Swarm intelligence : SI overview – Part 2


The role of diversity

In humans, SI is often utilised in the context of government panels, scientific committees and teams in companies. The size and composition of the group is strategically planned with the aim to maximize the benefits of group performance in terms of problem solving. The composition of such groups is often arranged in a way that utilises functional diversity (differences in the approach used for problem solving) and identity diversity (differences in race, gender, religion or ethnic background) to maximise the performance potential and perceived social equality of the group through representation. Studies that investigated the effect of identity diversity on performance (in terms of financial gain) through direct social interactions in teams or organisations, however, obtained mixed results. In some studies, increased diversity was not beneficial to group decision making, whereas in others a clear performance advantage was found. These results are at least partly owing to the fact that differences in perspective can create communication barriers and that identity diversity can result in a lack of trust and mutual respect. This means that there is a need to distinguish between the SI potential of a group and what can be realised, both of which are interesting areas of research in their own right. An important point regarding functional diversity is that a range of skills in itself within a group is not SI (in fact, this is a common misconception about SI) but it is the interaction between these functionally specialized individuals that has to produce new solutions.

Internet use for SI purposes

The electronic media have unlocked a hitherto largely untapped potential for SI in humans with relevance for company management, product development, prediction of elections and the media entertainment sector to name but a few. In fact, the strong interest from companies has already resulted in many meetings on this topic organised by industry. Information gathering from individuals has become possible on an unprecedented scale facilitated by the open and wide-spread access to electronic media that enables the process of tapping directly into the collective knowledge or memory and most importantly creativity of huge (potentially global) collections of people. Furthermore, every simple desk-top computer now has enough processing power to deal with such large data sets. But what should be done with the information that has been acquired in this way? Are there solid lines of approach that can be followed? A quick look at the internet gives the impression that SI has become the playground for both innovative research and wild speculations, generalisations and misunderstandings, a situation that partly motivated our decision to write this article.

Open-access interactive systems

Striking examples of recent SI use are internet-based platforms for discussions of problems that are characterised by open access (i.e. every user can contribute to a given problem) and their interactive format, which enables potentially vast numbers of people from all over the world to voice their opinions and make a contribution. The open and interactive qualities result in partially self-organised and partially decentralised systems. Examples of this form of SI use are the development of the Linux operating system (http://www.linux.org/info), The Apache Software Foundation (http://www.apache.org/), medical discussion forums and user-driven content for design problems. On the more experimental front is the collective management of football clubs. It is debatable whether Wikipedia (http://www.wikipedia.org/) should be considered as an example of SI because bits of information are mainly accumulated but not necessarily processed to provide a new cognitive solution to a problem. Admittedly, it is possible for people to edit what others have written and to combine ideas in new ways. Nevertheless, the very purpose of Wikipedia is largely incompatible with the SI concept because Wikipedia is meant to represent what is already known about a particular topic, rather than generate new insights. We identify a potential problem with SI; namely, that the simple averaging of all opinions will not make the best use of creative sparks by gifted individuals. however, shows that this is not necessarily a general weakness of SI because use-driven content can be a highly creative process making use of SI. In conclusion, whether SI stifles or promotes creativity probably depends on how opinions interact with each other.

Centralised information collection

We have already given an introduction to this category in the section on the possibilities and limitations of SI, with the example of the marbles in a glass jar. The aim of this type of SI use, which is known as a prediction market, is usually the assessment and prediction of developments. In this case, the interaction between the opinions occurs in form of trading virtual shares but the interpretation of the share value often requires further calculations. In all animal groups and many cases of human groups, social interactions will substitute for these calculations. Prediction markets are essentially a tool and, therefore, can be used for private gain (they were originally developed to provide companies with a competitive advantage) or public good depending on how the tool is used. However, prediction markets lack the partially self-organised quality of open-access systems.