An investigation of iterated multi-agent belief change

摘要

Multi-agent belief change is an area concerned with the belief dynamics of a network of communicating agents. A network is represented by a graph, where vertices represent agents that share information via a process of minimizing disagreements between themselves. Previous work by Delgrande, Lang, and Schaub addressed belief change through global minimization, with a weak notion of distance between agents. We extend it by applying iterative procedures that take distance into account. We have identified two approaches to iteration: in the first, a vertex incorporates information from its immediate neighbours only; in the second, a vertex incorporates information from progressively more distant neighbours. Our research has both theoretical and practical contributions: first, we define the iterative approaches, find relationships between them, and investigate their logical properties; then, we introduce a software system called Equibel that implements both the global and iterative approaches, using Answer Set Programming and Python.