Creative Influence and Complex Adaptive Systems
by Matt Garrod, University of Manchester
Think for a moment about your favourite piece of music or artwork. Now consider the fact that the artist who produced it almost certainly had their own favourite piece, which influenced their life just as much. The process of how artists and creators influence each other and draw influences from previous generations provides one example of a complex adaptive system in a social context.
Complex systems are typically composed of many entities known as agents. An agent could be a molecule in a chemical reaction, an animal in an ecosystem, someone in a crowd or a myriad of other possibilities. There is no single definition of a complex adaptive system, though some of their main features are neatly summed up by John Holland [1]. These include: the presence of many interacting agents, nonlinear interactions between these agents, diversity among agents, as well as “Internal models” employed by the agents. The latter can be thought of as a set of rules which an agent obeys when faced with a situation. For example, people moving in a large crowd will, on average, tend to stay a certain distance from each other.
The presence of non-linear interactions between agents in a system is often synonymous with chaotic behaviour. This can be described as “irregular behaviour in a deterministic system which persists over a long time and exhibits sensitivity on initial conditions.” [2] The point of this being: if you throw a stone of a given mass into a pond in which linear interactions occur, you’ll probably know how big the splash will be. However, in a chaotic system, a slight difference in the mass of the stone or the trajectory at which you throw it could mean the difference between a few ripples and a tsunami.
The result of all this uncertainty is that scientists often turn to computer models when studying complex systems. A good example of one of these models in the context of social influence is Robert Axlerod’s 1997 work on “The Dissemination of Culture” [3]. His model consisted of a series of agents placed at fixed sites on a two-dimensional grid, much like a chessboard. Each agent possessed a set of features which could take a series of different values. For example, if the feature was hair colour the values may correspond to blonde, brunette or black.
We can imagine the situation as a packed auditorium in which the audience have turned up with a random selection of clothes, hairstyles and other features. Each member of the audience can only talk to, or “interact” with, their immediate neighbours with a probability based on how much they have in common. The population of this world is easily swayed; the outcome of a conversation with someone will result in them changing one of their features to copy their neighbour.
The simulation was typically allowed to run until no more changes occurred to the features possessed by any of the agents. Axlerod often found that the end result would be a series of different homogeneous regions between which no interaction could occur due to a lack of any common features. In this case our audience has changed from being effectively random to something more like the stereotypical American high-school canteen, split into different cliques. This process is also comparable to the formation of different nations within the same continent over time.
To make an analogy between Axlerod’s model of social influence and the process of creative influence we could imagine giving everyone in our audience a canvas and paint. The formation of large homogeneous regions could now correspond to whole swathes of the population all deciding to paint boats. This concept emphasizes where our comparison between the cultural influence analysed in Axlerod’s model and creative influence breaks down. In creative endeavours, it is typically undesirable to copy others’ work (though this doesn’t stop some people trying). To compensate for this, one could imagine Axlerod’s model being adapted for use in the study of creative influence by placing a limit on the number of features shared by agents.
Another aspect Axlerod considered was the effect of the range of interactions on the size of the regions formed. It was discovered that the number of homogeneous regions decreased as the range of interaction increased [3]. In modern times the range of interactions between individuals in society can be thought of as effectively limitless; this is due to the presence of the internet and other remarkable means of communication. This implies that the effect of enhanced communication and transport on humanity is to promote increased homogeneity; this forms the basis of the concept of globalization.
Despite the importance of interactions between creative individuals in determining what they produce, it is important not to ignore the “internal model” within each agent. In the case of humans this is the human brain, which is arguably more complicated and more difficult to study than the interactions between its owners.
In summary, simple models such as that devised by Axlerod can provide a surprising amount of insight into how complex adaptive social systems behave. However, unravelling the creative process once and for all is sure to require thorough dialogue between neuroscience, computer science and the more general study of complex systems. As is often the case in science, analogies between models and reality should not be stretched too far. Nevertheless, through careful comparison with real measured data, these models have the potential to become relevant to the real world.
References
1. Morowitz, H and Singer, J. The Mind, The Brain and Complex Adaptive systems. Addison-Wesley (1995). pp54-47
2. Storgatz, S. Nonlinear Dynamics and Chaos. Perseus books Publishing (1994). p323
3. Axlerod, R. Complexity of Cooperation. Princeton University Press (1997). pp148-174