Bruce Edmonds and Kerstin Dautenhahn (2001)

Starting from Society - editorial

Journal of Artificial Societies and Social Simulation vol. 4, no. 1,
<https://www.jasss.org/4/1/0.html>

To cite articles published in the Journal of Artificial Societies and Social Simulation, please reference the above information and include paragraph numbers if necessary

Received: 25-Jan-01      Published: 31-Jan-01

1.1

Almost by definition simple systems are those that we understand. This contrasts with complex systems which seem have the ability to surprise us because we do not (or cannot) have adequate models of them. Researchers who attempt to understand complex systems by building e.g. computational models often face the dilemma that the model's behaviour might adequately match the behaviour that one attempted to replicate, but that the predictions the models give inevitably diverge from reality as soon as one attempts to apply the models outside the context they were developed in. This might be acceptable in cases where the model's behaviour is sufficient e.g. for usage in a particular application area, but it is not sufficient if the basic motivation is to understand a system by modelling. In these cases we have a choice of accepting the inaccuracy of our models or restricting their use to quite specific domains.

1.2

Human society is, by its nature, a very complex system. There are many reasons why this is so, just to name a few: human society is composed of individuals (complex systems on their own) who will act to effectively confound any model made of them; it is the ultimate medium for the transmission and operation of semantic constructs which can be based upon substantial chunks of historically and spatially distributed experience; it has been developing as a result of the interactions and artefacts of individuals that are themselves developed as a result of their social environment; it is composed of many complex, contingent and context-dependent processes; and our models of it affect our perception of it and hence its operation.

1.3

There have been many attempts made to use computational systems and algorithms that have been originally developed for other purposes or in other contexts, as models of some aspect of society or as optimisation techniques, for example Genetic Algorithms. Typically, in this sort of work, a fairly "clean" and straightforward "general purpose" basic algorithm is chosen and a computation is performed whose parts are labelled with suggestive terms to emphasise the analogy with biological or social systems (as normally perceived). The results of the computation are then analysed and conclusions drawn which are projected (implicitly or explicitly) upon the aspects of the social or biological system to which the analogy applies. Work along these lines in modelling societies can be very helpful in that it can give us new ways of thinking about society which may usefully inform our intuitions. However it is often mistakenly taken that such models tell us directly about social processes - a confusion that is not helped by the language used in many social simulation papers where the motivation of the paper is confused with the design and analysis of the model. Given the complexity of social systems that exist in the real world, we cannot suppose that such clean and simple constructs would inform us about the reality of messy and contingent social world we inhabit.

1.4

However the other direction seems less problematic: if we can identify processes that have developed and survived in society then they might well be suitable for application in computational systems which are too complex to completely control (and, in particular, too complex to control by means of deliberate design). In this way we can avoid over-enthusiastically projecting our system's behaviour onto society, and going cap in hand to society to learn something from it.

1.5

One can characterise the scientific method as one of collecting significant new information and deliberately trying to disconfirm existing models - the two factors, novelty and selection, being those necessary for an effective evolutionary process to occur (in this case the evolution of knowledge). In other words, the essence of good science is to continually 'seek to be surprised'. Thus, if one wants to do good science then complex systems are a suitable source of information.

1.6

Since societies (and especially human societies) are extremely complex systems, they are an ideal source for ideas: for processes and systems that work in the face of huge complexity as well as for counter-examples to over-hasty theories and designs. Biological systems are a similar source, whose influence has already started to be felt in computer science and artificial intelligence, informing e.g. learning, perception and sensori-motor control in robotic models. Now maybe it is the turn for social systems. To look at such cases was our intention for this special issue and the preceding workshop of the same name^[1].

1.7

None of the papers is a pure application of ideas taken from society to computational systems - all are to greater or lesser extent involved in the interplay between models and society, but the papers do reflect the emphasis of looking to society for ideas and resist the temptation of projecting over-simple models onto it.

1.8

Kerstin Dautenhahn and Steven Coles apply the idea of narrative intelligence to autonomous robotic agents. They develop a minimal computational framework based on stories, story-telling and autobiographic agents, and apply it to a computational model in a simulated environments. The article confirms the hypothesis that story-telling can be advantageous, i.e. that it can increase the survival of an autonomous, autobiographic, minimal agent in a simulated but quite 'realistic' environment. The proposed framework argues for the study of narrative from the bottom-up, namely starting with simple minimal agents as subjects and analysing the function and origin of narrative in computational and robotic studies.

1.9

Alexander Staller and Paolo Petta examine the bi-directional interrelationship between social norms and emotions. Taking the ideas that emotions play an instrumental role in the sustenance of social norms and that social norms being an essential element of regulation in the individual emotional system. Their computational studies draw heavily upon the functional appraisal theory of emotions. They describe a first implementation of a situated agent architecture that incorporates a simple appraisal mechanism and report on its evaluation for aggression control as a function of the prevalent norms. Their approach towards explaining and simulating human social behaviour is presented as a necessary alternative to many existing models that are based on logic.

1.10

Rosaria Conte and Mario Paolucci provide a contrast to the other contributions in this special issue, due to the top-down analytic approach it employs. It belongs to a philosophical tradition that has frequently been guilty of projecting over-simple ideas onto complex systems such as society. The article analyses social learning and imitation in terms of mental processes, although alternative approaches in biology, psychology, robotics and other fields that follow a different, bottom-up approach are acknowledged^[2]. However, this paper does draw heavily on real examples to motivate its analysis of social facilitation and imitation, it does consider these with respect to a continuum of cognitive complexity and it does return from its abstraction to draw some conclusions as to the advantages of intelligent social learning in agent systems applications.

1.11

Dietrich Fliedner discusses the topics of complexity and society from the point of view of a social geographer. He takes examples from his field (especially the settlement patterns of the abandoned Indian Pueblo Pecos and the area in New Mexico settled by the Spaniards) to draw conclusions about how such societies might be profitably modelled. In particular he re-examines the ideas self-organisation and Autopoiesis in the light of the different types and levels of complexity to be found in such systems.

1.12

The co-editors hope that these papers will further encourage work where the complexity of social systems is seen as an invaluable inspirational resource, rather than something to be simplified by the imposition of neat models formulated on an a priori basis. Our view is that any simplification and generalisation is likely to be entirely premature because it could only sensibly occur after a lot more field work and bottom-up descriptive modelling, testing and evaluation.

Notes

¹ AISB-00 Symposium Starting from Society - the application of social analogies to computational systems, a symposium at "Time for AI and Society" 2000 Convention of the Society for the Study of Artificial Intelligence and the Simulation of Behaviour , 17th-20th April 2000, University of Birmingham, England, http://www.cpm.mmu.ac.uk/~bruce/sfs/

² Non-mentalistic and experimental, interdisciplinary approaches and analyses of imitation can be found in Imitation in Animals and Artifacts, K. Dautenhahn and C. L. Nehaniv (Eds.), MIT Press, to appear 2001; Imitation in Natural and Artificial Systems, C. L. Nehaniv and K. Dautenhahn (Eds), Special Issue of Cybernetics and Systems, Volume 21 (1-2), Taylor and Francis.

Return to Contents of this issue

© Copyright Journal of Artificial Societies and Social Simulation, 1999