also demonstrated the stochastic foundation on the use of random parameters.

4.3. Up to my f-holes

When participants were asked to give musical examples of particular algorithmic techniques they
have used, two of three participants presented relevant examples. The first participant Ring shared
her composition Up to my f-holes (2008), see the score below (Figure 1).

24
 Findings and Reflection

Figure 1: Score of Up to my f-holes, page one (Ring 2008)

Up to my f-holes is a track from Ring’s debut album “What Was” produced by Ergodos records. It
is an electro-acoustic and truly experimental piece for solo cello and tape (Daylight Music 2008).
The performer is free to interpret the score in an individual manner by embellishing the timbre to
blend or contrast with the tape part. Ring wrote it in 2008 for cellist Laura Moody for the tape part
and it was performed by cellist Kate Ellis. “Each piece of mine is highly collaborative as I work
closely with musicians to find out what their instruments can do”, as Ring indicated (Ring 2008).
She has also explored a wide range of instruments as well as different styles. “Each player offers
something new to the mix, and each piece is unique to the player involved”, she added. The
duration of Up to my f-holes is six minutes and thirty seconds. There is a good balance and blend
between two parts: cello and tape. “I assembled the tape part out of samples I recorded with cellist
Laura Moody in London”, as Ring described her process of composing. Recordings were made of
the cellist exploring her instrument in imaginative ways, using extended techniques and additional
sound generation. Ring recorded all the sounds Moody could make on her cello from normal notes
to everything else they could think of. Then, Ring edited all the elements into single sound files and

25
 Findings and Reflection
layered them in different ways in Logic Pro X. The resulting sounds were composed into a tape part
that provides an exciting accompaniment to the energetic live part. The live part was written after
the tape part to compliment it and it aimed to provide a more interesting live experience. The
performer should become familiar with the tape part in order to interpret the score in a creative way,
enhancing both parts in the process.

Figure 2: Score of Up to my f-holes, page two (Ring 2008)

26
 Findings and Reflection

Figure 3: Score of Up to my f-holes, page three (Ring 2008)

As shown in the score (Figure 2), the distribution of notes was determined by randomized
procedures. At 0:50 of the piece, the note reads as “random rhythm and pauses, some straight and
some vibrato”. At 2:00 in the glissandi part, the note reads as “random dynamics”. The piece should
be played in “random speeds and lengths in the higher register interspersed with rapid up and down
glissandi on a particular string”, and it lasts for twenty seconds. At 2:30, a short fortissimo follows
with “random plucking against tape part”. At around 4:40, harsh and abrasive random accented
double stops are played, following a decrescendo of the descending screech (Figure 3). At 5:00, the
phrase changes to “random loud harmonics interspersed with col legno ricochet”. Throughout the
piece, the performer may feel free to embellish the score if a particular colouring of a segment
should come to mind. Ring concluded, “Every time it is played, it sounds slightly different as the
score is mainly gestural”. The whole processing of Up to my f-holes involves random choices. “My
music stems from the electro-acoustic world and its core technique is based on the practices of
musique concrète”, as Ring indicated.

27
 Findings and Reflection
4.4. Turing Test: Text to Music #1

Besides Ring’s Up to my f-holes, the other participant Adams also introduced his specialized
algorithmic composition system. Turing Test: Text to Music #1 (2016) is a generative notation
system. It was developed in 2016, using Max 6 and the Bach externals. The system is developed
based on text, improvisation and sight-reading, where a text could be converted to notation and used
as a tool for improvising. A screenshot of the system demo video is displayed below (Figure 4).

