Article Preview
TopIntroduction
The design of software-intensive systems may receive inspiration from diverse fields such as robotics, artificial intelligence, biology, economics and social sciences. For instance, these inspirations may help managing networks more effectively, and supplying services more reliably. Various simulation approaches of the past 15 years, mainly those of the multi-agent kind, dealt with modelling and simulating self-adapting and self-organising systems in quite different fields. Experience from these approaches can also be used in the development of new, artificial systems.
In this respect, self-organisation and self-adaptation have emerged as two promising facets of a paradigm shift. While self-organisation in a strict sense deals with the emergence of new structure out of a formerly unstructured aggregate, self-adaptation requires that a structured system already exists. Norm emergence and innovation is one of the processes that enable such a system to adapt to its environment and its element to adapt to each other.
Social and economic systems are usually endowed with self-adaptation as they seldom start from a “primordial soup”. If inspiration should come from the analysis of social and economic systems, this analysis must not come from more or less informal social and economic theory (as often in socionics, although socionics has brought about some first results in using patterns of self-adaptation in social and economic systems for design patterns in distributed systems engineering), but from the formal modelling of self-adaptive systems (as in sociophysics and econophysics, which on the other hand rather often use physical and biological metaphors instead of social ones).
The focus of the paper is on intelligently acting agents instead of particles just interacting in and reacting on an environment that is modelled in terms of fields and gradients. In different terms, the focus of the paper is on evolution of co-operation and communication in social and economic systems and what one can learn from these for setting up artificial systems. Real-world social actors are not only intelligent, but they communicate in a way no other living beings communicate, namely by a symbolic language instead by rather unspecified signals (e.g. pheromones), and even if the bees’ waggle dance can perhaps be interpreted as symbolic, the symbolic languages of human beings are far more elaborated. Thus learning from real-world social systems might be even more attractive for computer science than learning from other natural systems, as computer-based information systems are more often than not applied to fields in which they are part of socio-technical systems.
Successfully simulating self-adapting social or economic systems with multi-agent systems and distributed artificial intelligence (instead of “social force” and “social field” concepts) might lead to insights with which algorithms for self-adaptation can be constructed, as a (successful, validated) simulation model of a self-adaptive process in the real (social or economic) world is a self-adaptive software system.
The following sections first deal with human social systems which are more often than not nested systems or systems of systems. The third section gives an overview of traditional and more recent approaches to simulating human social systems, while the final section deals with the current development of a toolbox with the help of which normative behaviour of human actors and of software agents representing them can be described. The conclusion opens the view into new prospects.