On Computational Ecosystems in Media Arts Rui Filipe Nicolau Lima Antunes Thesis submitted for the degree of Doctor of Philosophy of the University of London Goldsmiths, University of London December, 2013 DECLARATION I hereby declare that I composed this thesis entirely myself and that it describes my own research. Rui Filipe Nicolau Lima Antunes 10th of October, 2013 2 This research is an exploration of issues surrounding the artistic production of Computa- tional Ecosystems. A Computational Ecosystem is a system of agents designed to emulate, in the computer, biological systems where autonomous individuals are organized in a hier- archical food chain and interact by trading units of energy. This thesis maps out this field and examines the modes of production and functions of these systems. The central claim is focused on how the narratives normally associated with these systems and their function- ing are two complementary, but separate entities. By virtue of considering these separately the computational ecosystem is argued to be an abstract generative engine for heterogeneity, spontaneity, and even novelty. It is contended that the set of methods of production developed by exploratory artists using these artefacts might be instrumentalized as generative methods for the animation of general purpose non-player characters in virtual worlds. Keywords: virtual worlds; computational ecosystems; non-player characters; computa- tional models of animation; generative art; evolutionary art; persistence. 4 Acknowledgments This project would be unthinkable without the routine support from supervisors, colleagues and friends, as well as the contribution of specialist knowledge by a host of professionals who were not only willing to be questioned about their experience and professional practice but were also enthusiastic about the prospect. I owe indebted to many people, far more than I can name here. First and foremost I owe gratitude to my supervisors, Professor Frederic Fol Leymarie, Professor William Latham and Professor Janis Jefferies, who guided this project with tremendous verve and thoughtfulness. Frederic was indefatigable in the endless meet- ings and revisions of the written material, William has never failed to produce provocative discussions, and Janis was always there when I needed her the most. I am very grateful to the ‘usual suspects’: my office mates, Mohammad al Rifaie and Patrick Tresset, and colleagues from seminars on Arts and Computational Technologies, who have been willing to discuss ideas, offer support and invariably lift my spirits with their energy and industry. I am equally grateful to those who initially supported and encouraged this project of research, such as art critic João Pinharanda and Professor Pedro Faria Lopes. João in particular provided an ex- traordinary impulse, inviting me to participate on the exhibition Lá fora. I owe thanks to Dr. Tim Taylor and Dr. Mick Grieson, who provided me with helpful pointers and offered most appreciated insights during the examination to transfer onto the PhD track. Among the circle of experts and practitioners that provided me with valuable information, I especially would like to thank Professor James Lovelock, Professor Jansen Vincent, Professor Iwan Jones, and Dr. Stefano Allesina, who have provided many valuable insights from their areas of expertise; and also thank Professor Jon McCormack, Dr. Alan Dorin, Professor Larry Yager, Dr. Tim Taylor, Dr. Alice Eldridge, Dr. Cefn Hoile and Dr. Graham Wakefield for the feedback on their work. Goldsmiths, and in particular the Computing Department, provided a comfortable base from which I was able to work on this project. Additionally, in the first year, the college awarded me a grant partially reducing the fees. This research was supported in the form of a PhD studentship by Fundação para a Ciência e Tecnologia in Portugal, contract reference SFRH /BD /61293 /2009. The work Where is Lourenço Marques? was also developed with the 5 financial assistance of Fundación Telefónica under the award Incentives for Ibero-American Production from the VIDA competition, and the technical assistance of Vy Phan, Rui Leitão, Teresa Cardoso, and several anonymous participants. The editing skills of Joanna Pylak were invaluable in all the videos for xTNZ, Senhora da Graça and Where is Lourenço Marques?. I owe a completely different debt to my family, and it is to them that I dedicate this thesis. Love and thanks to Joanna, in whom kindness and generosity are personified, for keeping me company and so much more. 6 Contents 1 Introduction 16 1.1 Motivations . 16 1.2 Hypothesis and questions . 19 1.3 Methodology: . 