On Generative Systems

On Generative Systems

Generative Systems essay,
UCL, the Bartlett school of architecture,
MSc Urban Design, UD 03.14,
Urban Evolution -
The Thames Gateway
Fabian Neuhaus 2006-03-24

Introduction - what are Generative Systems
Description and terminology gen-er-a-tive, adjective - of or relating to reproduction. 1) able to produce: the generative power of the life force. 2) Linguistics applying principles of generative grammar. Origin late Middle English: from late Latin generativus, from generare ‘beget’ sys-tem, noun - 1) a set of connected things or parts forming a complex whole, 1.1) a set of things working together as parts of a mechanism or an interconnecting network: the state railroad system | fl uid is pushed through a system of pipes or channels. 1.2) PHYSIOLOGY a set of organs in the body with a common structure or function: the digestive system. 1.3) the human or animal body as a whole: you need to get the cholesterol out of your system. 2) a set of principles or procedures according to which something is done; an organized scheme or method: a multi party system of government | the public school system. 2.1) orderliness; method: there was no system at all in the company. 2.2) a set of rules used in measurement or classifi cation: the metric system. 2.3) (the system) the prevailing political or social order, esp. when regarded as oppressive and intransigent: don’t try buckling the system [Defi nition from Dictionary Thesaurus] If these general defi nitions are now brought together, it can be read as reproduction of systems. Or more in the sense of the system itself, the system is able to reproduce itself or similar, identical systems. The generative aspect is not only focusing on reproduction in the sense of really produce a new system, it also means the system is generating it self new in parts of it and is able to organize itself according to any inputs or environment.
A generative system therefore is not only functioning, in terms of staying alive (not collapsing) or fulfi lling a function (target), it also is able to generate itself (self-control). This means it grows out of the technical term of systems like railroad or pipes, into a more nature like defi nition.
Therefore it could also be named as an artifi cial system with natural like behaviour.

General System Theory
The general theory of systems was fi rst introduced in 1968 by Bertalanffy.
He set out to replace the mechanistic foundations of science with a holistic vision: General system theory is a general science of “wholeness” which up till now was considered a vague, hazy, and semi-metaphysical concept. In elaborate form it would be a mathematical discipline, in itself purely formal but applicable to the various empirical science. For science concerned with “organized wholes”, it would be similar signifi cance to that which probability theory has for sciences concerned with “chance events” [Bertanalffy, 1968].
To make his point, Bertanalffy pinpointed a dilemma that had puzzled scientists since the nineteenth century, when the novel idea of evolution entered into scientifi c thinking. Whereas Newtonian mechanics was a science of forces and trajectories, evolutionary thinking - thinking in terms of change, growth and development - required a new science of complexity [Fritjof Capra, The web of Life - a new Synthesis of Mind and Matter, page 47].
Generative systems now bring together the reproduction and self organisation with the system approach. But does it goes as far as evolving systems. Does it bring in an evolutionary idea of growing systems that evolve, learn and decide?

Selforganising Systems
Pattern and structures
Throughout the history of Western science and philosophy there has been a tension between the study of substance and the study of form. The study of substance starts with the question, “What is it made of?” the study of form with the question, “What is its pattern?”
The key to a comprehensive theory of living systems lies in the synthesis of those two very different approaches, the study of substance (structure) and the study of form (or pattern). In the study of structure we measure and weigh things.
Patterns, however, cannot be measured or weighed; they must be mapped. To understand a pattern, we must map a confi guration of relationships. In other words, structures involve quantities, while pattern involves qualities [Fritjof Capra, The web of Life - a new Synthesis of Mind and Matter, page 81].
Is there a common pattern of organization that can be found in all living systems? This is the case. Its most important property is that it is a network pattern.
Whenever we encounter living systems - organisms, parts of organisms or communities of organisms - we can observe that their components are arranged in network fashion. Whenever we look at life, we look at networks.
The first and most obvious property of any network is its non linearity - it goes in all directions. Thus the relationships in a network pattern are non linear relationships. In particular, an infl uence, or message, may travel along a cyclical path, which may become a feedback loop. The concept of feedback is intimately connected with the network pattern.
Because networks of communication may generate feedback loops, they may acquire the ability to regulate themselves. For example, a community that maintains an active network of communication will learn from its mistakes, because the consequences of a mistake will spread through the network and return to the source along feedback loops. Thus the community can correct its mistakes, regulate itself, and organize itself.

The Aspects of Living Systems
In a machine such as a ipod the parts have been designed, manufactures, and then put together to form a structure with fi xed components. In a living system, by contrast, the components change continually. There is a ceaseless fl ux of matter through a living organism. Each cell continually synthesizes and dissolves structures, and eliminates waste products. Tissues and organs replace their cells in continual cycles. There is growth, development and evolution. Thus from the very beginning of biology, the understanding of living structure has been inseparable from the understanding of metabolic and developmental process.
This striking property of living systems suggests process as a third criterion for a comprehensive description of the nature of life. The process of life is the activity involved in the continual embodiment of the system’s pattern of organization. Thus process criterion is the link between pattern and structure [Fritjof Capra, The web of Life - a new Synthesis of Mind and Matter, page 255].
These three aspects can be seen as a description for living organisms. Recent computer simulation are able to simulate simplifi ed artifi cial versions including these aspects. From studying these simulation we are able to learn more about systems.