Figure 4: A screenshot of Turing Test system demo (Adams 2019)

“A fascinating aspect of real-time notated music is that it is a truly nascent field”, as Adams
explained the practical reasons for pursuing his invention, “the idea of live-generated notation sight-
read in real-time is musically compelling”. The interpretative faculties of a performer are seen in
their least cultured form during the performance of a piece written in Turing Test. The performer of
Turing Test has no time to think about the most musical way to approach a phrase, produces a very
different interpretation to the one they would of the same music with a written score and a week to
practice. The focus of Turing Test is left purely on the computer, with the composer providing no
melodic material and not participating in the performance in any way. The system follows in the
footsteps of the aleatoric, generative and algorithmic movements of musical composition. As
Adams stated, “John Cage is my main touchstone in all manners indeterminate”. His interest in
Cage was sparked through performing works like “four” for string quartet, where there is an almost

28
 Findings and Reflection
complete lack of determinacy and yet a very predictable end result. Cage’s varying levels of
indeterminacy all fascinate Adams and have informed on his own music.

Adams also described some technical details of the system. In the pitch generation process, creating
pitch material is one of the vital steps. He decided to ask the computer to generate a starting pitch
(the tonic, subdominant or dominant of the chosen “key”), and then generating the next pitch by
adding or subtracting a random interval from the previous note. It is essentially one basic operation
of arithmetic and yet altering the likelihood of different intervals radically altering the sound world
of the resulting music. Adams then described the most interesting element of the patch, the memory
system. “This is a collection of four memory banks and a series of transformers that alter pitch,
rhythm, tempo and dynamics of the material stored in the memory”, he explains. Each list of pitches
are inputted into index files called “colls” and then asking for a random item from the coll
whenever a new phrase is needed. The output from these colls then goes through the transformers
before being sent back out into Bach. The output from the memory is then routed back into the
memory as a new item, meaning that developments of phrases can then be developed further
themselves. The fourth and final memory bank receives everything that happens, whereas the first
three only receive information if they are selected. “Minimilate is my favourite transformer that I
have built into Turing Test”, Adams stated. Below is a screenshot of Minimilate’s video
demonstration (Figure 5).

Figure 5: A screenshot of minimilate object demo (Adams 2019)

29
 Findings and Reflection
In Minimilate, the phrase remembered by the computer can be minimilated. It takes a list of
numbers, copies N numbers from the list and then replaces N numbers elsewhere in the list with
those copied. Adams gave an example of the method: [before: {4 6 5 9 7 3 1 7 2 8 2} after: {4 3 1 7
7 3 1 7 2 8 2}].

Turing Test: Text to Music #1 aims at forcing performers to sight-read in public. The image below
shows flautist William Dowdall practices sight-reading in the Contemporary Music Centre’s Salon
series (Figure 6). The salon took place in Lutheran Hall, Dublin in 2016. It featured Adams’s solo
viola performance and Dowdall’s solo flute performance (CMC 2016). In a live situation, the
computer writes music during the concert, which then has to be sight-read by Dowdall. He could
not be able to practice the piece in advance. As Adams himself remarked, “the extraordinary tension
created by a musician sight-reading in the arena they normally reach only after careful preparation”.
The whole appeal of virtuoso music lies in the palpable demands and stress placed on a musician
when they and their instrument are pushed to the very edges of their capabilities. The demands
expected of a musician increase by expecting a player to sight-read in a concert setting thereby
increasing the demands and tension without abandoning notation, and allowing for audible thematic
development. Adams’s approach showed a unique perspective of the implementation of
sophisticated probabilistic models. The findings demonstrated that musical parameters can be
randomized in an indeterminate approach with stochastic elements.

Figure 6: An image of William Dowdall’s performance in Contemporary Music Centre’s Salon 2016 (CMC
2016)
30
 Findings and Reflection
4.5. Future Prospects

