This thesis demonstrates that many large, socio-technical systems fail to meet the goal of their sponsors and that one reason for this failure is that socio-technical systems are complex, co-evolutionary systems and that once such systems reach a particular level of complexity, they become as unpredictable as eddies in a stream.
Technology based systems are built by engineers; systems engineers, software engineers, telecommunication engineers, mechanical engineers, etc. Engineering is goal-directed and based on the laws of physics. There are goals to achieve and requirements to be met. Thus engineering, and systems engineering particularly, is concerned with plans for the future. These plans are based on predicting the global state description of the system of interest, or in other words the behaviour, at specified times in the future. In a truly complex, large socio-technical system, this planned predictability can break down into a morass of unanticipated behaviour.
This unpredictability comes from the propagation of the effects of change through coevolving subsystems. The effect of a change in one subsystem is often deterministic and in general predictable but when influencing links connect many subsystems, the effects become complex. The propagating changes originating from two or more initial changes interfere and alter the outcome. The subsystems' new states apply conflicting pressures for change on each other inducing them to enter a cyclic or chaotic set of states. These changes to individual subsystems or small groups of subsystems are local in nature but their aggregate effect is evident in the emergent behaviour of the system. The holistic system behaviour is unpredictable because it is dependent on a multitude of local effects.
At the local level the pressure one subsystem has on another is dependent on the degree of coupling between the subsystems. Within a subsystem, there is again conflict, the magnitude of alteration in one dimension being traded off against that of another so that the aggregate effect affects the pressure for change. The threshold computation and the propagation of change can be brought together in a protocol, i.e. a set of rules, that is described in the thesis. This protocol relates the pressure for change, the subsystem aims, the risks and costs of change, and the control of change. It also describes the effects of the propagation of change.
The ANZAC warship is used to illustrate the findings of the research. Each ship is a very complex socio-technical system so changes must be carefully considered in order to achieve the required result. Furthermore, each ship has an unending list of planned changes as capability requirements change.
A tool was needed so that multiple changes could be simulated concurrently or successively and the implications shown. The thesis describes a number of concepts that could be the basis for such a tool. The tool, named SeeChange, was designed and built as part of the research.
In a socio-technical system, the non-deterministic effect of the behaviour of humans also needs to be considered in depth. Human behaviour makes it extremely difficult to foresee the effect of changing one variable in a complex scenario. Building a large information system, for example, is a complex undertaking and the introduction of human "subsystems" raises the level of complexity. Often when change is contrived to achieve a particular effect the implementer is bewildered when the change achieves a different effect.
This thesis draws attention to the presence of chaotic states within relatively simple complex systems and extends that concept to real world systems. It highlights the certainty of change, the uncertainty in foreseeing the effects of change, indicates why some systems may fail and offers some suggestions that may help to alleviate the problem of complex socio-technical systems failure.
In summary, the thesis demonstrates that the effect of changes propagating within a complex, socio-technical system is inherently unpredictable and that there is a strong probability of the system failing to meet expectations. Therefore, there is not and, importantly, cannot be a single solution for preventing chaotic, socio-technical system failure, however, the SeeChange program coupled with the alleviative statements suggest a mechanism for improving the success rate and turning the nature of change to advantage.