Multi-Agent Belief Reasoning in a First-Order Logic Back-Chainer Technical Report Arnold Binas and Thomas R. Ioerger {a0b1767, ioerger}@cs.tamu.edu September 14, 2004 Abstract This report gives a brief overview of and a few pointers on belief reasoning for multi-agent systems. The modeling of beliefs is usually done in terms of modal logics which are fairly costly to reason in. We introduce our novel, more practical approach to belief reasoning in a first-order back-chaining system such as Prolog. At the end of the report, we justify the approach by relating some of its properties to those of weak-S5, a common modal belief logic, before concluding with a real-world example illustrating the approach. 1 Introduction Reasoning about other agents’ beliefs is widely recognized as an im- portant necessity in multi-agent systems. The motivations for such a capability are fairly obvious. Knowing or reasoning about others’ beliefs can benefit an agent both in competitive and cooperative envi- ronments. In competitive environments, knowledge about the beliefs of competitors enables an agent to anticipate their actions, which, in turn, helps it in determining its own actions. In a cooperative en- vironment, anticipating teammates’ actions can save communication time and enhance collaboration [5, 8].