Relating to the topic of further exploration of participants’ music approaches, three participants
responded that they would like to try new tools and approach differently on their music practices.
O'Neill pointed out that, “I’m always looking for new ways to approach composition but much of
my approach has developed for an initial interest in composing a particular way”. O'Neill is always
looking to extend the compositional technique but he is generally happy with the way he composes.
New tools such as Opusmodus offers a composing continuum, this is less a new approach than a
way to gather information, and access certain techniques all in a single place. O'Neill also
commented that “new tools are always worth exploring but sometimes one can be easily distracted
by technology”. Adams expressed that after Turing Test, he has tried various approaches in his
music practice, even creating a different application for Turing Test as an educational tool for
training sight-reading. It is possible using the existing architecture of the patch, however the
program would need to be tailored to produce much more conventional phrases before it could be
truly useful as an education program. Adams also mentioned his enthusiasm for machine-learning
system: “I would love to work with machine-learning, I tried but only very minimal, I have not
figured out an appropriate musical context for this yet”. Furthermore, the work of John Cage and
David Cope all fascinate Adams. “Something which has recently intrigued me is the work of David
Cope”, Adams responded. He suspects that taking Cope’s ideas on board will be incompatible with
his current application and will require a fresh start, however, “Cope’s work is something I hope to
delve into deeply in the future”, he indicated. From Ring’s point of view, she mentioned her
interests in touch screen technology. “I would love full touch screen capabilities on a massive
screen for Logic Pro X and Finale”, she expressed. Ring is happy with what she is working on at the
moment, but it would be great to be able to move and input elements very precisely with a mixture
of touch and trackpad. For example, to be able to zoom in close to a sample use a finger spread and
then to alter the volume envelope of that sample but tapping and sliding fingers in different ways. It
would also be great to be able to draw in the notes in Finale and have them automatically be typeset.
“There was a windows based tablet being especially designed for that at one stage, I believe it had
great potential”, she added. Findings showed two of three participants prefer to use their own
familiar software, and one expressed that new technology can be distracting. Nevertheless, the
whole sample highlighted the possibility of further development of advanced computational
approaches. Findings reflected that composers are slowly embracing computer-driven technology,
although some participants have not yet shown a keen interest to the software which they are not
familiar with.

31
 Findings and Reflection
When participants were asked their opinions on the future prospects of human-machine interaction,
AI and collaborative intelligence, they answered the question from different aspects. Adams believe
the work of many pioneers already demonstrates a future of the human-AI interaction, such as
Jennifer Walshe, Holly Herndon and David Cope. “There will be many more exciting ways to use
AI in music which have not been discovered yet”, as Adams stated. Likewise, Ring expressed her
thoughts too. She responded it would be exciting to see more opportunities and more pushing of
boundaries, and she imagines composition tools are just going to be more and more user-friendly
and capable of extraordinary things. Expectedly, apart from the positive prospects mentioned above,
findings also reflected the downsides and concerns. O'Neill responded there is no yes or no answer.
He sees AI as having a place in certain types of music and at certain stages in the compositional
process, as an aid to the composer or songwriter in the completion of certain tasks. However it is
important to see the distinction between music that is written by composers and songwriters and
that which is composed by an artificial intelligence. O'Neill believe it is a great trick but music
speaks of the human condition, whether it is a love song or a complex exploration of sound. He
indicated that “the day we relinquish control is the day we lose sight of the real function music
serves: a wordless communication between people and a way that we can connect with the deepest
parts of ourselves”. He would hate to think that he is listening to nothing more than an algorithm
that reduces human experience to ones and zeros. Similarly, Adams expressed he believes in a more
human-oriented approach. The composing process tends be a human-centred process incorporating
algorithmic approaches that include collaboration with humans, rather than attempting to replace or
outsell them. Adams also drew the concerns in terms of absolute music (instrumental music,
classical music etc) and functional music (video game music, film soundtrack etc). “There will be a
danger of shortcuts being taken with AI being used to generate unique but formulaic background
music as a way to avoid spending money on a composer, particularly in functional music”, Adams
pointed out. For absolute music, there is a challenge for AI to generate true harmony. It is likely that
there will be truly “creative” AI which produce work of note, but that the pressures of the market
will cause a major use of AI in music to be as a creator of bland, cheap (but original) stock music.
The response showed that participants believe computer is becoming the universal tool and the
development of computational algorithms and AI is inevitable. Despite the concerns raised, findings
suggested that composers acknowledge the power of human-AI interaction, and they are
maintaining an open mind on collaborative human and AI intelligence.

