Fracturing of small social networks
Well connected social networks forming a single component are an important factor for the success and persistence of organisations such as religious and political groups or companies. Therefore, any fracturing of the network should be avoided as it can lead to instability with the formation of sub-groups or sub-organisations or even organisations dissolving. Of special interest are networks that have been stable for long periods of time but have become destabilised as the result of a single disagreement between two individuals. While individuals explicitly "taking sides" for one of the two individuals involved in the disagreement provides a possible explanation for the fracturing of the network, we can observe the same result in cases where individuals not directly involved are trying to be impartial.
We investigate this phenomenon using an agent-based model of social network evolution in which agents interact based on a social network connecting them. This social network is subject to change, driven by the interaction dynamics it constrains. The interaction dynamics, in turn, is influenced by the topology. The structure of the network specifies the interactions that are possible as the weight of an edge between two individuals defines their probability of interaction. The network therefore determines the frequency of each agent's interactions with its neighbours and constrains the possible new edges that can be formed. In our model, in each time step each agent initiates a gathering with a certain probability. When an individual initiates a gathering, it invites all neighbouring agents. These neighbours decide whether to attend or not based on the weight on their edge to the host. If two individuals co-attend a gathering, this leads to an increase in the weight of the edge between them. All edges are subject to slow decay and therefore can only persist if they are continuously reinforced. Furthermore, we assume that each individual has a finite capacity for contacts and therefore assign a maximum capacity for the sum of all edge weights to each individual.
We study the influence of individual's behaviour as well as external influences such as the initial topology on the robustness of the social network to "fall-out events". This might enable us to detect critical points in the network and possible points of influence.
Algorithms and computational methods: Agents
Visualisation and data handling software: Gnuplot
Programming languages and libraries: Java
Transdisciplinary tags: Complex Systems