What is systems theory

Ludwig von Bertalanffy identified two essential features of systems theory. Firstly, it is transdisciplinary; secondly, it is a science of systems – indecomposable entities or ‘wholes’. According to Bertalanffy, there are general aspects, correspondences and isomorphisms that are common to very different systems. Even though such disciplines as physics, biology or psychology are ‘encapsulated’ in the boundaries of their domains, and communication between them is limited, independently of each other, similar concepts and problems appeared. Examining these concepts, problems and correspondences is a subject of general systems theory (Bertalanffy, 1968:

xx). The problems that appeared in one discipline have led to patterns of solutions that were applicable in other disciplines. The transdisciplinary character of system science finally found its more precise, mathematical formulation in computer science. According to Weinberg (2001: 46), the systems theory focuses on three activities:

1. General systems thinking, which is about methods and approaches;

2. General systems application;

3. General systems research, which is about creating new laws and refining old ones.

SYSTEMS THINKING

The systems thinking is a framework of thinking or an approach to thinking about systems. It is based on the principle that systems are wholes and that their elements should be examined in terms of their relationships with each other.

There are two central approaches within the systems theory: soft systems and hard systems. It is worthwhile to introduce these approaches, because architecture design includes elements of them both. A generative design system specifically, should be able to tackle problems typical for both soft and hard systems.

Soft systems are difficult to quantify, and they are often associated with social systems. The methodology of soft systems is aimed at providing a framework for tackling real-world situations, the ‘messy’ situations involving psychological, social and cultural elements – so called ‘wicked’ problems.

Hard systems on the other hand involve well-defined problems – these that have a single, optimum solution, that can be approached with analytical methods and these, in which technical factors tend to dominate. They are easier to quantify and their methodologies often use computational examination methods such as simulations, systems modelling, systems analysis or optimisation methods.

The essential difference between soft systems and hard systems lies in the definition of a system. The hard systems’ approach is ontological – it defines systems as objects existing in the real world or bounded entities with a physical existence. The soft system’s approach is epistemological – it understands systems as mental constructs. In this approach, a system is conceptualised from many perspectives, each of them providing a different understanding of it (e.g., the same man can be considered as a freedom fighter or as a terrorist). Soft systems investigate problems that are complex, i.e., where nonlinear relationships, feedback loops, hierarchies and emergent properties have to be taken into account. The soft system’s method postulates understanding of a system, by iterative learning process.

The elements of these two approaches are discussed more in-details in a number of places in the thesis, for example in the contexts of a design process in section IV, or in the context of building characteristics in section VI.

One of methods applied within system thinking is system dynamics.

System dynamics is universal approach, i.e., it is applicable to both hard and soft systems. Sterman (2000) defines system dynamics as follows:

System dynamics is a method to enhance learning in complex systems. Just as an airline uses flight simulators to help pilots learn, system dynamics is, partly, a method for developing management flight simulators, often computer simulation models, to help us learn about dynamic complexity, understand the sources of policy resistance, and design more effective policies. (Sterman, 2000: 4)

According to system dynamics, the complexity of the real-world systems exceeds human capacity to anticipate the implications of these systems.

Human beings operate within a complex network of positive and negative feedbacks, which are extremely difficult to model mentally. In examining complex systems, one needs to take into account phenomena such as time-delays, poor reasoning skills, defensive reactions and the like. Because mental simulations of complex behaviour are unreliable, more formal simulations, especially computer simulations, are useful.

SYSTEMS APPLICATION

Although systems theory was initially aimed at embracing a wide array of disciplines, it has a tendency to ‘dissolve’ when it is used in the context of a specific application. Mechanisms investigated by the systems theory are abstractions, and it is as such that they can be studied. Applied to specific problems, the mechanisms have to be carefully adjusted.

For example, the notion of natural evolution in its abstract, simplified form can be applicable to a large number of phenomena in many disciplines.

Broadly understood, the concept of evolution can be applied to language (e.g., how spellings have changed over time), society and culture (e.g., from simple to complex kinship systems), individual human beings (e.g.,

developmental changes). When the notion of evolution is applied to stellar development for instance, it both reduces its meaning (as it does not involve natural selection) and it extends it to a specific content (the description of a star’s life phases). The phases in which, through gradual warming, a star become a red giant, a white dwarf and a black hole are not constrained by natural selection, but by the initial mass of a star. Alternatively, applying the laws of biological evolution to society and culture, one can predict that stronger, more flexible or more fitted cultures survive. But the risk is that such a statement oversimplifies the intricacy of cultural phenomena.

The generalized mechanism of natural evolution is applied also in Evolutionary Computing as a general problem solving mechanism. In section VIII I applied this mechanism to the development of a building model.

SYSTEMS RESEARCH

General system research initially aimed at finding a central, unified theory (a general system theory), which could explain the behaviour of all systems in different disciplines. Nowadays, the aim of the general system theory is probably less ambitious. General system research transformed into field called complexity science, whose research interests, though interdisciplinary and broad, focus on mathematics and natural science.

Systems research – in terms of developing transdisciplinary principles or laws – is intertwined with systems application. Systems research affects the investigation of specific phenomena and reversely the specific application modifies the understanding and formulation of general mechanisms.

For example, the observation of a natural population growth led to a development of a mathematical model called ‘logistic map’. This

mathematical model was intensely examined, and resulted in the elaboration of interesting, new properties of systems, such as ‘deterministic chaos’¹⁴. The general mechanism of deterministic chaos can be in turn applied to other phenomena, such as dynamics of weather (a popular ‘butterfly effect’), economic and social systems dynamics, or bouncing ball dynamics.

Evolutionary Computing is an example of natural selection understood in the abstract sense. This computing technique arose from examination of biological phenomena – first represented mathematically, and then applied to computer science. As such, the mechanisms of EC are the subject of

distinctive research. At the same time, EC is a method for solving actual problems and have specific applications.

14 One of these remarkable properties of the ‘logistic map’ equation is that even though it is simply formulated and deterministic, there is certain range of parameters for which the equation produces series of

‘random’ numbers. Moreover, the equation yields different strings of random numbers for only slightly different initial parameters. This property has been called a ‘deterministic chaos’ (Mitchell, 2009). In fact, whatever the initial parameters were, the generated string of numbers was calculable (deterministic), but even a miniscule change in the initial parameters would result in a significant change in the generated numbers (chaos).