32
 Findings and Reflection
4.6. Reflection

Findings above illustrated that composers tend to use the musical software which can supplement
and complement their musical minds. They are seeking for a virtual workstation to develop their
ideas and organise pre-compositional materials. With the aid of specialized programs, composers
can adapt to a fast-paced environment in the radical-changed music industry. The music industry in
the age of digital transformation faces new challenges. The shift from handwriting scores to
composing on screen is a revolutionary change. Today’s composers go beyond the traditional
format of music production, and digital tools form an essential part of the modern musical
environment (Kiddle 2020, p.1). Laptop and studio composers are able to harness the powerful
software to implement the musical concepts. They can fully engage with technological capabilities
and greatly benefit from the practical applications.

Results from questionnaire recorded that the most used compositional tools are Max/MSP, Finale,
and Sibelius. Logic Pro X and Opusmodus are popular choices for some composers. Findings
suggested that composers often use more than two software with different focuses in terms of
functionality. They tend to use the software which do not heavily overlap with the musical functions.
In computer systems, having the programmatic and mathematical control over a piece are important
for composers. Musical software makes comprehensive compositional workflows possible with
built-in sound libraries, notation editors, and more. With the minds of computational thinking and
problem solving, composers are able to manipulate music just like writing code. Multi-paradigm
programming languages are popular among composers. Survey findings implied that compositional
approaches are vary depending on software and coding environments. For example, playback
engines, artificial intelligence sound library, and patch generation are compelling features.
Composers can work on compositional systems to control the parameters and algorithms.
Furthermore, findings demonstrated that composers believe that ultimately the creator of the music
is the human. A human composer adds the individual touch to a software or a program. In a
computer system where a human composer makes compositional decisions, intervenes and builds
algorithmic models, then the human legitimately claims to be the composer, or the designer of any
music or music generating systems developed. Findings also reflected that an important limitation
of an intelligent system is the generation of a real harmonic language, such as chordal progressions,
modulation and proper counterpoint. In indeterminate scores the musical information is randomized
and left by chance, but the result is not unpredictable. Human intervention is a compositional
technique in automatic systems.

33
 Findings and Reflection
The survey also demonstrated that the compositional process is a human-centred process, despite
the extensive use of digital tools and intelligent systems. “Composing has become a multi-faceted
process and takes ideas about structure and content from many disciplines” (Opusmodus 2020,
p.40). Musical materials can be transformed, restructured and extended with the assistance of the
computer. In composing processes, composers tend to work on screens, and finish a piece in one
single environment. However much of the work is still done by the human. Computer systems do
not provide the composing continuum, and the human approach is still dominant in the whole
process.

Furthermore, two musical examples illustrated the use of specific musical patterns and the
algorithmic techniques. The first example visualised how notes were assembled by chance and
determined by randomized procedures. The gestural score allows the performer to interpret the
piece in a creative way. The music performer is producing the music content as well. The second
example discussed how to handle probabilistic models and apply indeterminate technique to
composition. The memory system created easily audible developments of thematic material, and
scores could be read for improvising and sight reading. The musical examples demonstrated the
application of compositional strategies in algorithmic compositions in a contemporary context.
Modern composers continue to explore new approaches on their music practices. Many of them are
looking to extend the compositional technique by using new software and applying AI techniques.
Findings implied positive prospects on the further development of collaborative intelligence. The
combination of human and AI will guide a new direction of music generation. Creative AI is in its
infancy. The results of the questionnaire showed an enormous possibility that humans will go
beyond the present musical limits.

4.7. Conclusion

This chapter presents the analysis and interpretation of the findings, which are based on the
qualitative data collected from the questionnaire. The answers of the questionnaire were carefully
reviewed from different perspectives. In combination with the literature discussed in the
background chapter, there is significant grounding to assume the findings of this research
demonstrated the hypothetical relationship of the human and computer. The findings of the research
revealed a growing trend and focus on human-AI interaction in music creative practices.

