What Necessitate Multiply Sectioned Bayesian Networks?

Multiply sectioned Bayesian networks (MSBNs) provide a coherent framework for probabilistic reasoning in cooperative multi-agent distributed interpretation systems (CMADISs). Previous work on MSBNs fo-cuses on the suuciency of MSBNs for representation and inference with uncertain knowledge in CMADISs. Since several representation choices were made in the formation of a MSBN, it appears unclear whether certain choices were necessary. For example, it is unclear why a hypertree organization of agents was imposed. This study focuses on the necessity of MSBNs for representation of uncertain knowledge in CMADISs. We identify a small set of fundamental choices which logically implies a MSBN or some equivalent representations. We consider privacy of agents to be essential if we are to allow agents developed by independent vendors so that vendors' know-how can be protected. We found that the privacy of agents plays an important role in this necessity analysis. The study provides insights into the MSBN framework and valuable guid-ances to multiagent system researchers whether they are satissed with the framework or unsatissed with the restrictions imposed.