Cellular Automata
To simulate these self organising systems developers looked for the simplest way to simulate a network of cellular processes embodying an autopoietic pattern of organization. This meant that they had to design a computer program simulating a network process, in which the function of each component is to help produce or transform other components in the network. As a cell, this autopoietic network would also have to create its own boundary, which would participate in the network of process and the same time defi ne its extensions [Fritjof Capra, The web of Life - a new Synthesis of Mind and Matter, page 190].
A cellular automata is a rectangular grid of regular squares, or “cells”, like a chessboard. Each cell can take on a number of different values and has a defi nite number of neighbour cells that can infl uence it. The pattern, or “state”, of the entire grid changes in discrete steps according to a set of “transition rules” that apply simultaneously to every cell.
These “machines” or simulated systems are because of complexity reasons mostly depending on one time cycle. Whereas living structures are dealing with a large number of different time cycles. Some of these cycles even can be contrary. Maybe this is just a question of time in terms of computing power? Or are living systems complex enough that they can’t be simulated in all their complexity?
Maybe to use a picture for this we can refer to fractals. Images like the Mandelbrot form show an endless borderline. You can zoom in as much as you want and there is still a new picture with this pattern, self-similar to the previous. But the exiting thing to me is the understanding of infi nity. To be endless does not necessarily mean to be a straight line running a head forever. It is found in quite small patterns like the Mandelbrot form. Maybe this image can be transferred to complexity. Complex systems may not need to look like hard, dense packed patterns of knots. I can imagine them to be quite light and fluid.

Development and evolution
As it keeps interacting with its environment, a living organism will undergo a sequence of structural changes, and over time it will form its own, individual pathway of structural coupling. At any point on this pathway, the structure of the organism is a record of previous structural changes and thus of previous interactions. Living structure is always a record of previous development and ontogeny - the course of development of an individual organism - is the organism’s history of structural changes [Fritjof Capra, The web of Life - a new Synthesis of Mind and Matter, page 215].
However, rather than being determined by outside forces, it is determined by the organism’s own structure - a structure formed succession of autonomous structural changes. Thus the behaviour of the living organism is both determined and free. Moreover, the fact that the behaviour is structure-determined does not mean that it is predictable.

Conclusion
Simulation
It all seems to come back to “the rules” and “the point”. “The rules” to defi ne the structure, patterns and process, to defi ne the interaction and action the system can undertake.
“The point” on which someone (planner?) says stop, here we start building it. In this case the generative system is used only as a quite complex simulation of a process to fi nd structures to be built. Therefore a evaluation tool would be needed to pick up the possible solutions out of all produced solutions and evaluate out of these the one output that matches the criteria best. In actual examples developers use genetic algorithms to implement a kind of selection tool (see John Frazer, an Evolutionary Architecture). It turns out to be very important to evaluate the outcome. Maybe this is even more important than the input (the rules). Through to our recent computing power we are able to generate more than we can handle. Therefore the reduction of all outcomes down to these few possible ones is a new diffi cult task. It seems to me that this should be already considered a the very beginning.
Understanding
But could it be possible that we do not see generative systems as a simulation? Is there more into it? Maybe if we go more into the process and rather be far from the common thinking of planning and building.
To see a new development as a process is a must for this. The question may lays in the cycles. Where are they? What do they cause? How are they related to structure, pattern and process?
Especially in the fi eld of urban design, hardly a project is built in one go with only one planner. And then during the life time, projects turn out to have many similarities with living organisms, due to host, being used and adapted by people.
Of course in planning somewhere we have to start building, whether it is a house, a street or planting a three. But despite all planning and thinking rules the process goes on. Any introduced element becomes immediately part of a system whether the project is fi nished or not. Life takes over, time is moving, and they’re not waiting for the project to be finished.
The idea of anyone saying stop is only in theoretical simulation possible. But in the real world the people start using it and adapt it to their daily life. For me, at this point, this is the big issue in system thinking and knowing about selforganising structures. Every action we undertake is done in a living (in terms of changes and exchanges) environment. And as the system is able to organize itself it is going to deal with the input, in one way or another. This does not mean it doesn’t matter what we put in, the opposite would be through. A project sets up a structure but live’s (people) is going to fi ll, use and change it. This for, it has to be designed. The project has to a low the people to generate their networks but itself has to be integrated into a network and this means also it has to be able to change. There is no point of talking only about simulation. It is more about understanding the function and according to this to set up a design to allow structures to nest, act, maybe grow, but to communicate.
Project
In terms of our project the introduction of process, network and organization goes beyond individualization - introduced through the starting idea of mobility - and brings the elements back together. This may set up the city, may not necessary in common terms but in terms of system thinking. The overall idea turns out to be a system that can be read as a city.
City therefore doesn’t need to be a city in common terms, more in terms of conglomeration of single individuals forming a network that allows them to interact and through this (self-) organization occurs automatically. City is also about no isolation, being connected.
Size in this case matters only in terms of surviving. Therefore the limitations of size in both directions big and small have to be explored. Are there any such boundaries?



Bibliography

Fritjof Capra, The web of Life - a new Synthesis of Mind and Matter, London, Flamingo, 1997

John Frazer, an Evolutionary Architecture, London/Oxford, Architectural Association Publications, 1995

Kevin Kelly, Out of Control, Basic Books, 1994

Manfred Wolff Plottegg, Generative Systems, http://plottegg.tuwien.ac.at/vo030205.htm [accessed 2006-03-22]