34
 Conclusion

Chapter 5. Conclusion

5.1. Summary

Algorithmic compositional techniques have been extensively explored throughout history. Stepping
into the information age, the rapid shift in the traditional music genre to the common use of
computational technology become the driving forces of digital music evolution. The paper provides
a comprehensive view of algorithmic music and different approaches that have been used for
algorithmic composition. Many old models and frameworks from earlier musical traditions have
been widely used in today’s algorithmic compositional procedures. For example, Markov chain
models can be found in the process of Mozart's Musical Dice Game. David Cope’s musical
intelligence system is very similar with the structure of the Markov chain, which was developed in
the 1980s. Similarly, Bach’s approach to canonic composition, and Arnold Schoenberg’s twelve-
tone method were notable precedents of algorithmic practices. From the early efforts on formalism
to algorithmic music, composers imposed maximal complexity in terms of the development of more
advanced machine models. By the late-20th century, the well-defined compositional models were
established. The stochastic method, rule-structured technique and AI-based system are three major
approaches of algorithmic composition. In addition to the major approaches, genetic programming
(a branch of AI method) and the hybrid approach (combine several techniques) are recent
approaches.

Drawing from the literature and composer’s personal perspectives, the paper offers a broad
understanding on the role of AI in music creative practices in the contemporary context. Up to my f-
holes and Turing Test: Text to Music #1 are two musical examples representing the implementation
of formalized procedures. The paper also offers the critical view of computer-assisted composition
and its future prospects. Gaining profound insights into composers’ point of views of algorithmic
approaches, O'Neill, Ring and Adams highlighted the algorithmic nature of their music. Findings
showed the use of computer and musical software accelerating the development of new musical
forms.

In the modern landscape of algorithmic music and AI, the composing process integrates
randomization and recombination of information, indeterminate approaches, and more techniques
with minimal human intervention. Results of the survey largely represented composers’ clear needs
of digital tools. Each participant uses more than two software to create and manage the audio

35
 Conclusion
content. On the subject of ownership, many of the technical decisions indirectly involve
compositional decisions. Composers integrate different manners of compositional approaches with
the computer automatically producing the output. Despite the processing, the response of the survey
showed that participants believe music is more human-oriented. As Cope argues, musical
intelligence is an extension of the human mind. In the relationship between machines and musicians,
one should never replace the other. The survey findings also suggested a positive future outlook of
truly creative and original AI which could add to the spontaneity and subjectivity of music. The
combination of human and AI subverted the traditional means of how music is produced.
Furthermore, the increased capabilities of computers and algorithms continue to play an important
part in the musical discourse. For further exploration, there is a great opportunity to experiment
multi-dimensional approaches to music creative processes.

5.2. Contributions

The paper contributes in the field of computer-assisted composition by providing a deep

understanding of the complexities of algorithmic techniques. The multidisciplinary nature of
algorithmic composition brings together many intersecting fields of artificial intelligence, arts and
mathematics. The questionnaire results presented the idea that the computational techniques have
altered the working environment for composers permanently. The findings of the paper could
inspire programmers and software developers to design software products more attuned to
composers and their creative needs. Composers’ practical experience allows programmers to rethink
the framework of existing systems. The participants’ invaluable input to the study offered profound
insights of the composer user experience. One of the participants Ring implied, that in the future the
development of musical tools, the user-friendly interface and touch screen technology could provide
new focus on music technology development. This could provide composers a space whereby they
are able to work in an immersed and embodied environment, such as the use of touch screens.
Human-machine interfaces will make the interaction between the composer and computer more
streamlined and integrated. It will also allow the composer to interact with the music performer or
music interpreter.

Furthermore, the paper also contributes to the practical and theoretical aspects of algorithmic music.
In the composing process, the need for AI languages in musical applications is increasing. The
paper provides solutions on how to process AI-related musical problems. The intersection of music
and technology brings a different perspective to music composition. Findings demonstrated that
music generation should not be dependent on human effort alone. Many musicians benefit from the

36
 Conclusion
employment of intelligent instruments and programs. During the experimental process, composers
discover new methods on music practices and achieve more than expected.

Moreover, by analysing music theory and the knowledge from multiple disciplines, the paper
provides a scientific way to help the composer think algorithmically. Understanding and learning
algorithmic techniques is becoming an essential need for composers to access and engage with
compositional tools, and this paper explicitly responds to that need. In the larger context of the
modern digital landscape, the paper stimulates composers to adapt the algorithmic techniques in
order to further their individual compositional developments.