21 1.4 Presentation of the manuscript . 22 1.5 Public diffusion of the work . 25 2 From Animats to Crowds 28 2.1 Computational ecosystems . 30 2.1.1 The pioneers . 30 2.1.2 Evolution as art . 32 2.1.3 The artistic practice using CEs . 34 2.1.3.1 The new paradigm provided by sonic ecosystems . 37 2.1.3.2 The ‘remediation’ of the technology of the CE in Virtual worlds . 38 2.1.4 Behavioural animation of groups of non-player characters . 39 2.1.5 The problem of the persistence of virtual living populations . 41 2.1.6 The context of WisLM: Evolutionary Art and Relational Aesthetics . 43 2.1.7 Final notes . 44 3 The Materiality of Evolutionary Art 46 3.1 Introduction . 46 3.2 Biots and digital aquaria of life: AI techniques informing the construction of computational ecosystems . 46 3.2.1 The first class: Gtype-Ptype . 47 3.2.2 Cellular automata, the second class . 50 3.2.3 The third class: co-evolution in multi-agent-systems . 51 3.2.4 Other important techniques . 55 7 3.3 Interactive aquaria: A participative audience . 57 3.4 Final notes . 60 4 Computational Ecosystems as Cultural Forms 61 4.1 Introduction . 61 4.2 Aesthetic affiliations of Evolutionary Art . 61 4.2.1 Open systems . 62 4.2.2 An aesthetics of self-organization and feedback . 67 4.3 The meaning of life in the systems . 68 4.3.1 Life forms . 69 4.3.2 A narrative construct . 73 4.4 Final notes . 74 5 The Functioning of Computational Ecosystems 76 5.1 Introduction . 76 5.2 A narrative construction . 79 5.3 Case studies . 86 5.3.1 Senhora da Graça . 87 5.3.2 Vishnu’s Dance of Life and Death . 89 5.4 Final notes . 90 6 A Comparative Study Towards Persistent Computational Ecosystems 92 6.1 Introduction . 92 6.2 Characterization of the problem . 93 6.2.1 Homogeneous model (Hm) . 94 6.2.2 Chemostat model (C) . 94 6.2.3 Heterogeneous model (Ht) . 95 6.2.4 Food web model (F) . 95 6.2.5 Other forms of organization . 98 6.3 Methodology . 99 6.3.1 Questions . 99 6.3.2 Architecture . 100 6.3.2.1 The sub-models . 100 6.3.3 Experiments . 101 6.3.3.1 Algorithms . 101 6.3.3.2 Duration and dimension of the trials . 104 8 6.3.3.3 Sensitivity measurements . 105 6.4 Results . 105 6.4.1 Presentation of the results . 105 6.4.1.1 Comparison of approaches and methods . 108 6.4.1.2 Question 2: Is the predictive strategy better than the rescue ones? . 109 6.4.2 Discussion of the results . 110 6.4.3 Final notes . 113 6.5 Conclusions . 114 7 Where is Lourenço Marques? - I 116 7.1 Introduction . 116 7.2 Context: Mozambique 1974-76 . 117 7.3 Methodology: Chasing stories . 119 7.4 Discussion . 124 8 Where is Lourenço Marques? - II 130 8.1 Introduction . 130 8.1.1 The computational ecosystem as an AI engine . 132 8.2 Description of the system . 133 8.2.1 The agent representation . 133 8.3 Results . 142 8.4 Discussion and future work . 147 8.5 Final notes . 153 9 Conclusions 155 Bibliography 162 9 List of Figures 2.1 Still image from xTNZ, an ecosystem of evolutionary creatures whose physical features (shape and sounds) evolve in a generational time scale by means of a software architecture inspired by Mendelian genetics and Darwinian natural selection [19]. 29 2.2 Left: A reinterpretation of the famous work of Leonardo DaVinci the Vit- ruvian Man, using a CE. A community of dynamic and ‘starving’ pixels eat sections of a digital representation of the drawing based on the hue value of the other static pixels, which become resources. Right: A similar process is at the origin of the Halley series, a series of dynamic works where the digi- tal representations of the paintings from Peter Halley appear reinterpreted by means of the CE. 34 2.3 Some examples of population changes in natural ecosystems. a) Song Thrush in English Farmlands, b) Lynx in the Canadian Artic, c) Bupulus, a moth in German forests, d) Collared Dove in England, e) Spotted Flycatcher in English farmlands, and f) Whitethroat in English farmlands. Scales: the abscissa in all cases is in years. The units of density are relative ones in (a), (d), (e) and (f), and are set to 100 in either 1966 (a), (e) and (f) or in 1972, (d). In (b) the units are thousands trapped per year and in (c) numbers per hectare of forest floor. Source: [190] . 42 5.1 Still image from Senhora da Graça, where a population of plants, herbivores and carnivores evolves. 87 5.2 Still image from Vishnu’s Dance of Life and Death, a choreography for NPCs animated with a CE. 89 6.1 Four distinct models of population, from left to right: Homogeneous (Hm); Chemostat (C); Heterogeneous (Ht); and Food web (F). 94 10 6.2 From left to right: a) 25 runs with model pushing the demographics down to extinction, and b) 25 runs with model pushing the demographics up to over- population.