How to choose the optimal policy in multi-agent belief revision?

In multi-agent system, it is not enough to maintain the coherence of single agent's belief set only. By communication, one agent removing or adding a belief sentence may influence the certainty of other agents' belief sets. In this paper, we investigate carefully some features of belief revision process in different multi-agent systems and bring forward a uniform framework - Bereframe. In Bereframe, we think there are other objects rather than maintaining the coherence or maximize certainty of single belief revision process, and agent must choose the optimal revision policy to realize these. The credibility measure is brought forward, which is used to compute the max credibility degree of the knowledge background. In cooperative multi-agent system, agents always choose the combinative policy to maximize the certainty of whole system's belief set according to the welfare principle. And in semi-competitive multi-agent system, agents choose the revision policy according to the Pareto efficiency principle. In competitive multi-agent system, the Nash equilibrium criterion is applied into agent's belief revision process.