Value-Sensitive Design of Self-Organisation

Self-organization refers to the capacity of networks of agents to combine and recombine their activities towards some form of overall order without centralized coordination. Such global order is the result of each agent acting based on its own interests and values, whose combination results in a situation that is profitable for all. As such, these systems are particularly robust, because they adapt to changes, and are able to ensure their own survivability. The growing complexity of applications requires solutions based on autonomous, robust and adaptive components. An important challenge for self-organizing systems is to ensure that constituent agents are such that desired emergent phenomena are produced. In this paper, we propose an architecture for self-organising agents based on values. This architecture presents two main advantages. Firstly, the use of values as basis for agent coordination results in solutions that do not impinge directly on the autonomy of the agents, as values are more abstract than goals and plans and therefore leave enough freedom for agents to choose their own paths. Secondly, values also enable higher level adaptability to environmental changes by providing a common criteria for the adaptation, which is independent of the environment. i.e. Whereas goals refer to specific desirable states of the environment, which might change when the environment changes, values refer to preferences for types of situations. These preferences keep stable under most changes of the environment. Even though self-organisation often occurs for functional/self-interested reasons, the fact that agents form an organisation leads to satisfaction of affiliation motives which may in turn be the reason for maintaining the organisation. The proposed architecture extends our work on modelling social agents and social practices. We show that using these social aspects as foundation for the system leads to stable adaptation patterns while maintaining a level of autonomy and robustness of the overall system.