We presented a framework for structuring the knowledge of an agent
learning about agents, and then determined the complexities of
learning the different types of knowledge, and the advantages of
having lower complexity in a MAS where the other agents are also
learning. These results can be used by a designer of such an agent to
help decide which knowledge should be put into the agent and which one
learned. Our analysis also takes a first step into trying to elucidate
some of the emergent phenomena we might expect in these types
of MAS, e.g. the results predicted by Theorem 3
were previously observed in a MAS implementation by [7],
and are reminiscent of similar results in [3].
@finalversion
Jose M. Vidal
jmvidal@umich.edu
Thu Apr 24 15:00:31 EDT 1997