5.3. Limitations and Further Research

Considering to the limitation of the paper, the sample size is relatively small. Due to the scope of
the research, the small-scaled sample group has inherent limitations. Participants could not
contribute a musical database of various styles and genres. The analysis of musical genres is not the
central focus of the study, but algorithmic compositions generally cover all styles of music. For
example, electroacoustic music, contemporary classical music, jazz, interactive game music, dance
music and experimental music, they all belong to the domain of contemporary music activity.

Overall, the research experience was a rewarding practice, and such experiences helped the
researcher establish good research habits. Further investigation could focus on the exploration of
compositional approaches of different musical genres. The qualitative research would be conducted
on a larger scale to fill the gap of in-depth study in this field. The future study could also explore
the collaboration between composers, and how they work together in order to discover the musical
potential of different musical instruments.

37
 Appendices

Appendices
Appendix i. Sample Information Table

Appendix ii. Participant Information Sheet

38
 Appendices

39
 Appendices

40
 Appendices

41
 Appendices

Appendix iii. Sample of Questions

42
 References

References

Alpern, A. (1995). Techniques for Algorithmic Composition of Music. Available at:

https://www.semanticscholar.org/paper/Techniques-for-Algorithmic-Composition-of-Music-
Alpern/0c239a5c8a407088e5e01b2749dc75a83317b94b [Accessed 22 Mar. 2020].

Ashby, A. (2001). Schoenberg, Boulez, and Twelve-Tone Composition as "Ideal Type". Journal of
the American Musicological Society, 2001, vol. 54, no. 3, pp. 585–625.
AIVA. (2019). I am AI - AI Composed Music by AIVA. Available at:
https://www.youtube.com/watch?v=Emidxpkyk6o [Accessed 12 Apr. 2020].

Adams, S. (2019). Sebastian Adams’s website. Available at: https://www.sebastianadams.net/

[Accessed 25 Apr. 2020].

Biles, J. (1994).GenJam: A Genetic Algorithm for Generating Jazz Solos. International Computer
Music Conference, 1994, pp. 131-137.

Boyer, C. B. (1951). The Foremost Textbook of Modern Times. The American Mathematical
Monthly, 1951, vol. 58, no. 4, pp. 223. Available at: http://mathshistory.st-
andrews.ac.uk/Extras/Boyer_Foremost_Text.html [Accessed 15 Feb. 2020].

Cowell, H., and Nicholls, D. (1996). New Musical Resources. New York: Cambridge University
Press, 1996.
Cowell, H., and Godwin, J. (1969). New Musical resources: With a preface and notes by Joscelyn
Godwin. Something Else Press, Inc., 1969.

CMC. (2016). Contemporary Music Centre’s Salon. Available at:

https://www.cmc.ie/events/2016/may/contemporary-music-centre-salon-series-2016-sebastian-
adams [Accessed 29 Apr. 2020].

Cope, D. (2018). Portfolio website for David Cope. Artsites.ca. Available at:
http://artsites.ucsc.edu/faculty/cope/music.htm [Accessed 15 Feb. 2020].

Cope, D. (1996). Experiments in Musical Intelligence. Vol. 12 Computer Music and Digital Audio
Series. Madison, WI: A-R Editions, 1996.
Cope, D. (2000). The Algorithmic Composer. Wisconsin: A-R Editions, Inc., 2000.

Cope, D. (2001). Virtual Music: Computer Synthesis of Musical Style. Cambridge: MIT Press,
2001.

Collins, N. (2018). Origins of Algorithmic Thinking in Music. In The Oxford Handbook of

Algorithmic Music. Oxford University Press, 2018.
Caianiello, E. R. (1984). Cybernetic Systems. London: Research Studies Press.

Crocker, R. (1963). Pythagorean Mathematics and Music. The Journal of Aesthetics and Art
Criticism, 1963, vol. 22, no. 2, pp. 189–198.

43
 References
Collins, D. (1995). Musical Terminology in the Canonic Works of Bach: An Historical Context.
Bach, 1995, vol. 26, no. 1/2, pp. 91–101.

