Consensus-based integration of unordered partitions

In this paper a multi-agent system for the task of integrating individual knowledge stances represented in the form of unordered partitions of a given set of objects is presented. It is assumed that all individual agents are the autonomous and independent sources of initial stances (represented as unordered partitions) that are further integrated in order to give insights on the general stance of the entire society. The investigated approach towards the integration task follows the guidelines of consensus theory. For instance, the assumed criteria for centrality are capture in the form of postulates taken from the literature and lead to the notion of Kemeny's median. Further, two meta approaches to the integration mechanism were described - classical and non-classical heuristics. The former, involves three practical strategies of choosing and merging dominant groups within the partition in order to determine a single partition relatively close to the central stance. The latter, is a simple genetic algorithm for traversing the solution space. Additionally, several pragmatic aspects of each solution were presented and briefly discussed.