Collins, D. (1999). Bach and Approaches to Canonic Composition in Early Eighteenth-Century

Theoretical and Chamber Music Sources. Bach, 1999, vol. 30, no. 2, pp. 27–48.

Crawford, R., and Fox, R. (2016). Artificial Intelligence Perspectives in Intelligent Systems. Brno:
Springer, 2016.

Denscombe, M. (2008). The Length of Responses to Open-Ended Questions: A Comparison of

Online and Paper Questionnaires in Terms of a Mode Effect. Social Science Computer Review,
2008, vol. 26, no. 3, pp. 359–368.

Daylight Music. (2008). Kate Ellis - Up to my f-holes (Judith Ring). Available at:
https://www.youtube.com/watch?v=d6r6WlX1Ejg [Accessed 15 Apr. 2020].
Dawin, J. R. (2010). Markov Chain Analysis of Musical Dice Game. ResearchGate. Available at:
https://www.researchgate.net/publication/45913718_Markov_Chain_Analysis_of_Musical_Dice_G
ames [Accessed 16 Feb. 2020].

Dean. R. T. and McLean, A. (2018). The Oxford Handbook of Algorithmic Music. New York:
Oxford University Press, 2018.

Dannenberg, R.B. (2000). Artificial Intelligence, Machine Learning, and Music Understanding.
Computer Music, 2000. Available at: https://www.semanticscholar.org/paper/Artificial-Intelligence-
%2C-Machine-Learning-%2C-and-Dannenberg/9618ecdd92c85f2813dc83e3470405fb27552bc8
[Accessed 9 Apr. 2020].

Fiebrink, R., and Caramiaux, B. (2018). The Machine Learning Algorithm as Creative Musical
Tool. Oxford Handbook of Algorithmic Music. Oxford University Press, 2018. Available at:
https://www-oxfordhandbooks-
com.elib.tcd.ie/view/10.1093/oxfordhb/9780190226992.001.0001/oxfordhb-9780190226992-e-23
[Accessed 25 Mar. 2020].
Forte, A. (1967). Computer-Implemented Analysis of Musical Structure. In Computer Applications
in Music, West Virginia University, Morgantown, 1967.

Goldmark, D. (2007). Tunes for ’Toons: Music and the Hollywood Cartoon. London: University of
California Press, 2007.

Higgins, H., and Kahn, D. (2012). Mainframe Experimentalism: Early Computing and the
Foundations of the Digital Arts. London: University of California, 2012.

Hiller, L., and Baker, R. (1964). Computer Cantata: A Study in Compositional Method.
Perspectives of New Music, 1964, vol. 3, no. 1, pp. 62–90.

Jakubowska, J. H. (2011). Music perception of avant garde: Musical structure and time.
ResearchGate. Available at:
https://www.researchgate.net/publication/228953564_Music_perception_of_avant_garde_Musical_
structure_and_time [Accessed 5 Mar. 2020].

44
 References
jMusic. (2020). Music composition in Java. Available at:
https://explodingart.com/jmusic/ReferencesLinks.html [Accessed 1 Apr. 2020].
Kostka, S. (2016). Materials and Techniques of Post Tonal Music. New York: Routledge, 2016.

Koza, J. (1992). Genetic Programming: On the Programming of Computers by Means of Natural

Selection. Cambridge: MIT Press, 1992.

Keats, J. (2006). John Koza Has Built an Invention Machine. Popular Science, 2006, vol. 268, no.
5, pp. 66–92.

Kouritzin, S. G., Piquemal, N. and Norman, R. (2009). Qualitative Research: Challenging the
Orthodoxies in Standard Academic Discourse(s). New York: Routledge, 2009.

Kiddle, G. (2020). Learning Music Theory with Logic, Max, and Finale. New York: Routledge,
2020.

Loy, G. (2006). Musimathics, Volume 1: The Mathematical Foundations of Music. Cambridge: The
MIT Press, 2006.
Manning, P. (2004). Electronic and Computer Music. New York: Oxford University Press, 2004.
Milstein, S. (1992). Schoenberg's Serial Odyssey. Music & Letters, 1992, vol. 73, no. 1, pp. 62–74.

McBride, J. L. (2019). Advance Recordings: Innovative Label for American Twentieth-Century

Avant-Garde Music, ARSC Journal, 2019, vol. 50, no. 1, pp. 1–43.

Maurer, J. (1999). Interactive Electronics in Computer Music and Chris Chafe's "Push-Pull"
composition. Stanford University. Available at:
https://ccrma.stanford.edu/~blackrse/interactive_electronics.html [Accessed 9 Jan. 2020].

Maurer, J. (1999). A Brief History of Algorithmic Composition. Stanford University. Available at:
https://ccrma.stanford.edu/~blackrse/algorithm.html [Accessed 9 Jan. 2020].
Nierhaus, G. (2009). Algorithmic Composition: Paradigms of Automated Music Generation.
Mörlenbach: Springer Science & Business Media, 2009.

Nuhn, R., Eaglestone, B., and Ford, N. (2002).A Qualitative Analysis of Composers at Work.
ResearchGate. Available at:
https://www.researchgate.net/publication/228609098_A_Qualitative_Analysis_of_Composers_at_
Work [Accessed 9 Apr. 2020].
OpenAI. (2020). OpenAI site. Available at: https://openai.com/ [Accessed 28 Mar. 2020].
Opusmodus. (2020). Opusmodus quick guide. Available at:
https://opusmodus.com/media/pdf/Quick%20Guide.pdf [Accessed 19 Apr. 2020].
Rowe, R. (2001). Machine Musicianship. London: MIT Press, 2001.
Ring, J. (2008). Judith Ring’s website. Available at: http://www.judithring.com/about/ [Accessed
15 Apr. 2020].

Rogelberg, S. G. (2004). Handbook of Research Methods in Industrial and Organizational

Psychology. Oxford: Blackwell Publishing, 2004.

45
 References
Riedweg, C. (2008). Pythagoras: His Life, Teaching, and Influence. New York: Cornell University
Press, 2008.

Roads, C. (1980). Artificial Intelligence and Music. Computer Music Journal, 1980, vol. 4, no. 2,
pp. 13–25. Available at: https://www.jstor.org/stable/3680079 [Accessed 2 Mar. 2020].

Smith, L. (1973). Henry Cowell's Rhythmicana. Anuario Interamericano De Investigacion Musical,

1973, vol. 9, pp. 134–147.

Sitsky, L. (2002). Music of the Twentieth-century Avant-garde: A Biocritical Sourcebook. Westport:

Greenwood Press, 2002.
Schwanauer, S., and Levitt, D. (1993). Machine Models of Music. London: MIT Press, 1993.

Schwartz, E., and Godfrey, D. (1993), Music Since 1945: Issues, Materials, and Literature. New
York: Schirmer Books, 1993.

Simoni, M. (2003). Algorithmic composition: A gentle introduction to music composition using

LISP and common music. ResearchGate. Available at:
https://www.researchgate.net/publication/303911920_Algorithmic_composition_A_gentle_introduc
tion_to_music_composition_using_LISP_and_common_music [Accessed 10 Feb. 2020].

Srivastava, T. (2014). Introduction to Markov chain. Analytics Vidhya. Available at:

https://www.analyticsvidhya.com/blog/2014/07/markov-chain-simplified/ [Accessed 5 Mar. 2020].

Vogt, W. P., Gardner, D. C., and Haeffele, L, M. (2012). When to Use What Research Design. New
York: Guilford Press, 2012.

Winograd, T. (1968). Linguistics and the Computer Analysis of Tonal Harmony. Journal of Music
Theory, 1968, vol. 12, pp. 2-49.

Xenakis, I. (1992). Formalized Music: Thought and Mathematics in Composition. Stuyvesant:

Pendragon Press, 1992.

Zaripov, R. (1960). An Algorithmic Description of a Process of Musical Composition. Soviet

Physics Doklady, 1960, vol. 5, pp. 479-482.

46